Citation
Metapopulation dynamics and landscape ecology of the Florida scrub-jay, Aphelocoma coerulescens

Material Information

Title:
Metapopulation dynamics and landscape ecology of the Florida scrub-jay, Aphelocoma coerulescens
Creator:
Stith, Bradley M., 1956-
Publication Date:
Language:
English
Physical Description:
xxv, 383 leaves : ill. ; 29 cm.

Subjects

Subjects / Keywords:
Counties ( jstor )
Demography ( jstor )
Habitat conservation ( jstor )
Metapopulation ecology ( jstor )
Polygons ( jstor )
Population estimates ( jstor )
Population size ( jstor )
Protected areas ( jstor )
Simulations ( jstor )
Statistics ( jstor )
Dissertations, Academic -- Wildlife Ecology and Conservation -- UF ( lcsh )
Scrub jay -- Ecology -- Florida ( lcsh )
Scrub jay -- Geographical distribution -- Florida ( lcsh )
Wildlife Ecology and Conservation thesis, Ph. D ( lcsh )
Genre:
bibliography ( marcgt )
non-fiction ( marcgt )

Notes

Thesis:
Thesis (Ph. D.)--University of Florida, 1999.
Bibliography:
Includes bibliographical references (leaves 372-382).
General Note:
Printout.
General Note:
Vita.
Statement of Responsibility:
by Bradley M. Stith.

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
The University of Florida George A. Smathers Libraries respect the intellectual property rights of others and do not claim any copyright interest in this item. This item may be protected by copyright but is made available here under a claim of fair use (17 U.S.C. §107) for non-profit research and educational purposes. Users of this work have responsibility for determining copyright status prior to reusing, publishing or reproducing this item for purposes other than what is allowed by fair use or other copyright exemptions. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder. The Smathers Libraries would like to learn more about this item and invite individuals or organizations to contact the RDS coordinator (ufdissertations@uflib.ufl.edu) with any additional information they can provide.
Resource Identifier:
021580684 ( ALEPH )
43707737 ( OCLC )

Downloads

This item has the following downloads:


Full Text










METAPOPULATION DYNAMICS AND LANDSCAPE ECOLOGY OF THE
FLORIDA SCRUB-JAY, APHELOCOMA COERULESCENS











By

BRADLEY M. STITH













A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

1999

























Dedicated to my wife and best friend, Ellen Mary Thorns














ACKNOWLEDGMENTS

I wish to acknowledge each of my committee members for their unique

contributions to this dissertation. My chairman, Dr. Stephen R. Humphrey, stimulated my

interests in issues relating to philosophy, policy, and management. His broad interests

and ability to take on multiple careers were remarkable. In my times of need he provided

unfailing support while allowing me great freedom to pursue my own course of study. As

director of Archbold Biological Station, co-chairman Dr. John W. Fitzpatrick provided

me with a wonderful opportunity to study the Florida Scrub-Jay at this world-class

research facility. His ability to perform brilliantly as director, researcher, and

conservationist is inspirational. I am grateful to him for sending a generous U.S.F.W.S.

grant my way to support me towards the end of my program. I was also privileged to

work under the tutelage of the legendary pioneer of Florida Scrub-Jay research, Dr. Glen

E. Woolfenden, who showed me the world from a Florida Scrub-Jay's perspective. My

collaborative research with Dr. Lyn Branch on the Vizcacha and the Florida scrub lizard,

was a joy. She forced me to think about landscape ecology and the metapopulation

dynamics of "lesser" organisms. Her support, financial and otherwise, was extraordinary,

and a nicer person I have never met. Dr. Jon Allen exposed me to population modeling

from an entomologists perspective. His enthusiasm for teaching and helping graduate

students solve technical problems was exceptional.

The staff at Archbold Biological Station made my stay there rewarding beyond

words. Working with Dr. Reed Bowman at the Avon Park Bombing Range, and learning


iii








about his suburban Florida Scrub-Jays, was most stimulating. His invitation to participate

in the Brevard county Habitat Conservation Plan scientific committee was most

appreciated. Special thanks go to Steve Friedman and Roberta Pickert for their help with

GIS problems. Dan Childs, former manager of the affiliated MacArthur Agricultural

Research Center, and his staff provided unfailing assistance in many ways. Current

Archbold director Dr. Hilary Swain provided generous travel support for the 1997 AOU

meeting.

For collecting the bulk of the demographic data used in chapter 3, I am indebted

to Reed Bowman, Doug Stotz, Larry Riopelle, Natalie Hamel, and Mike McMillan. I

thank the staff at the Natural Resources Office at the Avon Park Air Force Range,

especially Bob Progulske, Paul Ebersbach, and Pat Walsh for their support.

Dave McDonald was instrumental in recruiting me to work on the statewide jay

survey and exposing me to the opportunities at Archbold. Dr. Keith Tarvin and Dr. Curt

Atkinson kindly lent me radiotelemetry equipment and provided much valuable advice.

Special thanks go to Dr. Ron Mumme who kindly allowed me to use color banded jays

from the south tract of Archbold which he has been monitoring for many years. His prior

work in banding, taming, and sexing these jays was critical to identifying candidates for

the displacement experiment. Bill Pranty worked tirelessly with me to digitize major

portions of the statewide survey. Steve Schoech provided occasional field support and

many hours of entertainment on the Archbold tennis court.

For stimulating discussions about jays and landscape rules, I am indebted to many

colleagues engaged in the study of Florida Scrub-Jays and their conservation. Participants

in the Habitat Conservation Planning group whom I have not yet mentioned include




iv








David Breininger, Grace Iverson, Michael O'Connell, Parks Small, Jon Thaxton, and

Brian Toland. The following people contributed data for the statewide survey: Reed

Bowman, Dave Breininger, Jack Dozier, Florida Game and Fresh Water Fish

Commission personnel, Grace Iverson, David McDonald, Ron Mumme, Ocala National

Forest personnel, Bill Pranty, Hilary Swain, Jon Thaxton, and Brian Toland. I thank

David Wesley and Dawn Zattau, of the U.S. Fish and Wildlife Service, who provided

lead funding for the statewide survey and helped stimulate our discussions of habitat

conservation planning.

For special assistance with site-specific questions related to the maps in chapter 5:

Mary Barnwell, Jim Beever, Reed Bowman, Dave Breininger, Mike Eagen, Mary

Huffman, Grace Iverson, Mike Jennings, Laura Lowry, Dan Pearson, Gary Popotnik, Bill

Pranty, Park Smalls, Hank Smith, Jon Thaxton, Brian Toland, Jane Tutin. For assistance

with GIS data: Reed Bowman, Dave Breininger, Kathy Bronson, Beth Needham, Bill

Pranty, Roberta Pickert, Katy NeSmith. For advice regarding modeling: Reed Bowman,

Dave Breininger, John Fitzpatrick, Glen Woolfenden. Bill Pranty deserves special thanks

for the exceptionally detailed information he has collated on scrub-jay locations around

the state (Pranty et al. manuscript). Input from members of the Recovery Team was most

helpful. Thanks go to the U.S.F.W.S. for funding the research in chapter 5, and especially

to Dawn Zattau for her support and patience.

I thank the support staff at the University of Florida Department of Wildlife

Ecology and Conservation. Joe Gasper provided extraordinary computer support.

Leonard Pearlstein and the entire USFWS coop unit helped with innumerable computer

problems. I thank support staff at Circa computing, especially Jiannong Xin for




v








programming resources, and John Dixon for statistical help. Dr. Ken Portier also

provided valuable statistical advice.

Last, but no least, I thank my family and especially my wife, Ellen Thorns.













































vi














TABLE OF CONTENTS


ag.e


ACKNOW LEDGM ENTS ........................................................................................... iii

LIST OF TABLES ...................................................................................................... xii

LIST OF FIGURES .................................................................................................... xv

ABSTRACT........................................................................................................... xxivv

CHAPTERS

1 INTRODUCTION ..................................................................................................... 1

Historical Background ............................................................................................ 1
Biological Background ........................................................................................... 3
Objectives...................................................................................................................... 4

2 CLASSIFYING FLORIDA SCRUB-JAY METAPOPULATIONS............................. 9

Introduction............................................................................................................. 9
Statewide Survey of the Florida Scrub-Jay......................................................... 11
Statewide Survey: M ethods ................................................ .......................... 11
Statewide Survey: Results................................................... .......................... 13
A M ethod for Classifying M etapopulations....................... ................................ 14
M etapopulation Structure of the Florida Scrub-Jay....................................... ....... 18
Dispersal Distances......................................................................................... 19
Patch Occupancy............................................................................................. 20
Population Viability Analysis................................................. ...................... 22
M etapopulation Structure................................................. ............................. 23
Caveats............................................................................................................ 25

3 REMOTE SENSING OF FLORIDA SCRUB-JAY HABITAT.................................. 40

Introduction........................................................................................................... 40
M ethods....................................................................................................................... 42
Image Source................................................................................................... 42
Image Scanning and Conversion.......................................... ......................... 43
Image Rectification and M osaicing ................................................ ....... .... 43


vii









Image Classification........................................................................................ 44
M anual Editing of Classification .................................................. .............. 45
Assessment of Classification Accuracy .......................................... ........... ... 46
Digitization of Territories and Background Features ......................................... 47
Tree Cover Buffering Procedure......................................... .......................... 47
Habitat Quality M odel ...................................................... ............................ 48
Collection of Demographic Data .................................................. .............. 49
Habitat-Demographic Analysis....................................................... ........... ... 49
Results................................................................................ ........................................ 50
Discussion ............................................................................................................. 54

4 MODELING DISPERSAL IN THE FLORIDA SCRUB-JAY..................................... 86

Introduction............................................................ ..................................................... 86
Dispersal Strategies......................................................................................... 89
Dispersal Traits of the Florida and W estern Scrub-Jay ................................ .... 90
M ethods....................................................................................................................... 92
General Approach ........................................................................................... 92
GIS Files ......................................................................................................... 93
Simulating Philopatric Dispersal .................................................. .............. 93
Short distance dispersal algorithm development and calibration................ 94
Simulating Long Distance Dispersal........................................................... 95
Estimating floater mortality and mobility................................... ........... . 96
Habitat attractiveness.................................................... .......................... 97
Floater detection radius............................................................. ........... .. 97
Estimating floater frequency............................................. ...................... 98
Floater algorithm development and calibration ............................................ 102
Jay displacement experiment...................................................................... 103
Constraint Analysis........................................................................................... 106
M odel Validation ........................................................................................... 108
Results....................................................................................................................... 108
Radiotelemetry Displacement Experiment ......................................................... 108
Floater Parameter Estimation............................................ .......................... 111
Constraint Analysis........................................................................................... 112
Calibration and Validation................................................................................ 112
Discussion............................................................................... ............................... 113

5 METAPOPULATION VIABILITY ANALYSIS OF THE FLORIDA SCRUB-
JAY ........................................................................................................................... 132

Introduction and Objectives................................................................................ 132
M ethods................................................................................................................... 135
Simulation M odel Description............................................................ ........... 135
Life Stages........................................................................................................... 136
Starting Population Stage Structure ............................................................... 137
Annual Life Cycle ............................................................................................ 137
Territories........................ .............................................................................. 140


viii









Background Landscape Im age ............................................................................ 141
M ap Production................................................................................................... 143
Statew ide m etapopulation m ap................................................................... 143
1992-1993 SM P m aps................................................................................... 143
Acquisition m aps .................................................................................... 144
GIS Database Preparation................................................................................... 146
Estim ation ofjay populations after restoration............................................. 146
Identification of protected areas................................................................... 147
Assessm ent of unprotected areas ................................................................ 147
Suburban jays ................................................................................................. 148
Sim ulation runs ................................................................................................... 149
Repetitions and duration of sim ulations ..................................................... 149
Reserve design configurations .................................................................... 150
Output statistics............................................................................................. 151
M odel Validation/Calibration ............................................................................. 152
Interpreting Sim ulation Results ........................................................................ 152
Results........................................ ............................................................................... 154
Levy (Cedar Key) (M l) ...................................................................................... 155
Citrus-S.W . M arion (M 2) ................................................................................... 164
Pasco-H em ando (M 3)......................................................................................... 176
M anatee-S. Hillsborough (M 4) ......................................................................... 186
Sarasota-W . Charlotte (M 5) ............................................................................... 196
N . W . Charlotte (M 6)........................................................................................ 204
Central Charlotte (M 7)........................................................................................ 212
Lee and N . Collier (M 8)...................................................................................... 220
Flagler-N .E. Volusia (M 9)................................................................................. 228
M erritt Island-S.E. Volusia and (M 10)............................................................. 236
N . Brevard (M 11)................................................................................................ 244
Central Brevard (M 12)........................................................................................ 253
S. Brevard-Indian River-N . St. Lucie (M 13).................................................... 261
St. Lucie - N . M artin (M 14)......... .. ................................................................... 270
M artin and N . Palm Beach (M 15)....................................................................... 278
South Palm Beach (M 16).................................................................................... 286
Ocala N ational Forest (M 17) .............................................................................. 294
N .E. Lake (M 18).............................................................................................. 302
S.W . Volusia (M 19).......................................................................................... 310
Central Lake (M 20)........................................................................................... 318
Lake W ales Ridge (M 21)............................................................................... 326
Other M etapopulations.................. ................................................................ 344
Brevard barrier island............................................................................... 344
Clay county .............................................................................................. 344
Osceola............................................... ................................................... 344
W estern Polk............................................................................................ 345
Bright Hour Ranch ................................................................................... 345
Recom m endations................................................................................................. .. 346
Ranking M etapopulation Vulnerability ........................................................ 346



ix









Sum m ary of Recom m endations ........................................................................ 349
D iscussion........................................................................................................... 359

6 SYN TH ESIS ................................................................................................................ 365

REFEREN CES ......................................................................................................... 372

BIO G RA PH ICA L SK ETCH .................................................................................... 383
















































x














LIST OF TABLES

Table page

2-1. Summary information for 42 Florida Scrub-Jay metapopulations, including
type of metapopulation, number of territories (pairs ofjays), and number of
subpopulations. ................................................................................................... 28

3-1. Demographic and habitat parameters for North and South Sandy Hill (1994
- 1995)...................................................................................................................... 8 1

3-2. Kolmogorov-Smimov test for normality of demographic and habitat
variables (* significantly different from normal).............................. ........... .. 84

3-3. Mann-Whitney_U test for differences in demographic and habitat variables
between North and South jay populations (* significantly)................................ 85

4-1. Landcover types from statewide habitat map (Kautz et al. 1993) used in
simulations and associated floater attractiveness values......................................... 120

4-3. Summary ofphilopatric dispersal rules showing sex differences and rules
used to implement the algorithm............. .................................................................. 122

4-4. Summary ofjay movement data obtained from displacement experiment
(distances in km). .............................................................................................. 123

4-5. Summary of constraint analysis for 9 simulation scenarios (50 years x 30
repetitions) showing number of colonizations from Lake Wales Ridge to
Bright Hour Ranch, DeSoto county, Florida......................................................... 124

5-1. Demographic and dispersal parameter settings for jays in optimal and
suburban conditions. ............................................................................................... 142

5-la. Levy county patch statistics (number of jay territories for different
configurations)..................................................................................................... 160

5-1b. Levy county (Cedar Key) simulation statistics................................................... 163

5-2a. Citrus and S. Marion county patch statistics (number of jay territories for
different configurations) .................................................................................... 172



xi








5-2b. Citrus and S. Marion county simulation statistics................................................. 175

5-3a. Pasco and Hemando county patch statistics (number ofjay territories for
different configurations) .................................................................................... 182

5-3b. Pasco county simulation statistics ....................................................................... 185

5-4a. Manatee and S. Hillsborough county patch statistics (number of jay
territories for different configurations) ................................................................. 191

5-4b. Manatee and S. Hillsborough county simulation statistics.................................. 195

5-5a. Sarasota and W. Charlotte county patch statistics (number of jay territories
for different configurations)............................................................................... 200

5-5b. Sarasota and W. Charlotte county simulation statistics....................................... 203

5-6a. N. W. Charlotte county patch statistics (number ofjay territories for
different configurations) ........................................................................................ 208

5-6b. N. W. Charlotte county simulation statistics ....................................................... 211

5-7a. Central Charlotte county patch statistics (number ofjay territories for
different configurations) ........................................................................................ 216

5-7b. Central Charlotte county simulation statistics.................................................. 219

5-8a. Lee and N. Collier county patch statistics (number ofjay territories for
different configurations) ......................................................................................... 224

5-8b. Lee and N. Collier county simulation statistics................................................... 227

5-9a. Flagler and N.E. Volusia county patch statistics (number of jay territories
for different configurations)..................................................................................... 232

5-9b. Flagler and N.E. Volusia county simulation statistics........................................... 235

5-10a. S.E. Volusia and Merritt Island county patch statistics (number ofjay
territories for different configurations) ................................................................. 240

5-10b. S.E. Volusia and Merritt Island county simulation statistics............................. 243

5-1 la. N. Brevard county patch statistics (number ofjay territories for different
configurations)..................................................................................................... 249

5-11 b. N. Brevard county simulation statistics............................................................. 252



xii










5-12a. Central Brevard county patch statistics (number ofjay territories for
different configurations) ..................................................................................... 257

5-12b. Central Brevard county simulation statistics.................................................... 260

5-13a. S. Brevard-Indian River-N. St. Lucie county patch statistics (number ofjay
territories for different configurations) ................................................................. 266

5-13b. S. Brevard-Indian River-N. St. Lucie county simulation statistics ................... 269

5-14a. St. Lucie - N. Martin county patch statistics (number of jay territories for
different configurations) ........................................................................................ 274

5-14b. St. Lucie county simulation statistics ................................................................ 277

5-15a. Martin and N. Palm Beach county patch statistics (number ofjay territories
for different configurations).................................................................................... 282

5-15b. Martin and N. Palm Beach county simulation statistics...................................... 285

5-16a. South Palm Beach county patch statistics (number ofjay territories for
different configurations) ........................................................................................ 290

5-16b. South Palm Beach county simulation statistics................................................... 293

5-17a. Ocala National Forest county patch statistics (number of jay territories for
different configurations) ........................................................................................ 298

5-17b. Ocala National Forest county simulation statistics...... ........................................ 301

5-18a. N.E. Lake county patch statistics (number ofjay territories for different
configurations) ........................................................................................................ 306

5-18b. N.E. Lake county simulation statistics .............................................................. 309

5-19a. S.W. Volusia county patch statistics (number ofjay territories for different
configurations) ........................................................................................................ 314

5-19b. S.W. Volusia county simulation statistics ......................................................... 317

5-20a. Central Lake county patch statistics (number ofjay territories for different
configurations)..................................................................................................... 322

5-20b. Central Lake county simulation statistics.......................................................... 325




xiii









5-21a. Lake Wales Ridge patch statistics (number ofjay territories for different
configurations) ........................................................................................................ 339

5-21b. Lake Wales Ridge simulation statistics............................................................. 343

5-22. Metapopulation viability statistics....................................................................... 350

5-23. Metapopulation vulnerability ranking - "no acquisition" (sorted by
decreasing quasi-extinction probability)............................................................ 351

5-23a. Metapopulation vulnerability ranking - "maximum acquisition" (sorted by
increasing percent protection)............................................................................... 352

5-24. Percent protected ranking - (sorted by increasing percent protection)................. 353

5-25. Metapopulation priority ranking (sorted by decreasing priority)........................ 354

5-26. Summary of recommendations (highest priority first) ........................................ 355

































xiv















LIST OF FIGURES

Figure page

2-1. 1993 distribution of Florida Scrub Jay groups (small black circles). Note the
discontinuous distribution and variability in patterns of aggregation................... 30

2-2. Classification scheme showing different types of metapopulations based on
patch size distribution (patches all small in size, mixture of small and large,
and all large in size) along the horizontal axis, and degree of patch isolation
(highly connected to highly isolated) on the vertical axis. Nonequilibrium,
classical, mainland-island, and patchy classes are named according to
H arrison (1991)................................................................................................... 31

2-3. Schematic depiction of different kinds of metapopulations, illustrating use of
dispersal-distance buffers to predict recolonization rates among
subpopulations. Dotted lines separate functional subpopulations, based on
frequency of dispersal beyond them. Solid lines separate metapopulations,
based on poor likelihood of dispersal among them. A. Patchy
metapopulation. B. Classical metapopulation. C. Nonequilibrium
metapopulations. D. Mainland-island metapopulation........................................... 32

2-4. Dispersal frequency curve. Dispersal distances from natal to breeding
territories for color-banded jays at Archbold Biological Station, 1970-1993.
About 85% of documented dispersals were within 3.5 km, and 99% within
8.3 km. The longest documented dispersal was 35 km............................................. 33

2-5. Proportion of suitable habitat patches occupied by Florida Scrub-Jays as a
function of their distance to the nearest separate patch of occupied habitat.
Occupancy rates are high (nearly 90 %) for patches up to 2 km apart and
decline monotonically to 12 km. Note the scale change after 16 km................... 34

2-6. Statewide jay distribution map with dispersal buffers. Shaded areas with thin,
solid lines depict subpopulations of jays within easy dispersal distance (3.5
km) of one another. Thick lines delineate demographically independent
metapopulations separated from each other by at least 12 km. ............................... 35

2-7. Frequency of Florida Scrub-Jay metapopulation sizes. Note that 21
metapopulations have 10 pairs or less of jays. These represent
nonequilibrium metapopulations. .................................. ............... ........... ...... 36


xv








2-8. Examples of non-equilibrium metapopulations from North Gulf coast. Each
of the six metapopulations contains fewer than 10 pairs ofjays, except for
the centrally located system that contains a single, midland-size
subpopulation........................................................................................................ 37

2-9. Example of a "classical" metapopulation from five counties in Central
Florida. Note the occurrence of jays in small islands of intermediate distance
from one another. ................................................................................................ 38

2-10. Portion of the largest mainland-midland-island metapopulation in the
interior, consisting of the Lake Wales Ridge and associated smaller sand
deposits. The large central subpopulation (enclosed by the thin black line)
contains nearly 800 pairs of jays. Small subpopulations to the south and east
are within known dispersal distance of the large, central mainland. A small
metapopulation to the west (in DeSoto County) contains a single
subpopulation of 21 territories. This small system qualifies as a patchy
metapopulation, since jays occur in two or more patches but the patches are
so close together that they function as a single demographic unit........................... 39

3-1. Map of Scrub-Jay Territories - Spring 1994. Dividing line between "North"
and "South" populations is the Kissimmee Rd......................................................... 60

3-2. Habitat Suitability Index graphs for percent bare sand, percent tree cover, and
distance to forest (modified from Duncan et al. 1995)........................................... 61

3-3. Illustrative map of 100, 200, and 400 m buffer zones around LOST territory.......... 62

3-4. Accuracy assessment correlation graph for bare sand based on quadrat vs.
classified image measurements (r-squared = 0.60).............................................. 63

3-5. Accuracy assessment correlation graph for tree cover based on transect vs.
classified image measurements (r-squared = 0.25).............................. ........ . 64

3-6. Measurements of sand, tree cover, and mixed vegetation for Spring 1994
territories on N. Sandy Hill.................................................. ........................... 65

3-7. Measurements of sand, tree cover, and mixed vegetation for Spring 1994
territories on S. Sandy Hill................................................ .................................... 66

3-8. Percent tree cover (mean and standard deviation) for 4 zones (inside
territories, 100, 200, 400 m buffer) - North vs. South Sandy Hill. South
population shows significantly higher tree cover within all zones compared
to N orth population............................................................................................... . 67

3-9. Percent tree cover for individual territories for 4 zones (inside territories, 100,
200, 400 m buffer) - North territories............................................. ............ .... 68



xvi









3-10. Percent tree cover for individual territories for 4 zones (inside territories,
100, 200, 400 m buffer) - South territories ............................................................. 69

3-11. Tree cover within jay territories vs. total tree cover in North vs. South
Sandy Hill. Note that jays select habitat with lower tree cover in both areas.......... 70

3-12. Percent bare sand (mean and standard deviation) for 4 zones (inside
territories, 100, 200, 400 m buffer) - North vs. South Sandy Hill.
Differences between two areas are not significant........................... ............. .. 71

3-13. Group size vs. percent tree cover within all territories (North and South
populations pooled). Trend towards smaller group size with higher tree
cover is not significant. ....................................................................................... 72

3-14. Group size vs. percent tree cover within 100 m buffer for all territories
(North and South populations pooled). Trend towards smaller group size
with higher tree cover is not significant.......................... ............. ............ ... 73

3-15. Group size (small = 2, medium = 3, large = 4 - 7 jays) vs. percent tree cover
within all territories (North and South populations pooled). Trend towards
smaller group size with higher tree cover is not significant. ................................... 74

3-16. Group size (small = 2, medium = 3, large = 4 - 7 jays) vs. percent tree cover
within 100 m buffer (North and South populations pooled). Trend towards
smaller group size with higher tree cover is not significant ................................. 75

3-17. Images and territories (black polygons) of North Sandy Hill. Right: color-
infrared image. Left: classified image (white = bare sand; green = trees;
brown = shrubs/grass; black = water)............................................. ............ .... 76

3-18. Images and territories (black polygons) of N. portion of South Sandy Hill.
Right: color-infrared image. Left: classified image (white = bare sand;
green = trees; brown = shrubs/grass; black = water) .............................................. 77

3-19. Images and territories (black polygons) of S. portion of South Sandy Hill.
Right: color-infrared image. Left: classified image (white = bare sand;
green = trees; brown = shrubs/grass; black = water) .............................................. 78

3-20. Habitat quality map of N. portion of South Sandy Hill......................................... 79

3-21. Habitat quality map of S. portion of South Sandy Hill.......................................... 80

4-1. Daily distances moved and number of days movements were tracked for 10
jays released at 3 sites in Highlands county, Florida.............................................. 125




xvii








4-2. Kaplan-meier survival curves of daily survival rates for 10 jays released at 3
sites in Highlands county, Florida. Upper curve: maximum possible
survival; middle curve: "best guess" survival; lower curve: minimum
possible survival............................................................................................... 126

4-3. Distribution of daily distances moved by released jays (solid line), and
inverse function fitted to observed movements (dashed line)............................... 127

4-4. Comparison of dispersal data from Archbold Biological Station and
simulated dispersal distances for male jays .......................................................... 128

4-5. Comparison of dispersal data from Archbold Biological Station and
simulated dispersal distances for female jays......................................................... 129

4-6. Comparison of stage-age data from Archbold Biological Station and
simulated stage-age data for breeders................................................................... 130

4-7. Comparison of stage-age data from Archbold Biological Station and
simulated stage-age data for helpers.......................... ................................................ 131

5-0. Delineations of 21 Florida Scrub-Jay metapopulations based on 1992 - 1993
statew ide survey............................................................................................... 145

5-la. Levy county maps - 1992 - 1993 jay and habitat distribution............................. 158

5-lb. Levy county acquisition map............................................................................... 159

5-1c. Levy county trajectory graphs. Top) no acquisition, Bottom) maximum
acquisition. ........................................................................................................ 161

5-id. Levy county quasi-extinction graphs. Top) no acquisition, Bottom)
maxim um acquisition.............................................................................................. 162

5-2a. Citrus county map - 1992-1993 jay and habitat distribution............................... 168

5-2d. S.W. Marion county acquisition map.................................................................. 171

5-2e. Citrus and S. Marion county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition....................................................................................... 173

5-2f. Citrus and S. Marion county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition .......................................................................... 174

5-3a. W. Pasco and Hernando county map - 1992 - 1993 jay and habitat
distribution. ....................................................................................................... 178




xviii








5-3b. E. Pasco and Hernando county map - 1992 - 1993 jay and habitat
distribution. ....................................................................................................... 179

5-3c. W. Pasco and Hemando county acquisition map.............................................. 180

5-3d. E. Pasco and Hemando county acquisition map.................................................. 181

5-3e. Pasco and Hemando county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition ....................................................................................... 183

5-3f. Pasco and Hemando county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition. ............................................................................ 184

5-4a. Manatee and S. Hillsborough county map - 1992 - 1993 jay and habitat
distribution. ....................................................................................................... 189

5-4b. Manatee and S. Hillsborough county acquisition map.......................................... 190

5-4c. Manatee and S. Hillsborough county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition. ............................................................................ 193

5-4d. Manatee and S. Hillsborough county quasi-extinction graphs. Top) no
acquisition, Bottom) maximum acquisition.......................................................... 194

5-5a. Sarasota and W. Charlotte county map - 1992 - 1993 jay and habitat
distribution. ....................................................................................................... 198

5-5b. Sarasota and W. Charlotte county acquisition map.............................. .............. 199

5-5c. Sarasota and W. Charlotte county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition. ............................................................................ 201

5-5d. Sarasota and W. Charlotte quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition............................................................................... 202

5-6a. N. W. Charlotte county map - 1992 - 1993 jay and habitat distribution .............. 206

5-6b. N. W. Charlotte county acquisition map. ............................................................ 207

5-6c. N. W. Charlotte county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition........................................................................................... 209

5-6d. N. W. Charlotte county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition. ............................................................................ 210

5-7a. Central Charlotte county map - 1992 - 1993 jay and habitat distribution........... 214



xix










5-7b. Central Charlotte county acquisition map ........................................................ 215

5-7c. Central Charlotte county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.......................................................... ................................ 217

5-7d. Central Charlotte county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition. ............................................................................ 218

5-8a. Lee and N. Collier county map - 1992 - 1993 jay and habitat distribution .......... 222

5-8b. Lee and N. Collier county acquisition map......................................................... 223

5-8c. Lee and N. Collier county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition....................................................................................... 225

5-8d. Lee and N. Collier county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition............................................................................... 226

5-9a. Flagler and N.E. Volusia county map - 1992 - 1993 jay and habitat
distribution. ....................................................................................................... 230

5-9b. Flagler and N.E. Volusia county acquisition map................................................. 231

5-9c. Flagler and N.E. Volusia county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition. ............................................................................ 233

5-9d. Flagler and N.E. Volusia county quasi-extinction graphs. Top) no
acquisition, Bottom) maximum acquisition.......................................................... 234

5-10a. Merritt Island and S.E. Volusia county map - 1992 - 1993 jay and habitat
distribution. ....................................................................................................... 238

5-10b. Merritt Island and S.E. Volusia county acquisition map................................... 239

5-10c. S.E. Volusia and Merritt Island county trajectory graphs. Top) no
acquisition, Bottom) maximum acquisition.......................................................... 241

5-10d. S.E. Volusia and Merritt Island county quasi-extinction graphs. Top) no
acquisition, Bottom) maximum acquisition.......................................................... 242

5-1 la. N. Brevard county map - 1992 - 1993 jay and habitat distribution................... 247

5-1 lb. N. Brevard county acquisition map................................................................... 248





xx








5-1 Ic. N. Brevard county trajectory graphs. Top) 30% acquisition, Bottom) 70%
acquisition. ........................................................................................................ 250

5-1 Id. N. Brevard county quasi-extinction graphs. Top) 30% acquisition, Bottom)
70% acquisition.................................................................................................... 251

5-12a. Central Brevard county map - 1992 - 1993 jay and habitat distribution .......... 255

5-12b. Central Brevard county acquisition map ........................................................... 256

5-12c. Central Brevard county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition............................................................................................ 258

5-12d. Central Brevard county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition. .......................................................................... 259

5-13a. S. Brevard-Indian River-N. St. Lucie Metapopulation county map - 1992 -
1993 jay and habitat distribution............................................................................. 264

5-13b. S. Brevard-Indian River-N. St. Lucie county acquisition map.......................... 265

5-13c. S. Brevard-Indian River-N. St. Lucie county trajectory graphs. Top) no
acquisition, Bottom) 30% acquisition by area ...................................................... 267

5-13d. S. Brevard-Indian River-N. St. Lucie county quasi-extinction graphs. Top)
no acquisition, Bottom) 30% acquisition by area................................................. 268

5-14a. St. Lucie - N. Martin county map - 1992 - 1993 jay and habitat
distribution. ....................................................................................................... 272

5-14b. St. Lucie - N. Martin county acquisition map ................................... ......... 273

5-14c. St. Lucie - N. Martin county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition. ............................................................................ 275

5-14d. St. Lucie - N. Martin county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition. ............................................................................ 276

5-15a. Martin and N. Palm Beach county map - 1992 - 1993 jay and habitat
distribution .......................................................................................................... 280

5-15b. Martin and N. Palm Beach county acquisition map.......................................... 281

5-15c. Martin and N. Palm Beach county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition. ............................................................................ 283




xxi








5-15d. Martin and N. Palm Beach county quasi-extinction graphs. Top) no
acquisition, Bottom) maximum acquisition.......................................................... 284

Fig 5-16a. Central Palm Beach county map - 1992 - 1993 jay and habitat
distribution ............................................................................................................. 288

5-16b. Central Palm Beach county acquisition map..................................................... 289

5-16c. South Palm Beach county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.............................................................................................. 291

5-16d. South Palm Beach county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition. ............................................................................ 292

5-17a. Ocala National Forest county map - 1992 - 1993 jay and habitat
distribution. ............................................................................................................. 296

5-17b. Ocala National Forest county acquisition map.................................................. 297

5-17c. Ocala National Forest county trajectory graphs. No acquisition........................ 299

5-17d. Ocala National Forest county quasi-extinction graphs. No acquisition ........... 300

5-18a. N.E. Lake county map - 1992 - 1993 jay and habitat distribution.................... 304

5-18b. N.E. Lake county acquisition map..................................................................... 305

5-18c. N.E. Lake county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.............................................................................................. 307

5-18d. N.E. Lake county quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition............................................................................................ 308

5-19a. S.W. Volusia county map - 1992 - 1993 jay and habitat distribution................ 312

5-19b. S.W. Volusia county acquisition map............................................................. 313

5-19c. S.W. Volusia county trajectory graphs. Top) no acquisition, Bottom)
m axim um acquisition.............................................................................................. 315

5-19d. S.W. Volusia county quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition........................................................................................ 316

5-20a. Central Lake county map - 1992 - 1993 jay and habitat distribution................ 320

5-20b. Central Lake county acquisition map................................................................ 321



y.xii









5-20c. Central Lake county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition....................................................................................... 323

5-20d. Central Lake county quasi-extinction graphs. Top) no acquisition, Bottom)
m axim um acquisition.............................................................................................. 324

5-21b. Lake Wales Ridge map - 1992 - 1993 jay and habitat distribution, Glades
County .............................................................................................................. 329

5-21e. Lake Wales Ridge map - 1992 - 1993 jay and habitat distribution, S.
central Polk county.................................................................................................. 332

5-21f. Lake Wales Ridge map - 1992 - 1993 jay and habitat distribution, N.E.
Polk and N.W. Osceola county............................................................................. 333

5-21h. Lake Wales Ridge acquisition map, S. Highlands county................................. 335

5-21i. Lake Wales Ridge acquisition map, N. Highlands and S. Polk county.............. 336

5-21j. Lake Wales Ridge acquisition map, S. central Polk county............................... 337

5-21k. Lake Wales Ridge acquisition map, N.E. Polk and N.W. Osceola county. ........ 338

5-211. Lake Wales Ridge trajectory graphs. Top) no acquisition, Bottom)
m axim um acquisition.............................................................................................. 341

5-21m. Lake Wales Ridge quasi-extinction graphs. Top) no acquisition, Bottom)
m axim um acquisition......................................................................................... 342




















xxiii














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy





METAPOPULATION DYNAMICS AND LANDSCAPE ECOLOGY OF THE
FLORIDA SCRUB-JAY, APHELOCOMA COERULESCENS

By

Bradley M. Stith

December 1999

Chairman: Dr. Stephen R. Humphrey
Major Department: Wildlife Ecology and Conservation


Florida's only endemic bird species, the Florida Scrub-Jay (Aphelocoma

coerulescens), is rapidly disappearing throughout much of its range. A 1992-1993

statewide survey shows that it has effectively gone extinct in 10 of 39 formerly occupied

counties in less than two decades. To characterize the spatial structure and vulnerability

of the Florida Scrub-Jay throughout the state, I developed and applied a new method to

describe the species' metapopulation structure. This method uses GIS-generated buffers

based on documented dispersal distances to identify separate metapopulations and highly

connected subpopulations called mainlands (extinction resistant), islands (extinction

prone), or midlands (vulnerable to extinction). Of the 42 jay metapopulations identified,

only five include mainlands; 21 consist only of extinction-prone islands. The resulting



xxiv






XXV


classification reveals key subpopulations requiring special attention to maintain the long-

term viability of the existing metapopulations.

I developed and applied a technique for measuring habitat features and estimating

habitat quality over large areas using image processing and GIS methods. The technique

showed that jays in central Florida had a strong preference for open sandy areas, and few

or no pine trees. A proximity analysis showed that demographic performance decreased

near forests. Measurement of habitat variables using this technique will be a valuable

technique for habitat management and conservation.

I developed a spatially explicit, individual-based model to simulate the

metapopulation dynamics of Florida Scrub-Jays. Special emphasis was placed on

realistically modeling dispersal. I conducted a small radio-tracking study and used data

from long-term studies to parameterize and validate the model. Stage-age structure and

dispersal distances generated by the model showed good fit to field data.

I used this simulation model to investigate the viability of 21 major Florida Scrub-

Jay metapopulations across the state. For each metapopulation I simulated 2 or more

hypothetical reserve designs, ranging from a minimal design with only currently

protected jays, to a maximal design containing all significant populations as of 1993. All

habitat was assumed to be restored and fully occupied. Model results indicated that only

3 of 21 metapopulations would be adequately protected without further habitat

acquisition. At least 4 metapopulations appear to be at great risk of extinction.














CHAPTER 1
INTRODUCTION


Historical Background


For most of geologic time, Florida lay under water, attached to what is now

northern Africa. With the breakup of the super-continent, Gondwanaland, in the

Mesozoic Florida rafted from Africa and become attached to North America. There she

remained under water throughout the entire Age of Dinosaurs and well into the Cenozoic.

Only recently, about 25 million years ago, Florida emerged from the sea. Never again

would she become completely inundated, despite huge global fluctuations in sea level.

She changed drastically in size and shape, however, due to the episodic transgressions

and regressions of the seas caused by the waxing and waning of great ice sheets on the

continents. As glaciers advanced southward, then retreated, they produced continental-

scale changes in the climate. During high sea levels, mesic forests in Florida prospered,

but with falling sea levels Florida became more and more desert-like. During such

periods, aid conditions prevailed across most of the southern continent, and fauna and

flora from the west were free to move to Florida along a Gulf coastal corridor. Among

these western immigrants was a xeric-adapted bird originating from a widespread species

known today as the Western Scrub-Jay (Aphelocoma californica). Exactly when this

species first arrived in Florida is uncertain, but it must have been at least several million

years ago. Having made it to Florida, this jay would become isolated from it's western

counterpart by the development of extensive wetlands associated with the Mississippi


1






2


delta during the mid-Pleistocene. As conditions became more mesic, xeric habitat became

reduced and isolated into desert-like islands within which a remarkable assemblage of

organisms evolved. Among these organisms that diverged from their western kin is the

Florida-Scrub Jay, the subject of this dissertation.

Not long after the most recent Wisconsin glacial retreat, a mere 12,000 years ago,

nomadic people known as the Clovis entered North America from Siberia and

encountered a continent largely or entirely devoid of humans. At that time the continent

teemed with giant animals such as mammoths, mastodons, ostrich-size flightless birds,

huge ground sloths, horses, camels, saber-toothed cats, and giant tortoises. This

megafaunal scene rivaled anything seen today in Africa. Within 1,000 to 2,000 years all

of these species and many more became extinct. Dozens of large vertebrates appear to

have made their last stand in Florida, their demise apparently coinciding with the arrival

of the Clovis. It remains uncertain whether humans were largely to blame for these

extinctions, as the megafauna also faced great changes in climate and landscape. Yet, it is

highly likely that humans contributed significantly to these megafaunal extinctions.

The Florida Scrub-Jay managed to survive this period of massive extinctions. But

by the mid-20t century, a new threat to the fauna and flora of Florida appeared.

Discoveries in applied sciences and engineering paved the way for the demise of the

Florida Scrub-Jay and its scrub habitat. Among these were the discovery that citrus could

be grown on the formerly worthless, sandy, infertile soils upon which Florida Scrub-Jay

habitat grew. The invention of air conditioning made tolerable the Florida summers,

ushering in an era of massive suburban sprawl, much of it devouring Florida Scrub-Jay

habitat.






3


In 1969, Glen E. Woolfenden turned his ornithological focus on the Florida

Scrub-Jay at the Archbold Biological Station. Thus began a continuous 31 year scientific

study of this single organism, making it one of the most thoroughly studied wild bird

species in the world. By 1975 his research on jays became a classic example of altruism

cited prominently in E.O. Wilson's (1975) influential Sociobiology. In the mid-1970s

Woolfenden teamed up with John W. Fitzpatrick, an ornithologist with a strong

background in population modeling. In 1984, they produced a highly-acclaimed

Princeton monograph (Woolfenden and Fitzpatrick, 1984) describing the demography

and cooperative behavior of this intriguing bird. Today, the number of publications on

this species approaches one hundred, and the Florida Scrub-Jay continues to be widely

cited as a classic example of altruism (e.g. Krebs and Davies 1998). No brief summary

can do justice to this large body of work. Nonetheless, a short review of the basic natural

history of the Florida Scrub-Jay follows (consult Woolfenden and Fitzpatrick 1996 for an

extensive list of references).

Biological Background


The Florida Scrub-Jay, Florida's only endemic bird species, is a disjunct, relict

taxon separated by more than 1600 km from its closest western relatives (Woolfenden

and Fitzpatrick 1984). This habitat specialist is restricted to a patchily distributed scrub

community found on sandy, infertile soils--mostly pre-Pleistocene and Pleistocene

shoreline deposits. The vegetation is dominated by several species of low-stature scrub

oaks (Quercus spp.). Jays rely heavily on acorns for food, especially during the winter,

when they retrieve thousands of acorns cached in open, sandy areas during the fall

(DeGange et al. 1989). Florida Scrub-Jays show a strong preference for low, open





4


habitats with numerous bare openings and few or no pine trees (Breininger et al. 1991).

These optimal habitat conditions are maintained by frequent fires (Abrahamson et al.

1984). Jays living in fire-suppressed, overgrown habitats have much poorer demographic

performance than jays in optimal conditions (Fitzpatrick and Woolfenden 1986), leading

rapidly to local extirpation unless the habitat is burned (Fitzpatrick et al. 1994).

Florida Scrub-Jays are monogamous, cooperative breeders that defend permanent

territories averaging 10 ha per family (Woolfenden and Fitzpatrick 1984). They have a

well-developed sentinel system, in which a family member watches for predators while

others in the group engage in other activities such as foraging (McGowan and

Woolfenden 1989). Young nearly always delay dispersal for at least one year, remaining

at home as helpers. Dispersal distances from natal to breeding territories are extremely

short for both sexes, and these movements within contiguous habitat average less than 1

territory for males and 3.5 territories for females (Woolfenden and Fitzpatrick 1984).

Dispersal behavior is associated with greatly elevated mortality even within optimal

habitat (Fitzpatrick and Woolfenden 1986), and many behavioral adaptations (e.g.,

cooperative breeding, sentinel system, delayed dispersal) suggest that predation is

extremely important to both resident and dispersing jays (Woolfenden and Fitzpatrick

1984; Fitzpatrick and Woolfenden 1986; Koenig et al. 1992).

Objectives


Despite the many studies and enormous amount of information available on

Florida Scrub-Jays, opportunities to contribution to this large body of knowledge are

afforded by new technologies, which allow new types of data to be collected and new

questions to be asked. I employ several of these new technologies in this dissertation,





5


including remote sensing, geographic information systems, radiotelemetry, and computer

modeling. Prior to the early 1990s, Florida Scrub-Jay research was not focused on

conservation issues. With the increasing destruction of scrub habitat, and the Federal

listing of the Florida Scrub-Jay as a threatened species in 1987, much more emphasis has

been placed recently on applied research, and a flurry of conservation-oriented

publications have since appeared. My hope is to fill in a few of the existing gaps.

Exemplifying a key goal of conservation biology, such efforts are made applicable to the

real world by the large body of basic knowledge already available for this species. This

prior body of knowledge also makes it possible to attempt to synthesize a variety of

information in the form of a spatially explicit computer model. More than half of my time

as a doctoral student has been spent developing this model.

Chapter 2 provides an analysis of the entire known distribution of the Florida

Scrub-Jay from a metapopulation perspective, based on a statewide census conducted in

1992-1993. A new classification technique is developed to describe the spatial structure

of the jay population. This technique is general in nature, and has implications for the

conservation of other species as well as the Florida Scrub-Jay. Portions of this chapter

have recently been published (Stith et al. 1996), and I wish to thank Island Press for

permission to include that publication in its entirety (with modifications).

Chapter 3 narrows the focus to a local level and examines the landscape ecology

of the Florida Scrub-Jay in central Florida. A team of researchers collected demographic

variables for color-banded jays at Avon Park Air Force Range (APAFR), in Highland and

Polk county. New computer technology--image processing and GIS--was used to

correlate physical features with habitat quality from a Florida Scrub-Jay's perspective in






6


this area. The relationship between demographic variables and the remotely-sensed

habitat variables were examined. At issue is the possibility of measuring habitat quality,

and potential demographic success remotely, and across large areas.

Chapter 4 examines the difficult subject of dispersal and describes an approach

used to simulate dispersal in an individual-based model (IBM) developed for this

dissertation. Two types of dispersal are simulated by the IBM. A close-distance dispersal

module mimics a "stay-home-and-foray" strategy that results in most dispersing jays

settling close to their natal territory. This module incorporates many details of Florida

Scrub-Jay biology documented by long-term color-band studies (Woolfenden and

Fitzpatrick 1984), including sex and age dominance relations. A long-distance dispersal

module simulates a "floater" strategy, which accounts for the infrequent, though

potentially important, tendency of some jays to abandon their natal territory and move

long distances, often between habitat patches and through hostile landscape matrices.

Empirical data on long-distance dispersal are poor, and a simple field experiment was

conducted with radiotelemetry to obtain information useful for modeling purposes. To

induce behavior that might be similar to long distance dispersal, radio-collared jays were

experimentally displaced kilometers away from their natal territories. Habitat

preferences, movement abilities, and mortality rates were recorded and incorporated into

the long-distance dispersal module. In combination, the close- and long-distance dispersal

modules produced a dispersal and stage-age curve that closely resembled results of long-

term data from Archbold Biological Station. A constraint analysis was used to place

plausible bounds on several of the poorly knowr long distance dispersal parameters. This





7


analysis relied on data that suggested where successful dispersal from Archbold

Biological Station could and could not take place.

Chapter 5 describes the complete individual-based, spatially explicit population

model, which incorporates the dispersal algorithms described in chapter 4. The model

provides a framework for integrating much of what is currently known about the Florida

Scrub-Jay. Simulations take place on a landscape provided by a geographic information

system (GIS) file. Non-dispersing jays occupy discrete territories. Both sexes are

modeled, and individual jays progress through 5 stages (juvenile, 1-year helper, older

helper, inexperienced and experienced breeder). Each territory has a separate set of

demographic parameters assigned to each sex and stage. Breeder experience and presence

of helpers may affect fecundity. Helpers monitor neighboring territories within their

"assessment sphere" and vie for breeder openings; the outcome of such competition is

determined by simple dominance rules. Helpers may leave on long distance dispersals,

during which time mortality and movement varies depending on landcover type.

The statewide population ofjays was divided into 21 metapopulations thought to

be demographically isolated from each other (fig. 5-0). Two series of maps were

developed for each metapopulation. One map type depicts jays and habitat as mapped in

1992-1993. A second map type, referred to as an "acquisition" map, depicts jays as they

might exist if all habitat patches were restored to optimal conditions, and distinguishes

among jays within protected areas, unprotected habitat patches, and suburban areas. Key

habitat patches are labeled on the acquisition maps, and are cross-referenced in the text

descriptions, tables, and recommendations.






8

A series of simulations were run for each metapopulation based on different

reserve design scenarios. These scenarios ranged from a minimal configuration consisting

of only currently protected patches (no acquisition option), to a maximal configuration

consisting of all significant patches (complete acquisition option). For all simulations, the

assumption was made that all protected areas were restored and properly managed, and

that jays had demographic performance and densities typical of high quality habitat.

These assumptions should be viewed as optimistic. Jays outside of protected areas were

assumed to have poor demographic performance typical of suburban areas.

The output from the simulation runs included estimates of extinction, quasi-

extinction (probability of falling below 10 pairs), and percent population decline.

Comparisons of these results provided the basis for ranking the vulnerability of different

metapopulations around the state. Metapopulations were ranked in terms of vulnerability

assuming no further acquisition, and in terms of potential for improvement through

acquiring all unprotected habitat. The proper uses and limitations of population modeling

are discussed.

Chapter 6 synthesizes previous chapters, focusing on some of the limitations of

metapopulation theory. The chapter closes by presenting a set of landscape rules that

provide guidelines for developing a statewide Habitat Conservation Plan for the Florida

Scrub-Jay. Adherence to these landscape rules would likely maintain the viability of

different jay populations across the state, while allowing for further loss of jays to human

development in some areas.














CHAPTER 2
CLASSIFYING FLORIDA SCRUB-JAY METAPOPULATIONS


Introduction


Metapopulation theory, now a major paradigm within conservation biology

(Harrison 1994; Doak and Mills 1994), can be viewed as island biogeography theory

applied to single species (Hanski and Gilpin 1991). Whereas application of island

biogeography to conservation followed shortly after its creation (MacArthur and Wilson

1967), application of metapopulation theory lagged far behind its formalization by Levins

(1969, 1970). Simberloff(1988) attributes its growing emergence to a shift in ecological

and conservation focus, from analysis of species turnover to analysis of extinction in

small populations of individual species. Describing real world metapopulations, however,

remains problematic.

Harrison (1991) pointed out ambiguities in the term "metapopulation," and

described four different configurations of habitat patches that could be called

metapopulations. Reviewing field studies of patchy systems, Harrison found few natural

examples that matched Levins' original concept of a metapopulation. Recently, Harrison

(1994) argued that metapopulation theory often is not applicable, such as cases where

populations are highly isolated, highly connected, or so large as to be essentially

invulnerable. She warned that: "the metapopulation concept is being taken seriously by

managers, and taken too literally could lead to the 'principles' that single, isolated

populations are always doomed, or that costly strategies involving multiple connected


9





10


reserves are always necessary" (p. 126). Doak and Mills (1994) reviewed the different

metapopulation classes described by Harrison and held that "it will often be difficult or

impossible to distinguish between these alternatives, and thus to assess the importance of

metapopulation dynamics" (p. 624). They also warned that spatially explicit population

models (SEPMs) simulating metapopulation dynamics typically use parameters that are

difficult to measure in the field. Their list of required data included within-patch

demographic rates and variances, temporal and spatial correlation of vital rates among

populations, and dispersal distances and success. Harrison (1994) also stressed the

difficulty of identifying all local populations and suitable habitat, and of estimating

extinction and colonization rates among patches.

Although still controversial (e.g., Harrison, Stahl and Doak 1993), the

metapopulation concept does provide a useful framework for describing the spatial

structure of real populations. The concept, after all, is grounded on two of the most robust

empirical generalizations in ecology and conservation biology: 1) extinction rates

decrease with increasing population size, and 2) immigration and recolonization rates

decrease with increasing isolation (MacArthur and Wilson 1967; Hanski 1994).

Our goal in this chapter is to illustrate how the above two generalizations can be

used 1) to characterize quantitatively the metapopulation structure of a species, and 2) to

develop "landscape rules" for conserving metapopulations of a declining species.



We begin with the results of a range-wide survey of the federally Threatened

Florida Scrub-Jay (Aphelocoma coerulescens), conducted in 1992-1993. This species is

patchily distributed, and thereby presents a challenging case study for describing








metapopulation structure. We offer a method for doing so using a detailed spatial

database, combined with existing biological information and GIS technology. The

technique uses computer-generated buffers, at several distances reflecting the dispersal

behavior of the species, to delineate subpopulations with differing degrees of

connectivity. Extinction vulnerability of each subpopulation is estimated via a PVA

model (in our case, that of Fitzpatrick et al. 1991). We propose a simple nomenclature for

classifying Florida Scrub-Jay metapopulations based on subpopulation size and

connectivity. We conclude by deriving a few, metapopulation-based "landscape rules"

that may be incorporated into a statewide framework for conservation plans affecting this

rapidly declining bird species.



Statewide Survey of the Florida Scrub-Jay



Statewide Survey: Methods

The Florida Scrub-Jay was listed by the U. S. Fish and Wildlife Service (USFWS)

as a Threatened species in 1987. In 1991 the USFWS notified landowners and county

governments that clearing scrub could violate the Endangered Species Act (ESA)

(USFWS 1991). At the same time, the USFWS began encouraging counties to develop

regional Habitat Conservation Plans (HCP) that could solve local permitting problems by

means of a single, biologically based, regional plan. To aid in this process, the USFWS

partially sponsored the authors, and their cooperators, to conduct an intensive survey

during 1992-1993 to document the range and sizes of subpopulations throughout the

state, and to inventory existing potential habitat, whether occupied or not.





12


Our methods were similar to those used for the Northern Spotted Owl (Sti

occidentalis caurina; Murphy and Noon 1992). Extensive prior information helped guide

us. Cox (1987) had documented numerous jay localities throughout the state in the early

1980s, and had compiled historic records from diverse sources such as museums and

Christmas Bird Counts. The Florida Breeding Bird Atlas (Kale et al. 1992) provided

valuable data on jay sightings made by hundreds of volunteers from 1986 to 1991.

Virtually all previously knownjay localities were revisited for this survey. Information

from the public was solicited through notices in magazines, newsletters, and newspapers.

Additional, potential habitat patches were identified from U.S. Soil Conservation Service

maps and aerial photographs, on which the white, sandy soil associated with jay habitat

forms a distinctive signature.

Standard surveying techniques based on tape playback of jay territorial scolds

(Fitzpatrick et al. 1991) were used to locate jays in habitat patches. Location and number

of individuals in each group were plotted on field maps. The following qualitative habitat

data were collected at most patches: occupancy by jays; estimated degree of vegetative

overgrowth (1-4 scale); extent of human disturbance (1-4 scale); ownership status with

respect to permanent protection from development. Time constraints prohibited making

quantitative habitat measurements at the thousands of habitat patches visited. Although

survey goals included attempting to find all known jay families outside of Federal lands,

we know that a few jays were missed because of limited access to certain private lands.

The total number missed, however, is not likely to exceed a few percent of the statewide

population.





13


Federally-owned lands were not surveyed for this project. Those with large

populations ofjays include: Cape Canaveral Air Force Station, Merritt Island National

Wildlife Refuge, Canaveral National Seashore, and Ocala National Forest. Florida Scrub-

Jays at each of these areas are currently under study, so for the statewide summary we

used estimates of numbers and locations ofjays provided by the respective biologists

conducting those studies.

To archive, map, and analyze the statewide data we developed a series of map

layers by means of a GIS at Archbold Biological Station. PC and Sun ARC/INFO were

used to input all GIS data (E.S.R.I. 1990). Habitat patches (both occupied and

unoccupied) and jay locations, originally hand-drawn on soil or topographic maps

(usually at 1:24000 scale), were digitized. Patch characteristics and jay family sizes were

entered into accompanying data files. Map layers included current and historic range of

jays, current distribution of suitable and potential habitat, and locations and numbers of

jay families encountered.


Statewide Survey: Results

We estimate that as of 1993 the total population of Florida Scrub-Jays consisted

of about 4,000 pairs (Fig. 2-1; Fitzpatrick et al. 1994). Both total numbers and overall

geographic range have decreased dramatically during this century (Cox 1987). In recent

decades the species has been extirpated from 10 of 39 formerly occupied counties, and it

is now reduced to fewer than 10 pairs in 5 additional counties (detailed tabulations in

Fitzpatrick et al., in prep.). Detailed, site-by site comparison of our survey with Cox's

(1987) suggests that the species may have declined as much as 25% to 50% during the

last decade alone.






14


Degraded quality of many currently occupied habitat patches suggests that

further, substantial declines in the jay population are inevitable. Specifically, those jays

occupying suburban areas (approximately 30% of all territories) are unlikely to persist as

these suburbs continue to build out, given the rapid rate at which Florida's human

population continues to expand. Furthermore, jays living in fire-suppressed, overgrown

habitat (at least 2,100 families, or 64% of all occupied scrub patches by area) already are

likely to be experiencing poor demographic performance (Fitzpatrick and Woolfenden

1986). These can be expected to decline further unless widespread restoration of habitat

is begun soon.



A Method for Classifying Metapopulations


The patchy distribution and variable clustering of territories throughout the range

of the Florida Scrub-Jay (Fig. 2-1) challenges us to expand upon traditional

metapopulation concepts in order to describe the spatial structure of this species. In this

section we describe our conceptual approach, and in the next we apply it to the Florida

Scrub-Jay data.

Harrison's (1991) four classes of metapopulations can be presented graphically

(Fig. 2-2) as different regions on a plot of degree of isolation against patch size

distribution. Thus, Harrison's "non-equilibrium" metapopulation is that set of small

patches in which each has a high probability of extinction, and among which little or no

migration occurs. Local extinctions are not offset by recolonization, resulting in overall

decline toward regional extinction. The "classical" model developed by Levins (1969,

1970) is a set of small patches which are individually prone to extinction, but which are






15


large enough and close enough to other patches so that recolonization balances

extinction. "Patchy" metapopulations consist of patches so close together that migration

among them is frequent, hence the patches function over the long run as a single

demographic unit. Finally, the "mainland-island" model has a mixture of large and small

patches close enough to allow frequent dispersal from an extinction-resistant mainland to

the extinction-prone islands.

The lower right side of Fig 2-2. portrays two classes not presented by Harrison

(1991). These large patches are either poorly-connected (i.e., "disjunct") or moderately-

connected (i.e., "mainland-mainland"). Such large populations tend to be less interesting

from a conservation standpoint, as they are essentially invulnerable to extinction.

Classifying metapopulations, therefore, requires species-specific information on

both connectivity (i.e., dispersal behavior and barriers) and extinction probabilities (i.e.,

population sizes in patches), across space. For example, a system of small habitat patches

might appear to support stable populations of certain organisms as "classical" or "patchy"

systems, while other species might be "nonequilibrial" in the same system because of low

density (hence small populations sizes) or limited dispersal ability.

Harrison's (1991) diagram of metapopulations represented connectivity among

patches by means of a dashed line around those among which dispersal is frequent

enough to 'unite the patches into a single demographic entity.' This boundary can be

viewed as a 'dispersal buffer': an isoline of equal dispersal probability. Any number of

patches may be included within a given dispersal buffer of a single subpopulation,

provided that fragmentation is sufficiently 'fine-grained' (sensu Rolstad 1991). However,

dispersal probability normally diminishes continuously (even if steeply) away from a






16


patch, and for most terrestrial species it asymptotically approaches zero at some point

farther away than Harrison's single, discrete dispersal buffer. Therefore, we extend

Harrison's diagrammatic approach by adding a second buffer to delineate the distance

beyond which dispersal is effectively reduced to zero. We maintain that this second

buffer functionally identifies separate metapopulations. We acknowledge that

connectivity actually should be represented graphically as a continuous surface of

dispersal probabilities. However, discrete boundaries, placed at biologically meaningful

(and empirically determined) distances, greatly simplify the description of

metapopulations. They also provide explicit, repeatable methodology for comparative or

modeling purposes.

Harrison's metapopulation types may be characterized using these two buffers

(Fig. 2-3). In "patchy" systems (Fig. 2-3a), every patch belongs to the same

subpopulation, so they are all enclosed within a single, inner dispersal buffer. "Classical"

systems (Fig. 2-3b) have small subpopulations separately encircled, representing the fact

that each may go extinct temporarily, or may be 'rescued' before going extinct (Brown

and Kodric-Brown 1977), both by way of colonization from another subpopulation

enclosed within the outer buffer. The simplest "nonequilibrium" systems (Fig. 2-3c) are

represented as bull's-eyes around small, isolated subpopulations. A "mainland-island"

metapopulation (Fig. 2-3d) has a large subpopulation and several small ones within a

single outer buffer.

The important point is that even more complicated patterns may be common in

nature, arising from combinations or intermediate cases, and many of these are not easily

fit into Harrison's (1991) four metapopulation classes. To deal with such complications,





17


we suggest characterizing metapopulations by describing the sizes of their constituent

subpopulations. We propose a simple nomenclature, based on three key words --"island,"

"mainland," and "midland" --to characterize the relative sizes of the subpopulations

within a metapopulation. Subpopulations small enough to be highly extinction prone in

the absence of significant immigration are called 'islands.' Those large enough to be

essentially invulnerable to extinction are called 'mainlands.' Intermediate sized

subpopulations are neither extinction prone nor invulnerable to extinction. For lack of a

better term, we refer to an intermediate size subpopulation as a 'midland.'

Distinctions among these categories need not be completely arbitrary. Species-

specific population viability analysis (PVA) provides an explicit, quantifiable approach

for describing subpopulations as extinction-prone, extinction-vulnerable, or extinction-

resistant. Our introduction of the 'midland' category helps clarify the importance of

turnover, which has been called the "hallmark of a genuine metapopulation dynamics"

(Hanski and Gilpin 1991). Specifically, turnover is expected in systems with island-size

subpopulations because they have high frequencies of extinction. But systems with

midlands rather than islands are perhaps more often characterized by rescue rather than

recolonization, as local extinctions will be rare. Thus, a system of midlands may exhibit

little or no turnover even though no real mainlands are present, while a system of islands

with the same degree of isolation may show high turnover. We agree with Sjogren (1991)

in emphasizing the importance of rescue in metapopulation dynamics. Traditional

emphasis on turnover probably resulted from the fact that rescue is much more difficult to

measure empirically, as turnover only requires presence-absence data.





18


Harrison's metapopulation classes can be described using this island-midland-

mainland nomenclature as follows: a "nonequilibrium" metapopulation is a system of

one or more islands (i.e., extinction-prone subpopulations), with a total population size

too small to persist. A "classical" metapopulation is a system of island-size

subpopulations large enough and close enough together and of sufficient total size to

allow persistence. Any system containing a midland or mainland (by definition) cannot

be a nonequilibrium or classical metapopulation, as all subpopulations in the latter

systems are extinction-prone. A "patchy" metapopulation is a set of patches close enough

together to form a single subpopulation of sufficient size to persist (i.e. a midland or

mainland). "Mainland-island" metapopulations are self-explanatory.

Explicit reference to 'midlands'--extinction-vulnerable patches of intermediate

population size--produces metapopulation types not described in Harrison (1991).

Systems with, for example, several midlands, or a mainland with several midlands, are

possible. We illustrate some of these configurations by applying our nomenclature,

quantitatively, to the Florida Scrub-Jay.



Metapopulation Structure of the Florida Scrub-Jay


Application of the above scheme to any species requires choosing two dispersal

buffer distances and two threshold values for extinction-vulnerability among single

populations. Here, for the Florida Scrub-Jay, we based each of these values on

empirically gathered biological data. Buffer distances were derived from long-term field

studies of marked individuals, and from information garnered on the statewide survey

regarding occupancy of habitat patches at various distances from source populations.





19


Extinction vulnerability was estimated using a single-population viability model

(Fitzpatrick et al. 1991). We then chose thresholds to delineate islands, midlands, and

mainlands, much as Mace and Lande (1991) used extinction probabilities to propose

IUCN threatened species categories.


Dispersal Distances

Between 1970 and 1993 we documented 233 successful natal dispersals from the

marked population under long-term study at Archbold Biological Station (Figure 2-4; see

also Woolfenden and Fitzpatrick 1984, 1986). Unlike the situation for most field studies

of birds (e.g., Barrowclough 1978), many characteristics of our study and the behavior of

jays themselves enhance our ability to locate dispersers that leave the main study area.

Once established as breeders, for example, Florida Scrub-Jays are long-lived and

completely sedentary. Furthermore, we have mapped in detail all scrub habitat within the

local range of the species, and we census these tracts periodically in search of dispersed

jays. (Such censuses reveal remarkably few banded dispersers among the many hundreds

of jays encountered.) Because banded Florida Scrub-Jays from our study usually are

tame to humans, both our own searches and casual encounters by local homeowners have

high likelihood of exposing any off-site dispersers to us once they become paired on a

territory. Indeed, if we assume that immigration and emigration rates are about equal in

our study area, evidence suggests that we have succeeded in locating all but a low

percentage of the jays that have departed over the 25-year period of our study. Therefore,

although some dispersers do escape our detection, our observed dispersal curve (Fig. 2-4)

can be only marginally biased toward the shorter distances.





20


About 80% of documented dispersals were within 1.7 km of the natal territory,

85% within 3.5 km, 97% within 6.7 km, and 99% within 8.3 km (Fig. 2-4). Data from

field studies elsewhere in Florida reveal the same, remarkably sedentary dispersal

behavior. The longest dispersal so far documented was a female we discovered pairing 35

km from her natal territory at Archbold, in 1994.

All dispersals we have documented around Archbold, including the longest one,

involved jays that had moved either through continuous habitat or across gaps no greater

than 5 km. To test the generality of this observation, we pooled dispersal information

from the seven other biologists currently color-banding Florida Scrub-Jays around the

state (D. Breininger, R. Bowman, G. Iverson, R. Mumme, P. Small, J. Thaxton, B.

Toland, unpubl. data). Their studies, along with ours, cumulatively have produced about

a thousand banded non-breeders that achieved dispersal age (Fitzpatrick et al., in prep.).

Collectively these studies have documented only about 10 dispersals of 20 km or more,

and only a few of these had crossed habitat gaps as large as 5 km. More important in the

present context, despite ample opportunity to observe longer-distance movements, not a

single example yet exists of a banded Florida Scrub-Jay having crossed more than 8 km

of habitat that does not contain scrub oaks. We suspect that this distance is close to the

biological maximum for the species.


Patch Occupancy

The observations just outlined suggest that for habitat specialists such as the

Florida Scrub-Jay, dispersal curves measured in relatively contiguous habitat actually

may overestimate the dispersal capabilities of individuals across fragmented systems.

Direct behavioral observations strongly indicate that Florida Scrub-Jays resist crossing






21


large habitat gaps. Still, few opportunities exist to observe jays in the act of dispersing,

hence the theoretical maximum dispersal-distance (i.e., the outer dispersal buffer) is

extremely difficult to establish directly.

Seeking an indirect measure of dispersal frequencies across habitat gaps, we

examined patch occupancy statewide as documented by our 1992-93 survey. We used

Fragstats software (McGarigal and Marks 1994) to measure distances between each

occupied patch of scrub habitat to its nearest neighboring, occupied patch. We then

measured (by hand, as Fragstats cannot measure distances between patches of different

attributes) the distances between each unoccupied suitable patch and the nearest occupied

patch. For each distance class, the ratio of the count of the occupied-to-occupied

distances to the total number of nearest neighbor distances yields the proportion of

patches that are occupied at that distance away from occupied habitat.

Presumably, declines in patch occupancy with increasing distance to the nearest

occupied habitat (Fig. 2-5) reflect diminishing recolonization rates following local

extinctions. Occupancy remains above zero even at great distances, probably because

larger isolated patches rarely experience extinction. This curve provides an empirical

approach for delineating subpopulations and metapopulations: a subpopulation buffer is

the maximum interpatch distance where occupancy rates remain high; the metapopulation

buffer is the smallest interpatch distance where occupancy rates reach their minimum.

For Florida Scrub-Jays (Fig. 2-5) patch occupancy is about 90% to at least 2 km

from a source, then declines monotonically to around 15% at 12 km. (Sample size of

isolated patches decreased rapidly beyond 16 km, necessitating lumping of classes at the

larger distances.) We infer from this occupancy curve that successful recolonization is a





22


rare event beyond about 12 km from an occupied patch of habitat. We use this distance to

identify metapopulations that have become essentially demographically independent from

one another (i.e., the outer dispersal-buffer).

We selected the distance of 3.5 km (about 2 miles) as an inner dispersal-buffer to

delineate subpopulations. We choose this figure because: 1) behavioral information from

a variety of sources, including radiotracking data (B. Stith, unpubl.), shows that jays

begin to show reluctance to crossing habitat gaps at about this size (and at much smaller

gaps where open water or closed-canopy forest are involved); 2) known dispersals of

many banded jays included habitat gaps up to 3.5 km, but their frequency declines

dramatically thereafter; 3) the observed dispersal curve from Archbold (Fig. 2-4) shows

that in good habitat, more than 85% of dispersals by females, and fully 97% of those by

males, are shorter than 3.5 km; 4) patch occupancy data (Fig. 2-5) show significant

decline in colonization rates at distances above 3.5 km.


Population Viability Analysis

PVA based on a simulation model incorporating demographic (but not genetic)

stochasticity and periodic, catastrophic epidemics (Fitzpatrick et al. 1991; Woolfenden

and Fitzpatrick 1991) provided a quantitative method for defining boundaries along the

island (extinction prone), midland (vulnerable), and mainland (extinction resistant)

continuum (but see Taylor 1995). Among the several methods for expressing extinction

vulnerability (e.g., Burgman et al. 1993; Boyce 1992; Caughley 1994) we elect the

simple approach of specifying time-specific probability of persistence of populations of a

given size.





23


Model results indicated that a population of jays with fewer than 10 breeding

pairs has about a 50% probability of extinction within 100 years, while a population with

100 pairs has a 2% to 3% probability of extinction in the same time period. These two

population sizes--10 and 100 pairs--provide convenient and biologically meaningful

values by which to classify subpopulations as "islands" (< 10 pairs), "midlands" (10-99

pairs), and "mainlands" (> 99 pairs). Although subjectively chosen, these values

effectively separate population sizes having fundamentally different levels of protection.

These values also receive empirical support from several long-term bird studies

(reviewed by Thomas 1990; Thomas et al. 1990; Boyce 1992).


Metapopulation Structure

We used a GIS buffering procedure (E.S.R.I. 1990) to generate dispersal-buffers

around groups ofjays occurring within 3.5 km (for subpopulations) and 12 km (for

metapopulations) of each other (Fig. 2-6). We buffered jay territories rather than habitat

patches because we strongly suspect that dispersing Florida Scrub-Jays cue on the

presence of other, resident jays even more strongly than on habitat, so the functional

boundaries of occupied patches may be determined by where actual jay families exist.

We modified the resulting buffers in the following areas to reflect the presence of hard

barriers to dispersal in the form of open water with forested margins: Myakka River,

Peace River, St. Johns River, St. Lucie River, and Indian River Lagoon.

Using a 3.5 km dispersal-buffer we delineated 191 separate Florida Scrub-Jay

subpopulations (Fig. 2-6). Over 80% (N=152 "islands") are smaller than 10 pairs (Fig. 2-

7), and 70 of these consist of only a single pair or family ofjays. Only 6 subpopulations

contain at least 100 pairs ("mainlands"), leaving 32 "midlands" (10-99 pairs).






24


Using a 12 km dispersal-buffer we delineated 42 separate Florida Scrub-Jay

metapopulations (Fig. 2-6). Again, most are small (Fig. 2-7). We tabulated the number

and type of subpopulations within each metapopulation (Table 2-1), and noted how each

metapopulation fits into Harrison's (1991) scheme.

Exactly half (21) contain fewer than 10 pairs, thereby constituting

"nonequilibrium" systems. Along the north-central Gulf Coast (Fig. 2-8), for example, a

group of non-equilibrium systems coincide with a heavily developed area containing a

burgeoning human population. Only three Florida Scrub-Jay systems have configurations

that may be "classical" metapopulations (i.e., contain only islands, but may be large

enough to support one another following extinctions; e.g., Fig. 2-9). However, this

technique provides no means of distinguishing "classical" systems from

"nonequilibrium". Chapter 5 addresses this shortcoming using a simulation model.

Another three systems represent "patchy" metapopulations (i.e., contain a single,

fragmented subpopulation large enough for long-term persistence).

Five systems approximate "mainland-island" metapopulations, but each of these

examples also contains at least one midland population (e.g., the large Lake Wales Ridge

system, with one mainland, 10 midland, and 39 island populations; Fig. 2-10). These 10,

plus 9 midland-island and one mainland-midland system, do not fit neatly any of

Harrison's (1991) metapopulation classes.

A total of 32 midland populations exist (mean size=30.7), and these occur in 18 of

the 42 separate metapopulations. Excluding the nonequilibrium systems, true islands are

present in 17 systems. Therefore, assuming that dispersal is not inhibited by habitat loss






25

expanding the distances among patches, rescue (on midlands) may be at least as

important as turnover (on islands) in Florida Scrub-Jay metapopulation dynamics.

Use of empirically derived dispersal-buffers and extinction probabilities provides

an explicit method for quantitatively describing metapopulation structure. Application of

this technique to the Florida Scrub-Jay demonstrates that a species can exhibit a variety

of metapopulation patterns across its range. Patterns of aggregation and isolation do not

conform to a single metapopulation class in the Florida Scrub-Jay. Such complex spatial

structure is probably common in nature, particularly among species with large and widely

dispersed populations restricted to a patchy habitat. Such patterns may be further

complicated by perturbations of the natural system caused by humans.


Caveats

We offer several caveats as to the generality of dispersal-buffer methodology in

conservation. (1) The technique is best suited for organisms occupying discrete

territories, home ranges, or habitat patches amenable to mapping. (2) The technique is

predicated on having a comprehensive survey. Missing data can lead to misleading

results, especially as regards connections among metapopulations or subpopulations. (3)

The technique presents a static, snapshot view of metapopulations. It does not easily

reveal important dynamics among subpopulations, such as those obtainable from an

SEPM. The viability of different configurations is best determined from SEPMs rather

than single population PVAs. (4) Populations in decline or in "sinks" can present an

overly optimistic picture (Thomas 1994). Indeed, we suspect that many of the "island"

and "midland" subpopulations of Florida Scrub-Jays currently are failing to replace

themselves demographically, as a result of habitat degradation from fire suppression.






26

Similarly, abnormally high densities may exist due to the "crowding effect" (Lamberson

et al. 1992) following recent habitat losses. (5) The technique relies on numerous

simplifying assumptions about dispersal behavior in defining connectivity among

patches. Most important, it assumes random movement between patches, equal

traversibility of interpatch habitats, absence of dispersal biases owing to habitat quality

differences at the origin or the destination, and absence of density-dependence in

behavior. More elaborate applications, of course, could incorporate alternative

assumptions about these and other factors.

Another important consideration are the kinds of data to buffer. To create

biologically meaningful--but very different--descriptions for the Florida Scrub-Jay we

could have buffered around jay territories (our choice), occupied patches, suitable habitat

patches both occupied and unoccupied, or all scrub habitat patches regardless of current

suitability. Organisms such as Florida Scrub-Jays that are reluctant to become established

in unoccupied, suitable habitat (e.g., Ebenhard 1991), or have high conspecific attraction

or an "allee" effect (Smith and Peacock 1990) are best buffered around actual territories

or occupied patches. This is because unoccupied sites have a low probability of becoming

occupied regardless of their degree of isolation, hence contribute little to the current

metapopulation dynamics of the species. On the other hand, excellent colonizers of empty

habitat or species adept at long-distance dispersals via unoccupied stepping stones

probably should be buffered around all habitat patches.

In summary, this method of classifying metapopulations provides a compact

means of describing both connectivity and local population size through the use of simple

terminology. Separate metapopulations are easily delineated using the maximum






27

dispersal buffer. The internal structure of each metapopulation is easily described using

the inner dispersal buffer to delineate islands, midlands, and mainlands. Enumerating all

metapopulations and describing their internal structure (table 2-1) reveals much about the

distribution and viability of Florida Scrub-Jays. The differing internal configurations of

metapopulations present different conservation problems and require different

management approaches. Discussion of these matters is deferred until chapter 6,

following the presentation of the modeling results (chapter 5) which analyze the viability

of metapopulations around the state.







Note: This chapter has been published as Stith et al., 1996, and is reproduced with some
modifications with the permission of Island Press.






28



Table 2-1. Summary information for 42 Florida Scrub-Jay metapopulations, including
type of metapopulation, number of territories (pairs ofjays), and number of
subpopulations.

Metapopulation type Metapopulation Type Size Number of
(after Harrison 1991) (Mainland, Midland, (pairs) subpopulations
and/or Island) a


Mainland-Island (?) Mn 10Md 391 1247 50

Mn Md 51 1036 7

Mn Md 51 466 7

Mn 2Md 61 237 9

Mn Md 51 179 7

Unknown Mn Md 126 2

4Md 111 120 15

2Md I 103 3

Md 51 94 6

Md I 58 2

2Md 31 55 5

Md I 50 2

Md 21 29 3

Md 31 22 4

Md 21 18 3

Patchy Md 26 1

Md 22 1






29


Table 2-1 cont.

Metapopulation type Metapopulation Type Size Number of
(after Harrison 1991) (Mainland, Midland, (pairs) subpopulations
and/or Island) a


Md 15 1

Classical 161 49 16

10I 24 10

61 21 6

Nonequilibrium 31 5 3

31 3 3

21 7 2

21 3 2

21 3 2

21 2 2

21 2 2

21 2 2

I 6 1

I 2 1b

I 1 1c



a Numerical prefix indicates number of Mainlands (Mn), Midlands (Md), and Islands (I).

See text for nomenclature.

b There was a total of 4 single Island systems composed of 2 pairs in one subpopulation.

C There was a total of 8 single Island systems composed of a subpopulation of one pair.






30












*, *
I1

S" *� .. * .


' * * *, . * .





S -











50 0 50 Kilometers
Fig. 2-1. 1993 distribution of Florida. (small black circles). Note the




discontinuous distribution and variability in patterns of aggregation.
e,







50 0 50 Kilometers









Fig. 2-1. 1993 distribution of Florida Scrub Jay groups (small black circles). Note the
discontinuous distribution and variability in patterns of aggregation.






31







Highly
Connected
Patchy




Patch Mainland- Mainland-
Isolation Classical Island Mainland





Nonequilibrium Disjunct
Highly
Isolated


Mixture of All
All
Small Small& Large
Large
Patch Size







Fig. 2-2. Classification scheme showing different types ofmetapopulations based on
patch size distribution (patches all small in size, mixture of small and large, and all large
in size) along the horizontal axis, and degree of patch isolation (highly connected to
highly isolated) on the vertical axis. Nonequilibrium, classical, mainland-island, and
patchy classes are named according to Harrison (1991).






32

















S------ -
S.. ... ...



6 .






























Fig. 2-3. Schematic depiction of different kinds of metapopulations, illustrating use of
dispersal-distance buffers to predict recolonization rates among subpopulations. Dotted
lines separate functional subpopulations, based on frequency of dispersal beyond them.
Solid lines separate metapopulations, based on poor likelihood of dispersal among them.
A. Patchy metapopulation. B. Classical metapopulation. C. Nonequilibrium
metapopulations. D. Mainland-island metapopulation.






33











45-
SFEMALES (N=109) F, MALES (N=124)
40

35.

S30-
o
uI.J
Z 25-

J 20







0 2 4 6 8 22
DISPERSAL DISTANCE (km)










Fig. 2-4. Dispersal frequency curve. Dispersal distances from natal to breeding territories
for color-banded jays at Archbold Biological Station, 1970-1993. About 85% of
documented dispersals were within 3.5 kmn, and 99% within 8.3 km. The longest
documented dispersal was 35 km.
10 - --- ' -| -----------------

5- - 1 | - -------
















documented dispersal was 35 km.






34











1.00
717
0.90 -

- 0.80

C 0.70 63

.! 0.60

o 0.50 27
0 22
0 0.40
0
; 0.30
0 s
o 0.20 7
IL I_
0.10

0.00
0 0 0 o 0 0
00 00 0
o 0 0 0 O 0 0 0 0 0 Q
S (0 O 0 0 0 0 M 0

Interpatch Distance (m)











Fig. 2-5. Proportion of suitable habitat patches occupied by Florida Scrub-Jays as a
function of their distance to the nearest separate patch of occupied habitat. Occupancy
rates are high (nearly 90 %) for patches up to 2 km apart and decline monotonically to 12
km. Note the scale change after 16 km.






35



















- --.---. ----

*




*O ..

- -- -------------

- -- --- --











50 0 50 Kilometers ) ---.....-- ....-----






Fig. 2-6. Statewide jay distribution map with dispersal buffers. Shaded areas with thin,
solid lines depict subpopulations of jays within easy dispersal distance (3.5 km) of one
another. Thick lines delineate demographically independent metapopulations separated
from each other by at least 12 km.















25



20




S15










2 2 2
o-
u. 10









m v 0 C r- 0 0 0 00 0 0














metapopulations have 10 pairs or less of jays. These represent nonequilibrium
metapopulations.



Fig.2-7 Freueny o FloidaScrb~aymetpoplatin szes Not tht i






37





















0 -









Pasco Q
Hillsborough








Fig. 2-8. Examples of non-equilibrium metapopulations from North Gulf coast. Each of
the six metapopulations contains fewer than 10 pairs of jays, except for the centrally
located system that contains a single, midland-size subpopulation.






38










































Fig. 2-9. Example of a "classical" metapopulation from five counties in Central Florida.
Note the occurrence ofjays in small islands of intermediate distance from one another.






39







** s














Char.ott


4W4.
















Fig. 2-10. Portion of the largest mainland-midland-island metapopulation in the interior,
consisting of the Lake Wales Ridge and associated smaller sand deposits. The large
central subpopulation (enclosed by the thin black line) contains nearly 800 pairs of jays.
Small subpopulations to the south and east are within known dispersal distance of the
large, central mainland. A small metapopulation to the west (in DeSoto County) contains
a single subpopulation of 21 territories. This small system qualifies as a patchy
metapopulation, since jays occur in two or more patches but the patches are so close
together that they function as a single demographic unit.














CHAPTER 3
REMOTE SENSING OF FLORIDA SCRUB-JAY HABITAT



Introduction


The importance of understanding relationships between wildlife and habitat has

been recognized for many decades (review in Morrison et al. 1992). The ultimate success

of wildlife management and conservation efforts depends to a large degree on our ability

to understand these relationships. Unfortunately, measuring habitat variables over large

areas, and with sufficient spatial resolution to capture the essential habitat features a

particular species responds to, is a difficult and time consuming task. In this chapter I

apply a technique that greatly assists in the measurement and analysis of wildlife-habitat

relationships across large areas. This technique, which relies heavily on recent advances

in computer hardware and software technology, uses image processing and GIS software

to measure habitat variables directly from scanned aerial photography.

Habitat requirements differ widely among most species, and choosing the

appropriate habitat variables to measure is greatly facilitated for species whose habitat

requirements are well understood. Long-term studies of the Florida Scrub-Jay have

revealed much about the habitat requirements of this species (Woolfenden and Fitzpatrick

1984; Breininger et al. 1991, 1995, 1996). Research indicates that scrub-jays have a fairly

simple "habitat template" consisting of low vegetation dominated by scrub oaks, open





40






41

sandy areas to cache acorns, and few or no pine trees. These habitat characteristics are

naturally created and maintained by fire.

Pine tree density may be a critical factor in determining Florida Scrub-Jay habitat

quality. Pine trees may affect scrub-jays indirectly in several fashions. Large trees

provide perch sites and cover used by accipiters, important predators ofjays. Pine trees

probably reduce the effectiveness of the jays' sentinel system against aerial predators. A

close competitor and nest predator, the Blue Jay, shows a strong preference for pine

forests through much of the scrub-jays's range (Tarvin 1997). Competition between these

two corvids may partially explain why Florida Scrub-Jays select open habitat (see

discussion in Woolfenden and Fitzpatric 1984).

Declines in nesting success, and survival of juveniles and adults have been

documented by Fitzpatrick and Woolfenden (1986) within a fire-suppressed habitat that

gradually becomes overgrown. Jays living in open habitat have higher survival, nesting

success, and larger mean group size than jays in less open habitat (Fitzpatrick and

Woolfenden 1986; Breininger et al. 1995). Differences in these demographic traits may

be due to habitat features immediately adjacent to jay territories (Breininger et al. 1996).

Thus, although the internal quality of a jay territory may be high, territory position within

the landscape mosaic may greatly influence demographic success.

In this chapter, I test some of these ideas using habitat features and demographic

variables measured at the Avon Park Air Force Range (APAFR) in Polk and Highlands

counties, Florida. I compare the habitat and demographic performance of jays occupying

a north-south trending scrub ridge some 15 km in length. Jays along this ridge occupy

habitat with different degrees of overgrowth and different adjacent habitat. The main






42

objective of this chapter is to test whether remote sensing can accurately measure habitat

features that explain Florida Scrub-Jay demographic performance.



Methods


All image processing and GIS work was completed on a Sun workstation or an

Intel 486-based PC using Arc/Info (ver. 4.3D) and Erdas (ver. 7.5) software.




Image Source

Regular color, black and white, and color infrared photography were evaluated for

use in this project. Color infrared photography was flown for much of the APAFR

through the NAPP program of the U.S.G.S. in March of 1994. I selected this photography

for three primary reasons: separation between sandy and grassy areas was more distinct,

wider coverage per frame made image mosaicing easier, and the photography is available

for study sites outside of the APAFR that are currently under investigation for Florida

Scrub-Jay conservation. Hence, results of this study could be extended to other areas.

Two photos covered the entire study area. Frame 6980-207 covered the area I

refer to as N. Sandy Hill, which is between Kissimmee and Smith roads, and part of S.

Sandy Hill south to Submarine Lake. Frame 6980-205 covered the remainder of S. Sandy

Hill south to the southern fence line.






43

Image Scanning and Conversion



A Umax 1200 color scanner with a transparency adapter (for scanning positive or

negative film; 600 dots-per-inch maximum hardware resolution) was used to scan the

1:40000 color transparencies at a resolution of 555 dots per inch, giving a ground

resolution of approximately 1.8 meters (6 feet). Image-In scanning software provided by

Umax for use with the scanner was used to scan all images on a PC.

Software from Earth Resource Data Analysis Systems (ERDAS) of Atlanta,

Georgia was used to convert the TIFF files to ERDAS 7.5 LAN files.




Image Rectification and Mosaicing



I used ERDAS software to rectify the two images to Zone 17 of the Universal

Transverse Mercator map projection using differentially corrected GPS points. I collected

these GPS control points at various road intersections throughout the study area using a

Trimble Pathfinder rover with a data logger. At each control point, 180 location

measurements were acquired. I differentially correc:ed these control points using base

files acquired from a Trimble Professional base station located on the APAFR. The mean

of the 180 measurements taken at each control poi:i was used as the final x and y

coordinate for rectifying the images.

During the rectification process, I discar-ed some GPS points because their

inclusion produced excessively high root mean square (RMS) errors. Several factors

explain these high RMS errors. First, the expected error from the differentially corrected






44

GPS points (Trimble claims 2 to 5 m for the units I used) is substantial relative to the

image resolution (2 m). Second, nonlinear distortions in the imagery owing to lens

curvature, tilt angle, etc. cannot be corrected with this type of linear rectification process.

Third, an error of one to several meters is introduced by inaccuracies in visually placing

each control point on the image. Four control points were common to both images, which

helped ensure that the two images matched well in the overlapping areas. The locational

accuracy of the rectification is unknown, and could not be determined without using

much more accurate and expensive ground-based surveying techniques. However, the

positional accuracy is probably within the range of 2 to 10 meters. This accuracy is more

than adequate for the purposes of this project.

The two rectified images were mosaiced together using Erdas software. Very little

displacement is present in the spliced region between the two images, indicating that the

rectification process was internally consistent.




Image Classification

The images were classified with unsupervised, isodata classifier in Erdas to

produce 27 statistically distinct spectral classes stored in a signature file. This signature

file was then used to reclassify the raw image with a maximum-likelihood classifier. I

examined each class from the resulting classification individually and visually compared

each to the original photography to evaluate the zcorrespondence of each spectral class to

known ground features. This comparative procedure consists of flashing each of the

classes on and off repeatedly while viewing the image on the screen, and simultaneously

consulting the original photographs. The most distinctive classes had either very low






45

reflectance or very high reflectance in all three spectral bands. Extremely low reflectance

values corresponded to tree crowns, the shadows cast by tree crowns, or standing water.

Although appearing in the same spectral class, water was easily distinguished from tree

crowns and shadows by pattern and texture. High reflectance values corresponded to bare

sand patches, and human disturbances such as dirt roads and excavations. Naturally

occurring bare sand patches were nearly always associated with xeric habitats, and had a

distinctive, fine-grained pattern and texture compared to ground features created by

humans. Spectral classes with intermediate reflectance were much more difficult to

associate with ground features. In general, grass-dominated prairies, such as occur

between scrub patches on N. Sandy Hill, had high reflectances that were only slightly less

than bare sand. Areas dominated by oak shrubs were spectrally similar to areas

containing various proportions of palmetto and wire grass. Discriminating among mixed

shrub classes was difficult and also was believed unlikely to affect jay dispersion at

APAFR. The dominant and most recognizable spectral classes corresponded to tree

canopies and the shadows they cast, and bare sand patches. All spectral classes were

recoded to the following landcover classes: 1 = tree cover, 2 = bare sand, 3 = mixed

grass/shrub, 4 = wetlands/seeps. These classes were intended to reflect key structural

components of the habitat rather than vegetative types.




Manual Editing of Classification

Manual editing of the final classification was necessary in some areas with very

dark signatures that were confused with tree shadows. These areas of confusion were all

wetlands or poorly drained areas with temporary standing water at the time the imagery





46

was acquired (March 1994). Most manual editing was accomplished using a single

"mask" file, which contained wetland polygons digitized from the original rectified

image. This mask file was used to recode all pixels inside of wetland polygons to the

wetland category. Special attention was given to areas immediately surrounding jay

territories. One cutthroat seep area adjacent to and just east of the NE territory (and

experimental plot 1) on S. Sandy Hill showed standing water and trees. In this area, two

dark pixel classes were found to have a good correspondence with tree cover, and two

dark pixel classes corresponded with standing water. Theses classes were recoded

accordingly.




Assessment of Classification Accuracy

Classification accuracy of percent tree cover was evaluated by comparison with

estimates from line transect data collected at five control plots on S. Sandy Hill as part of

the experimental manipulation experiment. A three pixel wide buffer was generated

around each transect using Arc/Info. This buffer was used to sample the imagery around

each transect using Erdas. Classification accuracy of bare sand was evaluated by

comparison with estimates from quadrat samples ( ;I m across) collected on the northern

section of S. Sandy Hill in areas largely devoid of -1ne trees. The measurements obtained

from the imagery and the transect measurements . re compared using a Paired T-Test

and a correlation analysis.






47

Digitization of Territories and Background Features

Polygons representing the boundaries ofjay territories and other background

features (e.g. roads) were digitized directly off of the computer screen with a mouse from

the mosaiced image using Arc/Info. This allowed the resulting coverages to be

automatically georeferenced to the aerial photography




Tree Cover Buffering Procedure

GIS buffers were generated around each jay territory at 100, 200, and 400 m

distances using Arc/Info. These buffers were used to characterize the habitat immediately

surrounding each territory. A major complication with this procedure was that the buffers

for each territory often overlapped with neighboring territories, and overlapping polygons

are not allowed in Arc/Info coverages. Therefore, each territory was kept in a separate

coverage and analyzed individually. The following procedure was used. Each coverage

was buffered at distances of 100, 200, and 400 meters. The resulting coverages were

converted to individual ERDAS ".dig" files. Each .dig file was overlaid on the classified

image to calculate the percent tree cover within each buffer using an ERDAS program

called POLYSTAT (developed by B. Stith and J. Richardson). The statistics generated

from this program were imported into spreadsheet :.nd statistical packages (Systat; SAS)

for further analysis.






48

Habitat Quality Model

A habitat quality model was developed from the habitat variables using a habitat

suitability index (HIS) approach similar to Duncan et al. (1995). The HSI model

combines three HSI values for percent bare sand within territories (BS), percent tree

cover within territories (TC), and distance to nearest forest (DF), to calculate a single

habitat quality value (HQ) for each territory. The equation used for this model is



HQ = BS * TC*DF



The HSI values for each of the three habitat variables was obtained from step

functions relating the habitat variable to an estimate of habitat quality (see Fig. 3-2

modified from Duncan et al. 1995). The shapes of these step functions were developed

subjectively by D. Breininger. The habitat quality values were mapped automatically

across the entire study area using the Spatial Modeler in the Imagine software package.

The BS variable was mapped using the "focal sum" operator to count the number of bare

sand pixels within 10 m of every point in the study area, and computing an HSI value

using the function in Fig. 3-2a. The TC variable was similarly mapped by measuring tree

pixels within a 60 m radius, and computir : an HSI alue using the function in Fig. 3-2b.

To compute the DF variable, the TC laye: was us ... and pixels surrounded by greater

than 30V o tree cover were coded as forest. The "search" operator was used to measure the

distance to nearest forest for all pixels. The DF - .,ue was then computed using the

function in Fig. 3-2c. The Imagine "summary" function was used to output HSI values






49

for each territory. Because "summary" requires integer values for input, the HSI values

were resampled to 5 equal intervals.


Collection of Demographic Data

Demographic data for jays on Sandy Hill were collected by a team of field

researchers (Brad Stith, Reed Bowman, Doug Stotz, Larry Riopelle, Natalie Hamel, and

Mike McMillan) during 1994 - 1995, as part of a larger, on-going project that now

monitors jays on the entire APAFR. All jays on Sandy Hill were captured, color-banded,

and monitoried quarterly using techniques similar to Woolfenden and Fitzpatrick (1984).

Nests were found and monitored during the spring and early summer. The raw

demographic data are presented in Table 3 - 1.


Habitat-Demographic Analysis

I compared the demographic performance and habitat characteristics between the

North and South Sandy Hill populations of jays. Owing to lack of normality for nearly all

parameters (Table 3-2; Kolmorov-Smirnov test for normality), the nonparametric Mann-

Whitney U statistic was used to test for demographic and habitat differences between the

North and South study areas (Table 3-3).

I searched for habitat-demographic relationships by performing multiple linear

regression (maximum R2 improvement technique for all combinations of variables) and

logistic regression (SAS Institute). Demographi :parameters served as dependent

variables, and habitat measurements as indepc- .,nt variables.






50

Results


The locations and names of Florida Scrub-Jay territories throughout the study area

are shown in Fig. 3-1. The division between the North and South population occurs at

Kissimmee Rd. Note the absence ofjays within the South population in the central part

of S. Sandy Hill. This area has high densities of pine trees and little bare sand, and has

three experimental plots where habitat restoration is underway.

A best-fit regression line, constrained to pass through the origin, showed a

significant relationship between percent bare sand measured from quadrats vs. imagery

(Fig. 3-4; r-squared = 0.60). The difference between the quadrat and image measurements

was not significant (Paired T-test; mean difference = 0.9372, n=12, p=0.256), indicating

no systematic bias in the measurements. Figure 3-5 shows the relationship between

percent tree cover measured from transect data vs. imagery. A best-fit logarithmic

regression line was drawn through the points (r-squared = 0.25). The differences between

the transect and image measurements was not significant (Paired T-test; mean difference

= 0.9372, n=40, p=0.824), indicating no systematic bias in the measurements. However,

the large scatter of points deviated considerably from the expected distribution which

would fall on a regression line intercept, the origin and having a slope of one. The

logarithmic trend line suggested that the :age n:. :surements give higher than expected

measurements at low tree cover values. :id low : i an expected measurements at high

tree cover values.

Percent tree cover ranged from 4% to % on S. Sandy Hill (Fig. 3-7), and only

0% to 4% on N. Sandy Hill (Fig. 3-6). The difference in tree cover between North and

South was highly significant (Table 3 - 3; Mann-Whitney U Test, Z = -6.291, P <






51


0.0000). Most of the territories with high tree cover occured in the south end of S. Sandy

Hill. Much greater variation was evident on S. Sandy Hill compared to N. Sandy Hill.

Percent bare sand ranged from 9% to 30% on S. Sandy Hill, and 2% to 34% on N. Sandy

Hill. The difference in bare sand cover between North and South was not significant

(Table 3 - 3; Mann-Whitney U Test, Z = -1.595, P = 0.111).

Comparison of tree cover in North vs. South territories and in the three buffer

zones (100, 200, and 400 m) showed striking differences (Fig. 3-8) and were highly

significant for all comparisons. South territories showed a dramatic increase in tree cover

with increasing distance from territories, while North territories showed only a slight

increase. In the North, tree cover increased only slightly if at all as distance from territory

increased (Fig. 3-9). In comparison, South territories showed large increases in tree cover

as distance from territory increases (Fig. 3-10). A strong correlation was found between

tree cover within territories and buffer zones in the South (e.g. r-squared = 0.88 for

territories vs. 100m buffer), suggesting that these variables had strong spatial

autocorrelation.

Comparison of tree cover within jay territories vs. total tree cover showed that jay

territories have fewer trees than the area available to them for both the North and South

territories (Fig. 3-11). Note that the measurements of available tree cover were made

across the entire North and South study area rather than from subsamples, hence no

standard deviations were calculated.

Measurements of bare sand within territories and the buffer zones showed a

decreasing amount of bare sand away from territories for both the North and South

territories (Figure 3-12), indicating selection of sandy areas by jays.






52

Comparison of demographic performance of jays in North vs. South for two years

(1994, 1995) showed significant differences only for nonbreeder survival (Table 3 - 3;

Mann-Whitney U Test, Z = -2.396, P = 0.017). Group size was nearly significant (P =

0.060). Number of fledglings produced, fledgling survival, yearling survival, and breeder

survival were not significantly different.

Relationships between group size and percent tree cover within all territories (Fig.

3-13), and group size and percent tree cover within 100 m buffers (Fig. 3-14) showed

decreasing group size with increasing tree cover. I lumped group size into 3 categories of

roughly equal size and looked for differences in tree cover at the territory (Fig. 3-15) and

100 m buffer (Fig. 3-16). Large families (4 to 7 jays per group; n = 13) had lower median

and variance in tree cover within and adjacent to their territories compared to medium (3

jays per group; n = 7) and small (2 jays per group; n=15) families. The differences,

however, were not significant (Kruskal-Wallis one-way analysis of variance; tree cover P

= 0.205; 100 m buffer P = 0.186). I lumped the 3 group categores into 2 group size

categories (2-3 jays per group; 4-7 jays per group) and performed the same analysis, but

the results were not significant.

Figures 3-17, 3-18, and 3-19 show side-by-side views of the classified and raw

images for N. Sandy Hill, the N. portion of S. Sandy Hill, and the S. portion of S. Sandy

Hill respectively. Jay territories are outlined in black. The names of jay territories are

shown in Fig. 3-I. Three colors on the classified images correspond to tree cover (green),

bare sand (white) and mixed shrubby or grassy vegetation (brown).

Figures 3-20 and 3-21 show presumed habitat quality as computed from the three

HSI variables for the N. and S. portion of S. Sandy Hill respectively. High quality habitat






53

is shown in red (HSI = 0.81 - 1.0), medium quality habitat is shown in blue (HSI = 0.61 -

0.80), and low quality habitat is shown as white (HSI < 0.61). Jay territories (outlined in

black) generally included substantial areas of low quality habitat in both the North and

South areas.

I searched for correlations between demographic (fledglings, independent young,

and yearlings produced, survival of fledglings, yearlings and breeders, and group size)

versus the habitat variables (percent sand, percent tree cover within territories, percent

tree cover within 100 m buffer, HSI for sand, HSI for distance to forest, HSI for trees

within habitat, and combined HSI). The clearest bivariate patterns were for group size

versus percent tree cover within territories (Fig. 3-13) and percent tree cover within the

100 m buffer (Fig. 3-14), but Kruskal-Wallis one-way analysis of variance results showed

no significant differences between different group size comparisons and percent tree

cover. Large group size variance existed in territories with low tree cover or adjacent to

low tree cover, but there was a strong, nonsignificant trend towards smaller groups as tree

cover increases. Kolmogorov-Smirov tests for normality showed that bare sand was the

only normally distributed habitat variable. Normality plots indicated that deviations from

normality could not be corrected by commonly used (e.g. arcsine, inverse, log, square

root) transformations. Nonparametric spearman rank correlation coefficients were low for

all pairings of demographic and habitat variables. Multiple regression models never

explained more than about 22% of the variation in demographic parameters using all

combinations of habitat variables. Similarly, no habitat variables in several logistic

regression models were significant.






54

Discussion


Accuracy of my remotely sensed habitat measurements is difficult to assess. Most

remote sensing studies are conducted at a much coarser scale, and they attempt to identify

discrete data classes (e.g. vegetation types). Such studies typically use a simple error

matrix analysis where percent correct classification is given. I have little precedence to

follow, since the goal of this classification was to provide continuous measurements (i.e.

percent cover) from structural classes (e.g. tree cover or bare sand) rather than discrete

vegetation classes. To assess the accuracy of these measurements quantitatively, I

compared them to ground based measurements using paired T-Tests and simple

correlation analysis. The bare sand measurements obtained from quadrats showed a fairly

good correlation with the image measurements (Fig. 3-4). In contrast, the transect

measurements of tree cover showed a weak correlation (Fig. 3-5). Some of these

differences resulted from classification errors noticeable in comparisons of the

photography with the classified image. Underestimates of tree cover were apparent in

some of the young or very dense pine plantations, which tended to form a uniform

canopy with few shadows. Overestimates were noted in areas with large scrub oaks, such

as in some of the long unburned scrub patches on the W. side of N. Sandy Hill. Larger

oaks spectrally may look very similar to pine trees. From a Florida Scrub-Jay standpoint,

stands of large oaks may be as unusable as pine forests, so for modeling jay habitat it may

be unnecessary to distinguish these tree cover types.

I suspect that many of the differences in tree cover estimates resulted from

differences in the locations of transects measured on the ground vs. the image. Because

transects are sampling vegetation intercepting a thin vertical plane, relatively small





55


difference in transect position can result in large differences in measurements. Quadrat

measurements are probably less sensitive to positional inaccuracy than transects.

It was surprisingly difficult to estimate the magnitude of the positional

inaccuracies of the transect locations. The locations of the transects were predetermined

by a program that generated random locations for transect endpoints. These transect

locations were plotted on high resolution photo maps which were taken into the field and

used to stake out the transects. Thus, the correspondence between the GIS-based location

and the actual ground location depended on the field person's ability to find the exact

location from the photo map. My qualitative impression was that accuracy of positioning

the transects depended on whether features visible on the photo map could be located on

the ground. In sparsely forested areas, individual trees and bare sand patches were

identifiable on the photo and ground, and served as good reference points. Under these

conditions, a transect could probably be placed within several meters of its position on

the photo. In heavily forested areas, good reference points were absent, and positional

accuracy probably decreased, to errors of 10 meters or more. It might seem that

differential GPS could solve this potential problem. Unfortunately, several factors make

this approach more problematic than anticipated. First, the stated accuracy of the

differential GPS approach available to us is 2 - 5 meters, which can misplace the ends of

a transect by 3 pixels in any direction. Also, I have occasionally encountered averaged,

differentially corrected points that are off by considerably more than 5 meters. Second,

current GPS units often are unable to pick up the necessary signals within forests;

precisely where they are most needed for this study. Third, the positional accuracy of the

imagery itself is unknown, but is probably on the order of 2 to 10 meters (see





56


georeferencing section). Since there is no reason to expect errors in georeferencing to

have the same bias as errors in GPS readings, the difference between the two could

compound to exceed 15 meters. An error of this magnitude probably exceeds the

expected error from a field person using a high resolution photo map to position a

transect. Unfortunately, this leaves us with no way of quantifying and correcting

positional inaccuracies. Recent advances in GPS technology may solve these problems as

sub-meter accuracy becomes increasingly affordable and practical.

Quadrat samples and visual comparison of the classified image with the aerial

photographs plus field knowledge suggest that the classification accurately reflected

biologically important differences among habitats. The image processing techniques,

combined with GIS files of the locations ofjay territories and buffer zones, provided

quantitative measurements ofjay habitat in a quick and efficient manner. The quantitative

results show dramatic differences in habitat structure between the North and South areas

corresponding well to impressions reported by field researchers (R. Bowman pers.

comm.). N. Sandy Hill jays have far fewer trees within and adjacent to their territories

compared to S. Sandy Hill jays.

The image processing results confirmed our general field impressions about

which jay families were living in good and poor quality habitat on Sandy Hill. Jays that

occupied the poorest habitat were in the southern part of S. Sandy Hill. Here, several

families were living in low quality habitat near a recent bum that was occupied by two or

three other families. The presence ofjays in poor habitat resulted from conspecific

attraction (Smith and Peacock 1990) or "queueing" behavior, where individuals stay near

high quality habitat to wait for breeding vacancies. Jays living in such poor habitat,





57


adjacent to good habitat, may create high variance in habitat-demographic relationships.

Nevertheless, it is clear from the comparison of used versus available habitat that jays

preferentially selected habitat with low tree cover (Fig. 3-11). These conditions exist

throughout much of the N. Sandy Hill area, but on S. Sandy Hill they generally occurred

only where recent fires were hot enough to kill most of the pines. These burs are

embedded in a pine forest matrix (Fig. 3-9) presumably full of jay predators and

competitors. In contrast, jays in N. Sandy Hill live in higher quality habitat embedded in

a less hostile habitat matrix (Fig. 3-10). I suspect that these differences in tree cover

explain the observed difference in nonbreeder survival between the North and South

subpopulations.

In contrast to the observed difference in nonbreeder survival between the North

and South subpopulations, and the nearly significant difference in group size, few clear

relationships are obvious when pooling the two subpopulations and looking at individual

territories. The locations of jay territories relative to the areas of highest habitat quality

(Figs. 3-20 and 3-21) show that most jay territories included substantial areas of poor

quality habitat. The scrub-jay model of Duncan et al. (1995) gave similar results; about

65% of their jay territories by area had HSI values below 0.5. Jay territories typically

incorporate unusable habitat types simply because their territories are large in size

relative to the patchiness of their habitat (Woolfenden and Fitzpatrick 1984). Both the

North and South study areas are lacking in large, contiguous, high quality habitat patches.

In the North, scrub patches occur on small lenses of well-drained soil surrounded by

poorly drained soil. Scrub vegetation characteristic of high quality habitat can only grow

on these well-drained soils. In the South, much larger areas of well-drained soil capable






58

of supporting high quality habitat are present, but high quality habitat occurs only in

small areas that were recently burned or cleared experimentally. Five of 6 experimental

plots were classified as relatively large patches of high quality habitat (see Fig. 3-21).

Only plot 6 was not classified as high quality; for unknown reasons it had a dark

reflectance when the area was photographed in March 1994. No jays have become

established in the three isolated experimental plots (2, 3, and 4), despite their appearance

as high quality. We suspect that conspecific attraction is extremely important to this

species, greatly reducing the likelihood that solitary jays will become established in

unoccupied, isolated patches.

The most important relationships I found for individual territories among the

habitat-demography variables is a decline in group size with increasing tree cover, both

within territories and within the 100 m buffer zones (Figs. 3-14 and 3-15). Large family

sizes only occur in territories with low tree cover and adjacent to low tree cover. Variance

in group size is high in the North because some groups accrue large size here, but not in

the South. Territories in or adjacent to habitat with moderate to high tree cover have

predictably small group sizes, presumably a result of successive years with poor

productivity and low survival. A similar pattern between group size and tree cover can be

seen in Figs. 3-16 and 3-17, where group size is lumped into 3 categories of roughly

equal size.

Demographic parameters other than group size showed no clear patterns with the

habitat variables at the individual territory level. Group size may be the least "noisy"

measure of demographic success, since it integrates past and current demographic

performance. Helper survival may be especially sensitive to habitat quality, since helpers






59


in poor quality habitat may have a greater tendency to emigrate than helpers in high

quality habitat.

The remote sensing techniques described above show significant potential for

evaluating Florida Scrub-Jay habitat. Tree cover and bare sand are important habitat

variables that are relatively easy to measure with these techniques. Oak cover is likely to

be important to scrub-jays, but the techniques I investigated could not discriminate oaks

from other low-lying vegetation such as palmettos. Because of low mast failure and high

acorn production, oak cover may not be a limiting factor for many scrub-jay populations.

Further investigation of the importance of oak cover is needed.

The results of this study suggest that tree cover exceeding 20% - 30% within or

adjacent to territories is associated with reduced demographic performance and may

create "sink" populations of Florida Scrub-Jays. Although I lack direct evidence, much

indirect evidence suggests that forest-dwelling predators and competitors explain the

negative relationship between tree cover and demographic success. Sink populations can

constitute a major proportion of a species population and may contribute to

metapopulation longevity (Howe and Davis 1991), but the loss of a single critical source

population may result in the extinction of all dependent sink populations (Pulliam 1988).

Thus, management practices should seek to convert sink populations to self-sustaining

source populations. The results of this study suggest that for the Florida Scrub-Jay, this

entails keeping tree cover within and adjacent to jay territories at relatively low levels.





60


QWK
DTCH. TRGT
North Sandy Hill
CLVT
SQAR XOVR

TOW C
GARD.
Kissimmee Rd.



Florida Scrub Jay Territories iOE
Spring 1994





South Sandy Hill







A LO G





1 0 1 2 Kilometers F G
JSUS


Fig. 3-1. Map of Scrub-Jay Territories - Spring 1994. Dividing line between "North" and
"South" populations is the Kissimmee Rd..







61



1.0



Suitability 05-
Index




0.0
20 40 60 80
Percent Tree Cover


1.0



Suitability
Index




0.0
100
Distance to Forest (m)


1.0




Suitability 0.5-
Index




0.0
20 40 60 80
Percent Bare Sand




Fig. 3-2. Habitat Suitability Index graphs for percent bare sand, percent tree cover, and
distance to forest (modified from Duncan et al. 1995).







62






r f



-- -- ----- ----
S i -------------------------



!' .----------
*; -I
























- -TRI -- -
^ I< H
i1





















Fi. 3-3. Illustrative map of 100200 and 400 m buffer zones around LOST territory.
, ,

Fig. 3 t00nS






















Fig. 3-3. Illustrative map of 100, 200, and 400 m buffer zones around LOST territory.






63













60



50







S30







10
S20
a-



10



0 - i
0 10 20 30 40 50 60
Percent Bare Sand (Quadrat)

















Fig. 3-4. Accuracy assessment correlation graph for bare sand based on quadrat vs.
classified image measurements (r-squared = 0.60).







64














100









S60-
80



*








01





0 20 40 60 80 100
Percent Tree Cover (Transect)

















Fig. 3-5. Accuracy assessment correlation graph for tree cover based on transect vs.
classified image measurements (r-squared = 0.25).
I U*

4, * ,
o.^






Pecn re oe (rnet








Fig.3-5 Acuacasssetcreaingphfrtecoebaeonrnetv.
classified img.esrmns(rsurd=02)







65






dNML





310Y





.LA1O





IjVDS


cons .











SW.Y























Fig. 3-6. Measurements of sand, tree cover, and mixed vegetation for Spring 1994
territories on N. Sandy Hill.







66


3UON


snsr


. " .. . -U . . : . �





dO.LH


001S
















St30
00I-











VIsw






SSdO




IdO,

- a a a . .. a a







Fig. 3-7. Measurements of sand, tree cover, and mixed vegetation for Spring 1994
territories on S. Sandy Hill.
:?:?'[::

>IIU


.;..: ..~~~~~~~~~~ I:"pp la~ 8��a --- :-.:.:- i" " .: /:: !: ::1 :!iii
: .. : .: :. . . . :... .: . . . . .. :: : : .. ..... .. .. .... .... . ....::: : : :" S d

: -: .. "" -: : "!% : ;" " :?' :'"L -.: "! :., '.::. :/::' , :


.: . ..: . . - . :. . �l�,- i: ,..:.-'~% s~a~k ... . :


o d o o o



Fi.3-.Mesrmet o anteecvean iedvgeaio o Srng19

territories on S. Sandy Hill.dO






67



















40

35

30-

0
a
20

1.. 15

10

5-
0


Territory 100 m 200 m 400 m

Buffer Distance














Fig. 3-8. Percent tree cover (mean and standard deviation) for 4 zones (inside territories,
100, 200, 400 m buffer) - North vs. South Sandy Hill. South population shows
significantly higher tree cover within all zones compared to North population.








68













0.6





0.5 -*-GARD

--RADI
--CURV
-N-CTRE

0.4 -4-ARDS
---TOWR
- FLIN
-XOVR
> --BUDD
o 0.3 -0--SOAR
& --NTRL
S--CLVT
---TRGT
--DTCH
0.2 -+-FRST
-TWNP
-DUMP



0.1





0
Territory 100m 200m 400 m
Buffer Distance












Fig. 3-9. Percent tree cover for individual territories for 4 zones (inside territories, 100,
200, 400 m buffer) - North territories.







69
















0.6




0.5 ---- LOST
-4- LOPI
6d --GPSS
-)- BRIK
0.4 - TRIA
--TRIS
-+--OEAD
-WPNO
- ECHO
O0.3 ---O-SNAG
--0-- LOGG
" -t- SLOG
---- HTOP
-M- SRNG
0.2 -- FARS
-- JSUS
- NORE


0.1






Teritory 100 m 200 m 400 m
Buffer Distance














Fig. 3-10. Percent tree cover for individual territories for 4 zones (inside territories, 100,
200, 400 m buffer) - South territories.






70





















45

40 -

35

* 30
0
u 25 0 Jays
* 25 4
0I Available
S 20

15 /

10

5


South North











Fig. 3-11. Tree cover within jay territories vs. total tree cover in North vs. South Sandy
Hill. Note that jays select habitat with lower tree cover in both areas.






71










30


25


20-




>
20








Te100200,er North






















not significant.
not significant.






72























7 -"-


6 *


5 . *


o

3.4 * **


2 -** * *** *



0 0.1 0.2 0.3 0.4 0.5
Prop.
Tree Cover












Fig. 3-13. Group size vs. percent tree cover within all territories (North and South
populations pooled). Trend towards smaller group size with higher tree cover is not
significant.






73


















7--





5 -. .


C 4 *�f


3 *o *


2 **** * * * *


1 . -
0 0.2 0.4 0.6
Prop. Tree Cover
















Fig. 3-14. Group size vs. percent tree cover within 100 m buffer for all territories (North
and South populations pooled). Trend towards smaller group size with higher tree cover
is not significant.






74















0.5


0.4
(1)
0
() 0.3
ao
) o
- 02 -


0.1 -


00 -





Large Med. Small

Group size








Fig. 3-15. Group size (small = 2, medium = 3, large = 4 - 7 jays) vs. percent tree cover
within all territories (North and South populations pooled). Trend towards smaller group
size with higher tree cover is not significant.






75














0.6


0.5


9 0.4 -
0

0.3 o

0
0
o- 0.2


0.1


0.0




Large Med. Small

Group size







Fig. 3-16. Group size (small = 2, medium = 3, large = 4 - 7 jays) vs. percent tree cover
within 100 m buffer (North and South populations pooled). Trend towards smaller group
size with higher tree cover is not significant.




Full Text
Habitat Quality N. Sandy Hill
Ten-94
/\/ 1994 Jay Territories
Roads
A / Roads
1.5 Kilometers
Fig. 3-20. Habitat quality map of N. portion of South Sandy Hill.


Population Size
307
a>
N
CO
Cl
5
Q.
20 40 60
Year
Fig. 5-18c. N.E. Lake county trajectory graphs. Top) no acquisition, Bottom) maximum
acquisition.


91
retreat to their natal territory when chased by resident breeders. By contrast, the floater
strategy prevails in Western Scrub-Jays in part owing to breeder tolerance of nonbreeders
in their territory (except briefly at the beginning of breeding season; Carmen 1989;
Koenig et al. 1992). This option of floating among breeders in high quality habitat is
unavailable to Florida Scrub-Jays because breeders are largely intolerant of floaters.
Thus, as floaters Florida Scrub-Jays would be forced to reside in marginal habitat where
they would suffer high predation rates as documented in Fitzpatrick and Woolfenden
(1986). Western Scrub-Jays also have an advantage here, since their survival rates in
marginal habitat are nearly the same as in high quality habitat (Carmen 1989; Koenig et
al. 1992).
The above considerations are nicely encapsulated by Fitzpatrick and Woolfenden
'V
(1986) in an evolutionary model of dispersal for the genus Aphelocoma. Their model
suggests that selection will favor individuals who attempt to breed immediately upon
maturation, rather than delaying dispersal, unless the cost of dispersal is high compared
to the cost of remaining home and not breeding. Delaying dispersal, foregoing early
breeding opportunities, and engaging in low risk forays are the predominant dispersal
behavior for Florida Scrub-Jays in natural, high quality habitat. Nevertheless,
documented cases of long distance dispersal by some Florida Scrub-Jays make it clear
that they sometimes engage in risky dispersal behavior by moving through matrix
habitat between high quality habitat patches. Such behavior, though not completely
analogous to Western Scrub-Jay dispersal, can be modeled as floater dispersal behavior.
These considerations suggest that a reasonable approach to modeling dispersal in the


Table 5-26 (cont.)
Metapopulation
Primary Acquisition
Targets
Secondary
Acquisition
Targets
Restoration (protected areas)
Other actions
Sarasota (M5)
Char2, Char3, Char4,
Sar3, Sar5, SarlO
Char5, Char6,
Sari, Sar2,
Sar7 (Casperson/Brohard), Sar4
(Lemon Bay Scrub C.P.), Chari
(Charlotte Harbor St. Buffer Pr.),
Sari4 (Myakka St. Forest)
Palm Beach
(Ml 6)
PB10 (Overlook Scrub),
PB13 (Tradewind/
Winchester)
PB14 (Yamato Scrub NAP), PB11
(Rolling Green Scrub Pr.), PB12
(Galaxy School Scrub Pr.)
Evaluate genetics;
consider translocation
program for educational
facilities
Central Lake
(M20)
Lake2, Lake3
Lake4, Lakel
Restore abandoned
citrus groves?
Flagler (M9)
Flagl, Flag2
Voll (N. Peninsula St. Rec. Area),
Evaluate genetics;
Investigate unoccupied
inland habitat for
translocation?
N.E. Lake
(M18)
Lake9, Lake 10
Lakd7 & Lake8 (Seminole St. Forest),
Ora2 & Lake6 (Rock Springs Run
S.Res.), Oral (Wekiwa Springs S.P.),
Martin (Ml5)
Marl 5, additions to
Marl 1, PB6, PB7
Marl2 (Jonathan Dickinson S.P.),
MarlO (Sea Branch S.P.), PB8 (Juno
Hills N.A.P.), all other protected areas
Merritt Island
(M10)
Brevl9, Brev20, Brev21,
Brev22, Brev23
Brev25 (Merritt Island N.W.R. &
Kennedy Space Center), Brev26
(Cape Canaveral Air Station)


215
Jay Territory Locations
(after restoration)
Scrub Polygons
CCC## Polygon ID
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Protection Status
/\f Protected
/\/ Unprotected
Water Bodies M7 Central Charlotte
County lines
9 Kilometers
1 : 250,000
Fig. 5-7b. Central Charlotte county acquisition map.


156
marsh and mesic flatwoods with high densities of pine. The estimated population sizes
after restoration (Table 5-la) may be overly optimistic.
Simulation results: This metapopulation ranked 5th in vulnerability (table 5-23),
17th in percent protected (22.7%; table 5-24), and 2nd in priority (table 5-25), with high
vulnerability and high potential for improvement. Simulations of the SMP configuration
indicate that the 1992-1993 configuration is extremely vulnerable to extinction (Table 5-
lb; extinction risk = 1.0; percent population decline = 100.0).
Simulations of the currently protected, restored configuration indicate that the
protected population would be vulnerable to extinction (Table 5-lb; extinction risk = 0.1;
Fig. 5-Id; percent population decline = 29.4%).
Because the habitat as mapped shows no fragmentation, no configurations
involving multiple patches were simulated. All acquisition configurations assume that
preference is given to acquiring contiguous habitat, but even if multiple patches were
created, the resulting interpatch distances would be small.
The 30% acquisition configuration was estimated to support about 34 jay families
(Table 5-la). Simulations of this configuration indicate that it has a small but significant
probability of falling below 10 families (Table 5-lb; quasi-extinction = 0.05). The mean
population trajectory shows an 8.8% decline.
The 70% acquisition configuration was estimated to support about 54 jay families.
Simulations of this configuration indicate that the population would not be vulnerable to
extinction or quasi-extinction (Table 5-lb and Fig. 5-Id; extinction risk = 0.0; quasi
extinction = 0.0). The mean population trajectory shows a 3.7% decline (Fig. 5-lc).


Table 4-2. Demographic parameters used to estimate the proportion of disappearing
helpers that become floaters.
121
Age Class
Dtotai
D b
i^eq
Enoniocai
Enoniocai/
Dtotai
P floater
Male
Yearling
0.225
0.22
0.005
0.022
0.022-0.75
(0.25)
Female
Yearling
0.42
0.35
0.07
0.167
0.167-0.75
(0.75)
Male 2nd
Year
Helper
0.25
0.15
0.10
0.40
0.40 0.75
(0.50)
Female 2nd
0.375
0.26
0.115
0.307
0.307-0.75
Year (0.75)
Helper
a Dtotai is the observed proportion of total disappearances (from Woolfenden and
Fitzpatrick 1984; Table 9.5; p. 275 ).
b Deq is the equilibrium mortality rate (from Woolfenden and Fitzpatrick 1984; Table 9.5;
p. 275 ).
c Enoniocai is the proportion of total disappearances due to dispersers becoming breeders
elsewhere as unobserved emigrants (from Woolfenden and Fitzpatrick 1984;
Appendix M) calculated as Dtotai Deq.
d Enoniocai/Dtotai is the proportion of Dtotai estimated to become breeders elsewhere as
emigrants. This value sets a lower limit for Pfioater-
e Pfioater is the proportion of Dtotai becoming floaters. A plausible range of values is listed
first, followed by a best guess in parentheses.


161
0)
N
W
c
c
TO
D
Q.
C
CL
Fj+H^ffTfU-f-j-l fH 11 -[ | ~f I 11 H~1 H H l-t-j-j
20
40
Year
60
Fig. 5-lc. Levy county trajectory graphs. Top) no acquisition, Bottom) maximum
acquisition.


290
Table 5-16a. South Palm Beach county patch statistics (number of jay territories for
different configurations)
Patch id
Status
1992-1993#
jay territories
No acquisition
(restored)
Maximum
acquisition
PB10
(Overlook Scrub -
proposed)
2
5
PB11
Rolling Green Scrub
Pr.
2
2
2
PB12
Galaxy School
Scrub Pr.
1
1
1
PB13
(Tradewind /
Winchester Site)
2
2
PB14
Yamato Scrub NAP
1
6
6
Totals
8
9
16


237
maximum acquisition option produced very similar results, with no extinction or quasi
extinction risk and very low percent population declines (Table 5-10b).
Recommendations: Although this is a large metapopulation, vulnerability to
hurricanes, habitat overgrowth and difficulties with habitat restoration pose serious
threats to this metapopulation. Modeling performed by Breininger et al. (in press) found
this metapopulation to be vulnerable to catastrophes associated with hurricanes, but
habitat degradation was a much more important risk factor. Years of fire suppression
have resulted in overgrown habitat which is difficult to restore compared to other areas,
apparently because the coastal soils and water table allow rapid regrowth of scrub oaks
and other vegetation, resulting in the closure of openings needed by jays for foraging and
predator detection. Preliminary results from a 1999 survey of MIN WR/KSC suggest that
the population may have declined as much as 50% compared to estimates made during
the SMP (Gary Popotnik, pers. comm.). Habitat restoration is urgently needed for this
metapopulation.


Population Size
225
Fig. 5-8c. Lee and N. Collier county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.


2-8. Examples of non-equilibrium metapopulations from North Gulf coast. Each
of the six metapopulations contains fewer than 10 pairs of jays, except for
the centrally located system that contains a single, midland-size
subpopulation 37
2-9. Example of a "classical" metapopulation from five counties in Central
Florida. Note the occurrence of jays in small islands of intermediate distance
from one another 38
2-10. Portion of the largest mainland-midland-island metapopulation in the
interior, consisting of the Lake Wales Ridge and associated smaller sand
deposits. The large central subpopulation (enclosed by the thin black line)
contains nearly 800 pairs of jays. Small subpopulations to the south and east
are within known dispersal distance of the large, central mainland. A small
metapopulation to the west (in DeSoto County) contains a single
subpopulation of 21 territories. This small system qualifies as a patchy
metapopulation, since jays occur in two or more patches but the patches are
so close together that they function as a single demographic unit 39
3-1. Map of Scrub-Jay Territories Spring 1994. Dividing line between North
and South populations is the Kissimmee Rd 60
3-2. Habitat Suitability Index graphs for percent bare sand, percent tree cover, and
distance to forest (modified from Duncan et al. 1995) 61
3-3. Illustrative map of 100, 200, and 400 m buffer zones around LOST territory 62
3-4. Accuracy assessment correlation graph for bare sand based on quadrat vs.
classified image measurements (r-squared = 0.60) 63
3-5. Accuracy assessment correlation graph for tree cover based on transect vs.
classified image measurements (r-squared = 0.25) 64
3-6. Measurements of sand, tree cover, and mixed vegetation for Spring 1994
territories on N. Sandy Hill 65
3-7. Measurements of sand, tree cover, and mixed vegetation for Spring 1994
territories on S. Sandy Hill 66
3-8. Percent tree cover (mean and standard deviation) for 4 zones (inside
territories, 100, 200, 400 m buffer) North vs. South Sandy Hill. South
population shows significantly higher tree cover within all zones compared
to North population 67
3-9. Percent tree cover for individual territories for 4 zones (inside territories, 100,
200,400 m buffer) North territories 68
xvi


360
are taken). Furthermore, the assumption that relative ranking is robust to model
inaccuracies apparently has yet to be tested, and it is not difficult to imagine situations
where relative rankings could change substantially depending on model parameter
settings. For example, when comparing the viability of two metapopulations, one
occurring in a single large patch, the other occurring in several small, isolated patches,
the assumptions made about dispersal, the spread of epidemics, the correlation of
environmental stochasticity, etc. could alter the relative ranking of the two
metapopulations viability.
I argue that these criticisms of models are excessive and stem from an unrealistic
view of models. It is a banal truism that all models are false; none are completely
accurate representations of real systems. Even simple, closed systems studied by
physicists cannot be accurately modeled except in a very restricted sense (Cartwright
1983; Giere 1999). The main issue is whether a given model is sufficiently accurate to
meet a particular need (Rykiel 1994). For this chapter, highly accurate predictions of the
fate of Florida Scrub-Jays were not needed. I am interested primarily in comparing gross
trends in the trajectories of hypothetical populations that might exist if present habitat
conditions were improved.
In all likelihood, jay habitat will not be restored as assumed by the reserve design
scenarios. Continued fire suppression and difficulties associated with habitat restoration
will result in much less restored habitat than assumed by the scenarios. Furthermore, even
if jay habitat were restored, the model results probably overestimate the persistence of
jays. This is because I chose slightly optimistic settings for two parameters to favor
small populations. Best estimates of the strength and frequency of epidemics measured at


Probability
173
10 20 30 40
Threshold Pop. Size
1 oT
r
-
08^
o
a!
0 4^
i i
50 100
Threshold Pop Size
Fig. 5-2e. Citrus and S. Marion county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.


119
other model (Letcher et al. 1998) that simulates both philopatric and floater dispersal. No
models were found that generated dispersal and stage-age distributions to compare with
field data. The level of detail and biological realism in this model are high due to the
availability of long term research data and crude telemetry data. The constraint analysis
proved valuable in placing bounds on acceptable parameter settings. Nonetheless,
numerous simplifying assumptions were made, as is true of any model. A full discussion
of the models assumptions is given in chapter 5, which describes the complete model.
The evaluation of the dispersal module presented in this chapter exemplifies a useful
technique of separately validating individual model components (Thomas et al. 1990).
Nonetheless, it is important to investigate the sensitivity of the overall model to potential
errors in the dispersal module and other model components. Such comparisons may
reveal that the model is crucially or inconsequentially affected by the dispersal module.


47
Digitization of Territories and Background Features
Polygons representing the boundaries of jay territories and other background
features (e.g. roads) were digitized directly off of the computer screen with a mouse from
the mosaiced image using Arc/Info. This allowed the resulting coverages to be
automatically georeferenced to the aerial photography
Tree Cover Buffering Procedure
GIS buffers were generated around each jay territory at 100, 200, and 400 m
distances using Arc/Info. These buffers were used to characterize the habitat immediately
surrounding each territory. A major complication with this procedure was that the buffers
for each territory often overlapped with neighboring territories, and overlapping polygons
are not allowed in Arc/Info coverages. Therefore, each territory was kept in a separate
coverage and analyzed individually. The following procedure was used. Each coverage
was buffered at distances of 100, 200, and 400 meters. The resulting coverages were
converted to individual ERDAS .dig files. Each .dig file was overlaid on the classified
image to calculate the percent tree cover within each buffer using an ERDAS program
called POLYSTAT (developed by B. Stith and J. Richardson). The statistics generated
from this program were imported into spreadsheet :.nd statistical packages (Systat; SAS)
for further analysis.


21
large habitat gaps. Still, few opportunities exist to observe jays in the act of dispersing,
hence the theoretical maximum dispersal-distance (i.e., the outer dispersal buffer) is
extremely difficult to establish directly.
Seeking an indirect measure of dispersal frequencies across habitat gaps, we
examined patch occupancy statewide as documented by our 1992-93 survey. We used
Fragstats software (McGarigal and Marks 1994) to measure distances between each
occupied patch of scrub habitat to its nearest neighboring, occupied patch. We then
measured (by hand, as Fragstats cannot measure distances between patches of different
attributes) the distances between each unoccupied suitable patch and the nearest occupied
patch. For each distance class, the ratio of the count of the occupied-to-occupied
distances to the total number of nearest neighbor distances yields the proportion of
patches that are occupied at that distance away from occupied habitat.
Presumably, declines in patch occupancy with increasing distance to the nearest
occupied habitat (Fig. 2-5) reflect diminishing recolonization rates following local
extinctions. Occupancy remains above zero even at great distances, probably because
larger isolated patches rarely experience extinction. This curve provides an empirical
approach for delineating subpopulations and metapopulations: a subpopulation buffer is
the maximum interpatch distance where occupancy rates remain high; the metapopulation
buffer is the smallest interpatch distance where occupancy rates reach their minimum.
For Florida Scrub-Jays (Fig. 2-5) patch occupancy is about 90% to at least 2 km
from a source, then declines monotonically to around 15% at 12 km. (Sample size of
isolated patches decreased rapidly beyond 16 km, necessitating lumping of classes at the
larger distances.) We infer from this occupancy curve that successful recolonization is a


METAPOPULATION DYNAMICS AND LANDSCAPE ECOLOGY OF THE
FLORIDA SCRUB-JAY, APHELOCOMA COERULESCENS
By
BRADLEY M. STITH
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1999

Dedicated to my wife and best friend, Ellen Mary Thoms

ACKNOWLEDGMENTS
I wish to acknowledge each of my committee members for their unique
contributions to this dissertation. My chairman, Dr. Stephen R. Humphrey, stimulated my
interests in issues relating to philosophy, policy, and management. His broad interests
and ability to take on multiple careers were remarkable. In my times of need he provided
unfailing support while allowing me great freedom to pursue my own course of study. As
director of Archbold Biological Station, co-chairman Dr. John W. Fitzpatrick provided
me with a wonderful opportunity to study the Florida Scrub-Jay at this world-class
research facility. His ability to perform brilliantly as director, researcher, and
conservationist is inspirational. I am grateful to him for sending a generous U.S.F.W.S.
grant my way to support me towards the end of my program. I was also privileged to
work under the tutelage of the legendary pioneer of Florida Scrub-Jay research, Dr. Glen
E. Woolfenden, who showed me the world from a Florida Scrub-Jays perspective. My
collaborative research with Dr. Lyn Branch on the Vizcacha and the Florida scrub lizard,
was a joy. She forced me to think about landscape ecology and the metapopulation
dynamics of lesser organisms. Her support, financial and otherwise, was extraordinary,
and a nicer person I have never met. Dr. Jon Allen exposed me to population modeling
from an entomologists perspective. His enthusiasm for teaching and helping graduate
students solve technical problems was exceptional.
The staff at Archbold Biological Station made my stay there rewarding beyond
words. Working with Dr. Reed Bowman at the Avon Park Bombing Range, and learning
iii

about his suburban Florida Scrub-Jays, was most stimulating. His invitation to participate
in the Brevard county Habitat Conservation Plan scientific committee was most
appreciated. Special thanks go to Steve Friedman and Roberta Pickert for their help with
GIS problems. Dan Childs, former manager of the affiliated MacArthur Agricultural
Research Center, and his staff provided unfailing assistance in many ways. Current
Archbold director Dr. Hilary Swain provided generous travel support for the 1997 AOU
meeting.
For collecting the bulk of the demographic data used in chapter 3,1 am indebted
to Reed Bowman, Doug Stotz, Larry Riopelle, Natalie Hamel, and Mike McMillan. I
thank the staff at the Natural Resources Office at the Avon Park Air Force Range,
especially Bob Progulske, Paul Ebersbach, and Pat Walsh for their support.
Dave McDonald was instrumental in recruiting me to work on the statewide jay
survey and exposing me to the opportunities at Archbold. Dr. Keith Tarvin and Dr. Curt
Atkinson kindly lent me radiotelemetry equipment and provided much valuable advice.
Special thanks go to Dr. Ron Mumme who kindly allowed me to use color banded jays
from the south tract of Archbold which he has been monitoring for many years. His prior
work in banding, taming, and sexing these jays was critical to identifying candidates for
the displacement experiment. Bill Pranty worked tirelessly with me to digitize major
portions of the statewide survey. Steve Schoech provided occasional field support and
many hours of entertainment on the Archbold tennis court.
For stimulating discussions about jays and landscape rules, I am indebted to many
colleagues engaged in the study of Florida Scrub-Jays and their conservation. Participants
in the Habitat Conservation Planning group whom I have not yet mentioned include
IV

David Breininger, Grace Iverson, Michael OConnell, Parks Small, Jon Thaxton, and
Brian Toland. The following people contributed data for the statewide survey: Reed
Bowman, Dave Breininger, Jack Dozier, Florida Game and Fresh Water Fish
Commission personnel, Grace Iverson, David McDonald, Ron Mumme, Ocala National
Forest personnel, Bill Pranty, Hilary Swain, Jon Thaxton, and Brian Toland. I thank
David Wesley and Dawn Zattau, of the U.S. Fish and Wildlife Service, who provided
lead funding for the statewide survey and helped stimulate our discussions of habitat
conservation planning.
For special assistance with site-specific questions related to the maps in chapter 5:
Mary Barnwell, Jim Beever, Reed Bowman, Dave Breininger, Mike Eagen, Mary
Huffman, Grace Iverson, Mike Jennings, Laura Lowry, Dan Pearson, Gary Popotnik, Bill
Pranty, Park Smalls, Hank Smith, Jon Thaxton, Brian Toland, Jane Tutin. For assistance
with GIS data: Reed Bowman, Dave Breininger, Kathy Bronson, Beth Needham, Bill
Pranty, Roberta Pickert, Katy NeSmith. For advice regarding modeling: Reed Bowman,
Dave Breininger, John Fitzpatrick, Glen Woolfenden. Bill Pranty deserves special thanks
for the exceptionally detailed information he has collated on scrub-jay locations around
the state (Pranty et al. manuscript). Input from members of the Recovery Team was most
helpful. Thanks go to the U.S.F.W.S. for funding the research in chapter 5, and especially
to Dawn Zattau for her support and patience.
I thank the support staff at the University of Florida Department of Wildlife
Ecology and Conservation. Joe Gasper provided extraordinary computer support.
Leonard Pearlstein and the entire USFWS coop unit helped with innumerable computer
problems. I thank support staff at Circa computing, especially Jiannong Xin for
v

programming resources, and John Dixon for statistical help. Dr. Ken Portier also
provided valuable statistical advice.
Last, but no least, I thank my family and especially my wife, Ellen Thoms.
vi

TABLE OF CONTENTS
Eige
ACKNOWLEDGMENTS iii
LIST OF TABLES xii
LIST OF FIGURES xv
ABSTRACT xxiw
CHAPTERS
1 INTRODUCTION 1
Historical Background 1
Biological Background 3
Objectives 4
2 CLASSIFYING FLORIDA SCRUB-JAY METAPOPULATIONS 9
Introduction 9
Statewide Survey of the Florida Scrub-Jay 11
Statewide Survey: Methods 11
Statewide Survey: Results 13
A Method for Classifying Metapopulations 14
Metapopulation Structure of the Florida Scrub-Jay 18
Dispersal Distances 19
Patch Occupancy 20
Population Viability Analysis 22
Metapopulation Structure 23
Caveats 25
3 REMOTE SENSING OF FLORIDA SCRUB-JAY HABITAT 40
Introduction 40
Methods 42
Image Source 42
Image Scanning and Conversion 43
Image Rectification and Mosaicing 43
vii

Image Classification 44
Manual Editing of Classification 45
Assessment of Classification Accuracy 46
Digitization of Territories and Background Features 47
Tree Cover Buffering Procedure 47
Habitat Quality Model 48
Collection of Demographic Data 49
Habitat-Demographic Analysis 49
Results 50
Discussion 54
4 MODELING DISPERSAL IN THE FLORIDA SCRUB-JAY 86
Introduction 86
Dispersal Strategies 89
Dispersal Traits of the Florida and Western Scrub-Jay 90
Methods 92
General Approach 92
GIS Files 93
Simulating Philopatric Dispersal 93
Short distance dispersal algorithm development and calibration 94
Simulating Long Distance Dispersal 95
Estimating floater mortality and mobility 96
Habitat attractiveness 97
Floater detection radius 97
Estimating floater frequency 98
Floater algorithm development and calibration 102
Jay displacement experiment 103
Constraint Analysis 106
Model Validation 108
Results 108
Radiotelemetry Displacement Experiment 108
Floater Parameter Estimation 111
Constraint Analysis 112
Calibration and Validation 112
Discussion 113
5 METAPOPULATION VIABILITY ANALYSIS OF THE FLORIDA SCRUB-
JAY 132
Introduction and Objectives 132
Methods 135
Simulation Model Description 135
Life Stages 136
Starting Population Stage Structure 137
Annual Life Cycle 137
Territories 140
Vlll

Background Landscape Image 141
Map Production 143
Statewide metapopulation map 143
1992-1993 SMP maps 143
Acquisition maps 144
GIS Database Preparation 146
Estimation of jay populations after restoration 146
Identification of protected areas 147
Assessment of unprotected areas 147
Suburban jays 148
Simulation runs 149
Repetitions and duration of simulations 149
Reserve design configurations 150
Output statistics 151
Model Validation/Calibration 152
Interpreting Simulation Results 152
Results 154
Levy (Cedar Key) (Ml) 155
Citrus-S.W. Marion (M2) 164
Pasco-Hemando (M3) 176
Manatee-S. Hillsborough (M4) 186
Sarasota-W. Charlotte (M5) 196
N. W. Charlotte (M6) 204
Central Charlotte (M7) 212
Lee and N. Collier (M8) 220
Flagler-N.E. Volusia (M9) 228
Merritt Island-S.E. Volusia and (M10) 236
N. Brevard (Mil) 244
Central Brevard (Ml2) 253
S. Brevard-Indian River-N. St. Lucie (Ml3) 261
St. Lucie N. Martin (Ml4) 270
Martin and N. Palm Beach (Ml5) 278
South Palm Beach (Ml6) 286
Ocala National Forest (Ml 7) 294
N.E. Lake (Ml 8) 302
S.W. Volusia (M19) 310
Central Lake (M20) 318
Lake Wales Ridge (M21) 326
Other Metapopulations 344
Brevard barrier island 344
Clay county 344
Osceola 344
Western Polk 345
Bright Hour Ranch 345
Recommendations 346
Ranking Metapopulation Vulnerability 346
IX

Summary of Recommendations 349
Discussion 359
6 SYNTHESIS 365
REFERENCES 372
BIOGRAPHICAL SKETCH 383
x

LIST OF TABLES
Table page
2-1. Summary information for 42 Florida Scrub-Jay metapopulations, including
type of metapopulation, number of territories (pairs of jays), and number of
subpopulations 28
3-1. Demographic and habitat parameters for North and South Sandy Hill (1994
- 1995) 81
3-2. Kolmogorov-Smimov test for normality of demographic and habitat
variables (* significantly different from normal) 84
3-3. Mann-Whitney_U test for differences in demographic and habitat variables
between North and South jay populations (* significantly) 85
4-1. Landcover types from statewide habitat map (Kautz et al. 1993) used in
simulations and associated floater attractiveness values 120
43. Summary of philopatric dispersal rules showing sex differences and rules
used to implement the algorithm 122
4-4. Summary of jay movement data obtained from displacement experiment
(distances in km) 123
4-5. Summary of constraint analysis for 9 simulation scenarios (50 years x 30
repetitions) showing number of colonizations from Lake Wales Ridge to
Bright Hour Ranch, DeSoto county, Florida 124
5-1. Demographic and dispersal parameter settings for jays in optimal and
suburban conditions 142
5-la. Levy county patch statistics (number of jay territories for different
configurations) 160
5-lb. Levy county (Cedar Key) simulation statistics 163
5-2a. Citrus and S. Marion county patch statistics (number of jay territories for
different configurations) 172
xi

5-2b. Citrus and S. Marion county simulation statistics 175
5-3a. Pasco and Hernando county patch statistics (number of jay territories for
different configurations) 182
5-3b. Pasco county simulation statistics 185
5-4a. Manatee and S. Hillsborough county patch statistics (number of jay
territories for different configurations) 191
5-4b. Manatee and S. Hillsborough county simulation statistics 195
5-5a. Sarasota and W. Charlotte county patch statistics (number of jay territories
for different configurations) 200
5-5b. Sarasota and W. Charlotte county simulation statistics 203
5-6a. N. W. Charlotte county patch statistics (number of jay territories for
different configurations) 208
5-6b. N. W. Charlotte county simulation statistics 211
5-7a. Central Charlotte county patch statistics (number of jay territories for
different configurations) 216
5-7b. Central Charlotte county simulation statistics 219
5-8a. Lee and N. Collier county patch statistics (number of jay territories for
different configurations) 224
5-8b. Lee and N. Collier county simulation statistics 227
5-9a. Flagler and N.E. Volusia county patch statistics (number of jay territories
for different configurations) 232
5-9b. Flagler and N.E. Volusia county simulation statistics 235
5-10a. S.E. Volusia and Merritt Island county patch statistics (number of jay
territories for different configurations) 240
5-10b. S.E. Volusia and Merritt Island county simulation statistics 243
5-1 la. N. Brevard county patch statistics (number of jay territories for different
configurations) 249
5-1 lb. N. Brevard county simulation statistics 252
xii

5-12a. Central Brevard county patch statistics (number of jay territories for
different configurations) 257
5-12b. Central Brevard county simulation statistics 260
5-13a. S. Brevard-Indian River-N. St. Lucie county patch statistics (number of jay
territories for different configurations) 266
5-13b. S. Brevard-Indian River-N. St. Lucie county simulation statistics 269
5-14a. St. Lucie N. Martin county patch statistics (number of jay territories for
different configurations) 274
5-14b. St. Lucie county simulation statistics 277
5-15a. Martin and N. Palm Beach county patch statistics (number of jay territories
for different configurations) 282
5-15b. Martin and N. Palm Beach county simulation statistics 285
5-16a. South Palm Beach county patch statistics (number of jay territories for
different configurations) 290
5-16b. South Palm Beach county simulation statistics 293
5-17a. Ocala National Forest county patch statistics (number of jay territories for
different configurations) 298
5-17b. Ocala National Forest county simulation statistics 301
5-18a. N.E. Lake county patch statistics (number of jay territories for different
configurations) 306
5-18b. N.E. Lake county simulation statistics 309
5-19a. S.W. Volusia county patch statistics (number of jay territories for different
configurations) 314
5-19b. S.W. Volusia county simulation statistics 317
5-20a. Central Lake county patch statistics (number of jay territories for different
configurations) 322
5-20b. Central Lake county simulation statistics 325
xiii

5-2 la. Lake Wales Ridge patch statistics (number of jay territories for different
configurations) 339
5-2lb. Lake Wales Ridge simulation statistics 343
5-22. Metapopulation viability statistics 350
5-23. Metapopulation vulnerability ranking no acquisition (sorted by
decreasing quasi-extinction probability) 351
5-23a. Metapopulation vulnerability ranking maximum acquisition (sorted by
increasing percent protection) 352
5-24. Percent protected ranking (sorted by increasing percent protection) 353
5-25. Metapopulation priority ranking (sorted by decreasing priority) 354
5-26. Summary of recommendations (highest priority first) 355
xiv

LIST OF FIGURES
Figure page
2-1. 1993 distribution of Florida Scrub Jay groups (small black circles). Note the
discontinuous distribution and variability in patterns of aggregation 30
2-2. Classification scheme showing different types of metapopulations based on
patch size distribution (patches all small in size, mixture of small and large,
and all large in size) along the horizontal axis, and degree of patch isolation
(highly connected to highly isolated) on the vertical axis. Nonequilibrium,
classical, mainland-island, and patchy classes are named according to
Harrison (1991) 31
2-3. Schematic depiction of different kinds of metapopulations, illustrating use of
dispersal-distance buffers to predict recolonization rates among
subpopulations. Dotted lines separate functional subpopulations, based on
frequency of dispersal beyond them. Solid lines separate metapopulations,
based on poor likelihood of dispersal among them. A. Patchy
metapopulation. B. Classical metapopulation. C. Nonequilibrium
metapopulations. D. Mainland-island metapopulation 32
2-4. Dispersal frequency curve. Dispersal distances from natal to breeding
territories for color-banded jays at Archbold Biological Station, 1970-1993.
About 85% of documented dispersals were within 3.5 km, and 99% within
8.3 km. The longest documented dispersal was 35 km 33
2-5. Proportion of suitable habitat patches occupied by Florida Scrub-Jays as a
function of their distance to the nearest separate patch of occupied habitat.
Occupancy rates are high (nearly 90 %) for patches up to 2 km apart and
decline monotonically to 12 km. Note the scale change after 16 km 34
2-6. Statewide jay distribution map with dispersal buffers. Shaded areas with thin,
solid lines depict subpopulations of jays within easy dispersal distance (3.5
km) of one another. Thick lines delineate demographically independent
metapopulations separated from each other by at least 12 km 35
2-7. Frequency of Florida Scrub-Jay metapopulation sizes. Note that 21
metapopulations have 10 pairs or less of jays. These represent
nonequilibrium metapopulations 36
xv

2-8. Examples of non-equilibrium metapopulations from North Gulf coast. Each
of the six metapopulations contains fewer than 10 pairs of jays, except for
the centrally located system that contains a single, midland-size
subpopulation 37
2-9. Example of a "classical" metapopulation from five counties in Central
Florida. Note the occurrence of jays in small islands of intermediate distance
from one another 38
2-10. Portion of the largest mainland-midland-island metapopulation in the
interior, consisting of the Lake Wales Ridge and associated smaller sand
deposits. The large central subpopulation (enclosed by the thin black line)
contains nearly 800 pairs of jays. Small subpopulations to the south and east
are within known dispersal distance of the large, central mainland. A small
metapopulation to the west (in DeSoto County) contains a single
subpopulation of 21 territories. This small system qualifies as a patchy
metapopulation, since jays occur in two or more patches but the patches are
so close together that they function as a single demographic unit 39
3-1. Map of Scrub-Jay Territories Spring 1994. Dividing line between North
and South populations is the Kissimmee Rd 60
3-2. Habitat Suitability Index graphs for percent bare sand, percent tree cover, and
distance to forest (modified from Duncan et al. 1995) 61
3-3. Illustrative map of 100, 200, and 400 m buffer zones around LOST territory 62
3-4. Accuracy assessment correlation graph for bare sand based on quadrat vs.
classified image measurements (r-squared = 0.60) 63
3-5. Accuracy assessment correlation graph for tree cover based on transect vs.
classified image measurements (r-squared = 0.25) 64
3-6. Measurements of sand, tree cover, and mixed vegetation for Spring 1994
territories on N. Sandy Hill 65
3-7. Measurements of sand, tree cover, and mixed vegetation for Spring 1994
territories on S. Sandy Hill 66
3-8. Percent tree cover (mean and standard deviation) for 4 zones (inside
territories, 100, 200, 400 m buffer) North vs. South Sandy Hill. South
population shows significantly higher tree cover within all zones compared
to North population 67
3-9. Percent tree cover for individual territories for 4 zones (inside territories, 100,
200,400 m buffer) North territories 68
xvi

3-10. Percent tree cover for individual territories for 4 zones (inside territories,
100,200,400 m buffer) South territories 69
3-11. Tree cover within jay territories vs. total tree cover in North vs. South
Sandy Hill. Note that jays select habitat with lower tree cover in both areas 70
3-12. Percent bare sand (mean and standard deviation) for 4 zones (inside
territories, 100, 200, 400 m buffer) North vs. South Sandy Hill.
Differences between two areas are not significant 71
3-13. Group size vs. percent tree cover within all territories (North and South
populations pooled). Trend towards smaller group size with higher tree
cover is not significant 72
3-14. Group size vs. percent tree cover within 100 m buffer for all territories
(North and South populations pooled). Trend towards smaller group size
with higher tree cover is not significant 73
3-15. Group size (small = 2, medium = 3, large = 4-7 jays) vs. percent tree cover
within all territories (North and South populations pooled). Trend towards
smaller group size with higher tree cover is not significant 74
3-16. Group size (small = 2, medium = 3, large = 4-7 jays) vs. percent tree cover
within 100 m buffer (North and South populations pooled). Trend towards
smaller group size with higher tree cover is not significant 75
3-17. Images and territories (black polygons) of North Sandy Hill. Right: color-
infrared image. Left: classified image (white = bare sand; green = trees;
brown = shrubs/grass; black = water) 76
3-18. Images and territories (black polygons) of N. portion of South Sandy Hill.
Right: color-infrared image. Left: classified image (white = bare sand;
green = trees; brown = shrubs/grass; black = water) 77
3-19. Images and territories (black polygons) of S. portion of South Sandy Hill.
Right: color-infrared image. Left: classified image (white = bare sand;
green = trees; brown = shrubs/grass; black = water) 78
3-20. Habitat quality map of N. portion of South Sandy Hill 79
3-21. Habitat quality map of S. portion of South Sandy Hill 80
4-1. Daily distances moved and number of days movements were tracked for 10
jays released at 3 sites in Highlands county, Florida 125
xvii

4-2. Kaplan-meier survival curves of daily survival rates for 10 jays released at 3
sites in Highlands county, Florida. Upper curve: maximum possible
survival; middle curve: best guess survival; lower curve: minimum
possible survival 126
4-3. Distribution of daily distances moved by released jays (solid line), and
inverse function fitted to observed movements (dashed line) 127
4-4. Comparison of dispersal data from Archbold Biological Station and
simulated dispersal distances for male jays 128
4-5. Comparison of dispersal data from Archbold Biological Station and
simulated dispersal distances for female jays 129
4-6. Comparison of stage-age data from Archbold Biological Station and
simulated stage-age data for breeders 130
4-7. Comparison of stage-age data from Archbold Biological Station and
simulated stage-age data for helpers 131
5-0. Delineations of 21 Florida Scrub-Jay metapopulations based on 1992 1993
statewide survey 145
5-la. Levy county maps 1992 1993 jay and habitat distribution 158
5-lb. Levy county acquisition map 159
5-lc. Levy county trajectory graphs. Top) no acquisition, Bottom) maximum
acquisition 161
5-ld. Levy county quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition 162
5-2a. Citrus county map 1992-1993 jay and habitat distribution 168
5-2d. S.W. Marion county acquisition map 171
5-2e. Citrus and S. Marion county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition 173
5-2f. Citrus and S. Marion county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition 174
5-3a. W. Pasco and Hernando county map 1992 1993 jay and habitat
distribution 178
XVlll

5-3b. E. Pasco and Hernando county map 1992 1993 jay and habitat
distribution 179
5-3c. W. Pasco and Hernando county acquisition map 180
5-3d. E. Pasco and Hernando county acquisition map 181
5-3e. Pasco and Hernando county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition 183
5-3f. Pasco and Hernando county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition 184
5-4a. Manatee and S. Hillsborough county map 1992 1993 jay and habitat
distribution 189
5-4b. Manatee and S. Hillsborough county acquisition map 190
5-4c. Manatee and S. Hillsborough county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition 193
5-4d. Manatee and S. Hillsborough county quasi-extinction graphs. Top) no
acquisition, Bottom) maximum acquisition 194
5-5a. Sarasota and W. Charlotte county map 1992 1993 jay and habitat
distribution 198
5-5b. Sarasota and W. Charlotte county acquisition map 199
5-5c. Sarasota and W. Charlotte county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition 201
5-5d. Sarasota and W. Charlotte quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition 202
5-6a. N. W. Charlotte county map 1992 1993 jay and habitat distribution 206
5-6b. N. W. Charlotte county acquisition map 207
5-6c. N. W. Charlotte county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition 209
5-6d. N. W. Charlotte county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition 210
5-7a. Central Charlotte county map 1992 1993 jay and habitat distribution 214
xix

5-7b. Central Charlotte county acquisition map 215
5-7c. Central Charlotte county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition 217
5-7d. Central Charlotte county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition 218
5-8a. Lee and N. Collier county map 1992 1993 jay and habitat distribution 222
5-8b. Lee and N. Collier county acquisition map 223
5-8c. Lee and N. Collier county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition 225
5-8d. Lee and N. Collier county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition 226
5-9a. Flagler and N.E. Volusia county map 1992 1993 jay and habitat
distribution 230
5-9b. Flagler and N.E. Volusia county acquisition map 231
5-9c. Flagler and N.E. Volusia county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition 233
5-9d. Flagler and N.E. Volusia county quasi-extinction graphs. Top) no
acquisition, Bottom) maximum acquisition 234
5-10a. Merritt Island and S.E. Volusia county map 1992 1993 jay and habitat
distribution 238
5-1 Ob. Merritt Island and S.E. Volusia county acquisition map 239
5-10c. S.E. Volusia and Merritt Island county trajectory graphs. Top) no
acquisition, Bottom) maximum acquisition 241
5-1 Od. S.E. Volusia and Merritt Island county quasi-extinction graphs. Top) no
acquisition, Bottom) maximum acquisition 242
5-1 la. N. Brevard county map 1992 1993 jay and habitat distribution 247
5-1 lb. N. Brevard county acquisition map 248
xx

5-1 le. N. Brevard county trajectory graphs. Top) 30% acquisition, Bottom) 70%
acquisition 250
5-1 Id. N. Brevard county quasi-extinction graphs. Top) 30% acquisition, Bottom)
70% acquisition 251
5-12a. Central Brevard county map 1992 1993 jay and habitat distribution 255
5-12b. Central Brevard county acquisition map 256
5-12c. Central Brevard county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition 258
5-12d. Central Brevard county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition 259
5-13a. S. Brevard-Indian River-N. St. Lucie Metapopulation county map 1992 -
1993 jay and habitat distribution 264
5-13b. S. Brevard-Indian River-N. St. Lucie county acquisition map 265
5-13c. S. Brevard-Indian River-N. St. Lucie county trajectory graphs. Top) no
acquisition, Bottom) 30% acquisition by area 267
5-13d. S. Brevard-Indian River-N. St. Lucie county quasi-extinction graphs. Top)
no acquisition, Bottom) 30% acquisition by area 268
5-14a. St. Lucie N. Martin county map 1992 1993 jay and habitat
distribution 272
5-14b. St. Lucie N. Martin county acquisition map 273
5-14c. St. Lucie N. Martin county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition 275
5-14d. St. Lucie N. Martin county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition 276
5-15a. Martin and N. Palm Beach county map 1992 1993 jay and habitat
distribution 280
5-15b. Martin and N. Palm Beach county acquisition map 281
5-15c. Martin and N. Palm Beach county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition 283
xxi

5-15d. Martin and N. Palm Beach county quasi-extinction graphs. Top) no
acquisition, Bottom) maximum acquisition 284
Fig 5-16a. Central Palm Beach county map 1992 1993 jay and habitat
distribution 288
5-16b. Central Palm Beach county acquisition map 289
5-16c. South Palm Beach county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition 291
5-16d. South Palm Beach county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition 292
5-17a. Ocala National Forest county map 1992 1993 jay and habitat
distribution 296
5-17b. Ocala National Forest county acquisition map 297
5-17c. Ocala National Forest county trajectory graphs. No acquisition 299
5-17d. Ocala National Forest county quasi-extinction graphs. No acquisition 300
5-18a. N.E. Lake county map 1992 1993 jay and habitat distribution 304
5-18b. N.E. Lake county acquisition map 305
5-18c. N.E. Lake county trajectory' graphs. Top) no acquisition, Bottom)
maximum acquisition 307
5-18d. N.E. Lake county quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition 308
5-19a. S.W. Volusia county map 1992 1993 jay and habitat distribution 312
5-19b. S.W. Volusia county acquisition map 313
5-19c. S.W. Volusia county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition 315
5-19d. S.W. Volusia county quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition 316
5-20a. Central Lake county map 1992 1993 jay and habitat distribution 320
5-20b. Central Lake county acquisition map 321
xxii

5-20c. Central Lake county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition 323
5-20d. Central Lake county quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition 324
5-2lb. Lake Wales Ridge map 1992 1993 jay and habitat distribution, Glades
County 329
5-2le. Lake Wales Ridge map 1992 1993 jay and habitat distribution, S.
central Polk county 332
5-2If. Lake Wales Ridge map 1992 1993 jay and habitat distribution, N.E.
Polk and N.W. Osceola county 333
5-2lh. Lake Wales Ridge acquisition map, S. Highlands county 335
5-2 li. Lake Wales Ridge acquisition map, N. Highlands and S. Polk county 336
5-2lj. Lake Wales Ridge acquisition map, S. central Polk county 337
5-21k. Lake Wales Ridge acquisition map, N.E. Polk and N.W. Osceola county 338
5-211. Lake Wales Ridge trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition 341
5-21m. Lake Wales Ridge quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition 342
xxm

Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
METAPOPULATION DYNAMICS AND LANDSCAPE ECOLOGY OF THE
FLORIDA SCRUB-JAY, APHELOCOMA COERULESCENS
By
Bradley M. Stith
December 1999
Chairman: Dr. Stephen R. Humphrey
Major Department: Wildlife Ecology and Conservation
Floridas only endemic bird species, the Florida Scrub-Jay (Aphelocoma
coerulescens), is rapidly disappearing throughout much of its range. A 1992-1993
statewide survey shows that it has effectively gone extinct in 10 of 39 formerly occupied
counties in less than two decades. To characterize the spatial structure and vulnerability
of the Florida Scrub-Jay throughout the state, I developed and applied a new method to
describe the species metapopulation structure. This method uses GIS-generated buffers
based on documented dispersal distances to identify separate metapopulations and highly
connected subpopulations called mainlands (extinction resistant), islands (extinction
prone), or midlands (vulnerable to extinction). Of the 42 jay metapopulations identified,
only five include mainlands; 21 consist only of extinction-prone islands. The resulting
xxiv

XXV
classification reveals key subpopulations requiring special attention to maintain the long
term viability of the existing metapopulations.
I developed and applied a technique for measuring habitat features and estimating
habitat quality over large areas using image processing and GIS methods. The technique
showed that jays in central Florida had a strong preference for open sandy areas, and few
or no pine trees. A proximity analysis showed that demographic performance decreased
near forests. Measurement of habitat variables using this technique will be a valuable
technique for habitat management and conservation.
I developed a spatially explicit, individual-based model to simulate the
metapopulation dynamics of Florida Scrub-Jays. Special emphasis was placed on
realistically modeling dispersal. I conducted a small radio-tracking study and used data
from long-term studies to parameterize and validate the model. Stage-age structure and
dispersal distances generated by the model showed good fit to field data.
I used this simulation model to investigate the viability of 21 major Florida Scrub-
Jay metapopulations across the state. For each metapopulation I simulated 2 or more
hypothetical reserve designs, ranging from a minimal design with only currently
protected jays, to a maximal design containing all significant populations as of 1993. All
habitat was assumed to be restored and fully occupied. Model results indicated that only
3 of 21 metapopulations would be adequately protected without further habitat
acquisition. At least 4 metapopulations appear to be at great risk of extinction.

CHAPTER 1
INTRODUCTION
Historical Background
For most of geologic time, Florida lay under water, attached to what is now
northern Africa. With the breakup of the super-continent, Gondwanaland, in the
Mesozoic Florida rafted from Africa and become attached to North America. There she
remained under water throughout the entire Age of Dinosaurs and well into the Cenozoic.
Only recently, about 25 million years ago, Florida emerged from the sea. Never again
would she become completely inundated, despite huge global fluctuations in sea level.
She changed drastically in size and shape, however, due to the episodic transgressions
and regressions of the seas caused by the waxing and waning of great ice sheets on the
continents. As glaciers advanced southward, then retreated, they produced continental-
scale changes in the climate. During high sea levels, mesic forests in Florida prospered,
but with falling sea levels Florida became more and more desert-like. During such
periods, arid conditions prevailed across most of the southern continent, and fauna and
flora from the west were free to move to Florida along a Gulf coastal corridor. Among
these western immigrants was a xeric-adapted bird originating from a widespread species
known today as the Western Scrub-Jay (Aphelocoma californica). Exactly when this
species first arrived in Florida is uncertain, but it must have been at least several million
years ago. Having made it to Florida, this jay would become isolated from its western
counterpart by the development of extensive wetlands associated with the Mississippi
1

2
delta during the mid-Pleistocene. As conditions became more mesic, xeric habitat became
reduced and isolated into desert-like islands within which a remarkable assemblage of
organisms evolved. Among these organisms that diverged from their western kin is the
Florida-Scrub Jay, the subject of this dissertation.
Not long after the most recent Wisconsin glacial retreat, a mere 12,000 years ago,
nomadic people known as the Clovis entered North America from Siberia and
encountered a continent largely or entirely devoid of humans. At that time the continent
teemed with giant animals such as mammoths, mastodons, ostrich-size flightless birds,
huge ground sloths, horses, camels, saber-toothed cats, and giant tortoises. This
megafaunal scene rivaled anything seen today in Africa. Within 1,000 to 2,000 years all
of these species and many more became extinct. Dozens of large vertebrates appear to
have made their last stand in Florida, their demise apparently coinciding with the arrival
of the Clovis. It remains uncertain whether humans were largely to blame for these
extinctions, as the megafauna also faced great changes in climate and landscape. Yet, it is
highly likely that humans contributed significantly to these megafaunal extinctions.
The Florida Scrub-Jay managed to survive this period of massive extinctions. But
by the mid-20th century, a new threat to the fauna and flora of Florida appeared.
Discoveries in applied sciences and engineering paved the way for the demise of the
Florida Scrub-Jay and its scrub habitat. Among these were the discovery that citrus could
be grown on the formerly worthless, sandy, infertile soils upon which Florida Scrub-Jay
habitat grew. The invention of air conditioning made tolerable the Florida summers,
ushering in an era of massive suburban sprawl, much of it devouring Florida Scrub-Jay
habitat.

3
In 1969, Glen E. Woolfenden turned his ornithological focus on the Florida
Scrub-Jay at the Archbold Biological Station. Thus began a continuous 31 year scientific
study of this single organism, making it one of the most thoroughly studied wild bird
species in the world. By 1975 his research on jays became a classic example of altruism
cited prominently in E.O. Wilsons (1975) influential Sociobioloev. In the mid-1970s
Woolfenden teamed up with John W. Fitzpatrick, an ornithologist with a strong
background in population modeling. In 1984, they produced a highly-acclaimed
Princeton monograph (Woolfenden and Fitzpatrick, 1984) describing the demography
and cooperative behavior of this intriguing bird. Today, the number of publications on
this species approaches one hundred, and the Florida Scrub-Jay continues to be widely
cited as a classic example of altruism (e.g. Krebs and Davies 1998). No brief summary
can do justice to this large body of work. Nonetheless, a short review of the basic natural
history of the Florida Scrub-Jay follows (consult Woolfenden and Fitzpatrick 1996 for an
extensive list of references).
Biological Background
The Florida Scrub-Jay, Florida's only endemic bird species, is a disjunct, relict
taxon separated by more than 1600 km from its closest western relatives (Woolfenden
and Fitzpatrick 1984). This habitat specialist is restricted to a patchily distributed scrub
community found on sandy, infertile soilsmostly pre-Pleistocene and Pleistocene
shoreline deposits. The vegetation is dominated by several species of low-stature scrub
oaks (Quercus spp.). Jays rely heavily on acorns for food, especially during the winter,
when they retrieve thousands of acorns cached in open, sandy areas during the fall
(DeGange et al. 1989). Florida Scrub-Jays show a strong preference for low, open

4
habitats with numerous bare openings and few or no pine trees (Breininger et al. 1991).
These optimal habitat conditions are maintained by frequent fires (Abrahamson et al.
1984). Jays living in fire-suppressed, overgrown habitats have much poorer demographic
performance than jays in optimal conditions (Fitzpatrick and Woolfenden 1986), leading
rapidly to local extirpation unless the habitat is burned (Fitzpatrick et al. 1994).
Florida Scrub-Jays are monogamous, cooperative breeders that defend permanent
territories averaging 10 ha per family (Woolfenden and Fitzpatrick 1984). They have a
well-developed sentinel system, in which a family member watches for predators while
others in the group engage in other activities such as foraging (McGowan and
Woolfenden 1989). Young nearly always delay dispersal for at least one year, remaining
at home as helpers. Dispersal distances from natal to breeding territories are extremely
short for both sexes, and these movements within contiguous habitat average less than 1
territory for males and 3.5 territories for females (Woolfenden and Fitzpatrick 1984).
Dispersal behavior is associated with greatly elevated mortality even within optimal
habitat (Fitzpatrick and Woolfenden 1986), and many behavioral adaptations (e.g.,
cooperative breeding, sentinel system, delayed dispersal) suggest that predation is
extremely important to both resident and dispersing jays (Woolfenden and Fitzpatrick
1984; Fitzpatrick and Woolfenden 1986; Koenig et al. 1992).
Objectives
Despite the many studies and enormous amount of information available on
Florida Scrub-Jays, opportunities to contribution to this large body of knowledge are
afforded by new technologies, which allow new types of data to be collected and new
questions to be asked. I employ several of these new technologies in this dissertation,

5
including remote sensing, geographic information systems, radiotelemetry, and computer
modeling. Prior to the early 1990s, Florida Scrub-Jay research was not focused on
conservation issues. With the increasing destruction of scrub habitat, and the Federal
listing of the Florida Scrub-Jay as a threatened species in 1987, much more emphasis has
been placed recently on applied research, and a flurry of conservation-oriented
publications have since appeared. My hope is to fill in a few of the existing gaps.
Exemplifying a key goal of conservation biology, such efforts are made applicable to the
real world by the large body of basic knowledge already available for this species. This
prior body of knowledge also makes it possible to attempt to synthesize a variety of
information in the form of a spatially explicit computer model. More than half of my time
as a doctoral student has been spent developing this model.
Chapter 2 provides an analysis of the entire known distribution of the Florida
Scrub-Jay from a metapopulation perspective, based on a statewide census conducted in
1992-1993. A new classification technique is developed to describe the spatial structure
of the jay population. This technique is general in nature, and has implications for the
conservation of other species as well as the Florida Scrub-Jay. Portions of this chapter
have recently been published (Stith et al. 1996), and I wish to thank Island Press for
permission to include that publication in its entirety (with modifications).
Chapter 3 narrows the focus to a local level and examines the landscape ecology
of the Florida Scrub-Jay in central Florida. A team of researchers collected demographic
variables for color-banded jays at Avon Park Air Force Range (APAFR), in Highland and
Polk county. New computer technology--image processing and GISwas used to
correlate physical features with habitat quality from a Florida Scrub-Jays perspective in

6
this area. The relationship between demographic variables and the remotely-sensed
habitat variables were examined. At issue is the possibility of measuring habitat quality,
and potential demographic success remotely, and across large areas.
Chapter 4 examines the difficult subject of dispersal and describes an approach
used to simulate dispersal in an individual-based model (IBM) developed for this
dissertation. Two types of dispersal are simulated by the IBM. A close-distance dispersal
module mimics a stay-home-and-foray strategy that results in most dispersing jays
settling close to their natal territory. This module incorporates many details of Florida
Scrub-Jay biology documented by long-term color-band studies (Woolfenden and
Fitzpatrick 1984), including sex and age dominance relations. A long-distance dispersal
module simulates a floater strategy, which accounts for the infrequent, though
potentially important, tendency of some jays to abandon their natal territory and move
long distances, often between habitat patches and through hostile landscape matrices.
Empirical data on long-distance dispersal are poor, and a simple field experiment was
conducted with radiotelemetry to obtain information useful for modeling purposes. To
induce behavior that might be similar to long distance dispersal, radio-collared jays were
experimentally displaced kilometers away from their natal territories. Habitat
preferences, movement abilities, and mortality rates were recorded and incorporated into
the long-distance dispersal module. In combination, the close- and long-distance dispersal
modules produced a dispersal and stage-age curve that closely resembled results of long
term data from Archbold Biological Station. A constraint analysis was used to place
plausible bounds on several of the poorly known long distance dispersal parameters. This

7
analysis relied on data that suggested where successful dispersal from Archbold
Biological Station could and could not take place.
Chapter 5 describes the complete individual-based, spatially explicit population
model, which incorporates the dispersal algorithms described in chapter 4. The model
provides a framework for integrating much of what is currently known about the Florida
Scrub-Jay. Simulations take place on a landscape provided by a geographic information
system (GIS) file. Non-dispersing jays occupy discrete territories. Both sexes are
modeled, and individual jays progress through 5 stages (juvenile, 1-year helper, older
helper, inexperienced and experienced breeder). Each territory has a separate set of
demographic parameters assigned to each sex and stage. Breeder experience and presence
of helpers may affect fecundity. Helpers monitor neighboring territories within their
assessment sphere and vie for breeder openings; the outcome of such competition is
determined by simple dominance rules. Helpers may leave on long distance dispersals,
during which time mortality and movement varies depending on landcover type.
The statewide population of jays was divided into 21 metapopulations thought to
be demographically isolated from each other (fig. 5-0). Two series of maps were
developed for each metapopulation. One map type depicts jays and habitat as mapped in
1992-1993. A second map type, referred to as an acquisition map, depicts jays as they
might exist if all habitat patches were restored to optimal conditions, and distinguishes
among jays within protected areas, unprotected habitat patches, and suburban areas. Key
habitat patches are labeled on the acquisition maps, and are cross-referenced in the text
descriptions, tables, and recommendations.

8
A series of simulations were run for each metapopulation based on different
reserve design scenarios. These scenarios ranged from a minimal configuration consisting
of only currently protected patches (no acquisition option), to a maximal configuration
consisting of all significant patches (complete acquisition option). For all simulations, the
assumption was made that all protected areas were restored and properly managed, and
that jays had demographic performance and densities typical of high quality habitat.
These assumptions should be viewed as optimistic. Jays outside of protected areas were
assumed to have poor demographic performance typical of suburban areas.
The output from the simulation runs included estimates of extinction, quasi
extinction (probability of falling below 10 pairs), and percent population decline.
Comparisons of these results provided the basis for ranking the vulnerability of different
metapopulations around the state. Metapopulations were ranked in terms of vulnerability
assuming no further acquisition, and in terms of potential for improvement through
acquiring all unprotected habitat. The proper uses and limitations of population modeling
are discussed.
Chapter 6 synthesizes previous chapters, focusing on some of the limitations of
metapopulation theory. The chapter closes by presenting a set of landscape rules that
provide guidelines for developing a statewide Habitat Conservation Plan for the Florida
Scrub-Jay. Adherence to these landscape rules would likely maintain the viability of
different jay populations across the state, while allowing for further loss of jays to human
development in some areas.

CHAPTER 2
CLASSIFYING FLORIDA SCRUB-JAY METAPOPULATIONS
Introduction
Metapopulation theory, now a major paradigm within conservation biology
(Harrison 1994; Doak and Mills 1994), can be viewed as island biogeography theory
applied to single species (Hanski and Gilpin 1991). Whereas application of island
biogeography to conservation followed shortly after its creation (MacArthur and Wilson
1967), application of metapopulation theory lagged far behind its formalization by Levins
(1969,1970). Simberloff (1988) attributes its growing emergence to a shift in ecological
and conservation focus, from analysis of species turnover to analysis of extinction in
small populations of individual species. Describing real world metapopulations, however,
remains problematic.
Harrison (1991) pointed out ambiguities in the term metapopulation, and
described four different configurations of habitat patches that could be called
metapopulations. Reviewing field studies of patchy systems, Harrison found few natural
examples that matched Levins original concept of a metapopulation. Recently, Harrison
(1994) argued that metapopulation theory often is not applicable, such as cases where
populations are highly isolated, highly connected, or so large as to be essentially
invulnerable. She warned that: the metapopulation concept is being taken seriously by
managers, and taken too literally could lead to the principles that single, isolated
populations are always doomed, or that costly strategies involving multiple connected
9

10
reserves are always necessary (p. 126). Doak and Mills (1994) reviewed the different
metapopulation classes described by Harrison and held that it will often be difficult or
impossible to distinguish between these alternatives, and thus to assess the importance of
metapopulation dynamics (p. 624). They also warned that spatially explicit population
models (SEPMs) simulating metapopulation dynamics typically use parameters that are
difficult to measure in the field. Their list of required data included within-patch
demographic rates and variances, temporal and spatial correlation of vital rates among
populations, and dispersal distances and success. Harrison (1994) also stressed the
difficulty of identifying all local populations and suitable habitat, and of estimating
extinction and colonization rates among patches.
Although still controversial (e.g., Harrison, Stahl and Doak 1993), the
metapopulation concept does provide a useful framework for describing the spatial
structure of real populations. The concept, after all, is grounded on two of the most robust
empirical generalizations in ecology and conservation biology: 1) extinction rates
decrease with increasing population size, and 2) immigration and recolonization rates
decrease with increasing isolation (MacArthur and Wilson 1967; Hanski 1994).
Our goal in this chapter is to illustrate how the above two generalizations can be
used 1) to characterize quantitatively the metapopulation structure of a species, and 2) to
develop landscape rules for conserving metapopulations of a declining species.
We begin with the results of a range-wide survey of the federally Threatened
Florida Scrub-Jay (Aphelocoma coerulescens), conducted in 1992-1993. This species is
patchily distributed, and thereby presents a challenging case study for describing

11
metapopulation structure. We offer a method for doing so using a detailed spatial
database, combined with existing biological information and GIS technology. The
technique uses computer-generated buffers, at several distances reflecting the dispersal
behavior of the species, to delineate subpopulations with differing degrees of
connectivity. Extinction vulnerability of each subpopulation is estimated via a PVA
model (in our case, that of Fitzpatrick et al. 1991). We propose a simple nomenclature for
classifying Florida Scrub-Jay metapopulations based on subpopulation size and
connectivity. We conclude by deriving a few, metapopulation-based landscape rules
that may be incorporated into a statewide framework for conservation plans affecting this
rapidly declining bird species.
Statewide Survey of the Florida Scrub-Jay
Statewide Survey: Methods
The Florida Scrub-Jay was listed by the U. S. Fish and Wildlife Service (USFWS)
as a Threatened species in 1987. In 1991 the USFWS notified landowners and county
governments that clearing scrub could violate the Endangered Species Act (ESA)
(USFWS 1991). At the same time, the USFWS began encouraging counties to develop
regional Habitat Conservation Plans (HCP) that could solve local permitting problems by
means of a single, biologically based, regional plan. To aid in this process, the USFWS
partially sponsored the authors, and their cooperators, to conduct an intensive survey
during 1992-1993 to document the range and sizes of subpopulations throughout the
state, and to inventory existing potential habitat, whether occupied or not.

12
Our methods were similar to those used for the Northern Spotted Owl (Strix
occidentals caurina; Murphy and Noon 1992). Extensive prior information helped guide
us. Cox (1987) had documented numerous jay localities throughout the state in the early
1980s, and had compiled historic records from diverse sources such as museums and
Christmas Bird Counts. The Florida Breeding Bird Atlas (Kale et al. 1992) provided
valuable data on jay sightings made by hundreds of volunteers from 1986 to 1991.
Virtually all previously known jay localities were revisited for this survey. Information
from the public was solicited through notices in magazines, newsletters, and newspapers.
Additional, potential habitat patches were identified from U.S. Soil Conservation Service
maps and aerial photographs, on which the white, sandy soil associated with jay habitat
forms a distinctive signature.
Standard surveying techniques based on tape playback of jay territorial scolds
(Fitzpatrick et al. 1991) were used to locate jays in habitat patches. Location and number
of individuals in each group were plotted on field maps. The following qualitative habitat
data were collected at most patches: occupancy by jays; estimated degree of vegetative
overgrowth (1-4 scale); extent of human disturbance (1-4 scale); ownership status with
respect to permanent protection from development. Time constraints prohibited making
quantitative habitat measurements at the thousands of habitat patches visited. Although
survey goals included attempting to find all known jay families outside of Federal lands,
we know that a few jays were missed because of limited access to certain private lands.
The total number missed, however, is not likely to exceed a few percent of the statewide
population.

13
Federally-owned lands were not surveyed for this project. Those with large
populations of jays include: Cape Canaveral Air Force Station, Merritt Island National
Wildlife Refuge, Canaveral National Seashore, and Ocala National Forest. Florida Scrub-
Jays at each of these areas are currently under study, so for the statewide summary we
used estimates of numbers and locations of jays provided by the respective biologists
conducting those studies.
To archive, map, and analyze the statewide data we developed a series of map
layers by means of a GIS at Archbold Biological Station. PC and Sun ARC/INFO were
used to input all GIS data (E.S.R.I. 1990). Habitat patches (both occupied and
unoccupied) and jay locations, originally hand-drawn on soil or topographic maps
(usually at 1:24000 scale), were digitized. Patch characteristics and jay family sizes were
entered into accompanying data files. Map layers included current and historic range of
jays, current distribution of suitable and potential habitat, and locations and numbers of
jay families encountered.
Statewide Survey: Results
We estimate that as of 1993 the total population of Florida Scrub-Jays consisted
of about 4,000 pairs (Fig. 2-1; Fitzpatrick et al. 1994). Both total numbers and overall
geographic range have decreased dramatically during this century (Cox 1987). In recent
decades the species has been extirpated from 10 of 39 formerly occupied counties, and it
is now reduced to fewer than 10 pairs in 5 additional counties (detailed tabulations in
Fitzpatrick et al., in prep.). Detailed, site-by site comparison of our survey with Cox's
(1987) suggests that the species may have declined as much as 25% to 50% during the
last decade alone.

14
Degraded quality of many currently occupied habitat patches suggests that
further, substantial declines in the jay population are inevitable. Specifically, those jays
occupying suburban areas (approximately 30% of all territories) are unlikely to persist as
these suburbs continue to build out, given the rapid rate at which Floridas human
population continues to expand. Furthermore, jays living in fire-suppressed, overgrown
habitat (at least 2,100 families, or 64% of all occupied scrub patches by area) already are
likely to be experiencing poor demographic performance (Fitzpatrick and Woolfenden
1986). These can be expected to decline further unless widespread restoration of habitat
is begun soon.
A Method for Classifying Metapopulations
The patchy distribution and variable clustering of territories throughout the range
of the Florida Scrub-Jay (Fig. 2-1) challenges us to expand upon traditional
metapopulation concepts in order to describe the spatial structure of this species. In this
section we describe our conceptual approach, and in the next we apply it to the Florida
Scrub-Jay data.
Harrisons (1991) four classes of metapopulations can be presented graphically
(Fig. 2-2) as different regions on a plot of degree of isolation against patch size
distribution. Thus, Harrisons "non-equilibrium" metapopulation is that set of small
patches in which each has a high probability of extinction, and among which little or no
migration occurs. Local extinctions are not offset by recolonization, resulting in overall
decline toward regional extinction. The classical model developed by Levins (1969,
1970) is a set of small patches which are individually prone to extinction, but which are

15
large enough and close enough to other patches so that recolonization balances
extinction. Patchy metapopulations consist of patches so close together that migration
among them is frequent, hence the patches function over the long run as a single
demographic unit. Finally, the mainland-island model has a mixture of large and small
patches close enough to allow frequent dispersal from an extinction-resistant mainland to
the extinction-prone islands.
The lower right side of Fig 2-2. portrays two classes not presented by Harrison
(1991). These large patches are either poorly-connected (i.e., disjunct) or moderately-
connected (i.e., mainland-mainland). Such large populations tend to be less interesting
from a conservation standpoint, as they are essentially invulnerable to extinction.
Classifying metapopulations, therefore, requires species-specific information on
both connectivity (i.e., dispersal behavior and barriers) and extinction probabilities (i.e.,
population sizes in patches), across space. For example, a system of small habitat patches
might appear to support stable populations of certain organisms as classical or patchy
systems, while other species might be nonequilibrial in the same system because of low
density (hence small populations sizes) or limited dispersal ability.
Harrisons (1991) diagram of metapopulations represented connectivity among
patches by means of a dashed line around those among which dispersal is frequent
enough to unite the patches into a single demographic entity. This boundary can be
viewed as a dispersal buffer: an isoline of equal dispersal probability. Any number of
patches may be included within a given dispersal buffer of a single subpopulation,
provided that fragmentation is sufficiently fine-grained (sensu Rolstad 1991). However,
dispersal probability normally diminishes continuously (even if steeply) away from a

16
patch, and for most terrestrial species it asymptotically approaches zero at some point
farther away than Harrisons single, discrete dispersal buffer. Therefore, we extend
Harrisons diagrammatic approach by adding a second buffer to delineate the distance
beyond which dispersal is effectively reduced to zero. We maintain that this second
buffer functionally identifies separate metapopulations. We acknowledge that
connectivity actually should be represented graphically as a continuous surface of
dispersal probabilities. However, discrete boundaries, placed at biologically meaningful
(and empirically determined) distances, greatly simplify the description of
metapopulations. They also provide explicit, repeatable methodology for comparative or
modeling purposes.
Harrisons metapopulation types may be characterized using these two buffers
(Fig. 2-3). In patchy systems (Fig. 2-3a), every patch belongs to the same
subpopulation, so they are all enclosed within a single, inner dispersal buffer. Classical
systems (Fig. 2-3b) have small subpopulations separately encircled, representing the fact
that each may go extinct temporarily, or may be rescued before going extinct (Brown
and Kodric-Brown 1977), both by way of colonization from another subpopulation
enclosed within the outer buffer. The simplest nonequilibrium systems (Fig. 2-3c) are
represented as bulls-eyes around small, isolated subpopulations. A mainland-island
metapopulation (Fig. 2-3d) has a large subpopulation and several small ones within a
single outer buffer.
The important point is that even more complicated patterns may be common in
nature, arising from combinations or intermediate cases, and many of these are not easily
fit into Harrisons (1991) four metapopulation classes. To deal with such complications,

17
we suggest characterizing metapopulations by describing the sizes of their constituent
subpopulations. We propose a simple nomenclature, based on three key words island,
mainland, and midland to characterize the relative sizes of the subpopulations
within a metapopulation. Subpopulations small enough to be highly extinction prone in
the absence of significant immigration are called islands. Those large enough to be
essentially invulnerable to extinction are called mainlands. Intermediate sized
subpopulations are neither extinction prone nor invulnerable to extinction. For lack of a
better term, we refer to an intermediate size subpopulation as a midland.
Distinctions among these categories need not be completely arbitrary. Species-
specific population viability analysis (PVA) provides an explicit, quantifiable approach
for describing subpopulations as extinction-prone, extinction-vulnerable, or extinction-
resistant. Our introduction of the midland category helps clarify the importance of
turnover, which has been called the hallmark of a genuine metapopulation dynamics
(Hanski and Gilpin 1991). Specifically, turnover is expected in systems with island-size
subpopulations because they have high frequencies of extinction. But systems with
midlands rather than islands are perhaps more often characterized by rescue rather than
recolonization, as local extinctions will be rare. Thus, a system of midlands may exhibit
little or no turnover even though no real mainlands are present, while a system of islands
with the same degree of isolation may show high turnover. We agree with Sjogren (1991)
in emphasizing the importance of rescue in metapopulation dynamics. Traditional
emphasis on turnover probably resulted from the fact that rescue is much more difficult to
measure empirically, as turnover only requires presence-absence data.

18
Harrisons metapopulation classes can be described using this island-midland-
mainland nomenclature as follows: a nonequilibrium metapopulation is a system of
one or more islands (i.e., extinction-prone subpopulations), with a total population size
too small to persist. A classical metapopulation is a system of island-size
subpopulations large enough and close enough together and of sufficient total size to
allow persistence. Any system containing a midland or mainland (by definition) cannot
be a nonequilibrium or classical metapopulation, as all subpopulations in the latter
systems are extinction-prone. A patchy metapopulation is a set of patches close enough
together to form a single subpopulation of sufficient size to persist (i.e. a midland or
mainland). Mainland-island metapopulations are self-explanatory.
Explicit reference to midlands-extinction-vulnerable patches of intermediate
population size-produces metapopulation types not described in Harrison (1991).
Systems with, for example, several midlands, or a mainland with several midlands, are
possible. We illustrate some of these configurations by applying our nomenclature,
quantitatively, to the Florida Scrub-Jay.
Metapopulation Structure of the Florida Scrub-Jav
Application of the above scheme to any species requires choosing two dispersal
buffer distances and two threshold values for extinction-vulnerability among single
populations. Here, for the Florida Scrub-Jay, we based each of these values on
empirically gathered biological data. Buffer distances were derived from long-term field
studies of marked individuals, and from information garnered on the statewide survey
regarding occupancy of habitat patches at various distances from source populations.

19
Extinction vulnerability was estimated using a single-population viability model
(Fitzpatrick et al. 1991). We then chose thresholds to delineate islands, midlands, and
mainlands, much as Mace and Lande (1991) used extinction probabilities to propose
IUCN threatened species categories.
Dispersal Distances
Between 1970 and 1993 we documented 233 successful natal dispersals from the
marked population under long-term study at Archbold Biological Station (Figure 2-4; see
also Woolfenden and Fitzpatrick 1984, 1986). Unlike the situation for most field studies
of birds (e.g., Barrowclough 1978), many characteristics of our study and the behavior of
jays themselves enhance our ability to locate dispersers that leave the main study area.
Once established as breeders, for example, Florida Scrub-Jays are long-lived and
completely sedentary. Furthermore, we have mapped in detail all scrub habitat within the
local range of the species, and we census these tracts periodically in search of dispersed
jays. (Such censuses reveal remarkably few banded dispersers among the many hundreds
of jays encountered.) Because banded Florida Scrub-Jays from our study usually are
tame to humans, both our own searches and casual encounters by local homeowners have
high likelihood of exposing any off-site dispersers to us once they become paired on a
territory. Indeed, if we assume that immigration and emigration rates are about equal in
our study area, evidence suggests that we have succeeded in locating all but a low
percentage of the jays that have departed over the 25-year period of our study. Therefore,
although some dispersers do escape our detection, our observed dispersal curve (Fig. 2-4)
can be only marginally biased toward the shorter distances.

20
About 80% of documented dispersals were within 1.7 km of the natal territory,
85% within 3.5 km, 97% within 6.7 km, and 99% within 8.3 km (Fig. 2-4). Data from
field studies elsewhere in Florida reveal the same, remarkably sedentary dispersal
behavior. The longest dispersal so far documented was a female we discovered pairing 35
km from her natal territory at Archbold, in 1994.
All dispersals we have documented around Archbold, including the longest one,
involved jays that had moved either through continuous habitat or across gaps no greater
than 5 km. To test the generality of this observation, we pooled dispersal information
from the seven other biologists currently color-banding Florida Scrub-Jays around the
state (D. Breininger, R. Bowman, G. Iverson, R. Mumme, P. Small, J. Thaxton, B.
Toland, unpubl. data). Their studies, along with ours, cumulatively have produced about
a thousand banded non-breeders that achieved dispersal age (Fitzpatrick et al., in prep.).
Collectively these studies have documented only about 10 dispersals of 20 km or more,
and only a few of these had crossed habitat gaps as large as 5 km. More important in the
present context, despite ample opportunity to observe longer-distance movements, not a
single example yet exists of a banded Florida Scrub-Jay having crossed more than 8 km
of habitat that does not contain scrub oaks. We suspect that this distance is close to the
biological maximum for the species.
Patch Occupancy
The observations just outlined suggest that for habitat specialists such as the
Florida Scrub-Jay, dispersal curves measured in relatively contiguous habitat actually
may overestimate the dispersal capabilities of individuals across fragmented systems.
Direct behavioral observations strongly indicate that Florida Scrub-Jays resist crossing

21
large habitat gaps. Still, few opportunities exist to observe jays in the act of dispersing,
hence the theoretical maximum dispersal-distance (i.e., the outer dispersal buffer) is
extremely difficult to establish directly.
Seeking an indirect measure of dispersal frequencies across habitat gaps, we
examined patch occupancy statewide as documented by our 1992-93 survey. We used
Fragstats software (McGarigal and Marks 1994) to measure distances between each
occupied patch of scrub habitat to its nearest neighboring, occupied patch. We then
measured (by hand, as Fragstats cannot measure distances between patches of different
attributes) the distances between each unoccupied suitable patch and the nearest occupied
patch. For each distance class, the ratio of the count of the occupied-to-occupied
distances to the total number of nearest neighbor distances yields the proportion of
patches that are occupied at that distance away from occupied habitat.
Presumably, declines in patch occupancy with increasing distance to the nearest
occupied habitat (Fig. 2-5) reflect diminishing recolonization rates following local
extinctions. Occupancy remains above zero even at great distances, probably because
larger isolated patches rarely experience extinction. This curve provides an empirical
approach for delineating subpopulations and metapopulations: a subpopulation buffer is
the maximum interpatch distance where occupancy rates remain high; the metapopulation
buffer is the smallest interpatch distance where occupancy rates reach their minimum.
For Florida Scrub-Jays (Fig. 2-5) patch occupancy is about 90% to at least 2 km
from a source, then declines monotonically to around 15% at 12 km. (Sample size of
isolated patches decreased rapidly beyond 16 km, necessitating lumping of classes at the
larger distances.) We infer from this occupancy curve that successful recolonization is a

22
rare event beyond about 12 km from an occupied patch of habitat. We use this distance to
identify metapopulations that have become essentially demographically independent from
one another (i.e., the outer dispersal-buffer).
We selected the distance of 3.5 km (about 2 miles) as an inner dispersal-buffer to
delineate subpopulations. We choose this figure because: 1) behavioral information from
a variety of sources, including radiotracking data (B. Stith, unpubl.), shows that jays
begin to show reluctance to crossing habitat gaps at about this size (and at much smaller
gaps where open water or closed-canopy forest are involved); 2) known dispersals of
many banded jays included habitat gaps up to 3.5 km, but their frequency declines
dramatically thereafter; 3) the observed dispersal curve from Archbold (Fig. 2-4) shows
that in good habitat, more than 85% of dispersals by females, and fully 97% of those by
males, are shorter than 3.5 km; 4) patch occupancy data (Fig. 2-5) show significant
decline in colonization rates at distances above 3.5 km.
Population Viability Analysis
PVA based on a simulation model incorporating demographic (but not genetic)
stochasticity and periodic, catastrophic epidemics (Fitzpatrick et al. 1991; Woolfenden
and Fitzpatrick 1991) provided a quantitative method for defining boundaries along the
island (extinction prone), midland (vulnerable), and mainland (extinction resistant)
continuum (but see Taylor 1995). Among the several methods for expressing extinction
vulnerability (e.g., Burgman et al. 1993; Boyce 1992; Caughley 1994) we elect the
simple approach of specifying time-specific probability of persistence of populations of a
given size.

23
Model results indicated that a population of jays with fewer than 10 breeding
pairs has about a 50% probability of extinction within 100 years, while a population with
100 pairs has a 2% to 3% probability of extinction in the same time period. These two
population sizes--10 and 100 pairsprovide convenient and biologically meaningful
values by which to classify subpopulations as islands (< 10 pairs), midlands (10-99
pairs), and mainlands (> 99 pairs). Although subjectively chosen, these values
effectively separate population sizes having fundamentally different levels of protection.
These values also receive empirical support from several long-term bird studies
(reviewed by Thomas 1990; Thomas et al. 1990; Boyce 1992).
Metapopulation Structure
We used a GIS buffering procedure (E.S.R.I. 1990) to generate dispersal-buffers
around groups of jays occurring within 3.5 km (for subpopulations) and 12 km (for
metapopulations) of each other (Fig. 2-6). We buffered jay territories rather than habitat
patches because we strongly suspect that dispersing Florida Scrub-Jays cue on the
presence of other, resident jays even more strongly than on habitat, so the functional
boundaries of occupied patches may be determined by where actual jay families exist.
We modified the resulting buffers in the following areas to reflect the presence of hard
barriers to dispersal in the form of open water with forested margins: Myakka River,
Peace River, St. Johns River, St. Lucie River, and Indian River Lagoon.
Using a 3.5 km dispersal-buffer we delineated 191 separate Florida Scrub-Jay
subpopulations (Fig. 2-6). Over 80% (N=152 islands) are smaller than 10 pairs (Fig. 2-
7), and 70 of these consist of only a single pair or family of jays. Only 6 subpopulations
contain at least 100 pairs (mainlands), leaving 32 midlands (10-99 pairs).

24
Using a 12 km dispersal-buffer we delineated 42 separate Florida Scrub-Jay
metapopulations (Fig. 2-6). Again, most are small (Fig. 2-7). We tabulated the number
and type of subpopulations within each metapopulation (Table 2-1), and noted how each
metapopulation fits into Harrisons (1991) scheme.
Exactly half (21) contain fewer than 10 pairs, thereby constituting
nonequilibrium systems. Along the north-central Gulf Coast (Fig. 2-8), for example, a
group of non-equilibrium systems coincide with a heavily developed area containing a
burgeoning human population. Only three Florida Scrub-Jay systems have configurations
that may be classical metapopulations (i.e., contain only islands, but may be large
enough to support one another following extinctions; e.g., Fig. 2-9). However, this
technique provides no means of distinguishing classical systems from
nonequilibrium. Chapter 5 addresses this shortcoming using a simulation model.
Another three systems represent patchy metapopulations (i.e., contain a single,
fragmented subpopulation large enough for long-term persistence).
Five systems approximate mainland-island metapopulations, but each of these
examples also contains at least one midland population (e.g., the large Lake Wales Ridge
system, with one mainland, 10 midland, and 39 island populations; Fig. 2-10). These 10,
plus 9 midland-island and one mainland-midland system, do not fit neatly any of
Harrisons (1991) metapopulation classes.
A total of 32 midland populations exist (mean size=30.7), and these occur in 18 of
the 42 separate metapopulations. Excluding the nonequilibrium systems, true islands are
present in 17 systems. Therefore, assuming that dispersal is not inhibited by habitat loss

25
expanding the distances among patches, rescue (on midlands) may be at least as
important as turnover (on islands) in Florida Scrub-Jay metapopulation dynamics.
Use of empirically derived dispersal-buffers and extinction probabilities provides
an explicit method for quantitatively describing metapopulation structure. Application of
this technique to the Florida Scrub-Jay demonstrates that a species can exhibit a variety
of metapopulation patterns across its range. Patterns of aggregation and isolation do not
conform to a single metapopulation class in the Florida Scrub-Jay. Such complex spatial
structure is probably common in nature, particularly among species with large and widely
dispersed populations restricted to a patchy habitat. Such patterns may be further
complicated by perturbations of the natural system caused by humans.
Caveats
We offer several caveats as to the generality of dispersal-buffer methodology in
conservation. (1) The technique is best suited for organisms occupying discrete
territories, home ranges, or habitat patches amenable to mapping. (2) The technique is
predicated on having a comprehensive survey. Missing data can lead to misleading
results, especially as regards connections among metapopulations or subpopulations. (3)
The technique presents a static, snapshot view of metapopulations. It does not easily
reveal important dynamics among subpopulations, such as those obtainable from an
SEPM. The viability of different configurations is best determined from SEPMs rather
than single population PVAs. (4) Populations in decline or in sinks can present an
overly optimistic picture (Thomas 1994). Indeed, we suspect that many of the island
and midland subpopulations of Florida Scrub-Jays currently are failing to replace
themselves demographically, as a result of habitat degradation from fire suppression.

26
Similarly, abnormally high densities may exist due to the crowding effect (Lamberson
et al. 1992) following recent habitat losses. (5) The technique relies on numerous
simplifying assumptions about dispersal behavior in defining connectivity among
patches. Most important, it assumes random movement between patches, equal
traversibility of interpatch habitats, absence of dispersal biases owing to habitat quality
differences at the origin or the destination, and absence of density-dependence in
behavior. More elaborate applications, of course, could incorporate alternative
assumptions about these and other factors.
Another important consideration are the kinds of data to buffer. To create
biologically meaningful--but very differentdescriptions for the Florida Scrub-Jay we
could have buffered around jay territories (our choice), occupied patches, suitable habitat
patches both occupied and unoccupied, or all scrub habitat patches regardless of current
suitability. Organisms such as Florida Scrub-Jays that are reluctant to become established
in unoccupied, suitable habitat (e.g., Ebenhard 1991), or have high conspecific attraction
or an allee effect (Smith and Peacock 1990) are best buffered around actual territories
or occupied patches. This is because unoccupied sites have a low probability of becoming
occupied regardless of their degree of isolation, hence contribute little to the current
metapopulation dynamics of the species. On the other hand, excellent colonizers of empty
habitat or species adept at long-distance dispersals via unoccupied stepping stones
probably should be buffered around all habitat patches.
In summary, this method of classifying metapopulations provides a compact
means of describing both connectivity and local population size through the use of simple
terminology. Separate metapopulations are easily delineated using the maximum

27
dispersal buffer. The internal structure of each metapopulation is easily described using
the inner dispersal buffer to delineate islands, midlands, and mainlands. Enumerating all
metapopulations and describing their internal structure (table 2-1) reveals much about the
distribution and viability of Florida Scrub-Jays. The differing internal configurations of
metapopulations present different conservation problems and require different
management approaches. Discussion of these matters is deferred until chapter 6,
following the presentation of the modeling results (chapter 5) which analyze the viability
of metapopulations around the state.
Note: This chapter has been published as Stith et al., 1996, and is reproduced with some
modifications with the permission of Island Press.

Table 2-1. Summary information for 42 Florida Scrub-Jay metapopulations, including
type of metapopulation, number of territories (pairs of jays), and number of
subpopulations.
28
Metapopulation type
(after Harrison 1991)
Mainland-Island (?)
Unknown
Metapopulation Type Size Number of
(Mainland, Midland, (pairs) subpopulations
and/or Island)a
Mn
lOMd
391
1247
50
Mn
Md
51
1036
7
Mn
Md
51
466
7
Mn
2Md
61
237
9
Mn
Md
51
179
7
Mn
Md
126
2
4Md
111
120
15
2Md
I
103
3
Md
51
94
6
Md
I
58
2
2Md
31
55
5
Md
I
50
2
Md
21
29
3
Md
31
22
4
Md
21
18
3
Md
26
1
Md
22
1
Patchy

29
Table 2-1 cont.
Metapopulation type
(after Harrison 1991)
Metapopulation Type
(Mainland, Midland,
and/or Island)a
Size
(pairs)
Number of
subpopulations
Md
15
1
Classical
161
49
16
101
24
10
61
21
6
Nonequilibrium
31
5
3
31
3
3
21
7
2
21
3
2
21
3
2
21
2
2
21
2
2
21
2
2
I
6
1
I
2
lb
I
1
lc
a Numerical prefix indicates number of Mainlands (Mn), Midlands (Md), and Islands (I).
See text for nomenclature.
b There was a total of 4 single Island systems composed of 2 pairs in one subpopulation.
c There was a total of 8 single Island systems composed of a subpopulation of one pair.

30
Fig. 2-1. 1993 distribution of Florida Scrub Jay groups (small black circles). Note the
discontinuous distribution and variability in patterns of aggregation.

31
Highly
Connected
A
Patch
Isolation

Highly
Isolated
Patchy
Classical
Mainland- Mainland-
Island Mainland
Nonequilibrium Disjunct
All
Small
Mixture of All
Small & Large
Large
Patch Size
Fig. 2-2. Classification scheme showing different types of metapopulations based on
patch size distribution (patches all small in size, mixture of small and large, and all large
in size) along the horizontal axis, and degree of patch isolation (highly connected to
highly isolated) on the vertical axis. Nonequilibrium, classical, mainland-island, and
patchy classes are named according to Harrison (1991).

32
/
Fig. 2-3. Schematic depiction of different kinds of metapopulations, illustrating use of
dispersal-distance buffers to predict recolonization rates among subpopulations. Dotted
lines separate functional subpopulations, based on frequency of dispersal beyond them.
Solid lines separate metapopulations, based on poor likelihood of dispersal among them,
A. Patchy metapopulation. B. Classical metapopulation. C. Nonequilibrium
metapopulations. D. Mainland-island metapopulation.

33
Fig. 2-4. Dispersal frequency curve. Dispersal distances from natal to breeding territories
for color-banded jays at Archbold Biological Station, 1970-1993. About 85% of
documented dispersals were within 3.5 km, and 99% within 8.3 km. The longest
documented dispersal was 35 km.

34
V)
0)
JZ
o
*->
re
Q.
o
0>
'5.
3
CJ
O
O
c
o
'
o
O.
O
0.
1.00
o
o
o
O
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
O
o
o
o
o
o
o
o
o
o
o
(N
CD
CO
o
CN
CD
o
o
o
o
T
T
r
T
CM
CO
kO
Interpatch Distance (m)
Fig. 2-5. Proportion of suitable habitat patches occupied by Florida Scrub-Jays as a
function of their distance to the nearest separate patch of occupied habitat. Occupancy
rates are high (nearly 90 %) for patches up to 2 km apart and decline monotonically to 12
km. Note the scale change after 16 km.

35
Fig. 2-6. Statewide jay distribution map with dispersal buffers. Shaded areas with thin,
solid lines depict subpopulations of jays within easy dispersal distance (3.5 km) of one
another. Thick lines delineate demograpnically independent metapopulations separated
from each other by at least 12 km.

25
21
Metapopulation Size
Fig. 2-7. Frequency of Florida Scrub-Jay metapopulation sizes. Note that 21
metapopulations have 10 pairs or less of jays. These represent nonequilibrium
1000

37
t
Fig. 2-8. Examples of non-equilibrium metapopulations from North Gulf coast. Each of
the six metapopulations contains fewer than 10 pairs of jays, except for the centrally
located system that contains a single, midland-size subpopulation.
i

38
Fig. 2-9. Example of a "classical" metapopulation from five counties in Central Florida.
Note the occurrence of jays in small islands of intermediate distance from one another.
L

39
Fig. 2-10. Portion of the largest mainland-midland-island metapopulation in the interior,
consisting of the Lake Wales Ridge and associated smaller sand deposits. The large
central subpopulation (enclosed by the thin black line) contains nearly 800 pairs of jays.
Small subpopulations to the south and east are within known dispersal distance of the
large, central mainland. A small metapopulation to the west (in DeSoto County) contains
a single subpopulation of 21 territories. This small system qualifies as a patchy
metapopulation, since jays occur in two or more patches but the patches are so close
together that they function as a single demographic unit.

CHAPTER 3
REMOTE SENSING OF FLORIDA SCRUB-JAY HABITAT
Introduction
The importance of understanding relationships between wildlife and habitat has
been recognized for many decades (review in Morrison et al. 1992). The ultimate success
of wildlife management and conservation efforts depends to a large degree on our ability
to understand these relationships. Unfortunately, measuring habitat variables over large
areas, and with sufficient spatial resolution to capture the essential habitat features a
particular species responds to, is a difficult and time consuming task. In this chapter I
apply a technique that greatly assists in the measurement and analysis of wildlife-habitat
relationships across large areas. This technique, which relies heavily on recent advances
in computer hardware and software technology, uses image processing and GIS software
to measure habitat variables directly from scanned aerial photography.
Habitat requirements differ widely among most species, and choosing the
appropriate habitat variables to measure is greatly facilitated for species whose habitat
requirements are well understood. Long-term studies of the Florida Scrub-Jay have
revealed much about the habitat requirements of this species (Woolfenden and Fitzpatrick
1984; Breininger et al. 1991, 1995, 1996). Research indicates that scrub-jays have a fairly
simple "habitat template" consisting of low vegetation dominated by scrub oaks, open
40

41
sandy areas to cache acorns, and few or no pine trees. These habitat characteristics are
naturally created and maintained by fire.
Pine tree density may be a critical factor in determining Florida Scrub-Jay habitat
quality. Pine trees may affect scrub-jays indirectly in several fashions. Large trees
provide perch sites and cover used by accipiters, important predators of jays. Pine trees
probably reduce the effectiveness of the jays' sentinel system against aerial predators. A
close competitor and nest predator, the Blue Jay, shows a strong preference for pine
forests through much of the scrub-jayss range (Tarvin 1997). Competition between these
two corvids may partially explain why Florida Scrub-Jays select open habitat (see
discussion in Woolfenden and Fitzpatric 1984).
Declines in nesting success, and survival of juveniles and adults have been
documented by Fitzpatrick and Woolfenden (1986) within a fire-suppressed habitat that
gradually becomes overgrown. Jays living in open habitat have higher survival, nesting
success, and larger mean group size than jays in less open habitat (Fitzpatrick and
Woolfenden 1986; Breininger et al. 1995). Differences in these demographic traits may
be due to habitat features immediately adjacent to jay territories (Breininger et al. 1996).
Thus, although the internal quality of a jay territory may be high, territory position within
the landscape mosaic may greatly influence demographic success.
In this chapter, I test some of these ideas using habitat features and demographic
variables measured at the Avon Park Air Force Range (APAFR) in Polk and Highlands
counties, Florida. I compare the habitat and demographic performance of jays occupying
a north-south trending scrub ridge some 15 km in length. Jays along this ridge occupy
habitat with different degrees of overgrowth and different adjacent habitat. The main

42
objective of this chapter is to test whether remote sensing can accurately measure habitat
features that explain Florida Scrub-Jay demographic performance.
Methods
All image processing and GIS work was completed on a Sun workstation or an
Intel 486-based PC using Arc/Info (ver. 4.3D) and Erdas (ver. 7.5) software.
Image Source
Regular color, black and white, and color infrared photography were evaluated for
use in this project. Color infrared photography was flown for much of the APAFR
through the NAPP program of the U.S.G.S. in March of 1994.1 selected this photography
for three primary reasons: separation between sandy and grassy areas was more distinct,
wider coverage per frame made image mosaicing easier, and the photography is available
for study sites outside of the APAFR that are currently under investigation for Florida
Scrub-Jay conservation. Hence, results of this study could be extended to other areas.
Two photos covered the entire study area. Frame 6980-207 covered the area I
refer to as N. Sandy Hill, which is between Kissimmee and Smith roads, and part of S.
Sandy Hill south to Submarine Lake. Frame 6980-205 covered the remainder of S. Sandy
Hill south to the southern fence line.

43
Image Scanning and Conversion
A Umax 1200 color scanner with a transparency adapter (for scanning positive or
negative film; 600 dots-per-inch maximum hardware resolution) was used to scan the
1:40000 color transparencies at a resolution of 555 dots per inch, giving a ground
resolution of approximately 1.8 meters (6 feet). Image-In scanning software provided by
Umax for use with the scanner was used to scan all images on a PC.
Software from Earth Resource Data Analysis Systems (ERDAS) of Atlanta,
Georgia was used to convert the TIFF files to ERDAS 7.5 LAN files.
Image Rectification and Mosaicing
I used ERDAS software to rectify the two images to Zone 17 of the Universal
Transverse Mercator map projection using differentially corrected GPS points. I collected
these GPS control points at various road intersections throughout the study area using a
Trimble Pathfinder rover with a data logger. At each control point, 180 location
measurements were acquired. I differentially corrected these control points using base
files acquired from a Trimble Professional base station located on the APAFR. The mean
of the 180 measurements taken at each control point was used as the final x and y
coordinate for rectifying the images.
During the rectification process, I discarded some GPS points because their
inclusion produced excessively high root mean square (RMS) errors. Several factors
explain these high RMS errors. First, the expected error from the differentially corrected

44
GPS points (Trimble claims 2 to 5 m for the units I used) is substantial relative to the
image resolution (2 m). Second, nonlinear distortions in the imagery owing to lens
curvature, tilt angle, etc. cannot be corrected with this type of linear rectification process.
Third, an error of one to several meters is introduced by inaccuracies in visually placing
each control point on the image. Four control points were common to both images, which
helped ensure that the two images matched well in the overlapping areas. The locational
accuracy of the rectification is unknown, and could not be determined without using
much more accurate and expensive ground-based surveying techniques. However, the
positional accuracy is probably within the range of 2 to 10 meters. This accuracy is more
than adequate for the purposes of this project.
The two rectified images were mosaiced together using Erdas software. Very little
displacement is present in the spliced region between the two images, indicating that the
rectification process was internally consistent.
Image Classification
The images were classified with unsupervised, isodata classifier in Erdas to
produce 27 statistically distinct spectral classes stored in a signature file. This signature
file was then used to reclassify the raw image with a maximum-likelihood classifier. I
examined each class from the resulting classification individually and visually compared
each to the original photography to evaluate the correspondence of each spectral class to
known ground features. This comparative procedure consists of flashing each of the
classes on and off repeatedly while viewing the image on the screen, and simultaneously
consulting the original photographs. The most distinctive classes had either very low

45
reflectance or very high reflectance in all three spectral bands. Extremely low reflectance
values corresponded to tree crowns, the shadows cast by tree crowns, or standing water.
Although appearing in the same spectral class, water was easily distinguished from tree
crowns and shadows by pattern and texture. High reflectance values corresponded to bare
sand patches, and human disturbances such as dirt roads and excavations. Naturally
occurring bare sand patches were nearly always associated with xeric habitats, and had a
distinctive, fine-grained pattern and texture compared to ground features created by
humans. Spectral classes with intermediate reflectance were much more difficult to
associate with ground features. In general, grass-dominated prairies, such as occur
between scrub patches on N. Sandy Hill, had high reflectances that were only slightly less
than bare sand. Areas dominated by oak shrubs were spectrally similar to areas
containing various proportions of palmetto and wire grass. Discriminating among mixed
shrub classes was difficult and also was believed unlikely to affect jay dispersion at
APAFR. The dominant and most recognizable spectral classes corresponded to tree
canopies and the shadows they cast, and bare sand patches. All spectral classes were
recoded to the following landcover classes: 1 = tree cover, 2 = bare sand, 3 = mixed
grass/shrub, 4 = wetlands/seeps. These classes were ;ntended to reflect key structural
components of the habitat rather than vegetative types.
Manual Editing of Classification
Manual editing of the final classification was necessary in some areas with very
dark signatures that were confused with tree shadows. These areas of confusion were all
wetlands or poorly drained areas with temporary standing water at the time the imagery

46
was acquired (March 1994). Most manual editing was accomplished using a single
mask file, which contained wetland polygons digitized from the original rectified
image. This mask file was used to recode all pixels inside of wetland polygons to the
wetland category. Special attention was given to areas immediately surrounding jay
territories. One cutthroat seep area adjacent to and just east of the NE territory (and
experimental plot 1) on S. Sandy Hill showed standing water and trees. In this area, two
dark pixel classes were found to have a good correspondence with tree cover, and two
dark pixel classes corresponded with standing water. Theses classes were recoded
accordingly.
Assessment of Classification Accuracy
Classification accuracy of percent tree cover was evaluated by comparison with
estimates from line transect data collected at five control plots on S. Sandy Hill as part of
the experimental manipulation experiment. A three pixel wide buffer was generated
around each transect using Arc/Info. This buffer was used to sample the imagery around
each transect using Erdas. Classification accuracy of bare sand was evaluated by
comparison with estimates from quadrat samples (70 m across) collected on the northern
section of S. Sandy Hill in areas largely devoid of pine trees. The measurements obtained
from the imagery and the transect measurements ere compared using a Paired T-Test
and a correlation analysis.

47
Digitization of Territories and Background Features
Polygons representing the boundaries of jay territories and other background
features (e.g. roads) were digitized directly off of the computer screen with a mouse from
the mosaiced image using Arc/Info. This allowed the resulting coverages to be
automatically georeferenced to the aerial photography
Tree Cover Buffering Procedure
GIS buffers were generated around each jay territory at 100, 200, and 400 m
distances using Arc/Info. These buffers were used to characterize the habitat immediately
surrounding each territory. A major complication with this procedure was that the buffers
for each territory often overlapped with neighboring territories, and overlapping polygons
are not allowed in Arc/Info coverages. Therefore, each territory was kept in a separate
coverage and analyzed individually. The following procedure was used. Each coverage
was buffered at distances of 100, 200, and 400 meters. The resulting coverages were
converted to individual ERDAS .dig files. Each .dig file was overlaid on the classified
image to calculate the percent tree cover within each buffer using an ERDAS program
called POLYSTAT (developed by B. Stith and J. Richardson). The statistics generated
from this program were imported into spreadsheet :.nd statistical packages (Systat; SAS)
for further analysis.

48
Habitat Quality Model
A habitat quality model was developed from the habitat variables using a habitat
suitability index (HIS) approach similar to Duncan et al. (1995). The HSI model
combines three HSI values for percent bare sand within territories (BS), percent tree
cover within territories (TC), and distance to nearest forest (DF), to calculate a single
habitat quality value (HQ) for each territory. The equation used for this model is
hq = Mbs *tc*df
The HSI values for each of the three habitat variables was obtained from step
functions relating the habitat variable to an estimate of habitat quality (see Fig. 3-2
modified from Duncan et al. 1995). The shapes of these step functions were developed
subjectively by D. Breininger. The habitat quality values were mapped automatically
across the entire study area using the Spatial Modeler in the Imagine software package.
The BS variable was mapped using the focal sum operator to count the number of bare
sand pixels within 10 m of every point in the study area, and computing an HSI value
using the function in Fig. 3-2a. The TC variable was similarly mapped by measuring tree
pixels within a 60 m radius, and computing an HSI value using the function in Fig. 3-2b.
To compute the DF variable, the TC layer was use ... and pixels surrounded by greater
than 30% tree cover were coded as forest. The search operator was used to measure the
distance to nearest forest for all pixels. The DF v.aue was then computed using the
function in Fig. 3-2c. The Imagine summary" function was used to output HSI values

49
for each territory. Because summary requires integer values for input, the HSI values
were resampled to 5 equal intervals.
Collection of Demographic Data
Demographic data for jays on Sandy Hill were collected by a team of field
researchers (Brad Stith, Reed Bowman, Doug Stotz, Larry Riopelle, Natalie Hamel, and
Mike McMillan) during 1994 1995, as part of a larger, on-going project that now
monitors jays on the entire APAFR. All jays on Sandy Hill were captured, color-banded,
and monitoried quarterly using techniques similar to Woolfenden and Fitzpatrick (1984).
Nests were found and monitored during the spring and early summer. The raw
demographic data are presented in Table 3-1.
Habitat-Demographic Analysis
I compared the demographic performance and habitat characteristics between the
North and South Sandy Hill populations of jays. Owing to lack of normality for nearly all
parameters (Table 3-2; Kolmorov-Smimov test for normality), the nonparametric Mann-
Whitney U statistic was used to test for demographic and habitat differences between the
North and South study areas (Table 3-3).
I searched for habitat-demographic relationships by performing multiple linear
regression (maximum R2 improvement technique for all combinations of variables) and
logistic regression (SAS Institute). Demography parameters served as dependent
variables, and habitat measurements as indepe- ,.ent variables.

50
Results
The locations and names of Florida Scrub-Jay territories throughout the study area
are shown in Fig. 3-1. The division between the North and South population occurs at
Kissimmee Rd. Note the absence of jays within the South population in the central part
of S. Sandy Hill. This area has high densities of pine trees and little bare sand, and has
three experimental plots where habitat restoration is underway.
A best-fit regression line, constrained to pass through the origin, showed a
significant relationship between percent bare sand measured from quadrats vs. imagery
(Fig. 3-4; r-squared = 0.60). The difference between the quadrat and image measurements
was not significant (Paired T-test; mean difference = 0.9372, n=12, p=0.256), indicating
no systematic bias in the measurements. Figure 3-5 shows the relationship between
percent tree cover measured from transect data vs. imagery. A best-fit logarithmic
regression line was drawn through the points (r-squared = 0.25). The differences between
the transect and image measurements was not significant (Paired T-test; mean difference
= 0.9372, n=40, p=0.824), indicating no systematic bias in the measurements. However,
the large scatter of points deviated considerably from the expected distribution which
would fall on a regression line intercept! 4 the origin and having a slope of one. The
logarithmic trend line suggested that the nage measurements give higher than expected
measurements at low tree cover values, id low-: nan expected measurements at high
tree cover values.
Percent tree cover ranged from 4% to ,% on S. Sandy Hill (Fig. 3-7), and only
0% to 4% on N. Sandy Hill (Fig. 3-6). The difference in tree cover between North and
South was highly significant (Table 3-3; Mann-Whitney U Test, Z = -6.291, P <

51
0.0000). Most of the territories with high tree cover occured in the south end of S. Sandy
Hill. Much greater variation was evident on S. Sandy Hill compared to N. Sandy Hill.
Percent bare sand ranged from 9% to 30% on S. Sandy Hill, and 2% to 34% on N. Sandy
Hill. The difference in bare sand cover between North and South was not significant
(Table 3-3; Mann-Whitney U Test, Z = -1.595, P = 0.111).
Comparison of tree cover in North vs. South territories and in the three buffer
zones (100, 200, and 400 m) showed striking differences (Fig. 3-8) and were highly
significant for all comparisons. South territories showed a dramatic increase in tree cover
with increasing distance from territories, while North territories showed only a slight
increase. In the North, tree cover increased only slightly if at all as distance from territory
increased (Fig. 3-9). In comparison, South territories showed large increases in tree cover
as distance from territory increases (Fig. 3-10). A strong correlation was found between
tree cover within territories and buffer zones in the South (e.g. r-squared = 0.88 for
territories vs. 100m buffer), suggesting that these variables had strong spatial
autocorrelation.
Comparison of tree cover within jay territories vs. total tree cover showed that jay
territories have fewer trees than the area available to them for both the North and South
territories (Fig. 3-11). Note that the measurements of available tree cover were made
across the entire North and South study area rather than from subsamples, hence no
standard deviations were calculated.
Measurements of bare sand within territories and the buffer zones showed a
decreasing amount of bare sand away from territories for both the North and South
territories (Figure 3-12), indicating selection of sandy areas by jays.

52
Comparison of demographic performance of jays in North vs. South for two years
(1994, 1995) showed significant differences only for nonbreeder survival (Table 3-3;
Mann-Whitney U Test, Z = -2.396, P = 0.017). Group size was nearly significant (P =
0.060). Number of fledglings produced, fledgling survival, yearling survival, and breeder
survival were not significantly different.
Relationships between group size and percent tree cover within all territories (Fig.
3-13), and group size and percent tree cover within 100 m buffers (Fig. 3-14) showed
decreasing group size with increasing tree cover. I lumped group size into 3 categories of
roughly equal size and looked for differences in tree cover at the territory (Fig. 3-15) and
100 m buffer (Fig. 3-16). Large families (4 to 7 jays per group; n = 13) had lower median
and variance in tree cover within and adjacent to their territories compared to medium (3
jays per group; n = 7) and small (2 jays per group; n=15) families. The differences,
however, were not significant (Kruskal-Wallis one-way analysis of variance; tree cover P
= 0.205; 100 m buffer P = 0.186). I lumped the 3 group categores into 2 group size
categories (2-3 jays per group; 4-7 jays per group) and performed the same analysis, but
the results were not significant.
Figures 3-17, 3-18, and 3-19 show side-by-side views of the classified and raw
images for N. Sandy Hill, the N. portion of S. Sandy Hill, and the S. portion of S. Sandy
Hill respectively. Jay territories are outlined in black. The names of jay territories are
shown in Fig. 3-1. Three colors on the classified images correspond to tree cover (green),
bare sand (white) and mixed shrubby or grassy vegetation (brown).
Figures 3-20 and 3-21 show presumed habitat quality as computed from the three
HSI variables for the N. and S. portion of S. Sandy Hill respectively. High quality habitat

53
is shown in red (HSI = 0.81 1.0), medium quality habitat is shown in blue (HSI = 0.61 -
0.80), and low quality habitat is shown as white (HSI < 0.61). Jay territories (outlined in
black) generally included substantial areas of low quality habitat in both the North and
South areas.
I searched for correlations between demographic (fledglings, independent young,
and yearlings produced, survival of fledglings, yearlings and breeders, and group size)
versus the habitat variables (percent sand, percent tree cover within territories, percent
tree cover within 100 m buffer, HSI for sand, HSI for distance to forest, HSI for trees
within habitat, and combined HSI). The clearest bivariate patterns were for group size
versus percent tree cover within territories (Fig. 3-13) and percent tree cover within the
100 m buffer (Fig. 3-14), but Kruskal-Wallis one-way analysis of variance results showed
no significant differences between different group size comparisons and percent tree
cover. Large group size variance existed in territories with low tree cover or adjacent to
low tree cover, but there was a strong, nonsignificant trend towards smaller groups as tree
cover increases. Kolmogorov-Smimov tests for normality showed that bare sand was the
only normally distributed habitat variable. Normality plots indicated that deviations from
normality could not be corrected by commonly used (e.g. arcsine, inverse, log, square
root) transformations. Nonparametric spearman rank correlation coefficients were low for
all pairings of demographic and habitat variables. Multiple regression models never
explained more than about 22% of the variation in demographic parameters using all
combinations of habitat variables. Similarly, no habitat variables in several logistic
regression models were significant.

54
Discussion
Accuracy of my remotely sensed habitat measurements is difficult to assess. Most
remote sensing studies are conducted at a much coarser scale, and they attempt to identify
discrete data classes (e.g. vegetation types). Such studies typically use a simple error
matrix analysis where percent correct classification is given. I have little precedence to
follow, since the goal of this classification was to provide continuous measurements (i.e.
percent cover) from structural classes (e.g. tree cover or bare sand) rather than discrete
vegetation classes. To assess the accuracy of these measurements quantitatively, I
compared them to ground based measurements using paired T-Tests and simple
correlation analysis. The bare sand measurements obtained from quadrats showed a fairly
good correlation with the image measurements (Fig. 3-4). In contrast, the transect
measurements of tree cover showed a weak correlation (Fig. 3-5). Some of these
differences resulted from classification errors noticeable in comparisons of the
photography with the classified image. Underestimates of tree cover were apparent in
some of the young or very dense pine plantations, which tended to form a uniform
canopy with few shadows. Overestimates were noted in areas with large scrub oaks, such
as in some of the long unbumed scrub patches on the W. side of N. Sandy Hill. Larger
oaks spectrally may look very similar to pine trees. From a Florida Scrub-Jay standpoint,
stands of large oaks may be as unusable as pine forests, so for modeling jay habitat it may
be unnecessary to distinguish these tree cover types.
I suspect that many of the differences in tree cover estimates resulted from
differences in the locations of transects measured on the ground vs. the image. Because
transects are sampling vegetation intercepting a thin vertical plane, relatively small

55
difference in transect position can result in large differences in measurements. Quadrat
measurements are probably less sensitive to positional inaccuracy than transects.
It was surprisingly difficult to estimate the magnitude of the positional
inaccuracies of the transect locations. The locations of the transects were predetermined
by a program that generated random locations for transect endpoints. These transect
locations were plotted on high resolution photo maps which were taken into the field and
used to stake out the transects. Thus, the correspondence between the GIS-based location
and the actual ground location depended on the field persons ability to find the exact
location from the photo map. My qualitative impression was that accuracy of positioning
the transects depended on whether features visible on the photo map could be located on
the ground. In sparsely forested areas, individual trees and bare sand patches were
identifiable on the photo and ground, and served as good reference points. Under these
conditions, a transect could probably be placed within several meters of its position on
the photo. In heavily forested areas, good reference points were absent, and positional
accuracy probably decreased, to errors of 10 meters or more. It might seem that
differential GPS could solve this potential problem. Unfortunately, several factors make
this approach more problematic than anticipated. First, the stated accuracy of the
differential GPS approach available to us is 2 5 meters, which can misplace the ends of
a transect by 3 pixels in any direction. Also, I have occasionally encountered averaged,
differentially corrected points that are off by considerably more than 5 meters. Second,
current GPS units often are unable to pick up the necessary signals within forests;
precisely where they are most needed for this study. Third, the positional accuracy of the
imagery itself is unknown, but is probably on the order of 2 to 10 meters (see

56
georeferencing section). Since there is no reason to expect errors in georeferencing to
have the same bias as errors in GPS readings, the difference between the two could
compound to exceed 15 meters. An error of this magnitude probably exceeds the
expected error from a field person using a high resolution photo map to position a
transect. Unfortunately, this leaves us with no way of quantifying and correcting
positional inaccuracies. Recent advances in GPS technology may solve these problems as
sub-meter accuracy becomes increasingly affordable and practical.
Quadrat samples and visual comparison of the classified image with the aerial
photographs plus field knowledge suggest that the classification accurately reflected
biologically important differences among habitats. The image processing techniques,
combined with GIS files of the locations of jay territories and buffer zones, provided
quantitative measurements of jay habitat in a quick and efficient manner. The quantitative
results show dramatic differences in habitat structure between the North and South areas
corresponding well to impressions reported by field researchers (R. Bowman pers.
comm.). N. Sandy Hill jays have far fewer trees within and adjacent to their territories
compared to S. Sandy Hill jays.
The image processing results confirmed our general field impressions about
which jay families were living in good and poor quality habitat on Sandy Hill. Jays that
occupied the poorest habitat were in the southern part of S. Sandy Hill. Here, several
families were living in low quality habitat near a recent bum that was occupied by two or
three other families. The presence of jays in poor habitat resulted from conspecific
attraction (Smith and Peacock 1990) or "queueing" behavior, where individuals stay near
high quality habitat to wait for breeding vacancies. Jays living in such poor habitat,

57
adjacent to good habitat, may create high variance in habitat-demographic relationships.
Nevertheless, it is clear from the comparison of used versus available habitat that jays
preferentially selected habitat with low tree cover (Fig. 3-11). These conditions exist
throughout much of the N. Sandy Hill area, but on S. Sandy Hill they generally occurred
only where recent fires were hot enough to kill most of the pines. These bums are
embedded in a pine forest matrix (Fig. 3-9) presumably full of jay predators and
competitors. In contrast, jays in N. Sandy Hill live in higher quality habitat embedded in
a less hostile habitat matrix (Fig. 3-10). I suspect that these differences in tree cover
explain the observed difference in nonbreeder survival between the North and South
subpopulations.
In contrast to the observed difference in nonbreeder survival between the North
and South subpopulations, and the nearly significant difference in group size, few clear
relationships are obvious when pooling the two subpopulations and looking at individual
territories. The locations of jay territories relative to the areas of highest habitat quality
(Figs. 3-20 and 3-21) show that most jay territories included substantial areas of poor
quality habitat. The scrub-jay model of Duncan et al. (1995) gave similar results; about
65% of their jay territories by area had HSI values below 0.5. Jay territories typically
incorporate unusable habitat types simply because their territories are large in size
relative to the patchiness of their habitat (Woolfenden and Fitzpatrick 1984). Both the
North and South study areas are lacking in large, contiguous, high quality habitat patches.
In the North, scrub patches occur on small lenses of well-drained soil surrounded by
poorly drained soil. Scrub vegetation characteristic of high quality habitat can only grow
on these well-drained soils. In the South, much larger areas of well-drained soil capable

58
of supporting high quality habitat are present, but high quality habitat occurs only in
small areas that were recently burned or cleared experimentally. Five of 6 experimental
plots were classified as relatively large patches of high quality habitat (see Fig. 3-21).
Only plot 6 was not classified as high quality; for unknown reasons it had a dark
reflectance when the area was photographed in March 1994. No jays have become
established in the three isolated experimental plots (2, 3, and 4), despite their appearance
as high quality. We suspect that conspecific attraction is extremely important to this
species, greatly reducing the likelihood that solitary jays will become established in
unoccupied, isolated patches.
The most important relationships I found for individual territories among the
habitat-demography variables is a decline in group size with increasing tree cover, both
within territories and within the 100 m buffer zones (Figs. 3-14 and 3-15). Large family
sizes only occur in territories with low tree cover and adjacent to low tree cover. Variance
in group size is high in the North because some groups accrue large size here, but not in
the South. Territories in or adjacent to habitat with moderate to high tree cover have
predictably small group sizes, presumably a result of successive years with poor
productivity and low survival. A similar pattern between group size and tree cover can be
seen in Figs. 3-16 and 3-17, where group size is lumped into 3 categories of roughly
equal size.
Demographic parameters other than group size showed no clear patterns with the
habitat variables at the individual territory level. Group size may be the least noisy
measure of demographic success, since it integrates past and current demographic
performance. Helper survival may be especially sensitive to habitat quality, since helpers

59
in poor quality habitat may have a greater tendency to emigrate than helpers in high
quality habitat.
The remote sensing techniques described above show significant potential for
evaluating Florida Scrub-Jay habitat. Tree cover and bare sand are important habitat
variables that are relatively easy to measure with these techniques. Oak cover is likely to
be important to scrub-jays, but the techniques I investigated could not discriminate oaks
from other low-lying vegetation such as palmettos. Because of low mast failure and high
acorn production, oak cover may not be a limiting factor for many scrub-jay populations.
Further investigation of the importance of oak cover is needed.
The results of this study suggest that tree cover exceeding 20% 30% within or
adjacent to territories is associated with reduced demographic performance and may
create sink populations of Florida Scrub-Jays. Although I lack direct evidence, much
indirect evidence suggests that forest-dwelling predators and competitors explain the
negative relationship between tree cover and demographic success. Sink populations can
constitute a major proportion of a species population and may contribute to
metapopulation longevity (Howe and Davis 1991), but the loss of a single critical source
population may result in the extinction of all dependent sink populations (Pulliam 1988).
Thus, management practices should seek to convert sink populations to self-sustaining
source populations. The results of this study suggest that for the Florida Scrub-Jay, this
entails keeping tree cover within and adjacent to jay territories at relatively low levels.

60
North Sandy Hill
dtch;
Florida Scrub Jay Territories
Spring 1994
South Sandy Hill
IN
A
2 Kilometers
Fig. 3-1. Map of Scrub-Jay Territories Spring 1994. Dividing line between North and
South populations is the Kissimmee Rd..

61
Fig. 3-2. Habitat Suitability Index graphs for percent bare sand, percent tree cover, and
distance to forest (modified from Duncan et al. 1995).

62
Fig. 3-3. Ilius.ra.ive map of 100,200. and 400 m buffer zones around LOST territory.

63
Fig. 3-4. Accuracy assessment correlation graph for bare sand based on quadrat vs.
classified image measurements (r-squared = 0.60).

64
Fig. 3-5. Accuracy assessment correlation graph for tree cover based on transect vs.
classified image measurements (r-squared = 0.25).

dwna
dNMl
ISdd
Hoia
10H1
1A10
IdlN
dVOS
aans
HAOX
NHd
dM 01
sad*
AAJkA
3H10
Aano
iavd
OdVO
a
6 o
to
6
to
6
to n -
6 o' 6 o'
o
uo|)iodaid
Fig. 3-6. Measurements of sand, tree cover, and mixed vegetation for Spring 1994
territories on N. Sandy Hill.
Tcrrltoriei

66
SNdS
dOlH
SOIS
SOOT
OVNS
0H03
NdM
0V3Q
sim
VldX
xiua
SSdO
idon
1S01
uo(podoid
Fig. 3-7. Measurements of sand, tree cover, and mixed vegetation for Spring 1994
territories on S. Sandy Hill.
Territories

67
o
>
O
O
*->
c
CD
O
u.
O
CL
M North
B South
Territory 100 m 200 m 400 m
Buffer Distance
Fig. 3-8. Percent tree cover (mean and standard deviation) for 4 zones (inside territories,
100, 200, 400 m buffer) North vs. South Sandy Hill. South population shows
significantly higher tree cover within all zones compared to North population.

Prop. Cover
68
GARO
--RADI
CURV
-K CTRE
*YYYY
ARDS
ITOWR
FUN
XOVR
O BUDD
aSOAR
6NTRL
-H-CLVT
*TRGT
*DTCH
)FRST
TWNP
DUMP
Buffer Distance
Fig. 3-9. Percent tree cover for individual territories for 4 zones (inside territories 100
200, 400 m buffer) North territories.

Prop. Cover
69
LOST
-B-LOPI
-A GPSS
XBRIK
*TRIA
--TRIS
iOEAD
WPND
ECHO
O SNAG
-O-IOGG
-A-SLOG
tt-HTOP
*SRNG
PARS
IJSUS
NORE
Buffer Distance
Fig. 3-10. Percent tree cover for individual territories for 4 zones (inside territories 100
200,400 m buffer) South territories.

70
O)
>
O
U
O)
O)
c
O)
u
a>
CL
Jays
Available
u n 3;!L TLree cover within jay territories vs. total tree cover in North vs. South Sandy
Hill. Note that jays select habitat with lower tree cover in both areas.

71
30
25
JD
a
Territory 100 m 200 m 400 m
Buffer Distance
Fig. 3-12. Percent bare sand (mean and standard deviation) for 4 zones (inside territories,
100, 200,400 m buffer) North vs. South Sandy Hill. Differences between two areas are
not significant.

72
Prop.
Tree Cover
Fig. 3-13. Group size vs. percent tree cover within all territories (North and South
populations pooled). Trend towards smaller group size with higher tree cover is not
significant.

73
Fig. 3-14. Group size vs. percent tree cover within 100 m buffer for all territories (North
and South populations pooled). Trend towards smaller group size with higher tree cover
is not significant.

Prop tree cover
74
0.5
0.4
0.3
0.2
0.1
0.0
o
I I
Large Med.
Group size
T
_i
Small
Fig. 3-15. Group size (small = 2, medium = 3, large = 4 7 jays) vs. percent tree cover
within all territories (North and South populations pooled). Trend towards smaller group
size with higher tree cover is not significant.

75
>
O
O
0)
0)
Q.
O
U-
Q-
0.6
0.5
0.4
0.3
0.2
0.1
0.0
Large Me(F
Group size
Small
Fig. 3-16. Group size (small = 2, medium = 3, large = 4 7 jays) vs. percent tree cover
within 100 m buffer (North and South populations pooled). Trend towards smaller group
size with higher tree cover is not significant.

Fig. 3-17. Images and territories (black polygons) of North Sandy Hill. Right: color-infrared image. Left: classified image (white -
bare sand; green = trees; brown = shrubs/grass; black = water).
Os

vm
Fig. 3-18. Images and territories (black polygons) ofN. portion of South Sandy Hill. Right: color-infrared image. Left: classified
image (white = bare sand; green = trees; brown = shrubs/grass; black = water).

Fig. 3-19. Images and territories (black polygons) of S. portion of South Sandy Hill. Right: color-infrared image. Left: classified
image (white = bare sand; green = trees; brown = shrubs/grass; black = water).
-j
OO

Habitat Quality N. Sandy Hill
Ten-94
/\/ 1994 Jay Territories
Roads
A / Roads
1.5 Kilometers
Fig. 3-20. Habitat quality map of N. portion of South Sandy Hill.

80
Habitat Quality
S. Sandy Hill
Ex_plotr
/V Experimental Plots
Terr94
A/ 1994 Jay Territories
Roads
A. / Roads
0.5 0 0.5 1 Kilometers
Fig. 3-21. Habitat quality map of S. portion ofSouth Sandy Hill.

Table 3-1. Demographic and habitat parameters for North and South Sandy Hill (1994 1995).
Location
Territory
Year Group
Size
Fledgling
Production
Yearling
Production
Breeder
Surv.
Helper
Surv.
Fledgl.
Surv.
Yearl.
Surv.
Bare
sand
Tree cover
(inside
territory)
Tree cover
(100 m
buffer)
NRIDGE
ARDS
1994
4
2
2
1
1
1
0
0.02
0.02
0.02
NRIDGE
BUDD
1994
2
2
2
0.5
1
0
0.15
0.01
0.01
NRIDGE
CLVT
1994
4
3
2
1
1
0.66
0
0.28
0.01
0.07
NRIDGE
CTRE
1994
4
2
0
0.5
0.5
0
0
0.04
0.03
0.02
NRIDGE
CURV
1994
2
4
1
0.5
0.25
0.25
0.16
0.03
0.03
NRIDGE
DTCH
1994
2
0
0
1
0.2
0.02
0.06
NRIDGE
DUMP
1994
3
0
0
1
1
0.23
0
0.06
NRIDGE
FLIN
1994
5
1
0
1
0.66
0
0
0.05
0
0.02
NRIDGE
FRST
1994
4
3
3
0.5
0.5
1
0.66
0.34
0
0
NRIDGE
GARD
1994
3
4
1
1
1
0.25
0
0.26
0.02
0.05
NRIDGE
NTRL
1994
2
2
1
1
0.5
0.5
0.19
0.02
0.02
NRIDGE
RADI
1994
6
0
0
0.5
0.33
0.16
0.03
0.03
NRIDGE
SQAR
1994
8
3
2
1
0.66
0.66
0
0.12
0.03
0.02
NRIDGE
TOWR
1994
5
4
3
0
0.66
0.75
0.25
0.15
0.04
0.06
NRIDGE
TRGT
1994
2
2
1
1
0.5
0
0.11
0.01
0.02
NRIDGE
TWNP
1994
4
3
3
1
0.5
1
0
0.25
0
0.02
NRIDGE
YYYY
1994
4
1
1
0.5
1
1
0
0.08
0.04
0.04
SRIDGE
BRIK
1994
2
4
3
0.5
0.75
0.25
0.3
0.05
0.07
SRIDGE
DEAD
1994
3
3
2
1
0
0.66
0
0.16
0.16
0.24
SRIDGE
ECHO
1994
3
0
0
1
0
0.25
0.16
0.26
SRIDGE
FARS
1994
4
5
3
1
0
0.6
0.2
0.21
0.28
0.28
SRIDGE
GPSS
1994
5
3
0
1
1
0
0
0.21
0.08
0.1
SRIDGE
HTOP
1994
2
3
2
0.5
0.66
0
0.15
0.33
0.53
SRIDGE
JSUS
1994
2
1
0
0.5
0
0
0.25
0.31
0.43

Table 3 1. Demographic and habitat parameters for North and South Sandy Hill (1994 1995).
Location
Territory
Year Group
Size
Fledgling
Production
Yearling
Production
Breeder
Surv.
Helper
Surv.
Fledgl.
Surv.
Yearl.
Surv.
Bare
sand
Tree cover
(inside
territory)
Tree cover
(100 m
buffer)
NRIDGE
ARDS
1994
4
2
2
1
1
1
0
0.02
0.02
0.02
NRIDGE
BUDD
1994
2
2
2
0.5
1
0
0.15
0.01
0.01
NRIDGE
CLVT
1994
4
3
2
1
1
0.66
0
0.28
0.01
0.07
NRIDGE
CTRE
1994
4
2
0
0.5
0.5
0
0
0.04
0.03
0.02
NRIDGE
CURV
1994
2
4
1
0.5
0.25
0.25
0.16
0.03
0.03
NRIDGE
DTCH
1994
2
0
0
1
0.2
0.02
0.06
NRIDGE
DUMP
1994
3
0
0
1
1
0.23
0
0.06
NRIDGE
FLIN
1994
5
1
0
1
0.66
0
0
0.05
0
0.02
NRIDGE
FRST
1994
4
3
3
0.5
0.5
1
0.66
0.34
0
0
NRIDGE
GARD
1994
3
4
1
1
1
0.25
0
0.26
0.02
0.05
NRIDGE
NTRL
1994
2
2
1
1
0.5
0.5
0.19
0.02
0.02
NRIDGE
RADI
1994
6
0
0
0.5
0.33
0.16
0.03
0.03
NRIDGE
SOAR
1994
8
3
2
1
0.66
0.66
0
0.12
0.03
0.02
NRIDGE
TOWR
1994
5
4
3
0
0.66
0.75
0.25
0.15
0.04
0.06
NRIDGE
TRGT
1994
2
2
1
1
0.5
0
0.11
0.01
0.02
NRIDGE
TWNP
1994
4
3
3
1
0.5
1
0
0.25
0
0.02
NRIDGE
YYYY
1994
4
1
1
0.5
1
1
0
0.08
0.04
0.04
SRIDGE
BRIK
1994
2
4
3
0.5
0.75
0.25
0.3
0.05
0.07
SR1DGE
DEAD
1994
3
3
2
1
0
0.66
0
0.16
0.16
0.24
SRIDGE
ECHO
1994
3
0
0
1
0
0.25
0.16
0.26
SRIDGE
FARS
1994
4
5
3
1
0
0.6
0.2
0.21
0.28
0.28
SRIDGE
GPSS
1994
5
3
0
1
1
0
0
0.21
0.08
0.1
SRIDGE
HTOP
1994
2
3
2
0.5
0.66
0
0.15
0.33
0.53
SRIDGE
JSUS
1994
2
1
0
0.5
0
0
0.25
0.31
0.43

Table 3-1 continued.
Location Territory Year Group Fledgling Yearling Breeder Helper
Size Production Production Surv. Surv.
SR1DGE
LOGG
1994
4
0
0
0.5
0.5
SRIDGE
LOPI
1994
2
3
1
0
SRIDGE
LOST
1994
3
2
2
1
0
SRIDGE
NORE
1994
2
1
0
0.5
SRIDGE
SLOG
1994
3
0
0
0.5
0
SRIDGE
SNAG
1994
2
4
1
1
SRIDGE
SRNG
1994
2
4
1
1
SRIDGE
TRIS
1994
2
3
1
1
SRIDGE
WPND
1994
2
3
1
1
NR1DGE
ARDS
1995
5
3
0
0.5
0.33
NRIDGE
BUDD
1995
6
2
0
0.5
0.66
NR1DGE
DTCH
1995
2
3
0
1
NRIDGE
DUMP
1995
2
4
1
1
NRIDGE
FLIN
1995
2
1
1
0.5
NRIDGE
FRST
1995
5
0
0
1
0.66
NRIDGE
GARD
1995
4
2
0
0
0.5
NRIDGE
JUVI
1995
2
1
0
1
NRIDGE
NTRL
1995
3
3
1
0.5
1
NRIDGE
PURP
1995
2
3
0
0
NRIDGE
SQAR
1995
5
3
1
1
0.66
NRIDGE
TOWR
1995
5
3
1
0.5
0.66
NRIDGE
TRGT
1995
3
1
1
0.5
1
NRIDGE
TWNP
1995
6
2
2
1
0.33
Fledgl. Yearl. Bare Tree cover Tree cover
Surv. Surv. sand (inside (100m
territory) buffer)
0.33
1
0
0.25
0.25
0.33
0.33
0
0
0
0.25
1
0
0
0.33
0
0.33
0.33
1
1
0.33
0
0
0
0.25
0.33
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0.14
0.28
0.21
0.24
0.21
0.11
0.29
0.17
0.1
0.01
0.19
0.23
0.29
0.09
0.24
0.32
0.08
0.26
0.1
0.15
0.21
0.21
0.16
0.22
0.04
0.08
0.07
0.13
0.46
0.31
0.14
0.28
0.02
0.07
0.01
0.01
0.03
0.04
0.09
0.02
0.01
0.02
0.03
0.01
0.06
0
0.28
0.08
0.01
0.12
0.25
0.44
0.35
0.19
0.24
0.02
0.01
0.01
0.01
0.01
0.02
0.06
0.02
0.01
0.02
0.02
0.04
0.02
0.02
oo
UJ

Table 3-1 continued.
Location Territory Year Group Fledgling Yearling Breeder Helper
Size Production Production Surv. Surv.
NRIDGE XOVR
1995
2
4
0
0
NRIDGE
YYYY
1995
5
0
0
0.5
0.66
SRIDGE
DEAD
1995
4
1
0
1
1
SRIDGE
ECHO
1995
2
0
0
1
SRIDGE
EXP5
1995
3
0
0
0.5
1
SRIDGE
FARS
1995
4
4
3
1
0
SRIDGE
GPSS
1995
5
0
0
0
0.33
SRIDGE
HTOP
1995
4
0
0
0
0
SRIDGE
JSUS
1995
2
4
2
1
SRIDGE
LOGG
1995
2
0
0
0
0
SRIDGE
LOPI
1995
4
0
0
0.5
0.5
SRIDGE
LOST
1995
3
0
0
1
0
SRIDGE
NORW
1995
3
1
0
0.5
1
SRIDGE
SNAG
1995
4
1
1
0
0.5
SRIDGE
SRNG
1995
3
3
0
1
1
SRIDGE
TRIS
1995
3
1
1
1
1
SRIDGE
WPND
1995
3
3
2
0.5
0
Fledgl. Yearl. Bare Tree cover Tree cover
Surv. Surv. sand (inside (100m
territory) buffer)
0
0
0.75
0.5
0
1
0
1
0.66
0
0
0.25
0
0
0
0
0
0
0.13
0.07
0.11
0.28
0.24
0.22
0.32
0.13
0.09
0.16
0.26
0.17
0.21
0.14
0.25
0.21
0.13
0.02
0.06
0.26
0.17
0.1
0.26
0.01
0.36
0.53
0.18
0.07
0.14
0.03
0.48
0.33
0.08
0.23
0.01
0.05
0.24
0.18
0.26
0.3
0.05
0.48
0.55
0.21
0.1
0.17
0.03
0.45
0.37
0.12
0.29
oo

Table 3-3. Mann-Whitney_U test for differences in demographic and habitat variables between North and South jay populations (*
significantly).
Group
size
Fledgling
production
Yearling
production
Breeder
survival
Nonbreeder
survival
Fledgling
survival
Yearling
survival
Bare sand
Tree cover
(inside
territory)
Tree cover
(100 m
buffer)
Mann-
Whitney
U
376.500
453.500
480.500
508.000
127.500
286.500
257.000
393.000
50.000
45.000
Z
-1.881
-.798
-.446
-.052
-2.396
-.155
-1.063
-1.595
-6.219
-6.308
P-value
.060
.425
.655
.959
.017*
.877
.288
.111
.000*
.000*
OO

CHAPTER 4
MODELING DISPERSAL IN THE FLORIDA SCRUB-JAY
Introduction
Dispersal is a fundamental biological process of great importance to many fields
of biology, including population biology, population genetics, behavioral ecology, and
conservation. Understanding dispersal is becoming increasingly important in
conservation biology, as dispersal ability may determine whether a given species can
survive in the face of ever-increasing habitat fragmentation. Spatially explicit population
models (SEPM) provide a formal framework for investigating the importance of dispersal
to the viability of populations within a given landscape. The use of SEPMs in
conservation is growing rapidly (Beissinger and Westfal 1998). Yet, little is known
about dispersal for most species, leading some to question the value of predictions
obtained from SEPMs. Thus, the quality of dispersal data and its use in population
modeling was a key issue in an important court case involving the Federally threatened
Northern Spotted Owl (Harrison, Stahl, and Doak 1993). More recently, Winnergren et
al. (1995) and Ruckelhaus et al. (1997) developed a simple dispersal model that was
extremely sensitive to relatively small differences in estimates of dispersal mortality.
They argued that dispersal data would never be known with enough accuracy to be used
reliably in SEPMs. In response to these criticisms, Mooij and DeAngelis (1999) and
South (1999) developed alternative simple models showing SEPMs to be relatively
86

87
insensitive to errors in dispersal parameters except under very limited conditions. The
debate over the reliability of SEPMs is likely to continue, but as South (1999) points out,
model sensitivity to dispersal parameters may be greatly reduced by increasing model
realism.
The Florida Scrub-Jay offers an excellent opportunity to develop an extremely
realistic SEPM. Extensive dispersal information is available for this intensively studied
species (e.g. Woolfenden and Fitzpatrick 1984). Furthermore, there is great concern for
this species, which is listed as a threatened species Federally and by the State of Florida.
Much is known about the spatial distribution of the species (e.g. chapter 2), making it
feasible to model the entire population. Root (1996,1998) used RAMAS GIS to develop
the first SEPM for scrub-jays. Roots research was focused on four somewhat isolated
populations within Brevard county, Florida. She modeled dispersal among these four
populations for female jays only, using the interpatch distances and the ABS dispersal
curve to estimate migration rates. Her results suggested that interpatch dispersal was
important for offsetting the deleterious effects of epidemics. Root (1998) stated that her
estimates of dispersal likely were optimistic and suggested that a better approach would
account for differential dispersal rates based on the interpatch matrix.
In this chapter I describe a SEPM I developed to account for the influence of
interpatch matrix on dispersal, as well as a host of other biological details documented in
the extensive scrub-jay literature (see Woolfenden and Fitzpatrick 1996 for a recent
literature review). The SEPM is an individual-based model; it tracks all individuals of
both sexes from birth to death, and simulates the daily movement of individuals during
dispersal within and between habitat patches.

88
Individual-based models are appealing because they allow the inclusion of almost
any biological detail, giving them unrivaled realism (Huston et al. 1988; Judson 1994).
Because they often incorporate behavioral mechanisms, the parameters in an individual-
based model tend to have clear biological meaning and typically are directly measurable
in the field. The increased level of detail inherent in an individual-based model also
provides more opportunities to compare model output with different types of field data
for validation purposes. Finally, the increased realism of individual-based models makes
them more likely to reveal dynamics that would otherwise be missed in a less detailed
model.
Individual-based models have been developed for a few avian species, including
Bachman Sparrow (Pulliam et al. 1992), Northern Spotted Owl (McKelvey et al. 1993),
American Wood Stork (Wolff 1994), Helmeted Honey Eater (McCarthy 1996), and Red-
cockaded Woodpecker (Letcher et al. 1998). The excellent field data and well-known
behavioral characteristics of the Florida Scrub-Jay make this species an excellent
candidate for an individual-based modeling approach.
The overall objectives of this chapter are as follows:
Develop a set of algorithms and parameters specifically for the Florida Scrub-
Jay to simulate dispersal in an individual-based model.
Calibrate the dispersal module using long term field data from Archbold
Biological Station data, and radiotelemetry data acquired from a displace-and-
release experiment.

89
Validate the dispersal module by comparing model output to Archbold
Biological Station dispersal and stage-age data.
Use constraint analysis to place plausible bounds on long distance dispersal
parameters. Using a realistic digital landscape for which dispersal data is
available, develop parameters that result in simulated dispersals between
patches with known exchanges, and no simulated dispersals between patches
with no observed exchanges.
Dispersal Strategies
The ornithological literature generally recognizes two extreme dispersal
strategies: delay and foray versus depart and float. The floater strategy, wherein
nonbreeders depart from their natal territory and wander in search of mates or unoccupied
habitat with little or no tendency to return to the natal territory, is common among non-
cooperative breeding birds (Zack 1990). From an optimal foraging standpoint, to move
through poor habitat efficiently floaters should travel in a fairly linear fashion, moving
with a low turning rate, weak turn bias, and strong locomotory rate (Bell 1992; Turchin
1998). Such straight line movements by floaters have been observed for Red-Cockaded
Woodpeckers (Letcher et al. 1998) and a variety of other organisms (Zollner and Lima
1999). This linear movement allows floaters to quickly leave unproductive areas. Once
good habitat or potential mates are encountered, a floater should move in a manner that
keeps it within that patch by moving with a high turning rate, strong turn bias, and small
movements (Bell 1992; Turchin 1998). This type of localized searching behavior is also
exhibited by delay and foray dispersers, who engage in temporary dispersal forays
away from their natal territories in search of breeder vacancies (Woolfenden and

90
Fizpatrick 1984). Because such forays usually are unsuccessful and end with a return to
the natal territory, the territories visited tend to fall within an assessment sphere around
the natal territory. When a successful dispersal occurs, the resulting dispersal distances
will be smaller on average than for floaters (Zack 1990).
Dispersal Traits of the Florida and Western Scrub-Jav
The floater and delay-and-foray dispersal strategies are well illustrated by the
contrasting behavior of Florida and Western Scrub-Jays. Formerly considered a
subspecies of the Western Scrub-Jay, the Florida Scrub-Jay was recently given full
species status (Woolfenden and Fitzpatrick 1996). The Florida Scrub-Jay is likely
descended from jays that colonized Florida from western North America during the
Pleistocene, perhaps as recently as 4 million years ago (Woolfenden and Fitzpatrick
1996). Subsequent isolation of jays in Florida from the western population resulted in the
evolution of significant genetic and behavioral differences (Woolfenden and Fitzpatrick
1996). Some of the most striking differences between the two species relate to dispersal
behavior, a fact which makes comparison of the two species helpful in developing a
model of dispersal behavior.
Several lines of evidence suggest that Florida Scrub-Jays saturate high quality
habitat, and do not occupy marginal habitat because survival and reproduction is poor
(Woolfenden and Fitzpatrick 1984; Fitzpatrick and Woolfenden 1986). To survive, young
Florida Scrub-Jays must occupy high quality habitat. But Florida Scrub-Jays are highly
territorial and vigorously chase away non-resident jays; dispersers cannot easily move
through occupied habitat. Instead, they employ a delay and foray strategy, wherein they
make short forays to investigate nearby breeder vacancies or unoccupied habitat, and

91
retreat to their natal territory when chased by resident breeders. By contrast, the floater
strategy prevails in Western Scrub-Jays in part owing to breeder tolerance of nonbreeders
in their territory (except briefly at the beginning of breeding season; Carmen 1989;
Koenig et al. 1992). This option of floating among breeders in high quality habitat is
unavailable to Florida Scrub-Jays because breeders are largely intolerant of floaters.
Thus, as floaters Florida Scrub-Jays would be forced to reside in marginal habitat where
they would suffer high predation rates as documented in Fitzpatrick and Woolfenden
(1986). Western Scrub-Jays also have an advantage here, since their survival rates in
marginal habitat are nearly the same as in high quality habitat (Carmen 1989; Koenig et
al. 1992).
The above considerations are nicely encapsulated by Fitzpatrick and Woolfenden
'V
(1986) in an evolutionary model of dispersal for the genus Aphelocoma. Their model
suggests that selection will favor individuals who attempt to breed immediately upon
maturation, rather than delaying dispersal, unless the cost of dispersal is high compared
to the cost of remaining home and not breeding. Delaying dispersal, foregoing early
breeding opportunities, and engaging in low risk forays are the predominant dispersal
behavior for Florida Scrub-Jays in natural, high quality habitat. Nevertheless,
documented cases of long distance dispersal by some Florida Scrub-Jays make it clear
that they sometimes engage in risky dispersal behavior by moving through matrix
habitat between high quality habitat patches. Such behavior, though not completely
analogous to Western Scrub-Jay dispersal, can be modeled as floater dispersal behavior.
These considerations suggest that a reasonable approach to modeling dispersal in the

92
Florida Scrub-Jay is to simulate both types of dispersal behavior (i.e., delay-and-foray
and float).
Methods
General Approach
I simulated dispersal in the Florida Scrub-Jay with two distinct algorithms, one
for jays that engage in short forays away from their natal territory (philopatric
algorithm), another for jays that become floaters and move long distances from their natal
territory (floater algorithm). The philopatric dispersal algorithm is intended, 1) to
simulate the prevalent mode of dispersal observed in real jay populations, 2) to produce
the majority of dispersals within a given simulation, and 3) to produce the modal
distribution of dispersal distances in simulated populations. The philopatric algorithm
produces no dispersals beyond a specified radius referred to as the assessment sphere.
The floater algorithm completely determines the tail of the distribution, as philopatric
jays settle only within the radius of the assessment sphere. Together, the two algorithms
produce the combined dispersal curve; the philopatric algorithm produces dispersals
ranging from the natal territory (i.e. inheritance) out to the radius of the assessment
sphere, the floater algorithm produces dispersals beyond the assessment sphere.
Considerable information from long term, color band studies (e.g. Woolfenden
and Fitzpatrick 1984) is available to aid in the simulation of philopatric dispersal. In
contrast, much less is known about jays that disperse as floaters. Although some long
distance dispersals have been documented, the number of observed movements is small

93
(< 20 birds from Archbold Biological Station; John W. Fitzpatrick, pers. comm.), and the
movement and behavior of such jays remains essentially unobserved. To acquire some
empirical data on floaters, a small radiotelemetry study was conducted for this
dissertation.
GIS Files
The GIS files used in the simulations for this chapter and chapter 5 were created
by overlaying the scrub patches obtained during the 1991-1992 statewide Florida Scrub-
Jay survey (see chapter 2) onto a statewide habitat classification map produced by the
Florida Game and Freshwater Fish Commission (FGFWFC) in 1992 (Kautz et al. 1993).
Spatial resolution of the GIS file was 30 m. The original landcover types coded in the
FGFWFC classification are shown in Table 4-1.
Simulating Philopatric Dispersal
I modeled the behavior of helpers searching for breeding vacancies near their
natal territory using a small set of behavioral rules. Helpers engaged in philopatric
dispersal are assumed to have perfect knowledge of the status of each territory within
their assessment sphere. Helpers compete for vacancies; older helpers out compete
younger helpers, and closer helpers out compete more distant helpers. Dispersers that
find no vacancies during this search return to their natal territory and remain as helpers
until the following year.
Philopatric disperser survival rate is not directly specified in the model, but is
related to the floater frequency rate described below.

94
Short distance dispersal algorithm development and calibration
To develop and calibrate the model I relied heavily on dispersal and stage-age
data obtained from Archbold Biological Station (raw data provided by J. Fitzpatrick and
G. Woolfenden). Because dispersal curves are highly sensitive to the spatial
configuration of territories, comparisons between Archbold data and model output were
made by running simulations that approximated the dispersion of territories and habitat in
the vicinity of Archbold Biological Station. These simulations included approximately
200 territories and encompassed the southern third of Highlands county, and parts of
Desoto and Glades counties.
Development and calibration of the model proceeded by iteratively running a
simulation of the Archbold scenario, comparing the resulting dispersal and stage-age
graphs with the Archbold data (see Figs. 4-4 through 4-7), then modifying the model
structure or parameter values to improve the fit between the model and field data. Initial
modeling attempts that used very simple algorithms failed to match the Archbold data.
An early implementation used the following simple rule: allow randomly selected
helpers to occupy randomly selected territories within their assessment sphere. Dispersal
behavior was modified by adding rules that altered the manner in which dispersers
departed, moved, and settled in new locations. Rules were added to the model only if they
made biological sense and increased the realism of the model. I relied heavily on
Woolfenden and Fitzpatrick (1984) for behavioral information that could be incorporated
into the model. The final set of rules is provided in Table 4-2. The two sexes were treated
differently to account for observed, sex-based differences in dominance, natal
inheritance, and dispersal behavior.

95
V
Simulating Long Distance Dispersal
Searching behavior of long distance dispersers is modeled with several simple
rules that are hardcoded into the module. The initial movement direction upon leaving the
natal territory is random. As dispersers move through the landscape, they see territories
or habitat within a user-specified detection radius. They process the objects they see in a
specific order: first breeder vacancies, then empty territories, and finally, habitat. When a
breeder vacancy or empty territory is detected, the decision to settle is determined by the
dispersers propensity to settle (set by the user default is always to settle). If the
disperser doesnt settle, it moves on towards the most attractive habitat that has not been
visited already. Dispersers move in a straight line only within homogenous habitat, and
deviate from a straight line when they detect a difference in habitat attractiveness (see
section below). They avoid habitat with low attractiveness, and move towards habitat
with high attractiveness. Dispersers remember previously visited locations, which makes
it less likely that they will backtrack unless alternative directions are very unattractive.
Values for several floater parameters can be modified, including the proportion of
helpers that become floaters, detection radius (maximum distance objects can be
detected), mobility (maximum daily distance moved), and daily survival. For some
parameters (mobility, mortality, floater formation), estimates could be derived from
empirical data; indirect methods were used for directionality and estimation of detection
radius.

96
Estimating floater mortality and mobility
Long distance dispersers have two daily survival rates: one for floaters within
scrub, another for floaters outside of scrub. Within scrub, the daily survival rate is
assumed to be higher than outside of scrub, and similar to survival rates of nondispersing
jays of similar age and same sex. The survival rate for dispersers outside of scrub is
drawn from the best-guess Kaplan-Meier curve (Fig. 4-2) derived from the
displacement experiment described below. This curve was hard-coded into the model and
daily survival rates were drawn from the distribution and applied to floaters moving in
the matrix between scrub patches. An option to use a constant daily survival rate (as in
scrub habitat) was also included in the model and evaluated in the constraint analysis.
Each disperser moves until it exceeds a daily-distance-moved threshold value
selected for each jay from a function that approximates the observed distribution of daily
move distances. This distribution was derived from field data obtained from the
displacement experiment described below. The function that approximates the field
distribution was generated by the curve fitting procedure of SPSS (ver. 7.5). The
distribution of distances excluded 0 distances (i.e. days when jays did not move see later
discussion on displacement experiment). Once the daily-distance-moved threshold is
exceeded, each jays daily mortality rate is used to determine if the jay survives to the
next day. These steps are repeated until each jay dies, finds a mate or vacant territory, or
leaves the simulation area. The order in which dispersers move is randomized each time
all jays have taken a step. Jays that leave the area are considered dead (i.e. there is no
immigration from outside the simulation area). In contrast to short distant dispersers, long
distance dispersers do not return home.

97
Habitat attractiveness
Five attractiveness values are used in the model: 0 for a repulsive landcover that
jays do not enter, 1 for an unattractive landcover, 2 for a neutral landcover, 3 for a
somewhat attractive landcover, and 4 for a highly attractive landcover. The attractiveness
values assigned to different landcover classes in the GIS files (described in the previous
GIS section) are provided in table 4-1.
Floater detection radius
The floater detection radius is the maximum distance at which a disperser can be
expected to detect another jay or vacant territory. A starting point for estimating this
parameter is the distance between tape playback census points, as recommended by
Fitzpatrick et al. (1991b p. 13): Adequate spacing between transects can be estimated
roughly as the distance at which a person listening to the tape directly in front of the
speaker perceives the bird to be no more than about 100 meters away. A distance of
100 to 200 meters between transects and between stations is generally adequate when
using a good-quality, hand-held cassette player broadcasting at full volume. Jays no
doubt see each other, especially during territorial display flights, at greater distances than
they can hear each other. A value of 450 meters was selected as the default value for the
detection radius.

98
Estimating floater frequency
For most bird species, all surviving young depart from their natal territory as
floaters upon reaching sexual maturity, often during the first year of life. From a
modeling standpoint, the proportion of young that become floaters is simply the
proportion that survive to dispersal age. Upon reaching dispersal age, all young
disappear permanently from their natal territory and become floaters searching for
breeder vacancies or unoccupied habitat. For Florida Scrub-Jays and other species that
delay dispersal, the situation is more difficult to model, as some dispersal age young may
return to their natal territory after making unsuccessful searches for nearby breeder
vacancies, while other young may depart permanently from their natal territory.
The objective of this section is to estimate the annual proportion of helpers that
become floaters (Dfioater)- The dispersal model uses Dfloater to establish the proportion of
jays that become floaters from the pool of jays that disappear. The starting point for
estimating Dfioater is D,otai, the total proportion of helpers that disappear annually. All
floaters must come from this pool of disappearing jays, so Dfioater must be less than or
equal to Dtotai. D,otai is calculated from field data as the number of helpers disappearing
during a year divided by the original number present at the start of the year. Dt0tai for
helpers is shown in Table 4-2 (taken from appendix M of Woolfenden and Fitzpatrick
1984).
From the modeling standpoint, Dtotai has two components: jays that disappear by
dying locally (i.e. within their assessment sphere), and jays that disappear by becoming
floaters who move beyond their assessment sphere and either die or become breeders.
This can be represented as equation 4-1:

99
Dfloater Dtotal D|ocal (Eq. 4-1)
Thus, the proportion of helpers that disappear and become floaters (Dfloater) is
equal to the total proportion of helpers observed to disappear (Dtotai) corrected downward
by subtracting the proportion of helpers expected to die locally (Diocai) within their
assessment sphere. Because Dtotai varies among sex and age class, to facilitate setting and
comparing these parameters we can normalize the values relative to Dtotai and designate
the new parameters with the prefix P. Thus, Pfloater = Dfioater /Dtotai, and normalizing
equation 4-1 we get:
Pflomer Dtot^/DtotBl Dlocal/Dtotai 1 Plocal (Eq. 4-2)
where Pfloater is the proportion of total disappearances due to floaters permanently leaving
their natal territory, and Piocbi is the proportion of total disappearances due to local death
within the assessment sphere. Note that the floater disappearance rate (Pnoater) determines
how many helpers disperse beyond their assessment sphere, but does not determine how
many floaters survive to become breeders. The fate of floaters (i.e. whether they die or
become breeders) is contingent on the runtime situation and landscape encountered by
floaters during simulations as they search for breeder vacancies.
The sum of Pnoater and P|0cai must equal 1 (equation 4-2), so setting the value of
one parameter completely determines the other. I know of no way to directly estimate
Pfioter and Piocbi- However, it is possible to set reasonable bounds on these parameters by
considering various aspects of Florida Scrub-Jay biology.

100
For example, Waser et al. (1994) describe a flexible approach for estimating from
census data the proportion of unobserved emigrants that survive, based on several types
of data such as the known number of successful immigrants, and the known survival rates
of non-dispersing sex and age classes. They cite Woolfenden and Fitzpatrick (1984) as
the earliest example of such an approach. Woolfenden and Fitzpatrick (1984 appendix M)
estimated the number of dispersers expected to successfully emigrate off their study area
and subtracted these estimates fr<&n the known disappearances to calculate more accurate
helper mortality rates. Their approach assumes that immigration and emigration are at
equilibrium. Table 4-2 shows their equilibrium mortality rates (column labeled Deq)
next to the total disappearance rates (Dtotai). The difference between D,0tai and Deq
(Enoniocai in table 4-2) is the proportion of disappearing helpers expected to become
breeders off the study area. If we assume that this emigration rate (En0niocai) is the
proportion of disappearing helpers that successfully became breeders by dispersing as
floaters, we can use En0niocai as a lower bound for Pfioater- That is, the proportion of
disappearing helpers becoming ffbaters must be at least as big as the proportion of helpers
that successfully emigrate off the study area. Thus, En0nioca]/ Dtotai, which is the proportion
of disappearing jays estimated to became breeders off Archbold Biological Station,
establishes a lower bound for Pnoater and because the proportion of floaters that die is
likely to be very high, Pfloaier is likely to be much greater than En0niocai/ Dtotai.
We can develop an upper bound for Pnoater by using the fact that the sum of Pfioater
and Pioci must equal 1 (equation 4-2) and estimating P|0cai, the proportion of
disappearances due to local death rather than floating. We begin with the assumption that
jays engaged in local forays are likely to die at the same or higher rate than jays of similar

101
age or experience who are not making frequent forays away from their natal territory (i.e.
breeders or older fledglings). As an example, consider yearling helpers that disappeared
at Archbold Biological Station (Dtotai female = 0.42, male = 0.22; table 4-2). The local
death rate of yearling helpers is almost certainly greater than that of breeders (0.18;
Woolfenden and Fitzpatrick 1984, table 9.2, p. 265), but may be lower than the death rate
of older fledglings (0.30; Woolfenden and Fitzpatrick 1984, fig. 9.1, p. 255). If we
assume that P|0Cai is less than 0.30, suppose 0.25, then the maximum value for Pfioater
would be 0.75 (Pfioater must be less than (1 Pi0Cai); eqn. 4-2).
In the real world, the proportion of helpers that become floaters may vary greatly,
depending on factors such as habitat quality or number of neighboring territories. Some
of these factors are reviewed below in the discussion. If we consider the Archbold
setting, for yearling male helpers, it is likely that most disappearances can be attributed to
local mortality within their assessment sphere, while the proportion of females dying
within their assessment sphere may be less because more females disperse as floaters.
However, since the female assessment sphere is larger than the male, the proportion of
females dying within their assessment sphere may be substantial. Furthermore, 2nd year
males are estimated to have the largest proportion of disappearances due to emigration
(Table 4-2; column En0niocai/ Dtotai), suggesting that Pfioater also might be considerable for
older male helpers.
Since the disappearance rates (Dtotai) of female helpers are considerably greater
than males, setting Pfioater equivalent for both sexes would produce substantially more
female than male floaters. The values for Pfioater likely are larger for females than males,
which further increases this bias in the number of female floaters. Although it may seem

102
that producing larger numbers of female floaters may inflate female survival rates
because many may survive if they find vacancies outside the assessment sphere, recall
that disappearance rates (Dtotai) for females are substantially higher than males, so more
female floaters must survive to achieve the equilibrium rate (Deq in Table 4-2).
Based on the above considerations, upper and lower bounds for Pfioater are listed in
Table 4-2, along with a best guess.. The sensitivity of the model to different Pfioater
settings likely will vary considerably with the configuration of the landscape, and will be
investigated at a later time. A constraint analysis was performed to assess model
sensitivity for the Archbold setting as described in a section below.
Floater algorithm development and calibration
Whereas a wealth of biological data is available to aid in the simulation of short
distance dispersal, much less is known about long distance dispersal. A few long distance
dispersals by Archbold birds have been documented during comprehensive off-station
surveys for color-banded birds, but these established birds provided little information
about the process by which they moved and became established. Radiotelemetry offers
the best hope for documenting the movement and interactions of long distance dispersers,
but such studies face serious logistic difficulties (Koenig et al. 1996). Foremost among
these is the low likelihood of tagging a reasonable number of jays that then become long
distance dispersers. Most birds will disperse a short distance, yielding little or no
information useful for modeling long distance dispersal. By knowing the sex, age, and
dominance of birds within a study area, the likelihood of tagging long distance dispersers
might be increased, but even then most of the prime candidates for dispersal will move
only a small distance. A further complication for the present study is that transmitters

103
appropriate for Florida Scrub-Jays currently have a battery life of only a few months,
necessitating frequent recapture and re-outfitting. An alternative approach to waiting for
jays to disperse is to artificially displace radiotagged jays and follow their movement and
survival.
Jav displacement experiment
Displacement experiments have a long history within ornithology, having been
conducted in numerous studies of homing ability (Papi 1992). The focus of my
displacement experiment was not homing ability, but to induce birds to make long
distance movements that might mimic natural movements of jays moving between
isolated scrub patches in different landscape matrices. Surprisingly, the application of this
technique in conservation biology to date is extremely rare.
Jav selection protocol. Owing to large differences among sexes and stages in
documented dispersal distances, I selected the sex and stage having the best chance of
dispersing long distances, viz. females rather than males, and experienced helpers rather
than breeders or juveniles (one exception was a female breeder taken from Orange
County in a mitigation deal). Jays of known sex and stage were selected for capture
from the color-banded population from the experimental tract of Archbold Biological
Station (then maintained and monitored by Ron Mumme). A total of 10 jays were
captured, including one from Orange county.
Handling protocol. Jays captured in the morning were immediately weighed,
measured, and outfitted with a transmitter, then observed in a large flight cage for several
hours to allow the bird to become accustomed to the backpack harness. Jays captured in
the afternoon were held overnight and processed the following morning. Jays were

104
transported to their release sites in cages covered with heavy material to prevent the jays
from seeing the passing landscape and thereby potentially developing a homing direction.
All jays were released in the late afternoon.
Radio-transmitters were mounted with a technique already in use by another
researcher (Keith Tarvin) on Blue Jays at Archbold Biological Station. A 2-g transmitter
(manufactured by Wildlife Materials of Carbondale, IL ) was mounted on the back of
each jay and secured with elastic cord. The transmitters had small tubes on the front and
back that the cord was run through to create a loose, independent loop of equal size
around each wing. A small loop of cord was run between the wing loops and across the
belly to make the harness snug. The cord was kept loose enough to avoid cutting into the
skin or restrict movement or breathing, but tight enough to eliminate slack cord that feet
or branches might catch on. Newly outfitted jays were observed in an outdoor aviary for
at least half a day, and sometimes overnight. When initially rigged, jays were
preoccupied with trying to remove the outfit, suggesting that initially they might be
especially vulnerable to predators. Properly fitting harnesses were ignored after a few
hours and more normal behave resumed including taking peanuts and drinking water.
My failure to tie the knots on the harness tightly occasionally allowed jays to untie and
remove the transmitter, but only while in the aviary. I used forceps to cinch knots tightly
to reduce their size; knots were placed away from areas where they might rub against
pressure points. The elastic cord, obtainable from many fabric stores, degrades and falls
apart within a year or so; one jay that returned to Archbold wore her transmitter for a year
before it finally fell off. Her antennae also broke off after several months (reinforcing
each antennae base with epoxy might have been beneficial). Since most electronics fail

105
during an initial bum-in period, I activated each transmitter for a 24 hour test period prior
to use. Nevertheless, several transmitters appear to have failed after only 1-2 days in the
field.
Release sites. Radio-tagged jays were released at three different sites: a ranch
setting, a citrus grove, and a suburban area. The ranch setting was the MacArthur
Agroecology Research Center. The citrus grove setting was adjacent to and west of U.S.
27, north of S.R. 70, and south of the city of Lake Placid, Highlands county, Florida. The
suburban setting was the center of the city of Lake Placid.
Sampling protocol. I followed jays continuously on the first day of release until
they went to roost. Thereafter jays were visited on a schedule dictated by their activities.
Jays were visited at least three times daily regardless of activity: in the early morning, in
the early afternoon, and near dusk. During periods of active movement, jays were
followed intensively until large movements ceased. Visual contact could be maintained
for many long movements, since jays typically flew above visual obstructions such as
shrubbery. Locations of moving jays were plotted in the field on 1:24000 aerial photos,
using landmarks visible in the field and on the photos. When visual contact was lost,
triangulation was ineffective for plotting locations, since multiple bearings could not be
taken before a jay had moved large distances. Visual contact was often reestablished by
moving to a position ahead of the jays last known trajectory and awaiting its
reappearance. One jay moved onto large private landholdings that could not be accessed
by land. Airplane overflights were used once or twice daily (N=6 days) to establish its
movement at a coarse scale.

106
Transmitter effects. Subsequent to my conducting the displacement experiments,
Reed Bowman (pers. comm.) compared the behavior of jays outfitted with 2 types of
transmitters: backpack harnesses like those used in this study and leg mounted
transmitters. He observed substantial behavioral differences between jays with backpacks
and jays with leg-mounted transmitters. Jays with backpacks showed periods of
immobility lasting hours or even days punctuated by periods of more normal movement.
This behavior was quite different from birds with leg mounts, which acted normally
compared to several untagged jays that served as controls. None of Bowmans jays were
artificially displaced from their natal territory. I also observed long periods of inactivity
(Fig. 4-1), but because my jays were displaced into completely unfamiliar settings, I
could not evaluate whether their inactivity resulted from the transmitter or was a
behavioral response to being displaced. I suspect the backpack outfit did reduce the
activity of displaced jays, but I have no reason to think that the backpack impeded their
movements once the jays became active. Therefore, to estimate the mobility of jays I
excluded days with no jay movement.
Constraint Analysis
Dispersal events recorded (or known to be absent) in the vicinity of Archbold
Biological Station were used to place constraints on several long distance dispersal
parameters. A small but stable population of jays occurs some 20 miles (33.32 km) due
west of Archbold Biological Station on the Bright Hour Ranch, in DeSoto county. The
intervening habitat consists of palmetto flatwoods, small wetlands and seeps, citrus
groves, and improved and unimproved pasture. Several surveys of this locality during
the 1980s and early 1990s (by Fitzpatrick et al.) have revealed no dispersals from

107
Archbold Biological Station. Furthermore, the vocalizations of these jays are noticeably
different from Archbold Biological Station jays, suggesting that this population is
genuinely isolated from the Lake Wales Ridge population. Therefore, a reasonable
constraint to place on the dispersal module is that few or no floaters from Archbold
Biological Station (or contiguous Lake Wales Ridge populations) can disperse
successfully as far as the Bright Hour Ranch. A second constraint in the opposite
direction can be placed on the model. A small number of dispersals have been
documented that are considerably longer than the distance separating Archbold
Biological Station from Bright Hour Ranch. These include jays color-banded at Archbold
Biological Station and Avon Park Air Force Range, several of which dispersed between
35 and 50 km. The dispersal module should produce at least a few colonizations by
Archbold floaters to distant patches; such colonizations will show up in the dispersal
curve generated by the model.
The objective of the constraint analysis was to identify parameter settings that
prohibit dispersals from Lake Wales Ridge populations (around Archbold Biological
Station) to Bright Hour Ranch, while allowing longer dispersals to areas where color-
banded jays are known to have become breeders. Simulations were run that included the
territories in and around Archbold Biological Station and at the Bright Hour Ranch. The
number of dispersers that successfully located the Bright Hour Ranch were tabulated for
each run. Several floater parameters were varied, including the detection radius, the
maximum daily distance traveled, the daily mortality rate, and the rate of floater
formation (Pnoateri varied by sex and age). Two types of daily mortality rates were used: a
Kaplan-Meier distribution, and a fixed mortality rate.

108
Model Validation
Model validation was performed by comparing model output with dispersal and
stage-age data from long-term research at Archbold Biological Station. Kolmogorov-
Smimov tests were used to measure the goodness of fit between the simulated frequency
data and those of the Woolfenden and Fitzpatrick study. The dispersal field data for male
jays that dispersed by territory budding were lumped with jays that dispersed one
territory. This was necessary because the model does not simulate budding. I also provide
a qualitative assessment of other aspects of the model.
Results
Radiotelemetry Displacement Experiment
Of 10 female helpers released in non-scrub habitat, 4 were apparently depredated
(3 probably by accipters, 1 probably by domestic cat), 1 became ill after 2 days and was
recaptured, 3 successfully returned to their Archbold Biological Station home territories,
and 2 disappeared for unknown reasons. Premature transmitter failure is suspected for
the latter 2 disappearances (after 2 and 6 days). Premature transmitter failure was known
to occur in 2 of 3 jays that returned to Archbold Biological Station (failures after 2 and 13
days). The 4 apparently depredated jays survived 1, 3,4, and 22 days. Jays generally
made large movements over a period of one or two days, followed by one or more days
of relative inactivity. During periods of large movements (defined as greater than 0.33
km), jays moved an average of 5.2 km/day (range = 0.42 to 19.9 km).

109
Of three jays released at a citrus site, one was depredated by an accipter on day 2
of its release after moving 1.25 km, and two jays used citrus and a scrubby railroad
corridor to return to Archbold Biological Station. All jays released at the citrus site
behaved secretively, mostly made small movements, and stayed in citrus rather than
crossing US-27 (a divided 4-lane highway) to nearby sandhill habitat. They showed no
tendency to use the edges of groves, and readily flew through and roosted in interior
portions of groves. Five jays crossed expanses of citrus of 1.66 km or more.
Of three jays released in a ranch setting (improved pasture with cabbage palm
/oak hammocks), one was apparently depredated by an accipiter, and two disappeared
after 2 and 6 days, possibly due to transmitter failure. All ranch jays made large
movements from hammock to hammock, never were observed crossing expanses of open
pasture greater than 200 meters, and never ventured far into the interior of any large
hammocks. While moving or resting around large hammocks, they were always at or
near the hammock edges. One ranch-site jay found and settled in a nearby citrus grove
for 2 days before disappearing. The other 2 jays roosted at edges of oak hammocks, and
seemed preferentially to select smaller, isolated oak hammocks (data were insufficient for
statistical analysis).
One jay released at the ranch (R-MF) engaged in highly unexpected roosting
behavior. On the morning following her release, the transmitter signal was originating
from below ground within decaying palm frond sheaths at the base of a small, dead
cabbage palm. After waiting until 08:15 seeing no sign of activity, I fully expected that
the jay had been captured and eaten by a snake, and began excavating the palm and
removing ground debris to retrieve the transmitter. Suddenly, R-MF flew out into my

110
face from a point at least 30 cm underground. She appeared to behave normally
thereafter, and roosted in cabbage palms high off the ground on at least one other
occasion.
All observed movements of ranch jays were relatively short, from tree to tree. No
continuous, sustained flights greater than about 500 m were observed. Jays making
movements larger than 500 m tended to move in a hop-scotch manner from treetop to
treetop, stopping frequently along the way for short periods (often less than 1 minute) to
survey their surroundings. If they flew to a large tree, either they would fly to the top of
the tree or, if landing lower on the tree, would usually work their way up to the tree top
and look around before moving on.
Of four jays released in downtown Lake Placid, one successfully returned to
Archbold Biological Station via a large tract of scrub and pine (her transmitter failed after
day 1 and she was discovered back home months later during a census), 1 was apparently
injured by an accipiter and died one day later, another was apparently killed by a cat on
day 4, and a fourth became noticeably ill (was seen regurgitating food) and was
recaptured on day 2. All downtown jays tended to move in a hopscotch manner from
treetop to treetop. Two downtown-site jays found jays within 2 days about 4.15 km to
the southwest in the heavily developed Lake Placid suburb after crossing about 2.5 km of
downtown Lake Placid and 166 km of citrus.
Table 4-4 summarizes the movement data for each radiotagged jay. Figure 4-1
shows the daily movement and the number of days each jay was alive. Several jays show
day-long periods of inactivity followed by large movements. Three jays were right-

Ill
censored in the figure because they successfully returned to their home territory. The
remaining jays either perished or were lost due to transmitter failure.
A Kaplan-Meier survival curve (Fig. 4-2) shows the daily probability of survival
for the pooled sample of jays. Three survival curves were generated by making different
assumptions about unknown disappearances. The upper curve assumes that all unknown
disappearances were caused by transmitter failure, the bottom all due to mortality, and
the middle curve makes an educated-guess about each disappearance based on
circumstantial evidence. The 3 jays that successfully returned to their home territory were
right-censored. All three curves show an early rapid decrease in survivorship caused by
known mortality of several jays shortly after release. The constant daily survivorship out
to day 21 reflects a single jay released at the ranch which was depredated at day 22.
Floater Parameter Estimation
Floater mobility was parameterized by fitting a function generated by the curve
fitting procedure of SPSS (ver. 7.5) to the distribution of daily distances (Fig. 4-3) moved
obtained from radiotelemetry field data (excluding days with no movement see
explanation above). An inverse function of the form x = bl / (y bO) produced a good fit
with the observed data (r2 = 0.851; F = 91.43; d.f. = 16; p = 0.000). Values for the two
coefficients were: bO = -0.48; bl = 5.57.
Daily survival of floaters within scrub was assumed to be only slightly less than
survival of non-dispersing stages, as measured at Archbold Biological Station. The
default daily survival value was set at 0.9988, which was computed from the estimated
annual survival rate of 2nd year female helpers (0.66365). Daily survival of floaters

112
outside scrub was parameterized with the best-guess Kaplan-Meier curve (Fig. 4-2)
generated from the radiotelemetry field data.
Other floater parameter values were estimated indirectly, and are listed in table 4-
5. The selection of these values is described in the Methods section.
Constraint Analysis
The constraint analysis provided a valuable upper bound for several of the floater
dispersal parameters (table 4-5). An unrealistically high number of dispersals from the
Lake Wales Ridge to the Bright Hour Ranch occured for parameter settings that are not
substantially higher than the default settings (table 4-5).
Calibration and Validation
Comparisons between model output and dispersal data obtained from long term
observations of marked jays (unpublished data provided by J.W. Fitzpatrick and G.E.
Woolfenden) are shown for male and female dispersal in Figs. 4-4 and 4-5. The general
shapes of the dispersal curves produced by the model closely resembled the field data.
Model output did not differ significantly for females and was marginally different for
males compared to the field data (Kolmogorov-Smimov: male Z = 1.373, p = 0.046;
female Z= 0.784, p = 0. 570). Comparisons between model and field data for age-stage
structure of helpers and breeders are shown in Figs. 4-6 and 4-7 (obtained from tables 9.8
and 9.9 in Woolfenden and Fitzpatrick 1984). Model and field data comparisons are not
significantly different for helpers or breeders (Kolmogorov-Smimov: breeder Z = 0.447,
p = 0.988; helpers Z= 0.267, p = 1.00).

113
Discussion
Each of the behavioral rules (table 4-3) implemented in the software to generate
dispersal and stage-age curves similar to the Archbold field data has clear biological
meaning and is readily matched to features of known jay biology. Early versions of the
model, incorporating simpler rules, produced dispersal curves quite different from the
Archbold data. Interestingly, some of these curves were more similar to dispersal curves
for non-cooperative species. The final set of rules seems to capture key elements of
dispersal behavior that may partly account for differences between cooperative and non-
cooperative species. For example, after reviewing bird studies with adequate dispersal
data, Zack (1990) found that cooperative species dispersal distances were strongly
skewed towards the natal territory compared to non-cooperative breeding species.
Dispersal curves generated by early versions of my model resembled non-cooperative
species in that the mode was several territories away from the natal territory. This pattern
arose when the probability of settling on a vacancy was nearly the same for any territory
within an individuals assessment sphere. To produce the strongly right-skewed pattern of
a cooperative breeder, rules must be added to increase the probability of acquiring nearby
vacancies rather than those that are more distant within the assessment sphere.
Algorithmically, I simulated this behavior by keeping a distance-sorted list of nearby
territories within each dispersers assessment sphere. When a breeder death creates a
vacancy, the helper closest to that territory was given first choice at the opening (ties
were broken by random draw). This dominance hierarchy is further modified to account
for age by processing older helpers before younger. These rules are summarized in table
4-3.

114
The biological consequence of these philopatric dispersal rules is that they
increase the probability that dispersers engaging in delay-and-foray behavior will acquire
a nearby vacancy. A fascinating controversy relating to this feature exists in the literature
on the evolution of delayed dispersal. Experts on the Florida Scrub-Jay attribute the
origins of this behavior to ecological constraints: under normal conditions high quality
jay habitat is fully occupied (saturated). Survival and fecundity of dispersers outside of
high quality habitat is so low that it is better to remain at the natal territory and maximize
opportunities for acquiring nearby vacancies in the future. While acting as resident
helpers, jays can engage in behavior that increases their likelihood of acquiring a
territory. Such behavior includes engaging in frequent short forays to investigate potential
vacancies due to a breeder death or illness, enlisting the help of parents and siblings to
bud a territory, establishing dominance over nearby helpers (potential competitors),
developing furtive relationships with neighboring opposite sex breeders, and learning
territory boundaries, refugia from predators, and roost sites. Besides these direct
benefits to staying home, certain indirect benefits accrue by helping relatives
(Woolfenden and Fitzpatrick 1984). Zack (1990), Stacey and Lign (1991), Brown
(1989) and others present these factors as benefits of philopatry, and argue that they are
more important than ecological constraints. Krebs and Davies (1998, p. 304) suggest
that the two views are not so much alternative hypotheses as different sides of the same
equation. They suggest that the pay-offs depend on the quality of breeding vacancies
available and how much competition exists.
Carmen (1989) showed that Western Scrub-Jays do not engage in delay-and-foray
dispersal behavior. In some cooperative breeders dispersers show considerable

115
behavioral flexibility. For example, in the Seychelles warbler (Acrocephalus
sechellensis), dispersers from high quality territories delay dispersal when high quality
territories are unavailable, while dispersers from low quality territories, where survival is
low, do not delay dispersal but instead settle for a low quality territory (Komdeur 1992).
Florida Scrub-Jays do exhibit some flexibility in dispersal, as suggested by the behavior
of suburban jays. For example, Bowman (1994), Thaxton and Hingten (1995), and
Breininger (1999) have documented traits in suburban jays that have much in common
with the floating behavior of the Western species, including reduced delays in dispersal
and much greater dispersal distances. Within good habitat floating behavior in Florida
Scrub-Jays is likely to be rare, but fire suppression may lower habitat quality sufficiently
to encourage floating and excessive numbers of helpers at a territory might encourage
subordinates to disperse early (so-called saturation dispersal). The possibility that
Florida Scrub-Jays may switch between these two dispersal strategies needs further
investigation.
The telemetry displacement study provided a number of useful findings that were
incorporated into the floater module. Jays clearly avoided open habitats lacking tree or
shrub cover, such as water bodies and pasture. In contrast, they moved freely through a
variety of landscapes with at least some tree or shrub cover, including citrus groves,
sandhill, pasture with small, scattered oak hammocks, and suburbs with trees. This
information was incorporated into the model by assigning different attractiveness values
to various landcover classes based on the amount cover each class is likely to provide.
Table 4-3 shows the original landcover classes encoded in the Florida Game and

116
Freshwater Fish Commissions statewide classification, and each classs respective
attractiveness value as assigned by me following the telemetry study.
The animation graphics displayed during a simulation run allows the movement
of long distance dispersers to be visually checked. Dispersers behaved as expected,
completely avoiding landcover types with highly repulsive values, moving in an
essentially straight line in habitat with neutral attractiveness, and preferentially turning
towards habitat with high attractiveness values. Simulated floaters could be observed
moving long distances, and the dispersal curves generated by the program showed that
they occasionally settled after moving long distances.
Quantitative comparison of movement patterns generated by the model with those
recorded for radiotagged jays was not feasible because the sample size, resolution of data
recorded, and movements of several jays all were too small for meaningful statistical
tests. Simple circular statistics could be used with a better data set to test jay movements
for directional bias (Batschelet 1981; Zar 1996). Turchin (1998) outlines several more
sophisticated analyses, but besides requiring better data sets, Turchins analyses are
designed to test for correlated random walk, which is more typical of insects than the
landscape-directed movements of vertebrates. Statistical tests appropriate for this model
probably do not exist, and would likely involve boot-strap tests based on simulations
developed for specific landscape settings.
Although Florida Scrub-Jays are fairly weak fliers (Woolfenden and Fitzpatrick
1996), jays that were experimentally displaced were capable of moving fairly long
distances, up to a maximum of 20 km per day, with an average daily movement of about
2.67 km. This information was used to parameterize the daily movement parameter in the

117
floater algorithm, and the constraint analysis suggested that actual movements should not
be much larger than the measured distribution.
The constraint analysis showed that simulated jays originating from Archbold or
nearby patches rarely colonized the Bright Hour Ranch for certain parameter settings,
even though the dispersal curve showed that some Archbold dispersers did succeed in
settling much greater distances within the Lake Wales Ridge than the distance to Bright
Hour. This bias in dispersal success is largely due to the floater mortality rates set in the
model for different habitat types; jays that moved the longest distances were those that
managed to stay in scrub habitat the longest.
The displacement experiment suggested that jays suffered much higher mortality
rates while moving through non-scrub landscapes than they would likely experience in
scrub habitat. Predators took several jays, and it is possible that several other jays that
vanished inexplicably were also predated. These losses suggest that dispersal outside
scrub habitat is a costly activity for Florida Scrub-Jays, which is consistent with other
findings and a suite of behavioral traits described earlier, including the sentinel system,
delayed dispersal, and unusually short dispersal distances. The Kaplan-Meier curve
developed from the radiotelemetry data provided strong evidence for the vulnerability of
jays to predation in unfamiliar, non-scrub habitat. The Kaplan-Meier curve also provided
a useful means of parameterizing the daily survival of floaters (fig. 4-2), and
demonstrated graphically the potentially high cost of long distance dispersal through
unfamiliar terrain. The mortality rates measured for displaced jays are extremely high,
and likely are higher than would be experienced by naturally dispersing, untagged jays
due to negative effects of the backpack harness and radio transmitter. Nevertheless, the

118
constraint analysis suggested that floater mortality rates outside of scrub must be high,
otherwise unrealistically high colonization rates to Bright Hour Ranch occurred. The
constraint analysis also indicated that a fixed daily survival rate of 0.75 or higher for
floaters outside of scrub produced excessive colonizations, assuming the default
movement function. In combination, the field derived Kaplan-Meier curve (Fig. 4-2) and
movement function (Fig. 4-3) produced satisfactory results in the constraint analysis.
The statistical comparison of model output with long-term field data (fig. 4-3 to 4-
6) gave mixed results, showing no significant differences for the stage-age data, but
marginal differences for the dispersal data. Even though the dispersal curves bordered on
being statistically different, the similarity of the simulated and field data is obvious,
showing the same mode and the appropriate differences between male and female
dispersers. A consistent bias is apparent in the simulated data, which shows too few
dispersers at the mode and too many dispersers just after the mode for both sexes.
Incorporating budding into the model might correct these biases.
Although statistical attempts to validate the model were equivocal, other means of
validation are available. In a comprehensive review of different ways of validating
ecological models Rykiel (1996) argues that qualitative assessments such as face
validation (the approval of an experienced field person) often are adequate. Lima and
Zolner (1996) suggest that qualitative assessments may have to suffice for evaluating
models of complex interactions between animal behavior and landscape ecology, at least
until the hard-to-acquire data sets become available and techniques such as those
proposed by Turchin (1998) are developed. Qualitatively, the dispersal module compares
favorably with other models in terms of complexity. A literature search revealed only one

119
other model (Letcher et al. 1998) that simulates both philopatric and floater dispersal. No
models were found that generated dispersal and stage-age distributions to compare with
field data. The level of detail and biological realism in this model are high due to the
availability of long term research data and crude telemetry data. The constraint analysis
proved valuable in placing bounds on acceptable parameter settings. Nonetheless,
numerous simplifying assumptions were made, as is true of any model. A full discussion
of the models assumptions is given in chapter 5, which describes the complete model.
The evaluation of the dispersal module presented in this chapter exemplifies a useful
technique of separately validating individual model components (Thomas et al. 1990).
Nonetheless, it is important to investigate the sensitivity of the overall model to potential
errors in the dispersal module and other model components. Such comparisons may
reveal that the model is crucially or inconsequentially affected by the dispersal module.

120
Table 4-1. Landcover types from statewide habitat map (Kautz et al. 1993) used in
simulations and associated floater attractiveness values.
Landcover Type
Attractiveness Value0
Coastal Strand 3
Dry Prairies 1
Pinelands 2
Sand Pine Scrub 2
Sandhill 2
Xeric Oak Scrub 4
Mixed Hardwood Pine Forest 1
Hardwood Hammocks and Forests 1
Tropical Hardwood Hammock 1
Coastal Salt Marshes 1
Freshwater Marsh and Wet Prairie 2
Cypress Swamp 1
Hardwood Swamp 1
Bottomland Hardwoods 1
Bay Swamp 1
Shrub Swamp 2
Mangrove Swamp 2
Aquatic 0
Grassland 1
Shrub and Brushland 2
Exotic Plant Communities 2
Barren 0
a Attractiveness values range from 0 to 4 (0 = repulsive; 1 = unattractive; 2 = neutral; 3
= attractive; 4 = highly attractive).

Table 4-2. Demographic parameters used to estimate the proportion of disappearing
helpers that become floaters.
121
Age Class
Dtotai
D b
i^eq
Enoniocai
Enoniocai/
Dtotai
P floater
Male
Yearling
0.225
0.22
0.005
0.022
0.022-0.75
(0.25)
Female
Yearling
0.42
0.35
0.07
0.167
0.167-0.75
(0.75)
Male 2nd
Year
Helper
0.25
0.15
0.10
0.40
0.40 0.75
(0.50)
Female 2nd
0.375
0.26
0.115
0.307
0.307-0.75
Year (0.75)
Helper
a Dtotai is the observed proportion of total disappearances (from Woolfenden and
Fitzpatrick 1984; Table 9.5; p. 275 ).
b Deq is the equilibrium mortality rate (from Woolfenden and Fitzpatrick 1984; Table 9.5;
p. 275 ).
c Enoniocai is the proportion of total disappearances due to dispersers becoming breeders
elsewhere as unobserved emigrants (from Woolfenden and Fitzpatrick 1984;
Appendix M) calculated as Dtotai Deq.
d Enoniocai/Dtotai is the proportion of Dtotai estimated to become breeders elsewhere as
emigrants. This value sets a lower limit for Pfioater-
e Pfioater is the proportion of Dtotai becoming floaters. A plausible range of values is listed
first, followed by a best guess in parentheses.

122
Table 4-3. Summary of philopatric dispersal rules showing sex differences and rules used
to implement the algorithm.
Male
Female
Algorithmic inmlementation
Smaller assessment
sphere
Larger assessment
sphere
Only move to vacancies within
the specified assessment sphere.
Natal Inheritance
No natal inheritance
If both breeders die, resident
male helper has priority over
nonresident male helpers.
Extreme competitive
advantage near home
Nominal competitive
advantage near home
Sort territories within
assessment sphere by distance.
Allow each female to select
nearest vacancy within
assessment sphere. For males,
divide sphere into 3 nested
subspheres. Process each
subsphere separately, innermost
to outermost. Allow each male
to select nearest vacancy within
each subsphere.
Prefer unpaired breeders
over vacant territories
Prefer unpaired breeders
over vacant territories
Complete search for unpaired
breeders for all dispersers
before searching for vacant
territories.
Older helpers dominate
yearling helpers
Older helpers dominate
yearling helpers
During search for unpaired
breeders or vacant territories,
process all older helpers before
yearling helpers.

Table 4-4. Summary of jay movement data obtained from displacement experiment (distances in km).
Jay
color
band id
Territory
of
origin a
Release
c
site
Max. daily
movement4*
Active mean
daily
movement
Inactive mean
daily movement
Cumulative
distance
traveled
Start-
end
dist8
Days
h. i
moving
Days
alive*k
G-MR
MIDR
1
1.25
1.25
1.25
1.25
1.0
1
1
GZ-O
EERR
1
8.33
2.97
1.22
12.2
6.67
4 [6]
[10]
{ABS}
YB-GM
Disney*5
2
15-20
5.33-5.83
2.33-2.67
53.3-58.3
3.33
10
22
R-MO
BYRD
2
(4.17)
(4.17)
(2.17)
(4.17)
(4.17)
(1)
(1)
R-MF
TP68
2
(4.17)
(2.0)
(1.03)
(6.17)
(1.66)
(3)
(6)
R-MA
CWHP
1
(1.67)
(0.77)
(0.30)
(3.83)
19.2
(13)
{ABS}
LZ-A
DEAD
3
4.17
4.17
1.38
4.17
4.17
1
3
Y-MW
MARE
3
2.58
2.42
1.22
4.83
1.00
2
4
Y-CR
DRUM
3
0.42
0.38
0.38
0.75
0.75
2
2
Z-YY
TRVN
3
(4.67)
(2.38)
(2.38)
(21.7)
16.6
(1)
{ABS}
Mean
(4.67-5.17)
(2.60 2.63)
(1.37- 1.40)
(11.2- 11.7)
5.83
3.3
3.9
a Territory of origin from the south (experimental) tract of Archbold Biological Station.
b Jay obtained in mitigation deal from Disney property in Orange Co., Florida.
c Release sites (Highlands county, Florida): 1 = Large citrus grove with nearby sandhill on W. side of Grassy Lake; 2 =
MacArthur Agroecology Research Center improved pasture and numerous cabbage palm/oak hammocks; 3 =
Downtown Lake Placid.
if Numbers in parentheses indicate distances or days measured for birds lost for unknown reasons.
e Days when birds were not moving were excluded in distance calculations.
/All days were included in distance calculation, including days with no movement.
g Straight line distance between release site and last known location.
h Number of days when bird moved more than 0.33 km.
/ Numbers in brackets include days when jays were in scrub habitat.
j Number of days jays were known to be alive. See d and /' for explanation of numbers in parentheses or brackets.
k {ABS} indicates jay successfully returned to its territory of origin at Archbold Biological Station.

124
Table 4-5. Summary of constraint analysis for 9 simulation scenarios (50 years x 30
repetitions) showing number of colonizations from Lake Wales Ridge to Bright Hour
Ranch, DeSoto county, Florida.
Search Radius
Survival
Max. Distance
Pfloater
Colonizations'
10
Kapl
400
0.25/0.75
2
10
Kapl
500
0.15/0.55
4
10
Kapl
500
0.25/0.75
3
10
Kapl
600
0.25/0.75
13
20
Kapl
500
0.25/0.75
7
10
0.50
500
0.25/0.75
2
10
0.70
500
0.25/0.75
22
10
0.90
500
0.25/0.75
65
10
1.00
500
0.25/0.75
225
a Distance in 30 m pixels dispersing jay sees.
b Daily survival rate: Kapl. = Kaplan-Meier data (fig. 4-2); Numbers = daily survival
c Maximum daily distance traveled by floater (units for inverse function see text).
d Probability of disappearing helpers becoming floater (male/female).
e Number of dispersers finding Bright Hour Ranch from Lake Wales Ridge

Distance Moved (km)
125
Fig. 4-1. Daily distances moved and number of days movements were tracked for 10 jays
released at 3 sites in Highlands county, Florida.
Ranch-3

126
Days Alive
Fig. 4-2. Kaplan-meier survival curves of daily survival rates for 10 jays released at 3
sites in Highlands county, Florida. Upper curve: maximum possible survival; middle
curve: best guess survival; lower curve: minimum possible survival.

127
Observed
Inverse
Fig. 4-3. Distribution of daily distances moved by released jays (solid line), and inverse
function fitted to observed movements (dashed line).

128
H field
H model
01 23456789 10 11
Territories Traversed
Fig. 4-4. Comparison of dispersal data from Archbold Biological Station and simulated
dispersal distances for male jays.

129
Territories Traversed
field
El model
Fig. 4-5. Comparison of dispersal data from Archbold Biological Station and simulated
dispersal distances for female jays.

130
Fig. 4-6. Comparison of stage-age data from Archbold Biological Station and simulated
stage-age data for breeders.

131
Fig. 4-7. Comparison of stage-age data from Archbold Biological Station and simulated
stage-age data for helpers.

CHAPTER 5
METAPOPULATION VIABILITY ANALYSIS OF THE FLORIDA SCRUB-JAY
Introduction and Objectives
The Florida Scrub-Jay is a Federally threatened bird species that occurs only in
Florida. Formerly found in 39 of 40 counties of peninsular Florida, by 1981 the Florida
Scrub-Jay was known to be extirpated from 7 counties (Broward, Dade, Duval, Gilchrist,
Hendry, Pinellas, and St. Johns; Cox 1987). A 1992-1993 statewide mapping project
(SMP) added Alachua and Clay to the list of extirpations, and estimated that less than 10
breeding pairs remained in 6 other counties (Flagler, Hardee, Hernando, Levy, Orange,
and Putnam; Fitzpatrick et al. 1994). In nearly all other counties jay populations have
declined drastically. Since 1993, a severe decline to near extinction was documented on
the barrier island south of Patrick Air Force Base in Brevard county (Breininger 1999),
and a huge decline exceeding 50% was documented on mainland S. Brevard county
during the same period (Breininger 1998). Undoubtedly, similar drastic declines are
occurring throughout the state. Humans are directly responsible for this dramatic
population reduction, primarily through suppression of natural fires which are necessary
to maintain high rates of reproduction and survival in scrub-jay populations, and through
outright destruction or degradation of jay habitat due to roads, housing developments, and
agriculture (e.g. citrus groves).
132

133
Although these negative population trends are severe, opportunities still exist to
acquire, restore, and manage occupied habitat in various portions of the scrub-jays
rapidly dwindling range. The 1992-1993 SMP, analyzed in chapter 2, showed that the
Florida Scrub-Jay has a highly variable spatial distribution around the state, resulting in
complex metapopulation structure associated with the wide range of local population
sizes and varying degrees of isolation. This variability in spatial structure undoubtedly
has strong effects on the ability of jays to persist within different landscapes. Faced with
this complexity, the question arises: which populations of scrub-jays are adequately
protected, and what additional land acquisitions are needed in areas where jays are not
adequately protected?
The answer to this question will depend not only on the influence of the spatial
distribution of jays and their habitat, but also the demographic success of the jays within
a given landscape context. Empirical research and population modeling has shown that
the demographic performance of jays varies considerably due to habitat features such as
the degree of vegetative overgrowth (Fitzpatrick and Woolfenden 1986; Woolfenden and
Fitzpatrick 1991; Breininger et al. 1995, 1996) and the extent of human disturbances such
as road mortality (Mumme et al. in press). Indeed, fire suppression alone has been shown
to guarantee extinction, even in large populations (Root 1998; Breininger et al., in press).
Thus, habitat acquisition alone will not ensure survival of jays; proper habitat
management is also critically important. Ascertaining which jay populations have the
greatest need for land acquisition is, therefore, complicated by the fact that many of the
currently protected areas have not been properly managed, and could support more jays.
Research and management experience has shown that local declines in scrub-jay

134
populations can be reversed in areas that are restored to optimal habitat conditions and
properly managed (J. Thaxton, pers. comm.; D. Breininger, pers. comm.). The critical
importance of optimal habitat to jays is becoming widely recognized, and initiatives to
manage protected areas for scrub-jays are rapidly gaining momentum. If these protected
areas were restored and jay populations increased to higher densities, how viable would
jay populations be with and without further land acquisition? Given the complex
distribution of scrub-jays, which areas are most in need of further acquisition, and which
habitat patches should be acquired?
These are the primary questions addressed in this chapter. The classification
technique developed and applied for metapopulations around the state (chapter 2), while
useful for describing the spatial structure, cannot provide viability estimates for the
various metapopulations. This chapter describes the application of a spatially explicit,
individual based population model (SEIBPM) to estimate the viability of Florida Scrub-
Jay metapopulations. SEIBPMs take into account site specific details of habitat
distribution and habitat quality, and provide a standardized, scientific framework for
answering questions relating to the viability of organisms in different reserve
configurations. SEIBPMs are a very new form of computer technology and are not
without their problems (Beissinger et al. 1998). Nevertheless, these models can produce
quantitative information that is obtainable in no other way, and when used properly can
provide invaluable information that supplements and informs more conventional
approaches to land acquisition and land management.

135
Objectives: The primary objectives of this chapter are:
Compare the relative vulnerability of different jay metapopulations around the state.
Use measures of vulnerability, such as quasi-extinction risk and percent population
decline, to rank the vulnerability of jays in different parts of their range.
Compare the relative vulnerability of different configurations of jay populations
within the same metapopulation. Model different spatial configurations of jay
metapopulations, ranging from a no additional acquisition configuration to a
maximum acquisition option.
Identify unprotected areas that would significantly increase the expected persistence
of each metapopulation if they were acquired, restored, and properly managed.
Methods
Simulation Model Description
All simulations were performed with a spatially explicit, individual-based
population model written by B. Stith specifically for the Florida Scrub-Jay. The model
was developed using Microsoft Visual C++ (ver. 5.0) for use on Intel-compatible
personal computers running Microsoft operating systems (Windows 95, 98, and NT 4.0).
Simulations of jay population dynamics take place on realistic landscapes provided by
geographic information system (GIS) files. The model incorporates general aspects
common to many population models, including demographic and environmental
stochasticity, as well as a host of details specific to scrub-jay biology.

136
Because of striking differences in the biology of male and female jays, both sexes
are modeled. Except when dispersing, simulated jays reside within discrete territories.
Individual jays of both sexes progress through 5 life stages (juvenile, 1-year helper, older
helper, novice breeder, and experienced breeder). Breeder experience and presence of
helpers affects breeder success. Helpers monitor neighboring territories and vie for
breeder openings; the outcome of such competition is determined by simple dominance
rules. Helpers may leave on long distance dispersals, during which time mortality and
movement varies depending on the type of landcover being traversed. During the
simulation, graphs of dispersal distances, stage-age structure, population trajectory, and
quasi-extinction probabilities are displayed, and text descriptions of various events
occurring during the simulation can be viewed and saved to an output file.
Life Stages
Each territory maintains a list of all individuals of both sexes in the following 5 stages:
Juvenile
1-Year Helper
Experienced Helper
Novice Breeder
Experienced Breeder
Helpers can transition into a sixth stage by departing from their natal territory and
becoming a long-distance disperser (floater).

137
Starting Population Stage Structure
At the start of each repetition of each simulation run, all territories are initialized
with a pair of inexperienced breeders (both 2 years old) and one inexperienced helper (1
year old; randomly selected sex). The location of each territory is obtained from an
ASCII file provided to the model (see GIS section below).
Annual Life Cycle
The scrub-jay annual life cycle is simulated by a series of events scheduled in an
event queue; each event is completed for the entire metapopulation before the next event
begins. The following is a summary of the major events in the annual cycle, which begins
with reproduction.
Reproduction. Each territory produces a poisson distribution of juveniles (see
Burgman et al. 1993) with 3 different means (see Table 5-1) for: 1) at least one
experienced breeder with at least one helper, 2) at least one experienced breeder but no
helpers, or 3) both novice breeders. The fecundity parameter values are set to the number
of one-year old offspring produced, rather than fledglings produced. From a software
efficiency standpoint, this greatly reduces the number of jays that must be created and
then destroyed in the mortality step that immediately follows (i.e. the model does not
subject juveniles to mortality during their first year the fecundity rate accounts for this
mortality). Demographic stochasticity of fecundity is implemented by randomly
selecting the sex of offspring. Environmental stochasticity of fecundity is not
implemented, but would be expected to have a negative effect on population persistence.

138
Field studies have shown that in the Florida Scrub-Jay, environmental stochasticity of
fecundity and mortality are positively correlated (Woolfenden and Fitzpatrick 1984).
Annual mortality. Breeder and helper annual mortality rates are based on the
current territory configuration (see Table 5-1). All breeder vacancies available to
dispersers are created in this step. Helpers do not actually die in this step, but are
considered to have disappeared. The floater frequency parameter later determines
whether they die or become floaters (see dispersal section below). Juveniles are not
subjected to mortality in this step since their annual mortality is already reflected in the
fecundity rate. Epidemics occur with an annual probability of 0.05, and increase the
mortality rate of juveniles by 100% and adult mortality by 20%. These percentages are
conservative; the actual mortality rates may be considerably higher, although the long
term values are unknown. Fitzpatrick and Woolfenden (1991) reduced adult survival to
0.55 and juvenile survival to 0.0 in their population model.
Promotion to next stage. Survivors of the mortality step are promoted to the
appropriate experienced stage: novice breeders to experienced breeders; 1-year helpers
to experienced helpers.
Dispersal. Two types of dispersal are modeled: philopatric dispersal forays
around the natal territory within an assessment sphere, and floater dispersal long
distance search in which a disperser permanently leaves its area of intimate knowledge
and moves through the landscape searching for breeder vacancies or empty territories.
All helpers that survive the mortality step engage in philopatric dispersal. The
order in which philopatric dispersal events occur mirrors the dominance hierarchy of
jays: males dominate females, older jays dominates younger, jays closer to their natal

139
territory dominate more distant jays. The first dispersal event allows male helpers to
inherit their natal territory if both breeders have died. Male helpers then search their
assessment sphere for unpaired females, and if successful they become novice breeders.
Older males search before younger males to simulate dominance relations, and jays
closer to their natal territory out compete more distant jays. Male helpers that fail to pair
up then search for empty territories within their assessment sphere. Female helpers then
search for unpaired males, and if unsuccessful search for empty territories, all within their
assessment sphere. Mortality associated with these short distance forays is assumed to be
already factored into the mortality rates for each stage.
Floater dispersal commences after philopatric dispersal; this ordering assumes
that philopatric dispersers dominate floaters. Helpers of both sexes who were marked-to-
disappear in the mortality step either die or leave their natal territory as floaters,
depending on whether a random uniform deviate drawn for each helper exceeds the
setting for the floater frequency parameter (see Table 5-1).
Searching behavior of floaters is modeled with several simple rules. Upon leaving
the territory, the initial movement direction is random. As dispersers move through the
landscape, they see territories or habitat within a user-specified detection radius. They
process the objects they see in a specific order, first looking for breeder vacancies, then
empty territories. If a breeder vacancy or empty territory is detected, the decision to settle
is determined by the dispersers propensity to settle. If the disperser doesnt settle, it
moves on towards the most attractive habitat that has not been visited already. In
completely homogeneous habitat dispersers move in a straight line, but in non-
homogeneous habitat their movement direction is affected by differences in habitat

140
attractiveness. They move away from habitat with low attractiveness, and towards habitat
with high attractiveness. Dispersers remember previously visited locations, which makes
it less likely that they will backtrack unless alternative directions are very unattractive.
Long distance dispersers have two mortality rates: one for dispersers within scrub,
another for dispersers outside of scrub. Within scrub, the disperser survival rate is higher
than outside of scrub (Table 5-1). Each disperser moves until it exceeds a random daily-
distance-moved threshold selected for each jay from a distribution of daily move
distances. Once the latter distance is exceeded, a daily mortality rate is used to determine
if the jay survives to the next day. These steps are repeated for each jay until it dies, finds
a mate or leaves the simulation area. Jays that leave the area are considered dead (i.e.
there is no immigration from outside the simulation area). In contrast to short distant
dispersers, long distance dispersers do not return home.
After all floaters have settled, died, or left the simulation area, the annual cycle is
repeated with the reproduction step. If all jays are extinct or the last year of the last
repetition is reached, the simulation terminates.
Territories
The model tracks individual territories, maintains a list of jays occupying each
territory, and graphically displays the occupancy status of each territory. In the
simulations completed for this project, all territories were assumed to be 9 hectares and
shaped as squares. Territory locations were read in from an ASCII file exported from
Arcview. Position, number, and habitat quality of territories did not change over time. All

141
territories were given parameter values for either high quality habitat or suburbs (see
parameter settings in Table 5-1).
Background Landscape Image
Bit-mapped GIS files provided the landscape setting upon which the population
dynamics and dispersal movements were simulated. These files were created by
overlaying the scrub patches in the 1992-1993 SMP database onto a statewide habitat
classification map produced by the Florida Game and Freshwater Fish Commission
(FGFWFC) based on 1985-1989 Landsat Thematic Mapper data (Kautz et al. 1993). All
GIS files had a spatial resolution of 30 m. The original landcover types coded in the
FGFWFC classification are shown in Table 4-2 (chapter 4), along with the associated
attractiveness values that affected the movement of floaters. In the simulations
completed for this chapter, the landscape was assumed to be static through time.

142
Table 5-1. Demographic and dispersal parameter settings for jays in optimal and
suburban conditions.
Optimal conditions
Suburban conditions
Parameter
Female
Male
Female
Male
Survival
1st Year Helper
0.580a
0.740a
0.480a
0.480a
Older Helper
0.625a
0.740a
0.480a
0.480a
Novice Breeder
0.740
0.740
0.480
0.480
Experienced Breeder without
0.770
0.770
0.770
0.770
Helper(s)
Experienced Breeder with
0.850
0.850
0.850
0.850
Helper(s)
Fedundity
Novice Breeder
0.50b
0.50b
Experienced Breeder without
Helper(s)
0.57b
0.57b
Experienced Breeder with
Helper(s)
0.77b
0.77b
Delay-and-foray Dispersal
Assessment sphere (radius no.
territories)
4
7
4
7
Floater Dispersal
1st Year Helper proportion
0.85
0.65
0.85
0.65
disappearing jays becoming
floaters
Older helper proportion
disappearing jays becoming
floaters
0.85
0.65
0.85
0.65
Detection radius (meters)
450
450
450
450
Daily Survival in scrub
0.9988
0.9988
0.9988
0.9988
Daily Survival in non-scrub
K-Mc
K-Mc
K-Mc
K-Mc
Daily movement distance
lnversed
Inverse11
Inverse0
Inverse1
a Includes disappearances (see chapter 4 for further explanation).
b Production of new 1st year helpers, not fledglings.
c Daily survival rates obtained from Kaplan-Meier curve (derived from radiotelemetry
data see chapter 4 for further explanation).
d Daily movement distances obtained from inverse function (derived from radiotelemetry
data see chapter 4 for further explanation).

143
Map Production
A statewide metapopulation map was produced to depict the 21 metapopulations
that were analyzed for this chapter (Fig. 5-1). For each of the 21 metapopulations, two
types of detailed maps were produced to depict the status of jays in 1992-1993 as
determined by the SMP, and to depict what jay populations might look like if all habitat
were restored and fully occupied by jays. The maps showing restored populations of jays
provided the basis for the simulations, as explained below.
Statewide metapopulation map
Metapopulations were delineated initially using a dispersal buffer approach
discussed in chapter 2. A GIS was used to generate a 12 km buffer around all jay
territories to enclose populations that are likely to be connected by dispersal. In certain
areas the buffers joined populations that probably are not connected due to physical
barriers to movement, such as a large river systems or cities. These physical barriers are
identified in the written accounts for each metapopulation (see Metapopulation Viability
Analysis section). Figure 5 -1 shows a map of the 21 metapopulations identified for the
entire state. Each of these 21 metapopulations were modeled as demographically
independent units as described below.
1992-1993 SMP maps
The status of jays and habitat as determined by the 1992-1993 SMP were
portrayed for each metapopulation on one or more maps. The jay data included some

144
miscellaneous sightings added since 1993. The maps show habitat polygons shaded with
coarse hatching representing different levels of human disturbance. Lightly shaded
polygons delineate protected areas or areas proposed for protection/acquisition as of the
end of 1997. These data were assembled from several agencies, and are now somewhat
outdated. Protected areas containing scrub jays are labeled in italics. Thin, circular lines
represent dispersal buffers connecting jay territories within 3 km of each other. A road
network, county boundaries, and major cities also are depicted.
Acquisition maps
Acquisition maps show the jay territory locations used as input to the simulations.
These maps show the estimated jay populations after habitat restoration and full
occupancy. Three types of jays are delineated on the acquisition maps. Jays in currently
protected habitat are enclosed in bold, solid polygons. The boundaries of the protected
areas were estimated from 1997 information and some updated sources (see
Identification of Protected Areas section below). Currently protected jays were
included in all simulations. Jays in currently unprotected potentially suitable habitat are
enclosed in bold, dashed polygons. These are the jays that may be included or excluded
in different simulations, reflecting various reserve design options. Jays that are not
enclosed in either type of polygon are considered to be suburban jays, and are included
in all simulations. The polygons delineating groups of jay territories are labeled with
alphanumeric identifiers consisting of a county-name prefix followed by an integer
number (e.g. Brevl2). These map labels are referred to in the text and tables to identify
groups of jays. A road network, county boundaries, and major cities also are depicted.

145
30 0 30 60 90 120 Kilometers
Fig. 5-0. Delineations of 21 Florida Scrub-Jay metapopulations based on 1992 1993
statewide survey.

146
GIS Database Preparation
Geographic information systems (Arc/Info, Arcview, and Imagine) were used to
produce raster images of landscapes used in the simulations, to provide jay locations and
territory quality information to the simulation model, and to produce maps described
above.
Estimation of jay populations after restoration
The starting point for estimating jay population size for use in the simulations was
the 1992-1993 SMP database produced at Archbold Biological Station in 1993
(Fitzpatrick et al. 1994), and subsequently updated by Bill Pranty, and now maintained by
Archbold Biological Station (Reed Bowman). Additional jay locations were added to a
copy of this database to reflect population sizes expected after habitat restoration and full
occupancy by jays.
The number of jays expected to occupy patches after being fully restored was
estimated by jay experts familiar with localities in the following counties: Jon Thaxton -
S. Manatee, Sarasota, Charlotte, N. Lee; Dave Breininger and Brian Toland Brevard,
Indian River, St. Lucie, Martin; Grace Iverson Palm Beach; Reed Bowman and Brad
Stith Highlands and Polk; Bill Pranty Pasco, Hernando, Citrus, Lake, Marion. These
experts were asked to examine maps showing the habitat patches and jay distributions
from the 1992-1993 SMP, and to give a subjective estimate of the number of jays an area
could support, based on their knowledge of the local conditions. In the remaining
counties, an estimate was made by B. Stith based on habitat attributes in the SMP
database, patch size, and supplemental habitat information where available (e.g. Pranty et

147
al. manuscript). Most of these estimates differed little from the densities determined by
the SMP.
Identification of protected areas
Boundaries were digitized around occupied, protected jay habitat using Arcview.
These boundaries appear on the acquisition maps for each metapopulation as heavy solid
lines. Protected areas were identified through the use of the 1998 F.N.A.I. publication
(Blanchard et al. 1998), annual C.A.R.L. reports (especially Anonymous 1999), Arc/Info
coverages obtained from water management districts, and from verbal updates provided
by individuals familiar with specific sites. The source data for the F.N.A.I. publication
and the Arc/Info coverages date from late 1997. A significant number of acquisitions
have been made since that time, some of which may not be identified as protected in this
document. In a few cases, local experts suggested that a particular tract of land be treated
as protected even though it was in private hands or the property had been only partially
purchased. In these cases, it is possible that areas designated as protected may be less
than shown on the maps. The ongoing process of land acquisition ensures that any map
will be obsolete as soon as it is published, and such errors will affect the outcome of
some simulations.
Assessment of unprotected areas
Unprotected, occupied patches of jay habitat delineated by the 1992-1993 SMP
were grouped into two categories: patches having sufficient potential to be considered for
acquisition, and patches with little or no acquisition value due to excessive human

148
disturbance (especially suburbanization). The determination of excessive degradation was
made by local experts for the following counties: Jon Thaxton S. Manatee, Sarasota,
Charlotte, N. Lee; Dave Breininger and Brian Toland Brevard, Indian River, St. Lucie,
Martin; Grace Iverson Palm Beach; Reed Bowman and Brad Stith Highlands and
Polk; Bill Pranty Pasco, Hernando, Citrus, Lake, Marion. In the remaining counties, a
determination was made by B. Stith based on habitat attributes in the SMP database and
the density of road networks as portrayed in the Florida Atlas and Gazetteer (1997).
Potential jay habitat found to be unoccupied by the 1992-1993 SMP generally was
not included in any of the modeling scenarios unless it was nearly adjacent to already
occupied habitat. Manatee county was an exception to this rule due to the fact that the
majority of patches in this area could not be surveyed during the SMP, and many patches
are likely to be occupied. Note that patches found to be occupied in the early 1980s by
Cox (1987) that were subsequently found to be unoccupied by the SMP were treated as
unoccupied patches. Whether jays were added to an unoccupied patch can be ascertained
by comparing the 1992-1993 SMP maps with the corresponding acquisition maps.
Suburban iavs
Jays living in suburban conditions are unlikely to persist in the long term
(Breininger 1999; Bowman et al. 1993). Nevertheless, suburban jays may play an
important role in the short term by providing colonists to restored or well-managed
habitat (Thaxton and Hingtgen 1996; Breininger 1999). To account for this potentially
important role in the simulations, suburban jays were included in model runs for areas
where they were originally present in the SMP. Suburban jays were assigned

149
demographic parameters corresponding to those measured by Breininger (1999) on
Satellite Beach, Brevard County (table 5-1). Also, suburban jays were given different
dispersal behavior compared to jays living in optimal habitat, based on the findings of
Thaxton and Hingtgen (1996). Simulated jays dispersing from optimal habitat could not
settle in suburban areas. Simulated jays from suburban areas could settle in optimal
habitat, and could settle with unpaired jays in suburbs, but once a suburban territory
became unoccupied, that territory could not be recolonized.
Simulation Runs
Each simulation run required two input files: a territory location/attribute file, and
a background landscape file. The input files for all simulation runs are ASCII files
exported from the Arcview database that was used to produce the acquisition maps for
each metapopulation.
Repetitions and duration of simulations
All simulations were run for 30 repetitions. Statistics generated from 30
repetitions were found to stabilize and adequately represent much lengthier simulations.
All simulations were run for a duration of 60 years. No standards currently exist for
choosing the duration of simulations, but recent research shows a tendency towards
shorter simulation times (Beissinger et al. 1998). The choice is arbitrary, and simulations
run for different lengths of time commonly produce very different absolute outcomes;
ultimately, all populations go extinct. However, relative outcomes are assumed to be the
same; the ranking of relative risk of extinction faced by different populations remains
unaffected by the length of the simulation (Beissinger et al. 1998).

150
Reserve design configurations
Two reserve design configurations played a key role in the analysis and
comparison of scrub-jay metapopulations around the state. These two configurations,
called the no acquisition and maximal acquisition options, represent the smallest and
largest possible reserve designs. Both configurations were simulated for all
metapopulations (except the Ocala National Forest).
The no acquisition option assumes that no more land will be acquired beyond
what is currently protected. Jays outside protected areas are treated as suburban jays with
the corresponding demographic characteristics (see table 5-1). The maximum
acquisition option assumes that all relatively undisturbed habitat with jays will be
acquired. For both configurations, all protected lands are assumed to be restored and
properly managed for scrub-jays, and jays are assumed to have the appropriate densities
and demographic performance for high quality habitat (see table 5-1).
Other configurations were simulated for metapopulations showing substantial
*
ir
differences between the no acquisition and maximal acquisition options. Typically,
the very small and very large metapopulations did not warrant evaluating additional
configurations because the results would be nearly the same as the two extreme
configurations. For metapopulations where intermediate configurations might be
substantially different, fixed percentages of jays were added to the already protected
populations. The percentages (30% or 70%) were applied to the difference between the
maximum and no acquisition option to determine the total size of the reserve.
In metapopulations with substantial spatial variability in unprotected jay
distributions, alternative reserve designs were simulated that emphasized maximizing

151
territory contiguity (i.e. favoring larger areas), or maximizing connectivity (i.e.
preserving small stepping stone populations).
Output statistics
Two main types of output statistics were generated by the model: quasi-extinction
probabilities, and population trajectory statistics. A quasi-extinction curve was generated
for each model run to show the cumulative probability (on the y-axis) that the population
fell below a range of population sizes (on the x-axis) at any time during the simulation.
Two statistics were extracted from each quasi-extinction curve: the probability of total
extinction, and the probability of falling below 10 pairs (referred to throughout this
document as the quasi-extinction probability). Prior research has identified a
population size of 10 pairs as an important threshold for assessing vulnerability. These
two statistics (extinction and quasi-extinction probability) were tabulated for all
simulations, and provide useful information for evaluating populations that are highly
vulnerable to extinction. Note that these two statistics do not provide any information
about the viability of populations that never fall below 10 pairs.
To evaluate larger populations, the trajectory statistics are likely to be more useful
than quasi-extinction statistics, as large populations may decline rapidly yet produce
quasi-extinction statistics that indicate no risk. A population trajectory curve is generated
for each model run which shows the mean population size (on the y-axis) for a given time
period (on the x-axis). Several statistics were extracted from each trajectory curve:
starting population size, mean ending population size, standard deviation, and percent
population decline. Percent population decline is calculated by subtracting the mean

152
ending population size from the starting population size, dividing this difference by the
starting population size, and multiplying this result by 100.
Model Validation/Calibration
Efforts to validate this model using long-term data from Archbold Biological
Station (ABS) (Woolfenden and Fitzpatrick 1984) are described in chapter 4. The
demographic parameters measured at ABS were used to parameterize jays in optimal
habitat (table 5-1). A constraint analysis and a small radiotelemetry study (chapter 4)
were used to develop parameters for the dispersal algorithm.
Interpreting Simulation Results
Two key assumptions have a large influence on the simulation results reported in
this chapter. First, the assumption has been made that the density of jays in all occupied
habitat is the maximum expected if the habitat were fully restored. The second
assumption is that the demographic performance of jays is maximal, corresponding to
measurements made in optimal habitat in the long-term study at Archbold Biological
Station (Woolfenden and Fitzpatrick 1984). Both of these assumptions are likely to be
very optimistic for most metapopulations around the state.
Many habitat patches, including those in public ownership, are not currently
managed properly for scrub-jays. In the absence of aggressive management, jay
demographic success decreases; small changes of 10% can produce dramatic declines in
population size and rapid extinction (Fitzpatrick and Woolfenden 1986; Breininger 1998;
Root 1998). The simulations results presented in this chapter assume continuous, optimal
habitat conditions. Unfortunately, even proper management of habitat may not guarantee

153
demographic success, especially for patches that are juxtaposed against landscapes that
subject jays to detrimental edge effects. These edge effects include human factors such as
road mortality (Mumme et al., in press), and predation by domestic cats, and natural
predators that occur at artificially high densities (e.g. racoons, grackles; Breininger 1999).
More subtle edge effects may occur in natural landscapes where jay habitat is located
next to forests or other habitats favorable to jay predators and competitors (Breininger et
al. 1998).
It is likely that these negative factors, which depress jay densities and
demographic performance, are the norm throughout much of the state. Fire management
programs are being developed and implemented on many public lands, but as
development continues in Florida many jays on these properties will have reduced
reproductive success simply because they are surrounded by human landscapes. A recent
study in well-managed jay habitat at Archbold Biological Station by Mumme et al. (in
press) documented substantial negative effects of road mortality in a rural setting. The
situation in suburban and urban settings is likely to be even worse (Breininger 1999;
Bowman 1993). Because the simulation model does not consider such edge effects, the
model results should be viewed as optimistic.
Given these caveats, the recommended use of the estimates reported below is to
compare the relative viability of jay metapopulations around the state as a guide for land
acquisition and to rank areas in terms of vulnerability. Probability and trajectory
estimates produced by the model should not be taken literally; they are best used for
comparative purposes.

154
Results
These results summarize the output of different simulations performed for each of
the 21 metapopulations. Results for each metapopulation are reported in separate
sections. Each section begins with a general description, lists the protected areas,
discusses restoration potential, summarizes the simulation results, and provides
recommendations. Maps are provided showing the distribution of jays and habitat during
the 1992-1993 SMP, and the distribution of jays in relation to the simulations and
acquisition possibilities. The acquisition maps depict what jay populations might look
like if all habitat were restored and fully occupied by jays. Tables are provided that
summarize the patch statistics (number of jays in each patch) for different reserve
configurations. The results from all simulations are presented in tables, and quasi
extinction and trajectory graphs are provided for at least 2 reserve design configurations.
The simulation results also include the statewide rankings developed and explained in the
last section (see Recommendations).

155
Levy (Cedar Key) (Mil
General description: The Levy county metapopulation is the most northerly
population of jays occurring along the Gulf Coast, and is highly isolated from other
metapopulations. The SMP delineated a single large scrub patch in this area, and found 8
groups of jays, 4 of which occurred in the Cedar Key Scrub State Reserve (see Fig. 5-la;
Table 5-la). The SMP found the condition of the scrub to be severely overgrown, and
noted that the number of jays present was only one-third the number found by Cox in
1980 (Pranty et al. manuscript). A 1997-1998 study of jays at Cedar Key (T. Webber in
F.D.E.P. 1998) found only 1 pair on the reserve, 2 pairs in a nearby junkyard, and 4
groups in the town of Rosewood 7-8 miles to the east. Estimated potential population size
after habitat restoration and full occupancy is 17 pairs in currently protected areas, and 75
pairs maximum.
Protected areas: The only protected jays in this metapopulation occur in the Cedar
Key Scrub State Reserve (Levyl).
Restoration potential: Some restoration has taken place at the Cedar Key Scrub
State Reserve, but many areas remain heavily overgrown. A recent report (F.D.E.P. 1998)
noted that a shortage of staff and difficulties associated with burning sand pine forests
have delayed restoration needed at this reserve. For modeling purposes, the currently
protected area is estimated to support about 17 jay families after restoration and full
occupancy (Fig. 5-lb; Table 5-la). A large, contiguous patch of unprotected habitat
(Levy2) was mapped by the SMP, but much of this habitat is not suitable for jays.
Restorable patches of scrub and scrubby flatwoods occur within a complex matrix of

156
marsh and mesic flatwoods with high densities of pine. The estimated population sizes
after restoration (Table 5-la) may be overly optimistic.
Simulation results: This metapopulation ranked 5th in vulnerability (table 5-23),
17th in percent protected (22.7%; table 5-24), and 2nd in priority (table 5-25), with high
vulnerability and high potential for improvement. Simulations of the SMP configuration
indicate that the 1992-1993 configuration is extremely vulnerable to extinction (Table 5-
lb; extinction risk = 1.0; percent population decline = 100.0).
Simulations of the currently protected, restored configuration indicate that the
protected population would be vulnerable to extinction (Table 5-lb; extinction risk = 0.1;
Fig. 5-Id; percent population decline = 29.4%).
Because the habitat as mapped shows no fragmentation, no configurations
involving multiple patches were simulated. All acquisition configurations assume that
preference is given to acquiring contiguous habitat, but even if multiple patches were
created, the resulting interpatch distances would be small.
The 30% acquisition configuration was estimated to support about 34 jay families
(Table 5-la). Simulations of this configuration indicate that it has a small but significant
probability of falling below 10 families (Table 5-lb; quasi-extinction = 0.05). The mean
population trajectory shows an 8.8% decline.
The 70% acquisition configuration was estimated to support about 54 jay families.
Simulations of this configuration indicate that the population would not be vulnerable to
extinction or quasi-extinction (Table 5-lb and Fig. 5-Id; extinction risk = 0.0; quasi
extinction = 0.0). The mean population trajectory shows a 3.7% decline (Fig. 5-lc).

157
The maximum acquisition configuration was estimated to support about 75 jay
families. Simulations of this configuration indicate that the population would not be
vulnerable to extinction or quasi-extinction (Table 5-lb; Fig. 5-Id; extinction risk = 0.0;
quasi-extinction = 0.0; percent decline = 1.3).
Recommendations: Jays in this metapopulation are in a very precarious state, and
their #2 priority rating perhaps should be upgraded to #1. Only one pair of jays is known
to occur within the park, and only two other pairs are nearby (T. Webber in F.D.E.P.
1998) and these jays are subject to predation by cats and mortality along increasingly
busy roads. A much greater level of support is needed to bolster current restoration
efforts. The simulation results suggest that even after full restoration, additional land
purchases beyond 30% are needed to reduce quasi-extinction risk to a low level. The
opportunity may still exist to acquire substantial pieces of unprotected habitat adjacent to
or near the Cedar Key Scrub State Reserve, but new housing developments are rapidly
destroying habitat along SR 24 and CR 347, and acquisition opportunities may soon be
foreclosed. Major efforts will be needed to restore any habitat that is acquired; removal of
extensive pine overstory will be needed in many areas. This metapopulation is in danger
of blinking out, and needs immediate attention.

158
Jay Territory Locations ^ 1.75 km dispersal buffer
Scrub Polygons
Lo Disturbance Protection Status
3 L Density Housing Protected
|WM Hi Density Housing Dr rmA
I Ranch/Ag
H Proposed
Water Bodies
/\/ Interstates
*\/ State highway
County roads
County lines
Ml Cedar Key
0 5 10 15 Kilometers
1 : 250,000 ^
Fig. 5-la. Levy county map 1992 1993 jay and habitat distribution.

159
jay Territory Locations
(after restoration)
Scrub Polygons
CCC## Polygon ID
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Protection Status
/\/ Protected
/\y Unprotected
Water Bodies
' County lines
Ml Cedar Key
0 5
10
15 Kilometers
1 : 250,000
Fig. 5-lb. Levy county acquisition map.

Table 5-la. Levy county patch statistics (number of jay territories for different configurations)
Patch id
Status
1992-1993#
jay territories
No
acquisition
(restored)
30%
preserved
by contiguity
70%
preserved
by contiguity
Maximum
acquisition
Levyl
Cedar Key
4
17
17
17
17
Scrub Reserve
Levy2
4
17
37
58
Totals
8
17
34
54
75

161
0)
N
W
c
c
TO
D
Q.
C
CL
Fj+H^ffTfU-f-j-l fH 11 -[ | ~f I 11 H~1 H H l-t-j-j
20
40
Year
60
Fig. 5-lc. Levy county trajectory graphs. Top) no acquisition, Bottom) maximum
acquisition.

Probability
162
5 10 15 20
Threshold Pop Size
20 40 60
Threshold Pop Size
Fig. 5-Id. Levy county quasi-extinction graphs. Top) no acquisition, Bottom) maximum
acquisition.

Table 5-lb. Levy county (Cedar Key) simulation statistics
Patch Name
Data type
Original
1992-1993
scenario
No
acquisition
30% acquisition
by area
70% acquisition
by area
Maximum
acquisition
Cedar Key
Scrub Reserve
starting
population size
4
17
34
54
75
x end pop. size
0
12
31
52
74
s.d.
0.8
5
6
5
5
percent
decline
100.0
29.4
8.8
3.7
1.3
extinction
risk
1.0
0.1
0
(8)
0
(26)
0
(45)
quasi-extinction
risk (10 pairs)

0.82
0.05
0
0

164
Citrus-S.W. Marion (M2)
General description: The Citrus and S.W. Marion county metapopulation consists
of small, scattered groups of jays living within about ten miles of the Gulf Coast, and
larger populations of jays concentrated mostly in the Big Scrub area of southwest Marion
and northwest Sumter counties. Most jays occur in small, somewhat isolated clusters; the
largest population is an unprotected group of 19 pairs at Mar2 (fig. 5-2b). Extensive
mosaics of scrub, scrubby flatwoods, sand pine, and sandhill occur throughout this
region. The habitat as mapped by the SMP makes no distinction among habitat types and
should be treated as very incomplete. Because of the severely overgrown habitat
conditions, many jays occur in marginal habitat, and the small, isolated populations along
the Gulf Coast are especially vulnerable to blinking out. The connection between the Gulf
Coast and Big Scrub area may be very poor; the 12 km dispersal buffer (fig. 5-1) shows
the tenuous connection occurring at the Twisted Oak golf course (Citr5 in fig. 5-2b).
The SMP documented about 108 jay territories, excluding suburban jays, in this
metapopulation. Estimated potential population size after habitat restoration and full
occupancy is 47 pairs in currently protected areas, and 145 pairs maximum.
Protected areas: The only protected jays along the coastal portion of this
metapopulation occur at the Crystal River State Buffer Preserve (Citrl). Protected jays
inland occur at Potts Preserve (Citr6), Cross Florida Greenway (Marl, Mar5,
Mar6), and Half Moon Wildlife Management Area (Sumtl). CARL sites with jays
include Mar8.
Restoration potential: Nearly all of the habitat in this metapopulation is heavily
overgrown and occurs as a complex mosaic of scrub, scrubby flatwoods, sandhill and

165
sandpine. Most of the larger habitat polygons mapped by the SMP include large amounts
of non-habitat, and the data are inadequate for making good estimates of restoration
potential. For modeling purposes, the only habitat polygon that was given significantly
more jays after restoration is the Crystal River State Buffer Preserve. Jays in the large
polygons in southwest Marion were kept at the densities determined by the SMP, which
may significantly underestimate the restoration potential.
Simulation results: This metapopulation ranked 13th in vulnerability (table 5-23),
13th in percent protected (37.6%; table 5-24), and 8th in priority (table 5-25), with high
vulnerability and moderate potential for improvement. No simulations were run for the
SMP configuration due to its similarity to the 70% connectivity configuration (see
results in Table 5-2b).
The fully restored, currently protected configuration was estimated to support 47
jay pairs. Simulations of this no acquisition configuration indicate that the
metapopulation would be vulnerable to extinction and quasi-extinction (Table 5-2b;
extinction risk = 0.17; quasi-extinction risk=0.47; % decline = 78%).
All intermediate simulations, except the 70% preserved by area configuration,
show a non-zero extinction probability and significant quasi-extinction risk (Table 5-2b).
Simulations of the latter configuration and the maximum acquisition configuration both
show a small quasi-extinction risk (Table 5-2b). The percent population decline is fairly
large for both configurations, reflecting the vulnerability of relatively small, somewhat
isolated subpopulations.
Recommendations: Improved habitat mapping is needed to better estimate the
restoration potential of the existing habitat. Additional surveys may be warranted

166
throughout this region, since jays often occur in atypical, unsurveyed habitat, as
evidenced by the recent discovery of jays at Marion 1, Ross Prairie and in Citr3 during
surveys for the proposed Suncoast Parkway II.
The restoration potential in this region is probably considerably higher than
reported here due to poor habitat dataconsiderably more habitat exists than is shown on
these maps. Because no large, contiguous populations of jays occur in this
metapopulation, even the maximum acquisition option is vulnerable to quasi-extinction
and shows a large mean percent population decline. These simulation results suggest that
as much habitat should be acquired and restored as possible, with an emphasis on
creating larger contiguous populations. Extensive areas of overgrown upland habitat
exists throughout both counties, and acquisition/restoration of parcels of unoccupied
habitat may be needed to attract jays in suboptimal habitat to restored habitat, especially
in Citrus county. A large sandpine forest occurs in N. central Citrus county with some
logged areas that may now be suitable for at least 5-25 jay territories (B.Stith pers. obs. -
see intersection of Citr3 with this habitat polygon in fig. 5-2b). The feasibility of
acquiring and restoring portions of this large forest should be investigated. Jays occurring
along the powerline corridor of the Crystal River nuclear power plant (Citr3) might
recolonize this patch.
Recent additions to the Crystal River State Buffer Preserve (Citrl in fig. 2-2b)
have increased the amount of protected scrub to nearly 400 acres (pers. comm. Randy
Martin), making this the most significant habitat patch along the Gulf Coast portion of
this metapopulation. Rapid restoration of this patch is crucial, as there may only be one or
two jay families persisting at this site (B.Stith pers. obs.). Citr5 (Twisted Oak golf

167
course) may be a potentially important stepping stone, and a local reserve at this site
should be considered.
In the Big Scrub area of S.W. Marion, large populations of unprotected jays
occurs near Gum Slough (Mar2) on the Rocking F Ranch, and just north of the Cross
Florida Greenway on the west side of 1-75 in the vicinity of the Marion Oaks D.R.I.
(Mar8). Acquisition and restoration of these patches is especially important to the
overall persistence of this jay metapopulation.

168
Scrub Polygons
^ Lo Disturbance
2¡ Lo Density Housing
Hi Density Housing
¡ggj Ranch/Ag
Protection Status
Protected
Proposed
Water Bodies
A/ Interstates
State highways
County roads
County lines
M2 Citrus
1 250,000
Kilometers
Fig. 5-2a. Citrus county map 1992-1993 jay and habitat distribution.

169
]ay Territory Locations
Scrub Polygons
^ Lo Disturbance
Lo Density Housing
tV/J Hi Density Housing
gggj Ranch/Ag
1.75 km dispersal buffer
Protection Status
i j Protected
~ Proposed
Water Bodies
Interstates
/\/ State highways
County roads
' County lines
M2 SW Mahon
0 5 10
1 : 250,000
15 Kilometers
Fig. 5-2b. S.W. Marion county map 1992-1993 jay and habitat distribution.

170
]ay Territory Locations
(after restoration) Protection Status
Scrub Polygons /\/ Protected
CCC## Polygon ID V* UnProtected
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Water Bodies
County lines
M2 Citrus
10
15
1 : 250,000
Kilometers
4-
Fig. 5-2c. Citrus county acquisition map.

171
Jay Territory Locations
(after restoration)
~~ Scrub Polygons
CCC## Polygon ID
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Protection Status
/\y Protected
/\/ Unprotected
Water Bodies
County lines
M2 SW Marion
10
15
1 : 250,000
Kilometers
+
Fig. 5-2d. S.W. Marion county acquisition map.

Table 5-2a. Citrus and S. Marion county patch statistics (number of jay territories for different configurations)
Patch id
Status
1992-1993#
No
30%
30%
70%
70%
Maximum
jay territories
acquisition
preserved
preserved
preserved
preserved
acquisition
(restored)
by contiguity
by connectivity
by contiguity
by connectivity
Citrl
Crystal River
St. But. Pr.
6
12
12
12
12
12
12
Citr 2
1
2
1
2
2
2
Citr 3
3
6
7
7
Citr 4
4
2
4
4
4
Citr 5
4
2
2
4
Citr 6
Potts Preserve
1
6
6
6
6
6
6
Citr 7
3
3
3
Marl
Cross FI.
Greenway
1
1
1
1
1
1
Mar2
19
10
5
19
12
19
Mar3
6
1
4
4
6
Mar4
3
3
3
3
3
Mar5
Cross FI.
Greenway
8
8
8
8
8
8
8
Mar6
Cross FI.
Greenway
3
3
3
3
3
3
3
Mar7
14
3
6
11
14
Mar8
14
7
2
14
10
14
Mar9
2
2
2
2
Sumtl
Half Moon
W.M.A.
17
17
17
17
17
17
17
Totals
108
47
69
69
101
101
125

Probability
173
10 20 30 40
Threshold Pop. Size
1 oT
r
-
08^
o
a!
0 4^
i i
50 100
Threshold Pop Size
Fig. 5-2e. Citrus and S. Marion county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.

Population Size
174
h
i
20 40 60
Year
20 40 60
Year
Fig. 5-2f. Citrus and S. Marion county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition

Table 5-2b. Citrus and S. Marion county simulation statistics
Data type
No
acquisition
30%
preserved by
connectivity
30%
preserved by
area
70%
preserved by
connectivity
70%
preserved by
area
Maximum
acquisition
starting
population
size
47
69
69
101
101
125
x end pop.
13.8
15.0
24.4
32.7
48.3
56.2
size
s.d.
8.5
11.4
16.1
20.1
20.1
17.7
percent
decline
70.6
78.3
65.2
67.6
52.2
55.0
extinction
risk
0.17
0.10
0.03
0.03
0.05
0.04
quasi
extinction
risk (10 pairs)
0.47
0.57
0.43
0.17
0.27
0.0

176
Pasco-Hemando (M31
General description: The Pasco-Hemando metapopulation (M3) consists of
several poorly connected subpopulations that occur near the coast, in the interior of
Pasco, and on the western edge of the Green Swamp. The SMP documented about 29 jay
territories, excluding suburban jays, in this metapopulation. Estimated potential
population size after habitat restoration and full occupancy is 63 pairs in currently
protected areas, and 69 pairs maximum.
Protected areas: This metapopulation ranked very high (3rd) in percentage of jays
protected. However, many unprotected areas are poorly surveyed (see below). Protected
jays have been found at: Weeki Wachee (Herl), Starkey & Serenova Wellfield
(Pasl), Cross-Bar Wellfield (Pas3)\ Alston Tract (Pas5), Green Swamp W.M.A.
(Pas6).
Restoration potential: The restoration potential of this metapopulation is large
(compare first and last data columns in Table 5-3a), especially at Weeki-Watchee
(Herl) and the Cross-bar/Al-bar Wellfields area (Pas3, Pas4).
Simulation results: This metapopulation ranked 14th in vulnerability (table 5-23),
3rd in percent protected (91.3%; table 5-24), and 11th in priority (table 5-25), with high
vulnerability and moderate potential for improvement.
Simulations of the SMP configuration, with 20 jay families, indicated that this
configuration is extremely vulnerable to extinction (Table 5-3a; extinction risk = 1.0).
The no acquisition configuration was estimated to support about 63 jay families
after restoration and full occupancy. Simulations of the currently protected, restored
configuration showed a small quasi-extinction risk (p=0.03) and a substantial quasi-

177
extinction risk (p=0.30; Table 5-3b). The maximum acquisition configuration, estimated
to support only 6 additional jay pairs (total of 69), had no extinction risk and a reduced
quasi-extinction risk (p=0.233).
Recommendations: The Pasco-Hemando metapopulation should be considered
poorly surveyed; fire suppression has forced jays to occupy atypical, unsurveyed habitat,
as evidenced by the recent discovery of jays at Pas2 (Pranty et al., manuscript). This
small population connects the 12 km dispersal buffer between Cross-bar and Serenova,
and may be an important acquisition. In the absence of new survey data, further
acquisition options appear very limited.
The potential for restoration of protected habitat in this metapopulation is large.
The Weeki Watchee State Park (Herl in Fig. 5-3b) is a large sand pine forest with a
dense oak understory that has had a single resident jay family residing in a small bum for
many years. This forest has the potential to support 17 or more pairs of jays, but the one
resident family may have recently disappeared (Pranty et al., manuscript). Portions of
this forest should be restored to scrub as soon as possible. The largest population of jays
in this metapopulation, and perhaps the 2nd largest jay population along the Gulf Coast,
occurs on the Cross-bar/Al-bar Wellfields (Pas3), which is currently being restored by
Pasco county (B. Pranty, pers. comm.). Habitat restoration is urgently needed at the
Starkey and Serenova properties (Pasl), as jays are nearly extirpated at this site.

178
Scrub Polygons
Lo Disturbance
\ Lo Density Housing
jgg Hi Density Housing
I R.anch/Ag
Protection Status
Protected
[j Proposed
Water Bodies
Interstates
State highways
County roads
County lines
M3 W. Pasco
1 : 250,000
Kilometers
Fig. 5-3a. W. Pasco and Hernando county map 1992 1993 jay and habitat distribution.

179
Scrub Polygons
1 1 Lo Disturbance
2] Lo Density Housing
iMfij Hi Density Housing
Ranch/Ag
Protection Status
HI Protected
Proposed
Water Bodies
/\f Interstates
State highways
County roads
' ' County lines
M3 E. Pasco
0 5 10 15
1 : 250,000
Kilom
-+
Fig. 5-3b. E. Pasco and Hernando county map 1992 1993 jay and habitat distribution.

180
Water Bodies
' County lines
W. Pasco
Jay Territory Locations
(after restoration)
Scrub Polygons
CCC## Polygon ID
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Protection Status
/\/ Protected
/\/ Unprotected
10
15 Kilometers
1 : 250,000
Fig. 5-3c. W. Pasco and Hernando county acquisition map.

181
Water Bodies
County lines
E. Pasco
jay Territory Locations
(after restoration) Protection Status
Scrub Polygons /\/ Protected
CCC Polygon ID A/ Unprotected
Jays outside of labeled, bold polygons are considered to be Suburban jays.
10
15
1 : 250,000
Kllometei
Fig. 5-3d. E. Pasco and Hernando county acquisition map.

182
Table 5-3a. Pasco and Hernando county patch statistics (number of jay territories for different
configurations)
Patch id
Status
1992-1993#
No
Maximum
jay territories
acquisition
(restored)
acquisition
Her1
Weeki Wachee
1
17
17
Pas1
Starkey Wellfield &
Serenova
3
13
13
Pas2
4
Pas3
Cross-Bar/AI-bar Wellfield
15
19
19
Pas4
2
2
Pas5
Alston Tract
2
2
2
Pas6
Green Swamp W.M.A.
6
12
12
Totals
29
63
69

Population Size
183
U
V)
L
20
Year
40
60
|-
i i 1 1 1 i 1 i ( i ¡ i |_
20 40 60
Year
Fig. 5-3e. Pasco and Hernando county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.

184
20 40 60 80
Threshold Pop. Size
20 40 60 80
Threshold Pop. Size
Fig. 5-3f. Pasco and Hernando county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition.

185
Table 5-3b. Pasco county simulation statistics
Data type
1992-1993
No
acquisition
Maximum
acquisition
starting
population size
29
63
69
x end pop. size
6.2
20.2
22.0
s.d.
4.5
10.5
percent
decline
78.6
67.9
68.2
extinction
risk
0.23
0.03
0.0
quasi-extinction
risk (10 pairs)
0.97
0.30
0.233

186
Manatee-S. Hillsborough (M41
General description: This metapopulation occurs predominantly in Manatee and
S. Hillsborough counties, with a few jays occurring in west Hardee and northeast
Sarasota county. The configuration of habitat in this region as mapped by the SMP is
very unusual, occurring as small patches isolated from each other by small to moderate
distances (1-10 km). During the 1992-1993 survey, a significant number of patches
were on private lands that could not be accessed. The SMP documented about 65 jay
territories, excluding suburban jays, in this metapopulation. Estimated potential
population size after habitat restoration and full occupancy is 36 pairs in currently
protected areas, and 145 pairs maximum.
Protected areas: Golden Aster Scrub Nature Preserve (Hill9), Balm-Boyette
Scrub (H118), Little Manatee River (H112), Little Manatee River State Recreation
Area (H113), Duette Park (Manl5), Lake Manatee Lower Watershed (ManlO),
Beker (Man 12, Man 16), Lake Manatee State Recreation Area (Man 17), Myakka
River State Park (Sari5), Verna Wellfield (Sarl9).
Restoration potential: Many of the patches in this metapopulation are heavily
overgrown, but the restoration potential of the numerous small patches may not be much
greater than the SMP population estimates. Nevertheless, habitat restoration is urgently
needed in many of the protected areas, since local populations of jays are small and very
vulnerable to local extinction.
Simulation results: This metapopulation ranked 6th in vulnerability (table 5-23),
14th in percent protected (24.8%; table 5-24), and 10th in priority (table 5-25), with high
vulnerability and moderate potential for improvement.

187
After restoration, the no acquisition configuration was estimated to support
about 36 jay territories in existing protected areas. This configuration had a very high risk
of extinction (p=0.97) and quasi-extinction (p=1.0).
Simulations for this metapopulation produced unexpected results that were
strikingly different from all other metapopulations. All scenarios for this metapopulation
declined rapidly and had high extinction and quasi-extinction risk. Even the maximum
acquisition configuration, with 145 pairs, had an extinction risk of 0.30, and a quasi
extinction risk of 0.90 (Table 5-4b).
These results are rather surprising. The model predicts that the long-term
probability of persistence is low for all configurations, yet there is no doubt that jays have
persisted in this area for many decades. Several factors may account for this apparent
discrepancy. First, this was the least thoroughly surveyed metapopulation; more than
67% of the patches were inaccessible. Undoubtedly, many more jays occur in this area
than were found during the SMP. For modeling purposes, some jays were added to
unsurveyed patches (compare Figs. 5-4a & b), but at lower densities than surveyed areas
(this was the only metapopulation in which substantial number of jays were added to
unsurveyed or unoccupied patches). Second, many jays probably occur in atypical habitat
that could not be identified on soil maps or aerial photographs used for the SMP. This is
evidenced by the recent discovery of jays in the northeast portion of this metapopulation
(R. Bowman, pers. comm.). Third, significant portions of the jay habitat in this area may
have been modified or lost recently (during the last couple of decades) due to vegetable
farms and ranch activities. Cox (1987) mentions records of abundant jays in Manatee
county along the coast which disappeared due to suburban development (e.g. near

188
Bradenton). Displacement of jays from developing areas and lags in population decline
are likely; model predictions of substantial declines in jay populations may be reasonable,
given the current landscape. Fourth, the habitat matrix may be less hostile to dispersers
than the model assumes. Many of the habitat patches occur in ranch settings with a fairly
open matrix and little or no suburbanization, which may create favorable conditions for
floater dispersal. However, simulations conducted by Stith et al. (manuscript in prep.)
suggest that even with unrealistically high floater dispersal ability and survivorship, jays
are unlikely to survive long-term in landscapes with small, somewhat isolated patches
unless landscape features such as corridors direct the movement of floaters towards
occupied patches.
Recommendations: A more comprehensive survey for jays is needed to fill in the
large data gaps. Besides mapping unsurveyed scrub patches, special effort should be
made to map atypical habitats with low densities of oak scrubs, as these may be important
to jays in this metapopulation. Clusters of contiguous jay territories are conspicuously
absent in this landscape. Acquisition efforts should emphasize larger patches that are as
close as possible to other large patches. Unprotected, unsurveyed patches near the north
(Mani) and south (Man 15) section of Beker, and further south at Man9, Man 10,
Man 1, and Man5are likely candidates. Acquisition of Sari 6 and Manl, which
are near protected jays at Verna Wellfield (Sari 9) and Myakka River State Park
(Sari 5) would benefit these jays. Besides restoring overgrown scrub, habitat
management should seek to remove dispersal barriers between patches. Creation of
partially cleared right-of-ways between patches may facilitate dispersal (research on this
subject is needed).

189
Balm-Boyettl
*P&¡wb Prm
SS#-
'^Goldenkster
Scrub Nature Pr.
'Hillsborough Cj>. _'Polk Co
Manatee Co
Hardee Co
Lake"Mariafee
S. Rec. Area
)ay Territory Locations
Scrub Polygons
^ Lo Disturbance
2] Lo Density Housing
[A/Z-j HI Density Housing
Ranch/Ag
1.75 km dispersal buffer
Protection Status
ifJHf] Protected
~ Proposed
Water Bodies
Interstates
State highways
County roads
County lines
M4 Manatee-S. Hillsborough
meter
+
12
18 Kilometers
1 : 320,000
Fig. 5-4a. Manatee and S. Hillsborough county map 1992 1993 jay and habitat
distribution.
i

190
Vqfa
:
Sarasota Co.
Man13¡ ',\^)'Har2
fe)9
,
^ 1
l_J 'Man l*Sa6 23 "7' ! I
MyaAj/ra River S.P.
x
Sfrsajio
]ay Territory Locations
(after restoration) Protection Status Water Bodies M4 Manatee-S. HillsborOUgt"
Scrub Polygons Protected .
CCC ## Polygon ID /V Unprotected ' V County l,nes , M,J ill iJ
Jays outside of labeled, bold polygons are considered to be Suburban jays. -| 320 000 t
Fig. 5-4b. Manatee and S. Hillsborough county acquisition map.

Table 5-4a. Manatee and S. Hillsborough county patch statistics (number of jay territories for different configurations)
Patch
Status
1992-
No
30%
30%
70%
70%
Maximum
id
1993# jay
acquisition
preserved
preserved
preserved
preserved
acquisition
territories
(restored)
by contiguity
by connectivity
by contiguity
by connectivity
Har1
8
3
8
8
8
Har 2
1
Har 3
1
Har 4
4
2
3
HilH
2
3
2
5
HII2
Little Manatee R.
2
2
2
2
2
2
2
HHI3
Little Manatee R. S.
Rec. Area
1
2
2
2
2
2
2
HII4
8
3
8
8
8
HHI5
1
1
1
1
1
HII6
1
HHI7
2
2
HII8
Balm-Boyette Scrub
Pr.
1
6
6
6
6
6
6
HII9
Golden Aster Scrub
Nature Pr.
2
2
2
2
2
2
2
HHI10
4
Him 1
2
2
2
2
HilH 2
1
HilH 3
3
Man1
4
5
3
5
5
5
Man2
4
4
2
4
4
4
Man3
3
5
Man4
3
2
5
5
5
Man5
1
6
5
10
Man6
2
5
3
5
3
5
Man7
2

Table 5-4a continued.
Patch id
Status
1992-1993#
No
30%
jay territories
acquisition
preserved
(restored)
by contiguity
Man8
Man9
Man10
Lake Manatee
Lower
Watershed
3
3
Man11
Man12
Beker
1
2
2
Man 13
1
Man14
1
Man15
Duette Park
6
11
11
Man16
Beker
6
8
Man17
Lake Manatee
St. Rec. Area
1
2
Man18
1
Sar15
Myakka River
St. Pk.
3
3
Sar16
3
Sar17
2
2
Totals
65
36
69
30%
preserved
by connectivity
70%
preserved
by contiguity
70% Maximum
preserved acquisition
by connectivity
3
3
3
1
1
3
1
2
1
1
11
6
2
1
2
4
11
15
2
1
2
4
1
12
15
2
2
2
4
3
12
15
2
3
3
3
3
2
2
4 4 4
2 2 2
69 112 112 145
vO
K)

Population Size
193
Fig. 5-4c. Manatee and S. Hillsborough county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition.

Probability
194
i oT
i
i-
0 8!
0 6~
r
04
l
0 2
r
50
100
Threshold Pop. Size
150
1 oT
|-
o.aT
i!
r!
611
0 4
0.2
50
100
Threshold Pop. Size
150
Fig. 5-4d. Manatee and S. Hillsborough county quasi-extinction graphs. Top) no
acquisition, Bottom) maximum acquisition.

Table 5-4b. Manatee and S. Hillsborough county simulation statistics
Data type
No
acquisition
70% acquisition
by connectivity
70% acquisition
by area
Maximum
acquisition
starting
population size
36
112
112
145
x end pop. size
0.2
5.5
2.4
4.9
s.d.
0.7
8.2
4.0
4.6
percent
decline
95.3
96.6
extinction
risk
0.97
0.43
0.60
0.30
quasi-extinction
risk (10 pairs)
1.0
0.90
0.97
0.90

196
Sarasota-W. Charlotte (M5)
General description: The Sarasota-W. Charlotte metapopulation occurs along the
Gulf coast from central Sarasota county south into Charlotte county, terminating at
Charlotte Harbor. It is separated from the N.W. Charlotte county metapopulation by the
Myakka River. The largest single population of jays along the Gulf coast occurs here in
Oscar Sherer State Park. The SMP documented about 64 jay territories, excluding
suburban jays, in this metapopulation. Estimated potential population size after habitat
restoration and full occupancy is 50 pairs in currently protected areas, and 89 pairs
maximum.
Protected areas: Only two protected areas have the potential to protect more than
10 pairs of jays: Oscar Sherer State Park (Sar8), and Casperson Beach County Park and
Brohard Park (Sar7). Other protected areas are very small: Lemon Bay Scrub County
Park (Sar4), Myakka State Forest (Sari4), Charlotte Harbor State Buffer Preserve
(Chari).
Restoration potential: Habitat acquistion and restoration at Oscar Sherer State
Park has increased the jay population from an estimated 19 pairs in 1992-1993, to 27
pairs in 1997, and when fully restored might support about 30 pairs (J. Thaxton, pers.
comm.). The Casperson Beach County Park and Brohard park complex was estimated to
support about 13 pairs of jays, compared to the 7 pairs found for the SMP.
Simulation results: This metapopulation ranked 16th in vulnerability (table 5-23),
8th in percent protected (56.2%; table 5-24), and 13th in priority (table 5-25), with
moderate vulnerability and moderate potential for improvement.

197
The no acquisition option is estimated to support about 50 jay families after
restoration and full occupancy. Simulation of this configuration indicated that the
metapopulation is vulnerable to extinction (p=0.03) and quasi-extinction (0.10).
Intermediate simulations (30% and 70%) did not produce any extinctions, and 2
simulations produced quasi-extinctions (Table 5-5b). Percent population declines were
fairly large for all simulations, reflecting the large number of jays residing in small,
somewhat isolated patches.
Recommendations: The relatively favorable ranking of this metapopulation is due
mainly to the Oscar Sherer State Park population, which is much larger than any other
local population and has a stable population trajectory. Considering that Oscar Sherer
makes up a large proportion of the total metapopulation, the 40 to 50 percent population
decline seen in the simulation results (Table 5-5b) is due to the collapse of the numerous
smaller, somewhat isolated populations.
Acquisition of larger patches that are near other patches probably will benefit this
population most. The best opportunity for acquisition appears to be the habitat patches in
Charlotte county (Char2), southwest of the Rotunda Circle. Nearby patches in Rotunda
Circle (Char3 and Char4) also should be investigated, as jays in this area could
provide dispersers to the protected but very isolated jays at Charlotte Harbor State Buffer
Preserve (Chari). The possibility of bolstering the small protected jay population at
Lemon Bay Scrub county park (Sar4) by acquiring nearby properties at Sar3 and
Sar5 should be investigated. The private reserve (Sar9) just north of Oscar Sherer
would benefit from the acquisition of Sari 0.

un
198
iviy.au na ni ver or
ohard Park
k Qaspeyson Beach
Charltt&tarbor
ftalUfuffe r "Pe.
pmon Bay
i erad C P 4
Charlotte HajtM
ISkgiP utfenJPr
Scrub Polygons
Lo Disturbance
Lo Density Housing
Hi Density Housing
Ranch/Ag
Protection Status
Protected
1 Proposed
Water Bodies
Intersutes
Sute highways
County roads
. County lines
H5 Sarasota
1 : 250,000
Kilometers
+
Fig. 5-5a. Sarasota and W. Charlotte county map 1992 1993 jay and habitat
distribution.

199
Jay Territory locations
(after restoration)
Scrub Polygons
CCC## Polygon ID
Jays outside of labeled, bold polygons are considered to be Suburban jays
Protection Status
/\y Protected
,\y Unprotected
Water Bodies Sarasota
County lines o !
10
15
1 : 250,000
Kilometers
Fig. 5-5b. Sarasota and W. Charlotte county acquisition map.

Table 5-5a. Sarasota and W. Charlotte county patch statistics (number of jay territories for different configurations)
Patch
Status
1992-1993#
No
30%
30%
70%
70%
Maximum
id
jay territories
acquisition
preserved
preserved
preserved
preserved
acquisition
(restored)
by contiguity
by connectivity
by contiguity
by connectivity
Chari
Charlotte Harbor
S.B.P.
3
3
3
3
3
3
3
Char2
13
9
1
13
7
13
Char3
2
2
2
2
2
Char4
5
1
5
3
5
Char5
2
2
2
Char6
4
1
3
4
Sari
1
1
2
Sar2
1
1
1
1
Sar3
1
1
1
1
1
1
Sar4
Lemon Bay Scrub
Cty. Pk.
2
2
2
2
2
2
2
Sar5
1
1
1
3
2
3
Sar6
1
1
3
3
Sar7
Casperson Beach
Cty. Pk./Brohard
Pk.
7
13
13
13
13
13
13
Sar8
Oscar Sherer S.P.
19
30
30
30
30
30
30
Sar9
Private reserve
1
1
1
1
1
1
1
Sano
2
1
3
2
3
Sari 4
Myakka S.F.
1
1
1
1
1
Totals
64
50
61
61
77
77
89
o
o

Population Size
201
a
w
a.
o
CL
I
40
10
20
l
30
Year
50
20 0
20
40
Year
60
Fig. 5-5c. Sarasota and W. Charlotte county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition.

Probability
202
r
0 2~[
i *
50 100 150
Threshold Pop Size
roT
0 8~
j-
(
os'!;
04
p
0 2~T
it
50 100
Threshold Pop. Size
150
Fig. 5-5d. Sarasota and W. Charlotte quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition.

Table 5-5b. Sarasota and W. Charlotte county simulation statistics
Data type
No
acquisition
30% acquisition
by connectivity
30% acquisition
by area
70% acquisition
by connectivity
70% acquisition
by area
Maximum
acquisition
starting
population size
50
61
61
77
77
89
x end pop. size
26.3
31.3
33.6
43.9
45.7
46.9
s.d.
10.7
96
8.0
11.1
13.2
11.6
percent
decline
47.4
48.7
44.9
43.0
40.7
47.3
extinction
risk
0.03
0.0
0.0
00
0.0
0.0
quasi-extinction
risk (10 pairs)
0.10
0.07
0.0
0.0
0.02
0.0
K>
o
u>

204
N. W. Charlotte M61
General description: The N.W. Charlotte metapopulation is isolated from the
Sarasota metapopulation to the west by the Myakka River, and is isolated from the
Central Charlotte metapopulation to the east by the Peace River. The SMP documented
about 44 jay territories, excluding suburban jays, in this metapopulation. Estimated
potential population size after habitat restoration and full occupancy is 28 pairs in
currently protected areas, and 56 pairs maximum.
Protected areas: Protected jays occur on the Charlotte Harbor State Buffer
Preserve (Char9; formerly known as Tippecanoe Scrub), and on the Myakka River
State Forest (Sari3), a private reserve (Chari2) and near 1-75 (Chari la).
Restoration potential: Most patches in this metapopulation are small, and jay
densities measured during the SMP were probably close to maximal. The Myakka River
State Forest (Sari 3), which had one pair of jays during the SMP, may have sufficient
habitat for 6 pairs.
Simulation results: This metapopulation ranked 10th in vulnerability (table 5-23),
9th in percent protected (50.0%; table 5-24), and 6th in priority (table 5-25), with high
vulnerability and high potential for improvement. All simulations had significant
extinction and quasi-extinction risk, and large percent population declines (see Table 5-
6b).
Recommendations: The risk estimates for the maximum acquisition
configuration are greatly improved compared to the no acquisition option. Acquisition
and restoration of as much habitat as possible is recommended. The most important
population of protected jays in this metapopulation probably occurs at the Charlotte

205
Harbor State Buffer Preserve (Char9). Acquisition of unprotected habitat adjacent to
this population (Char8, Char7) should be a high priority. Additional habitat adjacent
to the protected jays at Myakka State Forest (Sari 3) should be investigated for
acquisition.

Scrub Polygons
I 1 Lo Disturbance
2] Lo Density Housing
IrWSj Hi Density Housing
Ranch/Ag
Protection Status
f i Protected
2 Proposed
Water Bodies
/\y Interstates
State highways
, County roads
County lines
M6 N.W. Charlotte
0 4 8 12
1 : 200,000
Kilometers
Fig. 5-6a. N. W. Charlotte county map 1992 1993 jay and habitat distribution.

207
jay Territory Locations
(after restoration)
Scrub Polygons
CCC## Polygon ID
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Protection Status
/\f Protected
A/Unprotected
Water Bodies
County lines
M6 N.W. Charlotte
1 : 200,000
12
Kilometers
Fig. 5-6b. N. W. Charlotte county acquisition map.

Table 5-6a. N. W. Charlotte county patch statistics (number of jay territories for different configurations)
Patch id
Status
1992-1993#
No
70% preserved
70% preserved
Maximum
jay territories
acquisition
(restored)
by connectivity
by area
acquisition
Char7
11
8
11
11
Char8
4
3
4
4
Char9
Charlotte
Harbor State
Buffer Pr.
11
11
11
11
11
CharlO
3
2
1
3
Chari 1
2
2
3
3
Chari 1a
SWFMD?
2
5
5
5
5
Char12
Private Pr.
6
6
6
6
6
Sari 1
2
2
4
Sar12
2
2
4
Sar13
Myakka S.F
1
6
6
6
6
Totals
44
28
47
47
56
208

209
J ¡L
20 40 60
Year
Fig. 5-6c. N. W. Charlotte county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.

210
0 67
u
0 4~
r
[
4-
02
-
10 20
Threshold Pop Size
Fig. 5-6d. N. W. Charlotte county quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition.

211
Table 5-6b. N. W. Charlotte county simulation statistics
Data type
1992-1993
configuration
No
acquisition
70%
acquisition by
connectivity
70%
acquisition
by area
Maximum
acquisition
Starting
population size
44
28
47
47
56
Mean ending
population size
19.4
2.3
12.8
14.5
22.1
s.d.
11.3
3.9
9.1
8.6
11.3
% population
decline
55.9
91.8
72.8
69.1
60.7
Extinction
Risk
0.10
0.67
0.17
0.17
0.07
Quasi
extinction
Risk (10 pairs)
0.37
1.0
0.63
0.57
0.30

212
Central Charlotte (M7)
General description: The central Charlotte metapopulation is isolated from the
nearby northwest Charlotte metapopulation by the Peace River to the west. Most of the
habitat occurs along Prairie and Shell Creek, which drain into the Peace River. Two
somewhat isolated populations occur south of Punta Gorda (Char22) and into Lee
county (Lee5), the latter patch occurring near a proposed Carl addition to the Babcock-
Webb W.M.A. The SMP documented about 31 jay territories, excluding suburban jays, in
this metapopulation. Estimated potential population size after habitat restoration and full
occupancy is 5 pairs in currently protected areas, and 61 pairs maximum.
Protected areas: No jays are protected on public lands in this metapopulation; a
small population of jays occurs on a private reserve (Charl).
Restoration potential: The restoration potential for most patches in this
metapopulation probably is not significantly greater than the jay densities measured for
the SMP. However, two large incompletely surveyed patches occur along Prairie Creek
(Chari 7, Chari 8) and may have large restoration potential. For modeling purposes,
these patches were estimated to support considerably more jays than were found during
the SMP (Table 5-7a). Both patches probably have the potential to support substantially
more jays than the estimates used here. A large, unsurveyed patch east of Char20 along
Shell Creek may harbor jays, but no jays were included in any of the simulations.
Simulation results: This metapopulation ranked 2nd in vulnerability (table 5-23),
19th in percent protected (8.2%; table 5-24), and 3rd in priority (table 5-25), with high
vulnerability and high potential for improvement. The no acquisition option is

213
extremely vulnerable to extinction and quasi-extinction (Table 5-7b). The 70%
acquisition by area configuration is considerably improved, but still has a substantial
quasi-extinction risk (p=0.33). The maximum acquisition has a much lower quasi
extinction risk (p=0.07).
Recommendations: This metapopulation ranks 2nd in vulnerability due to the near
absence of jays on protected lands. It has a priority ranking of 3, with low protection and
high potential for improvement.
The private reserve (Charl) would benefit considerably by the acquisition of
nearby jay habitat, especially Chari 5. Substantial tracts of largely unsurveyed,
unprotected scrub occur along both sides of Prairie Creek (Chari 7, Chari 8), and jays
were documented in the western portions of these patches for the SMP. Acquisition and
restoration of these patches would greatly bolster this metapopulation. The large,
unsurveyed patch along Shell Creek also should be investigated. Consideration should be
given to adding the isolated scrub patch (Lee 5) to the proposed CARL addition to the
Babcock-Webb W.M.A.

214
Jay Territory Locations
Scrub Polygons
2 Lo Disturbance
2 Lo Density Housing
l/VYj Hi Density Housing
gggj Ranch/Ag
1.75 km dispersal buffer
Protection Status
Hifll- Protected
~ Proposed
Water Bodies
/V'interstates M7 Central Charlotte
A
State highways
County roads
County lines
9 Kilometers
1 : 250,000
Fig. 5-7a. Central Charlotte county map 1992 1993 jay and habitat distribution..

215
Jay Territory Locations
(after restoration)
Scrub Polygons
CCC## Polygon ID
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Protection Status
/\f Protected
/\/ Unprotected
Water Bodies M7 Central Charlotte
County lines
9 Kilometers
1 : 250,000
Fig. 5-7b. Central Charlotte county acquisition map.

Table 5-la. Central Charlotte county patch statistics (number of jay territories for different configurations)
Patch id
Status
1992-1993#
No
30%
30%
70%
70%
Maximum
jay territories
acquisition
acquisition by
acquisition
acquisition by
acquisition
acquisition
(restored)
connectivity
by area
connectivity
by area
Chari 3
1
1
2
3
Char14
2
2
3
5
5
Chari 5
7
2
11
7
11
11
Char16
private reserve
5
5
5
5
5
5
5
Chari 7
2
3
6
6
9
9
Chari 8
5
4
9
12
12
Chari 9
1
1
2
3
Char20
1
1
1
1
Char21
2
1
2
2
2
Char22
4
2
5
7
Lee5
1
1
2
3
Totals
31
5
22
22
44
44
61

217
20 40 60
Year
Fig. 5-7c. Central Charlotte county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.

218
Threshold Pop Size
1 0
0 8T
0 6~
I
04
0 2~
Threshold Pop Size
Fig. 5-7d. Central Charlotte county quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition.

219
Table 5-7b. Central Charlotte county simulation statistics
Data type
Original
1992-1993
scenario
No
acquisition
70% acquisition
by connectivity
70% acquisition
by area
Maximum
acquisition
starting
population size
31
5
44
44
61
x end pop. size
11.7
0.0
21.3
35.5
19.4
s.d.
6.3
0.0
10.4
10.4
11.3
percent
decline
62.2
100.0
51.6
19.3
65.4
extinction
risk
0.13
1.0
0.17
0.10
0.07
quasi-extinction
risk (10 pairs)
0.70
1.0
0.30
0.33
0.07

220
Lee and N. Collier (M8)
General description: The Lee metapopulation is isolated from the N.W. Charlotte
metapopulation by the heavily forested Babcock-Cecil Webb Wildlife Management Area.
The Lee metapopulation is just beyond the 12km dispersal buffer of the Lake Wales
Ridge metapopulation, but the intervening habitat may be conducive to some exchange
between these metapopulations. Jays within the Lee metapopulation are poorly
connected; jays occur in tiny patches along the Caloosahatchee River, a second cluster
occurs around the town of Immokalee, and a few jays occur near the Gulf Coast at Estero
Bay and into Collier county. A small experimental population of jays has been
translocated to Rookery Bay south of Naples (Mumme and Below 1999). The SMP
documented about 47 jay territories, excluding suburban jays, in this metapopulation.
Estimated potential population size after habitat restoration and full occupancy is 15 pairs
in currently protected areas, and 62 pairs maximum.
Protected areas: Protected jays occur on Estero Bay Aquatic Preserve (Lee3),
and at Rookery Bay National Estuarine Research Reserve (Coll2).
Restoration potential: Restoration potential of this metapopulation is quite limited
for jays except near the Gulf Coast. Estero Bay Aquatic Preserve (Lee3) might support
about 9 pairs of jays after restoration, and the unprotected habitat 7-8 km south at Colli
might support about 7 pairs. Some large habitat patches between Lee3 and Colli
appear to be unoccupied because of heavy overgrowth.
Simulation results: This metapopulation ranked 9th in vulnerability (table 5-23),
16th in percent protected (24.2%; table 5-24), and 9th in priority (table 5-25), with high
vulnerability and moderate potential for improvement. The no acquisition option has a

221
high risk of extinction (p=0.73) and quasi-extinction (p=1.0; Table 5-8b). The maximum
acquisition option shows a moderate risk of extinction (p=0.40) and a high risk of quasi
extinction (p=0.90).
Recommendations: The best opportunities for acquisition and restoration appear
to be at the Estero Bay Aquatic Preserve (Lee3) and south to Colli, including the
intervening unoccupied patches. Management of jay habitat at the Immokalee airport
(Coll3) and acquisition of nearby patches should be investigated. The acquisition of
unprotected jay habitat near the Caloosahatchee State Recreation Area (Lee2) and near
the Hickey Creek Mitigation Park (Leel) also should be investigated. Mumme and
Below (1999) state that more intensive habitat management is needed for the
experimental translocation at Rookery Bay to succeed.

222
Jay Territory Locations
Scrub Polygons
[=j Lo Disturbance Protection Status
Y \ Lo Density Housing LIJ Protected
Hi Density Housing Proposed
HIUII kanch/Ag Water Bodies
1.75 km dispersal buffer
/\J Interstates
State highways
County roads
County lines
M8 Lee
12
1 : 300,000
18 Kilometers
Fig. 5-8a. Lee and N. Collier county map 1992 1993 jay and habitat distribution.

Lee Co.
223
Protection Status
yAy/ Protected
/\/ Unprotected
Water Bodies
County lines
M8 Lee
jay Territory Locations
(after restoration)
Scrub Polygons
CCC## Polygon ID
Jays outside of labeled, bold polygons are considered to be Suburban jays.
12
18 Kilometers
1 : 300,000
Fig. 5-8b. Lee and N. Collier county acquisition map.

Table 5-8a. Lee and N. Collier county patch statistics (number of jay territories for different configurations)
Patch id
Status
1992-1993#
No
30%
70%
Maximum
jay territories
acquisition
acquisition
acquisition
acquisition
(restored)
by area
by area
Lee1
8
6
8
Lee2
15
8
15
Lee3
Estero Bay Aquatic Pr.
2
9
9
9
9
Lee4
1
1
Colli
2
6
6
7
Coll2
Rookery Bay Nat. Estuarine
Research Reserve
3
6
6
6
6
Coll3
Immokalee Airport
4
4
4
4
Coll4
2
2
2
Coll5
5
5
5
C0II6
3
3
Coll7
2
2
2
Totals
47
15
25
48
62
K>
K>

Population Size
225
Fig. 5-8c. Lee and N. Collier county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.

226
20
40
Threshold Pop Size
60
Fig. 5-8d. Lee and N. Collier county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition.

227
Table 5-8b. Lee and N. Collier county simulation statistics
Data type
Original
1992-1993
scenario
No
acquisition
30% acquisition
by area
70% acquisition
by area
Maximum
acquisition
starting
population size
47
15
25
48
62
x end pop. size
0.9
1.1
2.0
4.5
5.6
s.d.
2.4
2.2
3.2
6.1
4.7
percent
decline
98.0
92.7
92.0
90.1
91.0
extinction
risk
0.87
0.73
0.67
0.43
0.40
quasi-extinction
risk (10 pairs)
1.0
1.0
1.0
0.90
0.90

228
Flaeler-N.E. Volusia (M9)
General description: The Flagler-N.E. Volusia county metapopulation is the most
north-eastern population of jays occurring along the Atlantic Coast. This metapopulation
is isolated from the Volusia-Merritt Island metapopulation (mlO) to the south by the city
of Daytona Beach. All of the jays in this metapopulation occur near or along the beach;
consequently habitat loss due to oceanfront development has greatly reduced this
population. The SMP documented about 12 jay territories, excluding suburban jays, in
this metapopulation. Estimated potential population size after habitat restoration and full
occupancy is 5 pairs in currently protected areas, and 12 pairs maximum.
Protected areas: The only protected jays in this metapopulation occur in the N.
Peninsula State Recreation Area (Voll in Fig. 5-9b). Three protected areas had jays
prior to the SMP: Gamble Rogers Memorial State Recreation Area, Flagler Beach State
Recreation Area, and Washington Oaks State Gardens (jays still occasionally seen). Jays
occur near these parks, as well as near Marineland, but appear to occupy territories in
suburban or urban settings.
Restoration potential: Improved management of the unoccupied, protected areas
(Gamble Rogers, Flagler Beach, and Washington Oaks) might attract jays from nearby
suburban settings. The potential also exists to re-introduce jays to these parks, but for
modeling purposes jays were excluded from these areas. The population of jays at N.
Peninsula State Recreation Area as measured for the SMP was assumed to be close to
carrying capacity.
Simulation results: This metapopulation ranked 7th in vulnerability (table 5-23),
10th in percent protected (41.7%; table 5-24), and 16th in priority (table 5-25), with high

229
vulnerability and low potential for improvement. Simulations of the currently protected
configuration of 5 territories produced a high extinction (p=0.933) and quasi-extinction
risk (p=1.0) for the small population (Table 5-9b). The maximum acquisition option
produced a substantially reduced extinction risk (p=0.57; Table 5-9b), but the quasi
extinction risk remained high (p=l .0).
Recommendations: The habitat map for this metapopulation developed for the
SMP is based on old soil maps and is very outdated along the coast; it does not reflect the
extensive habitat destruction that has occurred subsequent to the production of the soil
maps. Recent aerial photographs should be used to update this habitat information.
Acquisition options are very limited along the coast in this area. The best
opportunity may be several small tracts of land near Marineland and Washington Oaks
(Flagl, Flag2). The prognosis for this small population of jays is not good, as they
probably face problems similar to those described by Breininger (1999) for the urban jays
on the the south Brevard county barrier island. Some large tracts of apparently
unoccupied scrub may still exist a few kilometers inland. Given the high risk faced by the
coastal jays and their potentially unique genetic traits, the possibility of acquiring and
restoring these unoccupied patches and translocating jays from nearby coastal areas
should be considered.

230
)ay Territory Locations 1 75 km dispersal buffer
Scrub Polygons
I 1 Lo Disturbance _
FI Lo Density Housing !!§ Pro,ecte^
Hi Density Housing L ropose
Ranch/Ag Water Bodes
Protection Status
Interstates
f\/ State highways
County roads
' County lines
M9 Flagler SC N.E. Volusia
12 Kilomeie
1 : 250,000
Fig. 5-9a. Flagler and N.E. Volusia county map 1992 1993 jay and habitat
distribution.

231
]ay Territory Locations
(after restoration)
Scrub Polygons
Protection Status
/\/ Protected
/\J Unprotected
CCC## Polygon ID
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Water Bodies M9 Flagler sc N.E. Volusia
' County lines 0 4 8 12 Kilomete!
_ N
1 : 250,000
Fig. 5-9b. Flagler and N.E. Volusia county acquisition map.

232
Table 5-9a. Flagler and N.E. Volusia county patch statistics (number of jay territories for
different configurations)
Patch id
Status
1992-
1993# jay
territories
No
acquisition
(restored)
Maximum
acquisition
Flagl
3
3
Flag2
2
2
Flag3
2
2
Vol1
N. Peninsula St. Rec. Area
5
5
5
Totals
12 5 12

Population Size
233
Fig. 5-9c. Flagler and N.E. Volusia county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition.

Probability
234
roT
0 8
%
CL
0.6~f
04
0 2
2 4
Threshold Pop. Size
Fig. 5-9d. Flagler and N.E. Volusia county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition.

235
Table 5-9b. Flagler and N.E. Volusia county simulation statistics
Data type
Original
1992-1993
scenario
No
acquisition
Maximum
acquisition
Starting
population size
12
5
12
Mean ending
population size
2.4
0.27
2.4
s.d.
2.91
0.995
2.91
Percent
population
decline
75.8
80.1
75.8
Extinction
Risk
0.57
0.933
0.57
Quasi
extinction
Risk (10 pairs)
1.00
1.00
1.00

236
Merritt Island-S.E. Volusia and (M10)
General description: This metapopulation includes a large number of protected
jays on the huge barrier island complex that includes Cape Canaveral Air Station
(CCAS), Kennedy Space Center (KSC), and Merritt Island National Wildlife Refuge
(MINWR). A few unprotected jays occur just north of MINWR in Volusia county, and
along the eastern and southern boundaries of MINWR and south of CCAS. This
metapopulation is isolated from the Flagler-N.E. Volusia metapopulation by the city of
Daytona Beach to the north, and from the N. Brevard metapopulation by the Indian River
and Turnbull Hammock to the west and southwest. The SMP documented about 536 jay
territories, excluding suburban jays, in this metapopulation. Estimated potential
population size after habitat restoration and full occupancy is 495 pairs in currently
protected areas, and 536 pairs maximum.
Protected areas: Protected jays occur within CCAS, KSC, MINWR, and several
small reserves (Table 5-10a).
Restoration potential: The restoration potential of KSC and MINWR is difficult
to estimate due to the heterogeneous nature of the habitat. The habitat information
provided by the SMP is much too coarse to attempt an assessment. Because the
population sizes estimated for the SMP are quite large, the simulation results would not
be affected significantly by increasing the population size above the SMP estimates.
Consequently, densities estimated by the SMP were used for all simulations.
Simulation results: This metapopulation ranked 19th in vulnerability (table 5-23),
2nd in percent protected (92.3%; table 5-24), and 19th in priority (table 5-25), with low
vulnerability and low potential for improvement.The no acquisition configuration and

237
maximum acquisition option produced very similar results, with no extinction or quasi
extinction risk and very low percent population declines (Table 5-10b).
Recommendations: Although this is a large metapopulation, vulnerability to
hurricanes, habitat overgrowth and difficulties with habitat restoration pose serious
threats to this metapopulation. Modeling performed by Breininger et al. (in press) found
this metapopulation to be vulnerable to catastrophes associated with hurricanes, but
habitat degradation was a much more important risk factor. Years of fire suppression
have resulted in overgrown habitat which is difficult to restore compared to other areas,
apparently because the coastal soils and water table allow rapid regrowth of scrub oaks
and other vegetation, resulting in the closure of openings needed by jays for foraging and
predator detection. Preliminary results from a 1999 survey of MIN WR/KSC suggest that
the population may have declined as much as 50% compared to estimates made during
the SMP (Gary Popotnik, pers. comm.). Habitat restoration is urgently needed for this
metapopulation.

238
)ay Territory Locations 1 75 km dispersal buffer
Scrub Polygons
^ Lo Disturbance
\7~7~\ Lo Density Housing
Hi Density Housing
B Ranch/Ag
Protection Status
[ I Protected
1 Proposed
Water Bodies
/\/ Interstates
State highways
County roads
. County lines
M10 Merritt Island St
S.E. Volusia
0 6 12 18
1 : 310,000
Kilometers
Fig. 5-10a. Merritt Island and S.E. Volusia county map 1992 1993 jay and habitat
distribution.

239
MIO
Jay Territory Locations
(after restoration) Protection Status
Scrub Polygons /\/ Protected
CCC## Polygon ID /V Unprotected
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Water Bodies
County lines
Merritt Island St
S.E. Volusia
12
1 : 310,000
18 Kilometers
Fig. 5-10b. Merritt Island and S.E. Volusia county acquisition map.

240
Table 5-10a. S.E. Volusia and Merritt Island county patch statistics (number of jay
territories for different configurations)
Patch id
Status
1992-1993# jay
territories
No
acquisition
(restored)
Maximum
acquisition
Vol2
3
3
Vol3
4
4
Vol4
7
7
Vol5
1
1
Vol6
6
6
Vol7
4
4
Vol8
1
1
Vol9
2
2
Brev19
4
4
Brev20
3
3
Brev21
3
3
Brev22
1
1
Brev23
1
1
Brev24
1
1
Brev25
Merritt Island N.W.R. & Kennedy
377
377
377
Space Center
Brev26
Cape Canaveral Air Station
118
118
118
Totals
536
495
536

Population Size
241
N
(/)
JTJ
Q.
c
CL
500 jiSi *i t-t i 1 1 ^ 1 i i i 1 i ^ | ^ i~~i^ ^ 1 i ^
3oo n
t
t
ioo L
t
20 40 60
Year
500
(- ,
20
I
40
Year
'H-hh+H+Hh
60
Fig. 5-1 Oc. S.E. Volusia and Merritt Island county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition.

Probability
242
02 !
L
h
I l
100
l|j. I i__J 1 I ; ( 1 L
I I
200 300 400
Threshold Pop Size
i.o7
8"t
i-
f
6
l
J-
04[
-
0.27
i i i i iii
100
r
i
j
j
i
r
J
J
J
r¡
200 300 400 500
Threshold Pop Size
Fig. 5-1 Od. S.E. Volusia and Merritt Island county quasi-extinction graphs. Top) no
acquisition, Bottom) maximum acquisition.

243
Table 5-1 Ob. S.E. Volusia and Merritt Island county simulation statistics
Data type
No
acquisition
Maximum
acquisition
starting
population size
495
536
x end pop. size
491.5
501.3
s.d.
5.7
14.5
percent
decline
0.7
6.5
extinction
risk
0.0
0.0
quasi-extinction
risk (10 pairs)
0.0
0.0

244
N. Brevard (MU')
General description: The N. Brevard metapopulation is separated from the Central
Brevard metapopulation (Ml 2) by the city of Cocoa, and from the S.E. Volusia and
Merritt Island metapopulation by the Indian River and Turnbull Hummock to the east and
northeast (see Fig. 5-1 la). The SMP documented about 101 jay territories, excluding
suburban jays, in this metapopulation. Estimated potential population size after habitat
restoration and full occupancy is 4 pairs in currently protected areas, and 110 pairs
maximum.
Protected areas: Few or no jays occur on protected lands in N. Brevard. For
modeling purposes, protected jays were assumed to occur only on recently acquired
property at South Lake (Brev7). Several properties that are targeted for acquisition
include Seminole Ranch (Brev8), Buck Lake (Brev4), and Tico (Brevl2,
Brevl 1, and BrevlO; see Fig. 5-1 lb). The Dicerandra Scrub Sanctuary falls within
BrevlO, and the Enchanted Forest Sanctuary falls within Brevl 1, but apparently
neither property has scrub-jays.
Restoration potential: At the time of the SMP, jay densities in most areas were
probably close to maximum (compare first and last data columns in Table 5-1 la).
Restored population sizes were increased slightly for Buck Lake (Brev4) and Seminole
Ranch (Brev8).
Simulation results: This metapopulation ranked 3rd in vulnerability (table 5-23),
20th in percent protected (3.6%; table 5-24), and 1st in priority (table 5-25), with high
vulnerability and high potential for improvement.

245
The small number of jays currently protected are extremely vulnerable to
extinction (p=l.0) and quasi-extinction (p=1.0; Table 5-1 la). The 30% acquisition
configuration, weighted by area, was estimated to support about 33 jay families
concentrated in only 4 patches currently protected or targeted for acquisition (Table 5-
11a). Simulations of this configuration indicated that the population is vulnerable to
quasi-extinction (Table 5-1 lb and Fig. 5-1 la; quasi-extinction = 0.40). The mean
population trajectory showed a 48.8% decline (Fig. 5-1 lb). The 30% acquisition
configuration, weighted by connectivity, was considerably worse than the area-weighted
configuration (Table 5-1 lb).
Simulations of the 70% acquisition configurations indicate that the population
would not be vulnerable to extinction and or quasi-extinction risk (Table 5-1 lb). The
mean percent population decline was considerably better for the area-weighted
configuration (20.8%) than the connectivity-weighted configuration 38.7%Table 5-1 lb).
The maximum acquisition configuration was estimated to support about 110 jay
families (Table 5-1 la). Simulations of this configuration indicate that the population
would not be vulnerable to extinction or quasi-extinction, and had a low mean percent
population decline (16.7%; Table 5-1 lb).
Recommendations: This high priority (#1) metapopulation is second-to-last in
percent protected jays, and needs substantial acquisition to adequately protect its
remaining jays. The acquisition and restoration of proposed properties listed for the 30%
area configuration (Table 5-1 la: Buck Lake, South Lake, Seminole Ranch, and portions
of Tico) is insufficient to secure this metapopulation (quasi-extinction risk = 0.40). The
viability of this metapopulation would be greatly increased by expanding the proposed

246
properties (especially around Tico and Seminole Ranch), and the acquisition and
restoration of some of the southern habitat patches (Brevl5, Brevl6, Brevl7,
Brevl8).

247
i-
Cape
Canaveral
Station
i
)ay Territory Locations
(after restoration) Protection Status
Scrub Polygons /S/ Protected -
A / UnDrotected 1 County lines o
CCC## Polygon ID / v ^Protected . ,
Jays outside of labeled, bold polygons are considered to be Suburban jays. -| 220 000
Fig. 5-1 la. N. Brevard county map 1992 1993 jay and habitat distribution.
Water Bodies Mil N. Brevard
5
10
15 Kilometi

Buck Lah
Prop.
MerrNsland N. W.R.
Cape
Canaveral
'Ar Station
Gocoa
1.75 km dispersal buffer
]ay Territory Locations
Scrub Polygons
Lo Disturbance Statuj
M Lo Density Housing [rotecte*
, Proposed
Hi Density Housing
I Ranch/Ag Water Bodies
/V Interstates
\/ State highways
County roads
.'County lines
Mil N. Brevard
1 : 220,000
12 Kilometers
Fig. 5-1 lb. N. Brevard county acquisition map.

Table 5-1 la. N. Brevard county patch statistics (number of jay territories for different configurations)
Patch id
Status
1992-1993#
No
30%
30%
70%
70%
Maximum
jay territories
acquisition
preserved
preserved
preserved
preserved
acquisition
(restored)
by contiguity
by connectivity
by contiguity
by connectivity
Brevl
2
1
1
2
Brev2
1
1
1
1
Brev3
1
1
1
1
Brev4
(Buck Lake -
6
13
6
13
13
13
proposed)
Brev5
1
1
Brev6
4
2
2
4
Brev7
South Lake
7
4
7
4
7
7
7
(partial)
Brev8
(Seminole
5
7
2
7
7
7
Ranch -
proposed)
Brev9
3
3
3
3
Brevl0
3
3
3
Brevl1
8
3
4
8
Brevl 2
(Tico partial/
21
6
4
21
12
21
proposed))
Brevl 3
3
3
3
Brevl4
1
1
Brevl5
16
3
14
9
16
Brevl 6
8
2
7
4
8
Brevl 7
5
1
5
3
5
Brevl8
6
6
Totals
101
4
33
33
77
77
110

250
Fig. 5-1 lc. N. Brevard county trajectory graphs. Top) 30% acquisition, Bottom) 70%
acquisition.

251
Fig. 5-1 Id. N. Brevard county quasi-extinction graphs. Top) 30% acquisition, Bottom)
70% acquisition.

Table 5-1 lb. N. Brevard county simulation statistics
Data type
No acquisition
30%
acquisition
by connectivity
30%
acquisition
by area
70%
acquisition
by connectivity
70%
preserved
by area
Maximum
acquisition
starting
population size
4
33
33
77
77
110
x end pop. size
5.5
16.9
47.2
61.0
91.5
s.d.
4.6
7.4
19.2
12.6
11.4
percent
decline
83.3
48.8
38.7
20.8
16.7
extinction
risk
0.37
0.0
0.0
0 0
0.0
quasi-extinction
risk (10 pairs)
0.90
0.40
0.0
0.0
0.0

253
Central Brevard (Ml2)
General description: The Central Brevard metapopulation is separated from the N.
Brevard metapopulation (Ml 1) by the city of Cocoa, from the S. Brevard- Indian River-
N. St. Lucie metapopulation (Ml 3) by the city of Melbourne, and from the Merritt Island
metapopulation (M10) to the east by the Indian River (see maps in Fig. 5-12a, b). The
SMP documented about 36 jay territories, excluding suburban jays, in this
metapopulation. Estimated potential population size after habitat restoration and full
occupancy is 5 pairs in currently protected areas, and 40 pairs maximum.
Protected areas: Rockledge Scrub Preserve (Brev40); Wickham County Park
(Brev43). Portions of the large, contiguous habitat patch (Brev41) just south of
Rockledge are targeted for acquisition (1999 Carl project), as is habitat (Brev42 -
CARL 1999 site) just north of Wickham County Park. The Melbourne regional airport
(Brev29), which lacks a habitat management plan, was not included as a protected area.
Restoration potential: At the time of the SMP, jay densities in most areas probably
were close to maximum (compare first and last data columns in Table 5-12a).
Simulation results: This metapopulation ranked 4th in vulnerability (table 5-23),
18th in percent protected (12.5%; table 5-24), and 4th in priority (table 5-25), with high
vulnerability and high potential for improvement. The no acquisition option has an
extremely high probability of extinction (p=1.0) and quasi-extinction (p=1.0). The 70%
acquisition by area has a high quasi-extinction risk (p=0.43) and a moderate extinction
risk (p=0.10). Risk estimates for the maximum acquisition option are substantially
reduced for both quasi-extinction (p=0.0) and extinction (p=0.10), as is percent
population decline (see Table 5-12b).

Recommendations: Although afforded little protection, the viability of this
metapopulation could be greatly increased through acquisition of the few remaining
254
habitat patches. The long term viability of this metapopulation and the small Rockledge
Scrub Preserve (Brev40) depends critically on substantial acquisition and restoration of
habitat at Brev41 (EELS/CARL 1999 site). The small population at Wickham County
Park (Brev43) would benefit greatly from proposed acquisition of habitat just to the
north (Brev42 1999 CARL). A habitat management plan is needed for the jays at
Melbourne Regional Airport (iBrev44).

Jay Territory Locations
1.75 km dispersal buffer
Scrub Polygons
3 Lo Disturbance
~7~\ Lo Density Housing 3 Prtected
Vm H¡ Density Housing ProPsed
kanch/Ag Water Bodies
Protection Status
A / Interstates
State highways
County roads
' County lines
Ml2 Central Brevard
1 : 160,000
9 Kilometers
Fig. 5-12a. Central Brevard county map 1992 1993 jay and habitat distribution.

256
Water Bodies
Jay Territory Locations
(after restoration)
_ Scrub Polygons
CCC## Polygon ID
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Protection Status
/V/ Protected .
s\/ Unprotected / V County lines
Ml2 Central Brevard
0 3 6 9 Kilometers
1 : 160,000 Hlh
Fig. 5-12b. Centra] Brevard county acquisition map.
Brev42
Brev43
Wickham
County
Brev44
Reg.

257
Table 5-12a. Central Brevard county patch statistics (number of jay territories for
different configurations)
Patch id
Status
1992-
No
70%
Maximum
1993# jay
acquisition
preserved
acquisition
territories
(restored)
by area
Brev40
Rockledge Scrub Pr.
3
3
3
3
Brev41
(DRI/EELS -
proposed)
15
16
19
Brev42
(Wickham Rd. CARL
site)
9
9
9
Brev43
Wickham County Pk.
2
2
2
2
Brev44
Melbourne regional
airport
7
7
Totals 36 5 30 40

258
Year
10 0
20 40 60
Year
Fig. 5-12c. Central Brevard county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.
I

Probability
259
ro ;
I*
L
0
>.
S L
15
n g-r
£ h
0- |.
0 4*f
0.2*1
"
(
2 4 6
Threshold Pop Size
Fig. 5-12d. Central Brevard county quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition.

Table 5-12b. Central Brevard county simulation statistics
Data type
No acquisition
70% acquisition
by area
Maximum
acquisition
starting
population size
5
30
40
x end pop. size
0
15.0
25.5
s.d.
0.0
6.4
8.8
percent
decline
100.0
50.0
36.3
extinction
risk
1.0
0.10
0.00
quasi-extinction
risk (10 pairs)
1.0
0.43
0.10

261
S. Brevard-Indian River-N. St. Lucie (Ml31
General description: The S. Brevard-Indian River-N. St. Lucie metapopulation is
separated from the Central Brevard metapopulation (Ml 2) by the city of Melbourne,
from the St. Lucie metapopulation (Ml4) to the south by Fort Pierce, and from the
Merritt Island metapopulation (M10) to the east by the Indian River (see maps in Fig. 5-
13a,b). The SMP documented about 153 jay territories, excluding suburban jays, in this
metapopulation. Estimated potential population size after habitat restoration and full
occupancy is 62 pairs in currently protected areas, and 165 pairs maximum.
Protected areas: Malabar Scrub Preserve (Brev30), Valkaria Scrub Preserve
(Brev32/32a), St. Sebastian River State Buffer Preserve (Brev35), private HCP
(InRi3), Wabassa Scrub Preserve (InRi3). Since the SMP, an additional 12 territories
were discovered in the N. Indian River county portion of the St. Sebastian River State
Buffer Preserve (Dave Breininger, pers. comm.). Two jay populations on airports
(Sebastian Municipal; St. Lucie County) were treated as unprotected due to lack of
habitat management.
Restoration potential: The restoration potential for most patches in this
metapopulation probably is not significantly greater than the jay densities measured for
the SMP (compare first and last data columns in Table 5-13a).
Simulation results: This metapopulation ranked 15th in vulnerability (table 5-23),
13th in percent protected (37.6%; table 5-24), and 11th in priority (table 5-25), with
moderate vulnerability and moderate potential for improvement. Simulations of the no
acquisition configuration indicate that currently protected jays are vulnerable to quasi-

262
extinction (Table 5-13b and Fig. 5-13f; quasi-extinction risk = 0.167) and show a
substantial population decline (Table 5-13b and Fig. 5-13e; mean ending population size
= 32.5; percent decline = 50.0). Intermediate configurations all show a substantial
reduction in quasi-extinction, but only the 70% preserved by area and maximum
acquisition have no quasi-extinction risk (Table 5-13b).
Recommendations: Comparison of simulations with equal population size but
different spatial configuration (area vs. connectivity) indicates that maintaining contiguity
of territories is more important than maintaining connectivity. Given this criteria,
unprotected patches such as Jordan (Brev31), Valkaria (Brev32), Babcock
(Brev38), and Brev36 should be high priority acquisition sites. South of Sebastian
along the coast the jay populations are in small, isolated populations that are extinction-
prone. Acquisition of habitat (e.g. InRi5) near the Wabasso Scrub Preserve (InRi4)
may bolster the long-term viability of that population. Significant numbers of unprotected
jays occur along the Ten Mile Ridge in Indian River county (InRi9 and InRilO), and
other populations likely occur nearby (Breininger 1998), making this an important area
for future acquisition. Habitat management plans are needed at the 3 airports known to
have jays within this metapopulation (St. Lucie County, Sebastian Municipal, and
Valkaria).
Recent surveys and color-band studies of Brev30, Brev31, Brev32,
Brev35, Brev36, Brev37, and Brev38 by Breininger (1998) documented an
alarming population decline exceeding 50% since 1993. This decline is due primarily to
habitat degradation resulting from fire suppression (Breininger 1998). An epidemic in
late 1997-early 1998 also may have had a significant effect in this region (Breininger

263
1998). This population decline is not predicted by the model, and illustrates clearly the
influence of the model parameter settings on the simulation results, which assume
optimal habitat conditions. Similar declines likely are occurring in many other parts of
the state, and highlight the importance of habitat restoration and management; land
acquisition alone is insufficient to preserve jay populations.

264
Matebcf^
Scrub Rr.
'aJkariaSScrub Sand
Micco Scrub Sartaii
Wabasso
Scrub Pr
T"111!!:!'--x'
SsSS$
Indian River
St.-H.ucie
)ay Territory Locations
1 75 km dispersal buffer
Scrub Polygons
^ Lo Disturbance
Lo Density Housing
tWj Hi Density Housing
Ranch/Ag
Protection Status
TW: Protected
Proposed
Water Bodies
/ \ f Interstates
State highways
County roads
County lines
Ml 3 S. Brevard-lndian River-
1 : 270.000
Kilometer!

Fig. 5-13a. S. Brevard-lndian River-N. St. Lucie Metapopulation county map 1992 -
1993 jay and habitat distribution.

265
Maiabaf^
Scrub Pr,
'fiT^lQrdan-)$rop
(alkari&gcrub Sana
Wabasso
Scrub Pr
mgtsQ ^
InRittS*
i)
lnR7(2
Indian River lnRi80'
St Lucie
)ay Territory Locations
(after restoration)
Scrub Polygons
CCC## Polygon ID /V
Jays outside of labeled, bold polygons are considered to be Suburban jays
Protection Status
/V Protected
Unprotected
Water Bodies
County lines
Ml3 S. Brevard-lndian River-
0 5 10 15 Kilometers
1 : 270,000 ^
Fig. 5-13b. S. Brevard-lndian River-N. St. Lucie county acquisition map.

Table 5-13a. S. Brevard-Indian River-N. St. Lucie county patch statistics (number of jay territories for different configurations)
Patch id
Status
1992-
No
30%
30%
70%
70%
Maximum
1993# jay
acquisition
preserved
preserved
preserved
preserved
acquisition
territories
(restored)
by connectivity
by area
by connectivity
by area
Brev30
Malabar Scrub Sanct.
10
10
10
10
10
10
10
Brev31
Jordan (proposed)
23
4
5
11
16
23
Brev32a
Valkaria (partial)
7
7
7
7
7
7
7
Brev32
(Valkaria proposed)
17
9
17
17
Brev33
5
3
5
2
5
Brev34
8
5
7
8
Brev35
St. Sebastian River
St. Pk. & Micco Scrub
Sanct.
28
28
28
28
28
28
28
Brev36
11
3
11
6
11
11
Brev37
2
2
2
2
2
Brev38
(Babcock proposed)
6
2
6
5
6
6
InRil
St. Sebastian River
State Park
12
12
12
12
12
12
lnR¡2
Sebastian municipal
airport
8
5
8
8
8
8
lnRi3
Private HCP
3
3
3
3
3
3
3
lnRi4
Wabasso Scrub Pr.
2
2
2
2
2
2
2
lnRi5
4
2
4
lnRi6
5
2
5
lnR¡7
1
1
1
lnRi8
2
2
2
lnRi9
5
3
5
5
5
InRilO
7
5
7
7
StLu1
St. Lucie airport
3
2
3
StLu2
1
1
1
Totals
153
62
91
91
129
136
165
266

Population Size
267
20
Year
40
60
j
Fig. 5-13c. S. Brevard-Indian River-N. St. Lucie county trajectory graphs. Top) no
acquisition, Bottom) 30% acquisition by area.

268
1 o
O 8
r
60
10
2
g
2
o
£
0 8
0 6
04 _
0 2
20
40
Threshold Pop. Size
60
80
Fig. 5-13d. S. Brevard-Indian River-N. St. Lucie county quasi-extinction graphs. Top) no
acquisition, Bottom) 30% acquisition by area.

Table 5-13b. S. Brevard-Indian River-N. St. Lucie county simulation statistics.
Data type
No
acquisition
30% preserved
by connectivity
30% preserved
by area
70% preserved
by connectivity
70% preserved
by area
Maximum
acquisition
Starting
population size
62
91
91
136
136
165
Mean ending
population size
28.5
44.3
71.9
91.1
107.3
124.1
s.d.
13.5
17.9
13.9
21.4
15.3
15.0
Percent
population
decline
54.0
51.3
21.0
33.0
21.1
24 8
Extinction
risk
0.07
0.0
0.0
0.0
0.0
0.0
Quasi
extinction
Risk (10 pairs)
0.20
0.07
0.03
0.03
0.0
0.0

270
St. Lucie N. Martin (Ml4)
General description: The St. Lucie-N. Martin metapopulation is separated from
the S. Brevard-Indian River-N. St. Lucie metapopulation (Ml 3) by the city of Fort Pierce
to the north, and the Martin-N. Palm Beach metapopulation (Ml5) by the St. Lucie Inlet
to the south (see map in Fig. 5-14a,b). The SMP documented about 28 jay territories,
excluding suburban jays, in this metapopulation. Estimated potential population size after
habitat restoration and full occupancy is 23 pairs in currently protected areas, and 33
pairs maximum.
Protected areas: Savannas State Park (Stl4), and portions of the S. Savannas
CARL site (Marl).
Restoration potential: The densities of jays measured by the SMP probably were
close to maximum, even though habitat conditions were not optimal. For modeling
purposes, the only population that was increased over the SMP was at Savannas State
Park (15 pairs increased to 20).
Simulation results: This metapopulation ranked 12th in vulnerability (table 5-23),
7th in percent protected (62.2%; table 5-24), and 7th in priority (table 5-25), with high
vulnerability and high potential for improvement. Quasi-extinction and extinction risk
was substantially higher for the no acquisition option (p=0.73 and 0.20 respectively)
compared to the the maximum acquisition option (p=0.27 and 0.03 respectively), even
though the difference in population size was small (14 territories; Table 5-14b).
Recommendations: Habitat restoration and proper management of the Savannas
State Park is crucial to this metapopulation. Acquisition of jay habitat within and south of
the S. Savannas CARL site (Marl, Mar2, Mar3, Mar4) will substantially

271
improve the long term prospects for this metapopulation. The status (and existence?) of
the habitat patch (Stl3) north of Savannas State Park and east of the county-owned
Savannas Outdoor Recreation Area should be investigated.

272
]ay Territory Locations 1.75 km dispersal buffer
Scrub Polygons
Lo Disturbance Protection Status A /Interstates
2] Lo Density Housing L......1 Protected /\/ State highways
Hi Density Housing I Pr0Psed County roads
M Ranch/Ag Water Bodies County lines
Ml 4 St. Lucie
0 4 8 12
1 : 210.000
Kilometers
+
Fig. 5-14a. St. Lucie N. Martin county map 1992 1993 jay and habitat distribution.

273
Pk.
Protection Status
/\y Protected
/\y Unprotected
Water Bodies Ml 4 St. Lucie
)ay Territory Locations
(after restoration)
Scrub Polygons
CCC## Polygon ID
Jays outside of labeled, bold polygons are considered to be Suburban jays.
County lines
12 Kilometers
1 : 210,000
Fig. 5-14b. St. Lucie N. Martin county acquisition map.

274
Table 5-14a. St. Lucie N. Martin county patch statistics (number of jay territories for
different configurations)
Patch id
Status
1992-1993# jay
territories
No acquisition
(restored)
Maximum
acquisition
StL3
5
5
StL4
Savannas St. Pk.
15
20
20
Marl
Private reserve
3
3
3
Mar2
1
1
Mar3
3
3
Mar4
1
1
Totals
28
23
33

275
0>
N
(/)
C
o
76
3
Q.
o
CL
*-
i-
10 0~
r
p
5 0 ^
L
20

40
60
Year
-
L
20
40
Year
60
Fig. 5-14c. St. Lucie N. Martin county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.
j

Probability
10
Threshold Pop. Size
20
30
Fig. 5-14d. St. Lucie N. Martin county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition.

277
Table 5-14b. St. Lucie county simulation statistics
Data type
No
acquisition
Maximum
acquisition
starting
population size
23
33
x end pop. size
9.8
19.4
s.d.
5.9
7.1
percent
decline
57.4
41.2
extinction
risk
0.20
0.03
quasi-extinction
risk (10 pairs)
0.73
0.27

278
Martin and N. Palm Beach (Ml 5)
General description: The Martin-N. Palm Beach metapopulation is isolated from
the St. Lucie-N. Martin metapopulation (Ml4) to the north by the St. Lucie Inlet, and is
isolated from the South Palm Beach metapopulation (Ml6) to the south by heavy
urbanization associated with West Palm Beach, Palm Springs, and Lake Worth. The SMP
documented about 115 jay territories, excluding suburban jays, in this metapopulation.
Estimated potential population size after habitat restoration and full occupancy is 85 pairs
in currently protected areas, and 120 pairs maximum.
Protected areas: Willoughby Dev. Preserve (Mar7), Seabranch State Park
(MarlO), Jonathan Dickinson State Park (Marl2), F.P.L. Hill Station Preserve
(PB1), St. Jude Scrub Jay Preserve (PB3), Jupiter Inlet Natural Area Preserve
(PB4), Carlin County Park (PB5), Jupiter Ridge Natural Area (PB6), Juno Hills
Natural Area Preserve (PB8), F.P.L. Universe Scrub Preserve (PB9). Portions of
Marl 1 just north of Jonathan Dickinson State Park have been preserved through an
HCP.
Restoration potential: The SMP estimates for Jonathan Dickinson State Park and
Sea Branch State Park probably were erroneously high, but correspond well with what
the densities would be after restoration and full occupancy. Restoration at Juno Hills
Natural Areas Preserve might increase this population from the SMP estimate of 9 pairs
to 14 pairs (Grace Iverson, pers. comm.).
Simulation results: This metapopulation ranked 18th in vulnerability (table 5-23),
4th in percent protected (70.8%; table 5-24), and 18th in priority (table 5-25), with low
vulnerability and low potential for improvement. Extinction and quasi-extinction risk was

279
low, even for the no acquisition option. However, the mean percent population decline
was substantially better for the maximum acquisition option (Table 5-15b).
Recommendations: The relatively favorable ranking of this metapopulation is due
mainly to the significant jay populations at Jonathan Dickinson State Park, Sea Branch
State Park, and Juno Hills Natural Area Preserve. The habitat quality for jays at all three
of these parks reportedly is poor; restoration and proper management at these sites is vital
to the viability of this metapopulation. Acquisition of unprotected habitat patches
(Marl 5, PB6, PB7) likely is important to the viability of nearby populations of
jays that are already protected (e.g. Jupiter Ridge Natural Area, St. Jude Scrub Jay
Preserve).

280
St.Luoie
Martin \
> \
Seabranch St Pk.
Dickinso
Preserve
Jay Preserve
County Pk.
Ridge Natural Area
Hills Natural Area
Universe Scrub Pr.
]ay Territory Locations
Scrub Polygons
Lo Disturbance
^ Lo Density Housing
Hi Density Housing
Ranch/Ag
1.75 km dispersal buffer
Protection Status
|||1-1 Protected
H Proposed
Water Bodies
/\J Interstates
State highways
County roads
County lines
Ml 5 Martin St N. Palm Beach
0 5 10 15 Kilometers
1 : 250,000
Fig. 5-15a. Martin and N. Palm Beach county map 1992 1993 jay and habitat
distribution.

281
St Lucie
Martin
Seabranch St. Pk.
Jonathan Dickinson St. Pk.
Hills Natural Area
L. Universe Scrub Pr.
Station Preserve
Scrub Jay Preserve
]ay Territory Locations
(after restoration) Protection Status
3 Scrub Polygons /\/ Protected
CCC## Polygon ID / V UnProtected
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Water Bodies
County lines
M 1 5 Martin SC N.
0 5 10
1 : 250,000
Palm Beach
15 Kilometers
Fig. 5-15b. Martin and N. Palm Beach county acquisition map.

282
Table 5-15a. Martin and N. Palm Beach county patch statistics (number of jay territories
for different configurations)
Patch id
Status
1992-1993#
No
Maximum
jay
acquisition
acquisition
territories
(restored)
Mar5
1
1
Mar6
2
2
Mar7
Willoughby Dev. Pr.
3
3
3
Mar8
4
4
Mar9
2
2
Mar10
Seabranch St. Pk.
15
15
15
Mar11
(HCP)
12
12
Mar12
Jonathan Dickinson
St. Pk.
40
40
40
Marl 3
1
1
Mar14
1
1
Marl 5
4
4
PB1
FPL Hill Station Pr.
1
1
1
PB2
(Tequesta Water
Dept. proposed)
1
1
PB3
St. Jude Scrub Jay
Pr.
3
2
2
PB4
Jupiter Inlet Natural
Area Pr.
3
4
4
PB5
Carlin County Pk.
3
3
3
PB6
Jupiter Ridge Nat.
Area
4
4
4
PB7
(Radnor proposed)
4
4
PB8
Juno Hills Natural
Area Pr.
9
14
14
PB9
FPL Universe Scrub
Pr.
2
2
2
Totals
115
85
120

Population Size
283
40 0
u
F
20 0
20 40 60
Year
h
t
20
:
40
Year
J
60
Fig. 5-15c. Martin and N. Palm Beach county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition.

284
z
1
£
1 O
O 8
O 6
O 4
02
/ i i
/
20
40
Threshold Pop. Size
60
-|
80
1 0
0 8
* 06
0 4
02
I
I
I
i
50 100
Threshold Pop Size
Fig. 5-15d. Martin and N. Palm Beach county quasi-extinction graphs. Top) no
acquisition, Bottom) maximum acquisition.

Table 5-15b. Martin and N. Palm Beach county simulation statistics
Data type
No
acquisition
Maximum
acquisition
starting
population size
85
120
x end pop. size
70.3
111.3
s.d.
10.0
8.6
percent
decline
17.3
7.5
extinction
risk
0.0
0.0
quasi-extinction
risk (10 pairs)
0.0
0.0

286
South Palm Beach (Ml6)
General description: The South Palm Beach metapopulation is the most southerly
metapopulation on the Atlantic coast. It is isolated from the Martin county
metapopulation (Ml5) to the north by heavy urbanization associated with West Palm
Beach, Palm Springs, and Lake Worth. All of the scrub patches are small and occur in
suburban or urban settings. The SMP found 8 groups of jays, and characterized the
condition of the scrub to be severely overgrown. The number of jays present during the
SMP was only one-third the number found by Cox in 1980 (Pranty et al. manuscript).
Estimated potential population size after habitat restoration and full occupancy is 9 pairs
in currently protected areas, and 16 pairs maximum. Grace Iverson, who has studied jays
in Palm Beach county for a number of years, provided invaluable information on this
metapopulation.
Protected areas: Rolling Green Scrub Preserve (PB11), Galaxy School Scrub
Preserve (PB12), Yamato Scrub NAP (PB14). A number of small scrub preserves
that did not have jays during the SMP were excluded from all simulations (Osborne
Scrub NAP, Gopher Tortoise Scrub NAP, Rosemary Ridge Scrub NAP, Leon Weeks
Scrub Preserve NAP, Seacrest Scrub NAP, Rosemary Scrub NAP).
Restoration potential: The Yamato Scrub NAP (PB14) had only 1 pair of jays
during the SMP, but is estimated to potentially support about 6 pairs of jays if fully
restored and managed.
Simulation results: This metapopulation ranked 8th in vulnerability (table 5-23),
6th in percent protected (69.2%; table 5-24), and 14th in priority (table 5-25), with high

287
vulnerability and low potential for improvement. Simulations of the no acquisition and
maximum acquisition option both show a high quasi-extinction risk (p=l .0 for both)
and a high extinction risk (p=0.90 and 0.77 respectively).
Recommendations: Because of the small size of this metapopulation and its
individual patches, and the heavily urbanized landscape which subjects these jays to
additional sources of mortality, the long-term prognosis for this metapopulation is poor.
An experimental program involving intensive human intervention might be necessary to
maintain this metapopulation. Such a program likely would involve intensive habitat
management, food supplementation, predator control, control of vehicular speed, and
translocation of jays to supplement local population declines. No such program has been
attempted for scrub-jays, but because of the huge human population in this area which
could support and benefit from such a progam, this metapopulation might be the best
candidate for such an experiment.
The two most significant habitat patches that remain unprotected include the
Overlook Scrub (PB14), and the Tradewind / Winchester Site (PB13). Both of these
patches occur near the already protected Rolling Green Scrub Preserve (PB11) and
Galaxy School Scrub Preserve (PB12). Acquisition and restoration of both of these
sites would benefit the two nearby protected areas.

288
Scrub Polygons
Lo Disturbance Prtectc>n Status A/ Interstates
3 Lo Density Housing i.l.'lD Protectec* /\/ Sute highways
If-V/j Hi Density Housing i ProPse^ County roads
Ranch/Ag Water Bodes ' ' County lines
Ml 6 S. Palm Beach
0 2 4 6
1 : 120,000
Kilometers
Fig 5-16a. Central Palm Beach county map 1992 1993 jay and habitat distribution.

289
tijok Scrub
Jay Territory Locations
(after restoration) Protection Status
~ Scrub Polygons /\/ Protected
CCC## Polygon ID /v Unprotected
Water Bodies Ml6 S. Palm Beach
' County lines 0 2 4 6 Kilometers
Jays outside of labeled, bold polygons are considered to be Suburban jays. -\ -|20 000
Fig. 5-16b. Central Palm Beach county acquisition map.

290
Table 5-16a. South Palm Beach county patch statistics (number of jay territories for
different configurations)
Patch id
Status
1992-1993#
jay territories
No acquisition
(restored)
Maximum
acquisition
PB10
(Overlook Scrub -
proposed)
2
5
PB11
Rolling Green Scrub
Pr.
2
2
2
PB12
Galaxy School
Scrub Pr.
1
1
1
PB13
(Tradewind /
Winchester Site)
2
2
PB14
Yamato Scrub NAP
1
6
6
Totals
8
9
16

291
Fig. 5-16c. South Palm Beach county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.

Probability Probability
292

-
0

0.27
f
r
5 10 15
Threshold Pop Size
Fig. 5-16d. South Palm Beach county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition.

Table 5-16b. South Palm Beach county simulation statistics
Data type
No
acquisition
Maximum
acquisition
starting
population size
9
16
x end pop. size
0.47
1.20
s.d.
1.34
2.40
percent
decline
94.8
91.0
extinction
risk
0.90
0.77
quasi-extinction
risk (10 pairs)
1.0
1.0

294
Ocala National Forest (Ml7)
General description: The Ocala National Forest (ONF) metapopulation occupies
most of W. Marion county and small portions of northeast Lake county. It is separated
from the Central Lake metapopulation (Ml 8) by major lakes (Lake Apopka, Lake Harris,
Lake Dora, Lake Eustis, Lake Griffm, Lake Yale). The northeast Lake metapopulation
(Ml8) to the southeast is separated from the ONF metapopulation by more than 30 km,
with dense forest stands in between.
Protected areas: Most of this metapopulation is protected, occurring within the
Ocala National Forest. During the SMP, an incomplete survey of the OCF population
estimated this population at about 448 pairs. More recent surveys have increased this
number considerably, to about 727 pairs (Laura Lowrie, pers. comm.).
Restoration potential: The restoration potential of the OCF population is
enormous, since most of the extensive sand pine forest that is currently unoccupied could
be restored to jay habitat. For modeling purposes, the population estimates given for the
SMP were used (448 pairs).
Simulation results: This metapopulation ranked 20th in vulnerability (table 5-23),
1st in percent protected (table 5-24), and 20th in priority (table 5-25), with low
vulnerability and low potential for improvement. Only a single simulation was run for
this metapopulation, which assumed a starting population size of 470 pairs (Table 5-17a).
This configuration had no risk of extinction or quasi-extinction, and showed a 25% mean
population decline.
Recommendations: Despite the unusual management practice on ONF of creating
temporary scrub jay habitat in small clearcuts within this extensive sandpine forest,

295
declines in jay populations have not been documented (Laura Lowrie, pers. comm.).
Efforts are now being made to locate new clearcuts adjacent to recent openings to reduce
fragmentation. The creation of a proposed 1900 acre parcel managed for scrub-jays and
other fire-dependent scrub species should be of great benefit to this metapopulation.
Three small, unprotected jay populations occur outside the southwest portion of
the ONF.

296
Jay Territory Locations 1 .75 km dispersal buffer
Scrub Polygons
Lo Disturbance
Lo Density Housing
Hi Density Housing
Ranch/Ag
Protection Status
| J Protected
j \ Proposed
Water Bodies
Interstates
State highways
County roads
' County lines
Ml 7 Ocala National Forest
0 6 12 18 Kilometers
1 : 320,000
Fig. 5-17a. Ocala National Forest county map 1992 1993 jay and habitat distribution.

297
]ay Territory Locations
(after restoration)
Scrub Polygons
CCC## Polygon ID / V
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Protection Status
/\/ Protected
A/ Unprotected
Water Bodies
County lines
Ml 7 Ocala National Forest
0 6 12 18 Kilometers
1 : 320,000
Fig. 5-17b. Ocala National Forest county acquisition map.

298
Table 5-17a. Ocala National Forest county patch statistics (number of jay territories for
different configurations)
Patch id Status 1992-1993#
jay territories
Marl 5
Ocala National
448
Forest
Mar16
9
Mar17
6
Lake5
7
Totals
470

Population Size
299
Fig. 5-17c. Ocala National Forest county trajectory graphs. No acquisition.

Probability
08
(-
o 6~r
04
-
i-
i i i i i j
100
I
;
200 300
Threshold Pop Size
400
Fig. 5-17d. Ocala National Forest county quasi-extinction graphs. No acquisition.

Table 5-17b. Ocala National Forest county simulation statistics
Data type
Original
1992-1993
scenario
starting
population size
470
x end pop. size
s.d.
352.5
65.7
percent
decline
25.0
extinction
risk
0.0
quasi-extinction
risk (10 pairs)
0.0

302
N.E. Lake (Ml8)
General description: The N.E. Lake metapopulation is separated from the W.
Volusia metapopulation (Ml9) by the heavily wooded St. Johns riverine system to the
west. The ONF metapopulation (Ml7) to the northwest is separated from the N.E. Lake
metapopulation by more than 30 km, with an intervening matrix of dense forest stands.
The SMP documented about 109 jay territories, excluding suburban jays, in this
metapopulation. Estimated potential population size after habitat restoration and full
occupancy is 67 pairs in currently protected areas, and 161 pairs maximum.
Protected areas: Ocala N.F.(Lakel 1), Seminole S. F. (Lake7, Lake8), Rock
Springs Run S.R. (Ora2, Lake6), Wekiwa Springs S.P. (Semi, Oral), Wekiva R.
Buffers C.A. (Semi), Yankee Lake Waste Water (Sem3).
Restoration potential: Habitat patches in this metapopulation are heavily
overgrown, and have an enormous potential for restoration. The largest patches of
occupied habitat currently are unprotected (Lake 10, Lake9), and could support many
more jays than were found during the SMP (compare the second and last data columns in
Table 5-18a).
Simulation results: This metapopulation ranked 17th in vulnerability (table 5-23),
11th in percent protected (41.6%; table 5-24), and 17th in priority (table 5-25), with low
vulnerability and low potential for improvement. The no acquisition option had a low
risk of quasi-extinction (p=0.03), no extinction risk, and a 33.7% mean percent
population decline (Table 5-18b). The maximum acquisition option has no risk of
extinction or quasi-extinction, and a 10.9% mean percent population decline (Table 5-
18b).

303
Recommendations: Habitat restoration is needed within most of the protected
areas, and could substantially increase the size of this metapopulation.
More than half of the jays found during the SMP were in two large unprotected
patches (Lake9, Lake 10), within or near the Royal Trails development. Acquisition
of major portions of these patches would substantially improve the stability of this
metapopulation.

304
Jay Territory Locations
Scrub Polygons
1.75 km dispersal buffer
Lo Disturbance
3 Lo Density Housing
Hi Density Housing
Ranch/Ag
Protection Status
mm Protected
Proposed
Water Bodies
/\/ Interstates
State highways
County roads
. County lines
Ml 8 N.E. Lake
1 : 200,000
12
Kilometers
+
Fig. 5-18a. N.E. Lake county map 1992 1993 jay and habitat distribution.

305
jay Territory Locations
(after restoration)
Scrub Polygons
CCC## Polygon ID
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Protection Status
/V Protected
/\y Unprotected
Water Bodies
' County lines
Ml 8 N.E. Lake
1 : 200,000
12
Kilometers
4"
Fig. 5-18b. N.E. Lake county acquisition map.

306
Table 5-18a. N.E. Lake county patch statistics (number of jay territories for different
configurations)
Patch id
Status
1992-1993# jay
territories
No
acquisition
(restored)
Maximum
acquisition
Lak6
Rock Springs Run S.R.
1
1
1
Lak7
Seminole S. F.
7
15
15
Lak8
Seminole S. F.
7
15
15
Lak9
56
71
Lak10
13
20
Lak11
Ocala N.F.
5
5
5
Oral
Wekiwa Springs S.P.
1
1
1
Ora2
Rock Springs Run S.R.
9
19
19
Semi
Wekiva R. Buffers C.A.
1
1
1
Sem2
2
2
Sem3
Yankee Lake Waste Water
6
10
10
Sem4
1
1
Totals
109
67
161

Population Size
307
a>
N
CO
Cl
5
Q.
20 40 60
Year
Fig. 5-18c. N.E. Lake county trajectory graphs. Top) no acquisition, Bottom) maximum
acquisition.

Probability
308
Threshold Pop Size
[
on]
02T
l-
(-
L
50
100
150
Threshold Pop. Size
Fig. 5-18d. N.E. Lake county quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition.

309
Table 5-18b. N.E. Lake county simulation statistics
Data type
No
acquisition
Maximum
acquisition
starting
population size
67
161
x end pop. size
44.4
143.8
s.d.
12.9
14.5
percent
decline
33.7
10.9
extinction
risk
0.0
0.0
quasi-extinction
risk (10 pairs)
0.03
0.0

310
S.W. Volusia (Ml91
General description: The S.W. Volusia metapopulation is separated from the N.E.
Lake metapopulation (Ml 8) to the west by the St. Johns riverine system. The SMP
documented about 54 jay territories, excluding suburban jays, in this metapopulation.
Estimated potential population size after habitat restoration and full occupancy is 17 pairs
in currently protected areas, and 70 pairs maximum.
Protected areas: The only protected jays occur on the Blue Springs State Park
(VollO in Fig. 5-19b); a single family was found during the SMP.
Restoration potential: For modeling purposes, Blue Springs State Park was
estimated to support 17 families of jays after restoration (probably an overly optimistic
estimate). The largest population of jays occurs on the unprotected Stewart Ranch
(Voll9), and this population likely could support more jays, but additional information
is needed. Other patches in this metapopulation occur in the rapidly developing Deltona
area south of Deland. The scrub in this area is heavily overgrown and the restoration
potential is unknown.
Simulation results: This metapopulation ranked 11th in vulnerability (table 5-23),
15th in percent protected (24.3%; table 5-24), and 4th in priority (table 5-25), with high
vulnerability and high potential for improvement. The no acquisition option had a high
risk of extinction (p=0.33) and quasi-extinction risk (p=0.90), and a 54.9% mean
population decline (Table 5-19b). The maximum acquisition option had no risk of
extinction or quasi-extinction, and a 20.2% mean population decline (Table 5-19b).

311
Recommendations: Near absence of protection for jays in this area combined with
high potential to increase the protected population make this metapopulation high on the
priority list. Three key responses to this situation are suggested First, since Blue
Springs is the only protected area with jays (one pair known from the SMP), all jay
habitat should be restored as quickly as possible. Improved habitat data is needed to
estimate the restoration potential of Blue Springs State Park. Second, unprotected,
contiguous scrub habitat occurs north and east of the park (Fig. 5-19a); acquisition and
restoration of these areas would bolster the local jay population.. Third, acquisition or
protection of the large population of jays on the Stewart Ranch (Voll9 in Fig. 5-19b) is
critically important to this metapopulation.

312
Jay Territory Locations
1.75 km dispersal buffer
Scrub Polygons
[ I Lo Disturbance
K 7 -1 Lo Density Housing IsHlIi ^rotectec*
IfV'/j Hi Density Housing Proposed
Water Bodies
Protection Status
/\/ Interstates
v State highways
County roads
County lines
Ml 9 W. Volusia
1 : 140,000
9 Kilometers
Fig. 5-19a. S.W. Volusia county map 1992 1993 jay and habitat distribution.

313
Water Bodies Ml9 W. Volusia
]ay Territory Locations
(after restoration) Protection Status
Scrub Polygons /\/ Protected
CCC## Polygon ID /''/ Unprotected
Jays outside of labeled, bold polygons are considered to be Suburban jays.
County lines
9 Kilometers
1 : 140,000
Fig. 5-19b. S.W. Volusia county acquisition map.

Table 5-19a. S.W. Volusia county patch statistics (number of jay territories for different configurations)
Patch id
Status
1992-1993#
No
30%
30%
70%
70%
Maximum
jay territories
acquisition
preserved
preserved
preserved
preserved
acquisition
(restored)
by contiguity
by connectivity
by contiguity
by connectivity
VoMO
Blue Springs
1
17
17
17
17
17
17
S.P.
Vol11
7
3
3
7
5
7
Vol12
2
1
1
2
2
2
Vol13
2
1
1
1
2
2
Vol14
5
2
2
4
5
Vol15
4
1
1
3
4
Vol16
2
1
1
1
2
Vol17
4
1
1
3
4
Vol18
3
2
2
3
2
3
Vol19
Stewart Ranch
(proposed)
24
4
4
24
15
24
Totals
54
17
33
33
54
54
70

Population Size
315
r
l
40
60
20
Year
Fig. 5-19c. S.W. Volusia county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition

316
-
o.2~r
5 10 15
Threshold Pop. Size
Fig. 5-19d. S.W. Volusia county quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition

317
Table 5-19b. S.W. Volusia county simulation statistics
Data type
No
acquisition
(restored)
Maximum
acquisition
Starting
population size
17
70
Mean ending
population size
7.67
55.87
s.d.
6.22
13.3
% population
decline
54.9
20.2
Extinction
Risk
0.33
0.00
Quasi
extinction
Risk (10 pairs)
0.90
0.00

318
Central Lake (M20)
General description: The Central Lake metapopulation (M20) is south of the
Ocala National Forest metapopulation (Ml 7) and west of the N.E. Lake metapopulation
(Ml8). Major lakes in this area (Lake Apopka, Lake Harris, Lake Dora, Lake Eustis,
Lake Griffin, Lake Yale) probably isolate these jays from the Ocala population to the
north. The once extensive native upland areas in this region were converted decades ago
to agriculture, and little native habitat remains. Only a few small patches of scrub remain,
and some jays are living in scattered groups within abandoned citrus groves. The SMP
documented about 13 jay territories, excluding suburban jays, in this metapopulation.
Estimated potential population size after habitat restoration and full occupancy is 0 pairs
in currently protected areas, and 20 pairs maximum.
Protected areas: No jays occur in currently protected lands, with the possible
exception of one group that may be using part of the Lake Apopka Restoration Area.
Some scrub occurs adjacent to this property (Lake2), but it was excluded from the
Water Management purchase. The habitat in Lake2 and nearby Lake3 potentially
may support up to 10 pairs of jays (Table 5-20b).
Restoration potential: The restoration potential of this metapopulation is low;
perhaps 20 pairs of jays could be protected under the maximum acquisition option (Table
5-20b).
Simulation results: This metapopulation is ranked 1st in vulnerability (table 5-23),
last in percent protected (0.0%, table 5-24), and 14th in priority (table 5-25), with high
vulnerability and low potential for improvement. Even the maximum acquisition option
has a high risk of extinction (p=0.70) and quasi-extinction (p=l .0; Table 5-20b).

319
Recommendations: The best opportunity for protecting a small population of jays
in this metapopulation probably lies in the acquisition and restoration of habitat near the
Lake Apopka Restoration Area (Lake2, Lake3). Jays found during the SMP at
Lakel are in habitat proposed for acquisition; a search for additional nearby habitat
may be worthwhile. A few unoccupied, heavily overgrown patches exist around lake
margins in the northern portion of this metapopulation; jays potentially could be
translocated to these sites if they were restored.

320
]ay Territory Locations ) 75 km dispersal buffer
Scrub Polygons
^ Lo Disturbance
3 Lo Density Housing
IfMj Hi Density Housing
Ranch/Ag
Protection Status
IlSagj Proteaed
]] Proposed
Water Bodies
/\f Interstates
State highways
County roads
' County lines
M20 Central Lake
0 5 [0 15
1 : 250,000
Kilometers
Fig. 5-20a. Central Lake county map 1992 1993 jay and habitat distribution.

321
Jay Territory Locations
(after restoration) Protection Status
~ Scrub Polygons A/ Protected
CCC## Polygon ID A/;Unprotected
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Water Bodes
. County lines
M20 Central Lake
10
1 : 250,000
15
Kilometers
Fig. 5-20b. Central Lake county acquisition map.

322
Table 5-20a. Central Lake county patch statistics (number of jay territories for different
configurations)
Patch id Status
1992-1993#
jay territories
No
acquisition
(restored)
Maximum
acquisition
(restored)
Lakel
2
2
Lake2
4
8
Lake3
1
4
Lake4
6
6
Totals
13
0
20

323
20 40 60
Year
Fig. 5-20c. Central Lake county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.
J

Probability
324
1 OT
0 6
0 6
0 4
0.2
5 10
Threshold Pop Size
Fig. 5-20d. Central Lake county quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition.

325
Table 5-20b. Central Lake county simulation statistics
Data type
Original
1992-1993
scenario
No
acquisition
Maximum
acquisition
starting
population size
13
0
20
x end pop. size
0.10

1.93
s.d.
0.60
2.9
percent
decline
99.2

90.4
extinction
risk
0.97
0.70
quasi-extinction
risk (10 pairs)
1.0

1.0

326
Lake Wales Ridge (M21)
General description: The Lake Wales Ridge metapopulation is the largest
metapopulation both in numbers of jays and in geographic extent. Its northern limit
reaches into Orange county and extends southward through Polk, Highlands, and Glades
county. The SMP documented about 565 jay territories, excluding suburban jays, in this
metapopulation. Estimated potential population size after habitat restoration and full
occupancy is 535 pairs in currently protected areas, and 858 pairs maximum.
Protected areas: Tiger Creek & Lake Wales S.F. (Polk6), Lk. Kissimmee S.P.
(PolklO), Catfish Creek(Polkl 1), Disney Wilderness Pr. (Polkl2), Platt Branch Mit. Pk.
(Highl), Archbold Biol. St. & Lake Placid W.E.A. (High4), Lake Wales Ridge W.E.A.
(High7), Lake June Scrub S.P. (Highl 1), Lake Wales Ridge W.E.A. (Highl3), Highlands
Hammock S.P. (Highl 5), Carter Creek(Highl7), Sun n Lakes (Highl9), Avon Park Air
Force Range(Polk3). See Table 5-2 la for tabulation of jay numbers.
Restoration potential: Much of the habitat in protected areas of this
metapopulation has been or is being restored, but unprotected habitat is becoming
increasingly overgrown. At the time of the SMP, jay densities in most areas were
probably close to maximum (compare first and last data columns in Table 5-2la).
Simulation results: This metapopulation ranked 21 st (last) in vulnerability (table
5-23), 5th in percent protected (62.3%; table 5-24), and 21st in priority (table 5-25), with
low vulnerability and low potential for improvement. The no acquisition and
maximum acquisition option both had no extinction or quasi-extinction risk, and had
mean population declines of 18.5% and 17.4% respectively (Table 5-2lb).

Recommendations: This metapopulation has been the focus of intensive
acquisition efforts, and most major occupied habitat patches that are relatively
327
undeveloped appear to be acquired or in the process of being acquired. However, none of
the jays in Glades county are protected; many occur on the extensive landholdings of the
Lykes Brothers Corporation. In Highlands county, the Hendrie Ranch (High2) is an
important unprotected population that doesnt appear on most acquisition lists. The jay
population at Highlands Hammock State Park (Highl5) is very small and somewhat
isolated. Habitat restoration and additional acquisition is needed for this population. In
Polk county, unprotected jay habitat (Polk7) exists that would help connect Tiger
Creek (Polk6) and Catfish Creek (Polkl 1). The tiny population of jays at Lake
Kissimmee State Park (Polk 10) would benefit from the acquisition of jays and habitat
at Polk8 and Polk9. The most significant northerly population of jays occurs on the
northeast margin of Lake Marion (Polk 13) on unprotected habitat, and doesnt appear
on most acquisition lists.

328
< HTAUisney^iOOT^sf-r y-^
rr \% P^\K
*>\
' rrL>
]ay Territory Locations
(after restoration) Protection Status Water Bodies Lake Wales Ridge
Scrub Polygons /\/ Protected .
BC Polygon ID .-V Unprotected V ^ ' *
Jays outside of labeled, bold polygons are considered to be Suburban jays. -| 720 000
30 Kilometers
Fig. 5-2la. Lake Wales Ridge map overview.

329
)ay Territory Locations
1.75 km dispersal buffer
Scrub Polygons
^ Lo Disturbance
3 Lo Density Housing
IMftj Hi Density Housing
Ranch/Ag
Protection Status
: ) Protected
J Proposed
Water Bodies
/\/ Interstates
State highways
County roads
County lines
M21 Glades
10
1 : 250,000
15 Kilometers
-+
Fig. 5-2lb. Lake Wales Ridge map 1992 1993 jay and habitat distribution, Glades
County.

330
]ay Territory Locations
Scrub Polygons
^ Lo Disturbance
2 Lo Density Housing
Hi Density Housing
Ranch/Ag
1.75 km dispersal buffer
Protection Status
jlgill Protected
H Proposed
Water Bodies
Interstates
Ay/ State highways
County roads
' ' County lines
M21 S. Highlands
10
1 : 250,000
15 Kilometers
Fig. 5-2lc. Lake Wales Ridge map 1992 1993 jay and habitat distribution, S.
Highlands county.

331
)ay Territory Locations 1.75 km dispersal buffer
Scrub Polygons
F==| Lo Disturbance Protection Status yy Interstates M2 1 N. Highland SC S. Polk
f7/"') Lo Density Housing ***** /y Sw* hl|hw¡o 0 5 10 15 Kilometers
Hi Density Housing PropOSed County roads i
I Ranch/Ag Water Bodies County lines 1 250 000
Fig. 5-21(1. Lake Wales Ridge map 1992 1993 jay and habitat distribution, N
Highlands and S. Polk county.
i

332
Jay Territory Locations 1.75 km dispersal buffer
Scrub Polygons
I 1 Lo Disturbance
7/j Lo Density Housing
|MM Hi Density Housing
Ranch/Ag
Protection Status
lijjsgj Protected
]] Proposed
Water Bodies
Interstates
f\? State highways
' County roads
County lines
M21 S. Cen Polk
10 15 Kil
1 : 250,000
Fig. 5-21 e. Lake Wales Ridge map 1992 1993 jay and habitat distribution, S. central
Polk county.
j

333
]ay Territory Locations 1.75 km dispersal buffer
Scrub Polygons
2 Lo Disturbance
^ Lo Density Housing
Hi Density Housing
Ranch/Ag
Protection Status
E '"j Protected
Proposed
Water Bodies
/\/ Interstates
/\/ State highways
\y County roads
' County lines
M21 NE Polk/NW Osceola
0 5 10 15
1 : 250,000
Kilometer
i
Fig. 5-2If. Lake Wales Ridge map 1992 1993 jay and habitat distribution, N.E. Polk
and N.W. Osceola county.

334
]ay Territory Locations
(after restoration)
Scrub Polygons
CCC## Polygon ID
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Protection Status
/\/ Protected
/\y Unprotected
Water Bodies
County lines
M21 Glades
10
1 : 250,000
15 Kilometers
njh
Fig. 5-2 lg. Lake Wales Ridge acquisition map, Glades county.

335
Jay Territory Locations
(after restoration)
| Scrub Polygons
CCC## Polygon ID
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Protection Status
/\/ Protected
/\/ Unprotected
Water Bodies
County lines
M 21 S. Highlands
O
1 : 250,000
i 5 Kilometers
V
4*
Fig. 5-2lh. Lake Wales Ridge acquisition map, S. Highlands county.

336
Jay Territory Locations
(after restoration) Protection Status
Scrub Polygons /\/ Protected
CCC Polygon ID A/1Unprotected
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Water Bodies
County lines
M 21 N. Highlands SC S. Polk
0 5 10 15 Kilometers
1 : 250,000
Fig. 5-2 li. Lake Wales Ridge acquisition map, N. Highlands and S. Polk county.

337
Jay Territory Locations
Scrub Polygons
^ Lo Disturbance
[ ' \ Lo Density Housing
l/VM Hi Density Housing
ggj Ranch/Ag
Protection Status
liSgg:;; Protected
H Proposed
Water Bodies
yW Interstates
/\ / State highways
County roads
County lines
M21 S. Cen Polk
0 5 10
1 : 250,000
15 Kilometers
4*
Fig. 5-2lj. Lake Wales Ridge acquisition map, S. central Polk county.

338
]ay Territory Locations
(after restoration)
i Scrub Polygons
CCC## Polygon ID /V
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Protection Status
/\/ Protected
A/ Unprotected
Water Bodies
County lines
M21 NE Polk/NW Osceola
10
) 5 Kilometei
1 : 250,000
Fig. 5-2lk. Lake Wales Ridge acquisition map, N.E. Polk and N.W. Osceola county.

339
Table 5-2 la. Lake Wales Ridge patch statistics (number of jay territories for different
configurations)
Patch id
Status
1992-1993#
iay territories
No acquisition
(restored)
Maximum
acquisition
Char22
2
2
Gladel
33
33
Glade2
2
2
Glade3
11
11
Glade4
41
41
Glade5
1
1
Highl
Platt Branch Mit. Pk.
14
14
14
High2
40
40
High3
4
4
High4
Archbold Biol. St. & Lake
141
134
141
Placid W.E.A.
High6
(proposed)
8
8
8
High7
Lake Wales Ridge W.E.A.
20
24
26
High8
Lake Wales Ridge W.E.A.
23
25
27
High9
8
8
HighIO
7
7
Highl 1
Lake June Scrub S.P.
18
19
21
High12
30
28
30
Highl 3
Lake Wales Ridge W.E.A.
10
8
10
High14
11
11
Highl 5
Highlands Hammock S.P.
8
9
12
High16
7
12
High17
Carter Creek
36
48
Highl 8
7
7
Highl 9
Sun 'n Lakes
10
11
Okeel
11
11
Osc2
2
2
2
Polkl
2
5
Polk3
Avon Park Air Force Range
114
152
153
Polk4
4
4
4
Polk5
2
6

340
Table 5-21a continued.
Patch id
Status
1992-1993#
jay territories
No acquisition
(restored)
Maximum
acquisition
Polk6
Tiger Creek & Lake Wales
11
17
17
S.F.
Polk7
9
13
Polk8
6
8
Polk9
6
6
PolklO
Lk. Kissimmee S.P.
6
6
6
Polk11
Catfish Creek
34
41
41
Polk12
Disney Wilderness Pr.
39
39
39
Polk13
17
20
Totals
655
535
858

341
Fig. 5-211. Lake Wales Ridge trajectory graphs. Top) no acquisition, Bottom) maximum
acquisition.

342
1 O
I*
-Q
n
a
o
£1
0 6
F
0 A~~
02
L
L
100
200 300
Threshold Pop Size
400
500
1 0
A
1
c
£
04
rJ
[
L
500
Threshold Pop Size
1000
Fig. 5-2 lm. Lake Wales Ridge quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition.

343
Table 5-2lb. Lake Wales Ridge simulation statistics
Data type
No
acquisition
Maximum
acquisition
starting
population size
535
858
x end pop. size
435.7
708.5
s.d.
56.4
65.4
percent
decline
18.5
17.4
extinction
risk
0.0
0.0
quasi-extinction
risk (10 pairs)
0.0
0.0

344
Other Metapopulations
Brevard barrier island
Jays on the Brevard barrier island have been studied intensively by Breininger
(1999), who documented a precipitous decline since the SMP. These jays live in an
urbanized matrix, and have very poor demographic performance. The demographic
parameters measured by Breininger (1999) were used to parameterize suburban jays in all
simulations (see Methods).
Clay county
Scrub-jays occur sporadically in Clay county, and have been seen most recently at
Camp Blanding Military Reservation and along a powerline near Goldhead Branch State
Park (Pranty et al. manuscript). These are the most northerly known jays in Floridas
interior. Because they are fairly isolated from the nearest jay population in Ocala
National Forest, they may be of interest genetically. Camp Blanding and Goldhead
Branch have scrub that has been or is being restored. Natural recolonization may be a
problem at these sites as the nearest significant population of jays is at Ocala National
Forest. Translocation of jays might be worth investigating.
Osceola
A few jays are known to occur in scrub patches near Yeehaw junction and at
scattered sites on the Osceola prairie to just north of St. Cloud. The SMP was not able to
survey quite a few patches in this region, and it is likely that significant numbers of jays

345
occur in these patches and in unmapped habitat. Additional surveys are urgently needed
in this area.
Western Polk
A few jays are known to occur in small scrub patches in southern Polk county to
the west of the Lake Wales Ridge, several of which have been preserved for endemic
plant species (e.g. Lake Blue and Lake McLeod), but are lacking scrub-jays. The western
portion of Polk county has been heavily modified by the phosphate industry, and partially
successful attempts have been made by industry to restore scrub or physically move scrub
soils and vegetation. A few scrub-jays occur on these sites, and other jays undoubtedly
occur on areas unsurveyed by the SMP.
Bright Hour Ranch
A substantial and apparently viable population of jays, estimated at 21 pairs
during the SMP, exists in western DeSoto county at the Bright Hour Ranch. Now a
conservation easement, this property is only 20 miles west of Archbold Biological
Station. The jays at Bright Hour Ranch have a distinctive hiccup call (J. Fitzpatrick,
pers. comm.; B. Stith, pers. obs.) which suggests that they may be highly isolated from
the nearby jays on the Lake Wales Ridge.

346
Recommendations
Ranking Metapopulation Vulnerability
A summary of the primary metapopulation viability statistics produced by the
simulations for each metapopulation is provided in Table 5-22. The metapopulations are
not ranked in any meaningful order in this table. These are the raw values used in the
analyses that follow.
Quasi-extinction risk (the probability of falling below 10 pairs) is used in this
chapter as the primary statistic for ranking metapopulation vulnerability. A precedence
exists for preferring quasi-extinction risk over extinction risk (the probability of total
extinction) (see Stith et al. 1996; Breininger et al. in press). Other statistics such as
population size and percent population decline also are used in these analyses, primarily
to break ties for metapopulations having identical quasi-extinction risk.
Metapopulation viability ranking no acquisition option. Table 5-23 provides a
ranking of metapopulations based on vulnerability to quasi-extinction under the no
acquisition option. This table provides a simple ranking of how vulnerable each
metapopulation is if no further land acquisition occurs. The first 11 metapopulations on
this list have a high quasi-extinction risk of 0.90 or greater (Central Lake, Central
Charlotte, N. Brevard, Central Brevard, Levy, Manatee, Flagler, Palm Beach, Lee, N.W.
Charlotte, and W. Volusia). These metapopulations warrant immediate attention, since
without further acquisitions they are extremely vulnerable to quasi-extinction. At the
bottom of the list, the last 5 metapopulations have a low probability of quasi-extinction,
even without further acquisitions (N.E. Lake, Martin, Merritt Island, Ocala N.F., and

347
Lake Wales). However, interpretation of these low quasi-extinction probabilities must
take into account the fact that this statistic is only sensitive to populations that decline
below 10 pairs. Populations that have declined catastrophically, but not below 10 pairs,
will not appear problematic using this statistic. Therefore, for these larger populations it
is important to evaluate other statistics, such as percent population decline, and the mean-
variance of the ending population size, which may reveal large declines in population
size. One of the metapopulations that has a low quasi-extinction risk, N.E. Lake, has a
large decline in population size (N.E. Lake: % decline = 33.7). Furthermore, none of the
statistics presented here show which populations become extinct within each
metapopulation. Examining the population persistence of specific key patches also may
be important before concluding that a given metapopulation will not benefit significantly
from further acquisitions.
Metapopulation viability ranking maximum acquisition option. Although table
5-23 is helpful for identifying metapopulations most in need of further acquisition, it does
not provide information on how much improvement in viability is potentially possible for
a given metapopulation. Table 5-23a provides a ranking of metapopulation viability for
the maximum acquisition option. Comparison of Tables 5-23 and 5-23a shows that the
number of metapopulations with a quasi-extinction risk of 0.90 or greater is reduced from
11 to 5. The top three metapopulations appear to have little room for improvement.
However, it may be inappropriate to write-off even these metapopulations, since
unoccupied, restorable habitat may exist nearby, or the jays may warrant extreme
intervention due to unique genetics, educational opportunities, or strong local support.

348
Percent population protected. Table 5-24 shows the percent of each population
that is currently protected (assuming all habitat is restored and fully occupied). Eleven of
the 21 metapopulations have less than 50% of their potential population protected.
Priority ranking. Table 5-25 provides a priority ranking of metapopulations based
on a simple classification scheme I devised. The highest priority ranking is given to those
metapopulations that are most vulnerable and have the highest potential for improvement.
The lowest ranking is given to metapopulations with the lowest vulnerability and least
potential for improvement. I arbitrarily defined three vulnerability categories based on
the quasi-extinction estimate for the no acquisition scenarios (low vulnerability: p =
0.0 0.05, moderate vulnerability : p = 0.5 0. 20, and high vulnerability: p > 0.20). I
arbitrarily defined three potential for improvement categories based on the difference
between the quasi-extinction estimates for the maximum acquisition option and the no
acquisition option (low improvement: p = 0.0 0.05, moderate improvement: p = 0.5 -
0.20, and high improvement: p > 0.20).
This classification scheme indicates that there are 13 metapopulations of moderate
or high vulnerability that also have moderate or high potential for improvement. These 13
metapopulations (N. Brevard, Levy, Central Charlotte, Central Brevard. W. Volusia,
N.W. Charlotte, St. Lucie, Citrus, Lee, Manatee, Pasco, S. Brevard, and Sarasota) score
highest on the priority list. Three metapopulations (Palm Beach, Central Lake, and
Flagler) have high vulnerability, but low potential for improvement. The remaining 5
metapopulations have low vulnerability and low potential for improvement, at least as
measured by quasi-extinction estimates. As discussed earlier, other statistics such as
percent population decline should be used to evaluate these large metapopulations.

349
Summary of Recommendations
Table 5-26 lists each metapopulation in the order given by the priority ranking
(table 5-25), and summarizes the recommendations provided earlier in each
metapopulation section. Unprotected habitat patches with a high priority for acquisition
are listed in the Primary Acquisition Target column; patches that may be of lower
priority are listed in the Secondary Acquisition Target column. I decided which patches
to place into the two acquisition categories subjectively, based on the results of different
reserve design simulations and my overall impression from the modeling results that
maintaining habitat contiguity is much more important than maintaining habitat
connectivity. A systematic analysis of this contiguity vs. connectivity issue is needed.
Habitat patches that are already protected but are in immediate need of restoration and
management are listed in the Restoration column. The last column, Other actions,
lists miscellaneous activities that are recommended, such as additional surveys.

Table 5-22. Metapopulation viability statistics.
Metapopulation
Protected
population
size
Maximum
population
size
Extinction
prob. (no
acquisition)
Extinction
prob
(maximum
acquisition)
Quasi-ext.
prob
(no
acquisition)
Quasi-ext.
prob.
(maximum
acquisition)
% decline
(no
acquisition)
% decline
(maximum
acquisition)
M1 Levy
17
75
1.0
0.0
1.0
0.0
100.0
1.3
M2 Citrus
47
125
0.17
0 0
047
0.33
70.6
55.0
M3 Pasco
63
69
0.03
00
0.30
0.233
67.9
68.2
M4 Manatee
36
145
0.97
0 30
1.0
0 90
95.3
96.6
M5 Sarasota
50
89
0 03
0 0
0.10
0.0
47.4
47 3
M6 N.W. Charlotte
28
56
067
0.07
1.0
0.30
91.8
60.7
M7 Cen. Charlotte
5
61
1.0
0.07
1.0
0 07
100.0
65.4
M8 Lee
15
62
0.73
040
1.0
0 90
92.7
90.9
M9 Flagler
5
12
0.93
0.57
1.0
1.0
80.1
75.8
M10 Merritt Island
495
536
0.0
0.0
00
0.0
0.70
6.5
M11 N. Brevard
4
110
1.0
0.0
1.0
0.0
100.0
14.5
M12 Cen. Brevard
5
40
1.0
0.0
1.0
0.10
100.0
36.3
M13 S. Brevard
62
165
0.07
0.0
0.20
0.0
54 0
24.8
M14 St. Lucie
23
37
0.20
0.03
0.73
0.27
57.4
19.4
M15 Martin
85
120
0.0
00
00
0.0
17.3
7.5
M16 Palm Beach
9
13
0.90
0.77
1.0
1.0
94.8
91.0
M17- Ocala N.F.
448
???
0 0
0.0
roo
0.0
25.0
???
M18-N.E. Lake
67
161
0.0
0.0
003
0.0
33.7
10.9
M19-W. Volusia
17
70
0.33
0.0
0.90
0.0
54 9
20.2
M20 Cen. Lake
0
20
1.0
0.70
1.0
1.0
100.0
90.4
M21 Lake Wales
535
858
0.0
0.0
0.0
0.0
18.5
17.4

Table 5-23. Metapopulation vulnerability ranking no acquisition (sorted by decreasing quasi-extinction probability).
Rank
Meta population
Quasi-ext.
prob. (no
acquisition)
Extinction
prob. (no
acquisition)
Protected
population
size
Maximum
population
size
1
M20 Cen. Lake
1.0
1.0
0
20
2
M7 Cen. Charlotte
1.0
1.0
5
61
3
M11 N. Brevard
1.0
1.0
4
110
4
M12 Cen. Brevard
1.0
1.0
5
40
5
M1 Levy
1.0
1.0
17
75
6
M4 Manatee
1.0
0.97
36
145
7
M9 Flagler
1.0
0.93
5
12
8
M16 Palm Beach
1.0
0.90
9
13
9
M8 Lee
1.0
0 73
15
62
10
M6 N.W. Charlotte
1.0
0.67
28
56
11
M19 W. Volusia
0.90
0.33
17
70
12
M14 St. Lucie
0.73
0.20
23
37
13
M2 Citrus
047
0.17
47
125
14
M3 Pasco
0.30
0.03
63
69
15
M13 S. Brevard
0.20
0.07
62
165
16
M5 Sarasota
0.10
0.03
50
89
17
M18-N.E. Lake
0.03
0.0
67
161
18
M15 Martin
0.0
0.0
85
120
19
M10 Merritt Island
0.0
0.0
495
536
20
M17-Ocala N.F.
0.0
0.0
448
???
21
M21 Lake Wales
0.0
0.0
535
858

Table 5-23a. Metapopulation vulnerability ranking maximum acquisition (sorted by increasing percent protection).
Rank
Metapopulation
Quasi-ext.
prob (maximum
acquisition)
Extinction
prob.
(maximum
acquisition)
Protected
population
size
Maximum
population
size
1
M16 Palm Beach
1.0
0.77
9
13
2
M20 Cen. Lake
1.0
0.70
0
20
3
M9 Flagler
1.0
0.57
5
12
4
M8 Lee
0.90
0.40
15
62
5
M4 Manatee
0 90
0.30
36
145
6
M2 Citrus
0.33
0.0
47
125
7
M6-N.W. Charlotte
0.30
0.07
28
56
8
M14 St. Lucie
0.27
0.03
23
37
9
M3 Pasco
0.23
0.0
63
69
10
M12 Cen. Brevard
0.10
0.0
5
40
11
M7 Cen. Charlotte
0.07
0.07
5
61
12
M19-W. Volusia
0.0
0.0
17
70
13
M1 Levy
0.0
0.0
17
75
14
M5 Sarasota
0.0
0.0
50
89
15
M11 N. Brevard
0.0
0.0
4
110
16
M15 Martin
0.0
0.0
85
120
17
M13 S. Brevard
0.0
0.0
62
165
18
M18-N E. Lake
0.0
0.0
67
161
19
M10 Merritt Island
0.0
0.0
495
536
20
M21 Lake Wales
0 0
0.0
535
858
21
M17-Ocala N.F.
0.0
00
448
???

Table 5-24. Percent protected ranking (sorted by increasing percent protection).
Rank
Metapopulation
Percent
protected
(after
restoration)
Protected
population
size
Maximum
population
size
Quasi-ext.
prob.
(maximum
acquisition)
Extinction
prob.
(maximum
acquisition)
21
M20 Cen. Lake
0.0
0
20
1.0
0.70
20
M11 N. Brevard
3.6
4
110
0.0
0.0
19
M12 Cen. Brevard
12.5
5
40
0.10
0.0
18
M1 Levy
22.7
17
75
0.0
0.0
17
M8 Lee
24.2
15
62
0 90
0.40
16
M19 W. Volusia
24.3
17
70
0.0
0.0
15
M4 Manatee
24 8
36
145
0.90
0.30
14
M2 Citrus
37.6
47
125
0.33
0.0
13
M13 S. Brevard
37.6
62
165
0.0
0.0
12
M18-N.E. Lake
41.6
67
161
0.0
0.0
11
M9 Flagler
41.7
5
12
1.0
0.57
10
M6-N.W. Charlotte
50.0
28
56
0.30
0.07
9
M5 Sarasota
56.2
50
89
0.0
0.0
8
M14 St. Lucie
62.2
23
37
0.27
0.03
7
M16 Palm Beach
69.2
9
13
1.0
0.77
6
M21 Lake Wales
62.3
535
858
0.0
0.0
5
M15-Martin
70.8
85
120
0.0
0.0
4
M7 Cen. Charlotte
82.0
5
61
0.07
0.07
3
M3 Pasco
91.3
63
69
0.23
0.0
2
M10 Merritt Island
92.3
495
536
0.0
0.0
1
M17-Ocala N.F.
???
448
???
0.0
0.0
UJ

Table 5-25. Metapopulation priority ranking (sorted by decreasing priority).
Rank
Metapopulation
Vulnerability
Potential for
improvement
Quasi-ext. prob. -
no acquisition
Difference in quasi-
ext. prob. (max no
acquisition)
1
M11 N. Brevard
high
high
1.0
1.0
2
M1 Levy
high
high
1.0
1.0
3
M7 Cen. Charlotte
high
high
1.0
0 93
4
M12 Cen. Brevard
high
high
1.0
0 90
5
M19-W. Volusia
high
high
0.90
0.90
6
M6 N.W. Charlotte
high
high
1.0
0.70
9
M14 St. Lucie
high
high
0.73
0.43
7
M2 Citrus
high
mod.
0.47
0.14
8
M8 Lee
high
mod.
1.0
0.10
10
M4 Manatee
high
mod
1.0
0.10
11
M3 Pasco
high
mod.
0.30
0.07
12
M13 S. Brevard
mod.
mod.
0.16
0.16
13
M5 Sarasota
mod.
mod.
0.10
0.10
14
M16 Palm Beach
high
low
1.0
0.0
15
M20 Cen. Lake
high
low
1.0
0.0
16
M9 Flaqler
high
low
1.0
0.0
17
M18-N.E. Lake
low
low
0.03
0.03
18
M15 Martin
low
low
0.0
0.0
19
M10 Merritt Island
low
low
0.0
0.0
20
M17-Ocala N.F.
low
low
0.0
0.0
21
M21 Lake Wales
low
low
0.0
0.0

Table 5-26. Summary of recommendations (highest priority first).
Metapopulation
Primary Acquisition
Targets
Secondary
Acquisition
Targets
Restoration (protected areas)
Other actions
N. Brevard
(Mil)
Brev4 (Buck Lake),
Brev7 (addition to South
Lake), Brev8 (Seminole
Ranch), Brev9, BrevlO,
Brevl 1, Brevl2, Brevl5,
Brev 16, Brev 17, Brev 18
Brev6, Brev 13
New jay surveys needed
at most acquisition sites
Cedar Key (Ml)
Levy2 (as much as
possible)
Levyl (Cedar Key Scrub Preserve)
Central
Charlotte (M7)
Chari 7, Chari 8, Chari 5,
Chari4, Chari9
Lee5
Charl
Surveys needed along
Prairie and Shell Creek
(Char 17,18,20)
Central Brevard
(Ml 2)
Brev41, Brev42
Brev40 (Rockledge Scrub Pr.), Brev43
(Wickham County Pk.), Brev44
(Melbourne regional airport)
W. Volusia
(Ml 9)
Voll9, Voll8, Volll
VollO (Blue Springs)
Additions to VollO
(Blue Springs)
N.W. Charlotte
(M6)
Char8, Char7, Sari 3
(expand), Sari 1
CharlO, Chari 1,
Sari 2
Char9 (Charlotte Harbor S.B. Pr.)
St. Lucie-N.
Martin (Ml4)
Marl (additions), Mar2,
Mar3, Mar4
Stl4 (Savannas S.P.), Marl
Investigate status of Stl3

Table 5-26 (cont.)
Metapopulation
Primary Acquisition
Targets
Secondary
Acquisition
Targets
Restoration (protected areas)
Other actions
Citrus-S.W.
Marion (M2)
Mar2, Mar8, Citr2
(Re-evaluate after new
survey completed)
Citr5, Mar7
(Re-evaluate after
new survey
completed)
Citrl (Crystal R. St. Buffer Pr.), Citr6,
Mar5, Sumtl
Improved survey.
Purchase/restore
unoccupied patches e.g.
N. side of Citr3 & Mar3
Lee-Collier
(M8)
Colli, Coll4, Coll5,
Coll7,
Lee3 (Estero Bay Aquatic Pr.), Coll2
(Rookery Bay N. Estuarine Research
R.), Coll3 (Immokalee airport)
Purchase/restore
unoccupied patches
south of Lee3
Manatee (M4)
Sari6, Man, Mani,
Man 15, Man9, Man 10,
Manl 1, Man5, Harl
(Re-evaluate after new
survey completed)
(Re-evaluate after
new survey
completed)
Sari 5 (Myakka S.P.), Sari 9 (Verna
Wellfield), Manl5 (Duette),
Manl2&16 (Beker), Manl7-18 (Lake
Manatee), H112-3 (Little Manatee),
H118 (Balm-Boyette Scrub Pr.), HH19
(Golden Aster Scrub Nature Pr.)
New survey (emphasize
atypical habitat)
Pasco (M3)
Pas2
(Re-evaluate after new
survey completed)
(Re-evaluate after
new survey
completed)
Herl (Weeki Wachee), Pasl
(Starkey/Serenova), Pas3 (Cross-bar/
Al-bar), Pas5 (Alston Tract), Pas
(Green Swamp W.M.A.)
New survey (emphasize
atypical habitat)
S. Brevard
(Ml 3)
Brev31, Brev32, Brev38
Brev36, InRi9, InRilO,
InRi5, InRi6
Brev30 (Malabar Scrub Sanct.),
Brev31 (Valkaria), Brev35 (St.
Sebastian River S.P.& Micco Scrub
Sanct.), InRil (Sebastian airport),
lnRi2 (private), InRi4 (Wabasso Scrub
Pr.), StLul (St. Lucie airport)
Survey for additional
jays on south end of Ten
Mile Ridge

Table 5-26 (cont.)
Metapopulation
Primary Acquisition
Targets
Secondary
Acquisition
Targets
Restoration (protected areas)
Other actions
Sarasota (M5)
Char2, Char3, Char4,
Sar3, Sar5, SarlO
Char5, Char6,
Sari, Sar2,
Sar7 (Casperson/Brohard), Sar4
(Lemon Bay Scrub C.P.), Chari
(Charlotte Harbor St. Buffer Pr.),
Sari4 (Myakka St. Forest)
Palm Beach
(Ml 6)
PB10 (Overlook Scrub),
PB13 (Tradewind/
Winchester)
PB14 (Yamato Scrub NAP), PB11
(Rolling Green Scrub Pr.), PB12
(Galaxy School Scrub Pr.)
Evaluate genetics;
consider translocation
program for educational
facilities
Central Lake
(M20)
Lake2, Lake3
Lake4, Lakel
Restore abandoned
citrus groves?
Flagler (M9)
Flagl, Flag2
Voll (N. Peninsula St. Rec. Area),
Evaluate genetics;
Investigate unoccupied
inland habitat for
translocation?
N.E. Lake
(M18)
Lake9, Lake 10
Lakd7 & Lake8 (Seminole St. Forest),
Ora2 & Lake6 (Rock Springs Run
S.Res.), Oral (Wekiwa Springs S.P.),
Martin (Ml5)
Marl 5, additions to
Marl 1, PB6, PB7
Marl2 (Jonathan Dickinson S.P.),
MarlO (Sea Branch S.P.), PB8 (Juno
Hills N.A.P.), all other protected areas
Merritt Island
(M10)
Brevl9, Brev20, Brev21,
Brev22, Brev23
Brev25 (Merritt Island N.W.R. &
Kennedy Space Center), Brev26
(Cape Canaveral Air Station)

Table 5-26 (cont.)
Metapopulation
Primary Acquisition
Targets
Secondary
Acquisition
Targets
Restoration (protected areas)
Other actions
Ocala National
Forest (Ml7)
Creation of scrub-jay preserve in
Marl 5
Lake Wales
(M21)
High2 (Hendrie Ranch),
High 15 (additions to
Highlands Hammock
S.P.), Polk7, Polk8,
Polk9, Polk 13
all patches in
Glades county?
High 15 (Highlands Hammock S.P.),
High 17 (Carter Creek), High 19 (Sun
n Lakes)

359
Discussion
Population viability analysis (PVA) models are coming under increasing criticism
due to their sensitivity to questionable assumptions on model structure and their
dependence on parameters for which field data are inadequate to define meaningful
values (Karieva et al. 1997; Ludwig 1999). Examples of problematic model assumptions
include density dependence (Mills et al. 1996) and correlation of environmental variation
(spatial and temporal) (Burgman et al. 1993). Problems more specific to spatially explicit
models were discussed recently in a series of papers (Bart 1995; Conroy et al. 1995;
Dunning et al. 1995; Turner et al. 1995). Strong criticisms of spatially explicit models
were made recently by Wennergren et al. (1995) and Ruckeslhaus et al. (1997), with
replies by Mooij and DeAngelis (1999) and South (1999). Beissinger and Westphal
(1998) provide a careful review of PVA models, and propose important qualifiers for the
proper use of such models. Chief among these is the admonition to "place very limited
confidence in the extinction estimates generated by these models. (p. 832-833). Instead,
modelers are advised to "evaluate relative rather than absolute rates of extinction and
concentrate on how well potential management actions perform relative to the baseline
of current conditions. (p. 833). Beissinger and Westphal (1998) seem to imply that all
such viability estimates are unreliable, but this is hard to square with their admonition to
evaluate relative rather than absolute rates of extinction. The use of relative rates
presupposes that any differences in the predicted rates are significant to the viability of a
population. If the model predictions differ greatly from what actually would happen in
the real population, the predicted differences may amount to nothing from a management
standpoint (e.g. the population will go extinct regardless of which management actions

360
are taken). Furthermore, the assumption that relative ranking is robust to model
inaccuracies apparently has yet to be tested, and it is not difficult to imagine situations
where relative rankings could change substantially depending on model parameter
settings. For example, when comparing the viability of two metapopulations, one
occurring in a single large patch, the other occurring in several small, isolated patches,
the assumptions made about dispersal, the spread of epidemics, the correlation of
environmental stochasticity, etc. could alter the relative ranking of the two
metapopulations viability.
I argue that these criticisms of models are excessive and stem from an unrealistic
view of models. It is a banal truism that all models are false; none are completely
accurate representations of real systems. Even simple, closed systems studied by
physicists cannot be accurately modeled except in a very restricted sense (Cartwright
1983; Giere 1999). The main issue is whether a given model is sufficiently accurate to
meet a particular need (Rykiel 1994). For this chapter, highly accurate predictions of the
fate of Florida Scrub-Jays were not needed. I am interested primarily in comparing gross
trends in the trajectories of hypothetical populations that might exist if present habitat
conditions were improved.
In all likelihood, jay habitat will not be restored as assumed by the reserve design
scenarios. Continued fire suppression and difficulties associated with habitat restoration
will result in much less restored habitat than assumed by the scenarios. Furthermore, even
if jay habitat were restored, the model results probably overestimate the persistence of
jays. This is because I chose slightly optimistic settings for two parameters to favor
small populations. Best estimates of the strength and frequency of epidemics measured at

361
Archbold Biological Station (Woolfenden and Fitzpatrick 1991) drive all small
populations rapidly to extinction, at least according to my model (manuscript in prep.)
and the models of Root (1996) and Breininger et al. (in press). I used less draconian
values (see Methods section) for epidemics, thus assuming that some small populations
will not be subject to epidemics as severe as those observed at Archbold Biological
Station. I can see no other explanation for the decades-long persistence of certain
isolated, small populations such as the Bright Hour Ranch (DeSoto county). A second
factor which I set to favor small populations is the proportion of nonbreeders that become
floaters (see Methods section). I assumed that in small populations, jays are more likely
to become floaters than in larger populations, as fewer breeding opportunities exist (e.g.
Breininger 1999).
Other reasons exist for viewing the results of this chapter as underestimating
extinction risk. Several potentially important factors were excluded from the model,
including genetic factors (e.g. inbreeding effects), hurricanes, fire suppression, and road
mortality. Relevant genetic data are mostly unavailable for Florida Scrub-Jays, and no
previous model of the species has included genetic factors. Breininger et al. (in press)
examined the influence of hurricanes, and concluded that they could have a substantial
effect on coastal populations. Although fires are thought to have strong negative effects
for many species (e.g. McCarthy 1996), Florida Scrub-Jay populations actually depend
on fire to maintain high habitat quality (Fitzpatrick et al. 1996). Thus, the reduction of
demographic success in scrub-jays due to fire suppression (Fitzpatrick and Woolfenden
1986) is a major problem that is ignored in the simulations performed for this chapter.
Mumme et al. (in review) documented significant effects of road mortality on scrub-jays,

362
which may create source-sink dynamics along roads. The absence of these factors from
my model provides further reason to view' the model predictions as underestimating
extinction risk.
Thus, the results presented in this chapter must be viewed with these biases and
assumptions in mind. One of the important steps in the modeling process is to present the
structure of a model and its assumptions explicitly, so that others can decide whether the
results are useful (Rykiel 1994). Models are assumption analyzers (Bart 1995) they
provide a means of integrating empirical data, hypotheses, theories, and intuition into a
formal framework that reveals the consequences of the underlying assumptions. In my
estimation, the assumptions I made for the simulations presented in this chapter should be
viewed as optimistic with regard to Florida Scrub-Jay viability. Therefore, the simulation
results which are summarized below should be viewed as optimistic scenarios for Florida
Scrub-Jay metapopulations.
Assuming that no additional scrub-jay habitat is protected, 11 of 21
metapopulations were estimated to be highly vulnerable to quasi-extinction (N. Brevard,
Levy, Central Charlotte, Central Brevard, W. Volusia, N.W. Charlotte, St. Lucie, Citrus,
Lee, Manatee, Pasco). Of these 11 metapopulations, the risk of quasi-extinction could be
greatly reduced for 7 metapopulations by acquiring all or major portions of the remaining
jay habitat (N. Brevard, Levy, Central Charlotte, Central Brevard, W. Volusia, N.W.
Charlotte, St. Lucie). However, even after total acquisition Central Charlotte and N.W.
Charlotte showed large mean population declines (65% and 61% respectively). The other
4 metapopulations (Citrus, Lee, Manatee, and Pasco) showed high quasiextinction
vulnerability, and moderate potential for improvement through acquisition. However,

363
each of these metapopulations showed large mean population declines ( 55% 97%),
even after complete acquisition of remaining habitat. Considering only the population
trajectory data for the 11 highly vulnerable metapopulations, only 4 (N. Brevard, Levy,
W. Volusia, and St. Lucie) had mean population declines of 20% or less after total
acquisition.
Two metapopulations were classified as moderately vulnerable with a moderate
potential for improvement (S. Brevard and Sarasota). Both of these metapopulations had
one or more fairly stable subpopulations in protection, but had substantial mean
population declines under the no acquisition option (54% and 47% respectively),
indicating that without further acquisition the rest of the metapopulation might collapse,
leaving both metapopulations vulnerable to epidemics or catastrophes.
Three metapopulations were classified as highly vulnerable to quasi-extinction,
but had low potential for improvement (S. Palm Beach, Central Lake, and Flagler). These
small populations are embedded in urban or agricultural landscapes with little or no
habitat to acquire or restore. The peripheral Flagler and S. Palm Beach metapopulations
are on the extreme north and south ends of the scrub-jays range and may warrant special
attention due to genetic considerations. Unoccupied inland scrub may be suitable for
translocating coastal jays in Flagler. The S. Palm Beach metapopulation may have special
educational value due to its proximity to the huge S. Florida human population.
The remaining five metapopulations (N.E. Lake, Martin, Merritt Island, Ocala
National Forest, and Lake Wales Ridge) were classified as having low risk of quasi
extinction. Two of these metapopulations (Martin; N.E. Lake) have significant mean

364
population declines under the no acquisition option (17% and 34% respectively); these
declines could be improved considerably by additional acquisitions.

CHAPTER 6
SYNTHESIS
The technique developed in chapter 2 to classify the metapopulation structure of
the Florida Scrub-Jay provided qualitative expectations about the viability of different
types of metapopulations. For example, systems composed only of islands are more
vulnerable than systems with midlands, which in turn are more vulnerable than systems
with mainlands. However, the technique provides no quantitative estimates of the
viability of different configurations. The individual-based model described in chapter 5
permits such viability estimates, and allowed an assessment to be made of the viability of
the major scrub-jay metapopulations around the state.
The reserve design scenarios simulated in chapter 5 did not allow the landscape to
change over time. In a theoretical paper, Fahrig (1992) argues that temporal changes in
habitat (patch lifespan) are likely to be much more important than spatial factors. She
found that the temporal scale of dispersal (dispersal frequency) far outweighed the spatial
scale (dispersal distance) in affecting population recovery from patch disturbance. The
most applicable implication of this finding for Florida Scrub-Jays today is that given the
relatively short life span of scrub patches under the current human-dominated regime of
fire suppression, large numbers of dispersing jays exploring many areas are needed to
find the few recently burned, unoccupied patches. The ability to move long distances is
much less unimportant than the ability to send out many dispersers to canvass a large
365

366
area. Given the dependence of jays on fire, scrub may seem like an ephemeral habitat, but
relative to the average life span of a jay high quality scrub patches actually have a long
life span (Woolfenden and Fitzpatrick 1984, chapter 10), especially under natural fire
regimes. Consequently, jays need not rely on frequent, long distance dispersal to locate
new patches of high quality scrub created by fire. In the spatial model developed for
Florida Scrub-Jays by Root (1998), recovery of newly restored habitat was slow, and
maintaining high habitat quality was much more important for population persistence.
Indeed, sophisticated modeling is unnecessary to show that a population of jays in poor
habitat cannot persist; it is a mathematical necessity that a population will decline unless
reproductive rates offset mortality rates. An examination of the demographic
performance of jays living in poor habitat made this clear over a decade ago (Fitpatrick
and Woolfenden 1986).
As habitat patches change over time, two factors become difficult to separate:
habitat loss and habitat fragmentation. Based on a general simulation model, Fahrig
(1997) argues that details of how habitat is arranged cannot usually mitigate the effects of
habitat loss, and that current emphasis on spatial pattern of habitat may be misplaced and
overly optimistic. She recommends that conservation efforts be directed foremost at
stopping habitat loss and at habitat restoration. Decisions still must be made about which
habitat losses to stop and which habitat patches to restore. The simulation model
developed for this dissertation can address such questions. However, the use of
simulation to guide all such decisions often may be unnecessary. In the next and final
section of this dissertation, a set of principles are presented to help guide conservation of
the Florida Scrub-Jay. These principles are landscape rules that encapsulate much of

367
the information gleaned from modeling performed for this dissertation and by previous
modeling efforts.
Conserving Florida Scrub-Jay Metapopulations
Human-induced fragmentation and habitat loss already have split the Florida
Scrub-Jay into numerous metapopulations that are now effectively isolated from one
another. Further habitat loss will have the inevitable effect of driving each
metapopulation down an ever-steepening gradient of endangerment: mainland-midland
configurations will become midland-island ones, classical configurations will become
nonequilibrium ones, which in turn are headed inexorably to extinction. At different
stages of this process, conservation strategies should vary. For systems still containing
mainlands, preserving the mainlands usually overrides other concerns, as these large
subpopulations have the greatest role in persistence of the system. As mainlands are lost,
and subpopulations shift towards configurations of islands and midlands, conservation
emphasis should shift from maintaining area to maintaining connectivity. In this phase,
priority should be placed on keeping contiguous territories together, preserving centrally
located patches, and minimizing distances among patches, thereby facilitating philopatric
dispersal and maintaining opportunities for recolonization or rescue (Hanski 1994). As
patch size and connectivity both become problematic (i.e., approaching nonequilibrium
configuration), drastic measures are appropriate, such as intensive habitat restoration,
perhaps coupled with translocation and reintroduction as a substitute for natural dispersal.

368
To stave off the Florida Scrub-Jays current slide down the endangerment
gradient, the following landscape rules should be applied to each metapopulation
independently, as appropriate (modified from Stith et al. 1996):
1) Preserve the cores. Three large, geographically separate subpopulations still
have sufficient size as of 1993 (Fig. 2-6) to be highly invulnerable to extinction except in
the face of a major catastrophe. Habitat protection should be undertaken to ensure that
these large subpopulations are not split into two or more smaller ones. Two of these core
populations occur on federal land (Ocala National Forest; Merritt Island and Cape
Canaveral), the third is largely on private land in the southern part of the Lake Wales
Ridge. We emphasize that even these core populations are not invulnerable to extinction.
Epidemics among Florida Scrub-Jays are known to occur, and can be severe
(Woolfenden and Fitzpatrick 1991). Furthermore, the entire Merritt Island-Cape
Canaveral population exists only a few meters above sea level, and the effects of a large
hurricane or sea level changes could be devastating (Breininger et al. in press).
2) Preserve all potentially viable metapopulations. The most effective long-term
insurance against extinction is to make every effort to spread the risk of catastrophe as
widely as possible. Certain nonequilibrium metapopulationsthose with few remaining
jays and lacking restorable habitatprobably are not viable in the long run. These may
not warrant expensive conservation efforts, unless they have special genetic uniqueness,
or geographic or educational importance (e.g., Lesica and Allendorf 1995). Permitting the
ultimate destruction of these metapopulations should, of course, be accompanied by
commensurate mitigation measures carried out in more viable metapopulations. Although
the focus here is on Florida Scrub-Jays, it also should be pointed out that the jay co-

369
occurs with numerous narrowly adapted, range-restricted scrub endemics (e.g.,
Christman and Judd 1990). Many of these species will automatically be preserved if the
full jay distribution is maintained; others, however, will require habitat preservation in
areas deemed nonviable for the jay.
3) Favor preservation of contiguous territories. Jays that exist in clusters of
contiguous territories are less extinction-prone than populations of equivalent size that
occur as noncontiguous territories. The risks associated with floater dispersal are high
(chapter 4), and breeder vacancies that arise within contiguous territories can be found
using the much less risky philopatric dispersal strategy. Currently, I can offer no
minimum population size or distance thresholds as quantitative guidelines for what
constitutes a stable cluster of contiguous territories. I suspect that epidemics are the
critical factor for determining this threshold (assuming habitat quality is high), but further
modeling is needed to come up with appropriate guidelines.
4) Prohibit the splitting of metapopulations. Habitat gaps larger than 12 km
represent barriers to natural dispersal and recolonization (chapter 2). To maintain all
existing metapopulations, therefore, all habitat gaps must be kept well below this 12 km
threshold. Failure to do so would effectively split the system apart and create two
smaller, hence less viable, systems. Because coastal populations of Florida Scrub-Jays are
distributed in narrow strips parallel to the coast line (dune and shoreline deposits), they
are especially vulnerable to being split as a result of elimination of small habitat patches.
The emphasis of this dissertation has been on the influence of spatial factors on
metapopulation viability. Landscape rules integrate much of the information derived from
previous chapters and prior research, in a form useful for conservation. There are,

370
however, additional rules of a non-spatial nature that should be added to the landscape
rules. Among the possible candidates, nothing is more important than the effects of
habitat quality. Florida Scrub-Jay populations will decline drastically and
deterministically as habitat quality declines (Woolfenden and Fitzpatrick 1991; Root
1997). It is unfortunate for the Florida Scrub-Jay that fire suppression by humans has
greatly reduced habitat quality in many areas. Thus, a rule to the effect that habitat
quality should be maintained at a high level is an obvious addition to the landscape rules.
Restoration efforts at Kennedy Space Center, Florida, suggest that the use of fire alone to
restore densely overgrown coastal scrub oaks may be ineffective (D. Breininger, pers.
comm.). Swain et al. (1995) suggest that unbumed scrub close to public roads or with
high pine canopy cover will be difficult to restore. The restoration experiment mentioned
in chapter 3 suggests that jays may be reluctant to colonize newly restored, vacant
habitat. Modeling results of Breininger et al. (in press) suggest that effective restoration
depends on having a surplus of local helpers. Although much remains to be learned about
habitat restoration, the selection of habitat patches to be restored clearly should take into
account spatial factors, especially the proximity of healthy jay populations that can
provide colonists to the habitat being restored.
The spatially explicit, individual-based modeling approach provides a powerful
framework for investigating conservation issues in a repeatable, quantifiable fashion.
The model I developed for this dissertation could be modified to incorporate features
such as dynamically changing landscapes, fire, road effects, genetics, etc. The
technology and programming tools are capable of this and much more. What is lacking is
the necessary field data to calibrate and validate the model. Perhaps the biggest problem

371
facing these types of models is ascertaining the reliability of model predictions. Rykiel
(1996) provides a general review of various means of model validation, and concludes
that model validation has many different meanings and no standard methods. The use of
postdiction or retrospective testing to compare historic data with model predictions has
been used only occasionally for PVA models. Brook et al. (1997) documented the 10-
year population trajectory of the Lord Howe Island wood hen following the release of 86
captive-bred individuals. Model predictions were unreliable unless accurate post facto
estimates of carrying capacity were used. Intuitively, the strongest form of validation
occurs when model predictions are not falsified by future events. This strong form of
model validation, the statistical comparison of predicted and actual trajectories, requires
long term data replicated in different areas for different landscapes. Obtaining such data
may be impossible for most species, but substantial data sets for at least 6 different color-
banded populations of Florida Scrub-Jays (Archbold Biological Station, Lake Placid,
Avon Park Air Force Range, Sarasota county, S. Brevard county, and Kennedy Space
Center) are already available, and coarser survey data for other parts of the state also are
available. Stable and declining population trajectories have been documented in these
areas, and offer an excellent opportunity to further test and refine this model. It is my
hope to continue working on this model and intriguing species, the Florida Scrub-Jay, for
which so much is known, but so much more remains to be discovered.

REFERENCES
Abrahamson, W.G., A.F. Johnson, J.N. Layne, and P.A. Peroni. 1984. Vegetation of the
Archbold Biological Station, Florida: an example of the southern Lake Wales
Ridge. Florida Scientist 47: 209-250.
Akcakaya, H.R. 1994. RAMAS/GIS: Linking Landscape Data with Population Viability
Analyses (version 1.01. Setauket, New York: Applied Biomathematics. 230 pp.
Akcakaya, H.R. and J. Atwood. 1997. A habitat-based metapopulation model of the
California Gnatcatcher. Conservation Biology 11: 422-434.
Akcakaya, H.R. and S. Ferson. 1990. RAMAS/space User Manual: Spatially-structured
Population Models for Conservation Biology. Setauket, New York: Applied
Biomathematics. 83 pp.
Anderson, P.K. Dispersal in Rodents: A Resident Fitness Hypothesis. Special publication
no. 9. The American Society of Mammalogists. 139 pp.
Anonymous, 1999. Conservation and recreation lands program annual report. Office of
Land Use Planning and Biological Services, Florida Department of Natural
Resources, Tallahassee, Florida.
Barrowclough, G.F. 1978. Sampling bias in dispersal studies based on finite area. Bird-
Banding 49:333-341.
Bart, J. 1995. Acceptance criteria for using individual-based models to make
management decisions. Ecological Applications 5: 411-420.
Batschelet, E. 1981. Circular Statistics in Biology. Academic Press, New York. 371 pp.
Beier, P. 1996. Metapopulation models, tenacious tracking, and cougar conservation.
Pages 293-324 in D. McCullough (ed.), Metapopulations and Wildlife
Conservation. Washington DC: Island Press. 429 pp.
Bell, W.J. 1991. Searching behaviour: the Behavioural Ecology of Finding Resources.
New York: Chapman and Hall. 358 pp.
Beissinger, S.R. and M.I. Wesphal. 1998. On the use of demographic models of
population viability in endangered species management. J. Wildlife Management
62: 821-841.
372

373
Best, L.B., and D.F. Stauffer. 1986. Factors confounding the evaluation of bird-habitat
relationships. Pages 209-217 in J. Vemer, M.L. Morrison, and C.J. Ralph (eds.),
Wildlife 2000: Modeling habitat relationships of terrestrial vertebrates. Madison:
University of Wisconsin Press.
Blanchard, J., S. Jue, and A. Crook. 1998. Florida Conservation Lands 1998.
Bowman, R. and L. Averill. 1993. Demography of a suburban population of Florida
Scrub Jays. Annual Progress Report to the U.S. Fish and Wildlife Service.
Boyce, M.S. 1992. Population viability analysis. Ann. Rev. Ecol. Syst. 26: 481-506.
Breininger, D.R. 1999. Florida Scrub-Jay demography and dispersal in a fragmented
urban landscape. Auk 116: 520-527.
Breininger, D.R. 1997. Florida Scrub-Jay demography in an urban metapopulation. 1997
Final Report to Endangered Species Office, U.S. Fish and Wildlife Service,
Jacksonville, FL.
Breininger, D.R., V.L. Larson, D.M. Oddy, R.B. Smith, M.J. Barkaszi. 1996.
Demography of Florida scrub-jays in different landscapes. Auk: 113: 617-625.
Breininger, D.R., V.L. Larson, R.W. Duncan. R.B. Smith, D.M. Oddy, M.F. Goodchild.
1995. Landscape patterns of Florida Scrub Jay habitat use and demographic
success. Conservation Biology 9: 1442 1453.
Breininger, D.M., M.J. Provancha, and R.B. Smith. 1991. Mapping Florida Scrub Jay
habitat for purposes of land-use management. Photogr. Eng, and Remote Sensing
57: 1467-1474.
Breininger, D.M., M.A. Burgman, and B.M. Stith. In press. Influence of habitat quality,
catastrophes, and population size on extinction risk of the Florida Scrub-Jay.
Wildlife Society Bulletin
Brook, B.W., L. Lim, R. Harden, and R. Frankham. 1997. Does population viability
analysis software predict the behaviour of real populations? A retrospective study
on the Lord Howe Island woodhen Tricholimnas sylvestris (Sclater). Biological
Conservation 82: 119-128.
Brown, J.H., and A. Kodric-Brown. 1977. Turnover rates in insular biogeography: effect
of immigration on extinction. Ecology 58: 445-449.
Buechner, M. 1987. A geometric model of vertebrate dispersal: tests and implication.
Ecology 68: 310-318.

374
Burgman, M., S. Ferson, and H.R. Akcakaya. 1993. Risk Assessment in Conservation
Biology. New York: Chapman & Hall. 314 pp.
Carmen, W. 1989. Behavioral ecology of the California Scrub Jay: a noncooperative
breeder with close cooperative relatives. Ph.D. dissertation. University of
California, Berkeley, CA. 134 pp.
Caughley, G. 1994. Directions in conservation biology. J. Animal Ecology 63: 215-244.
Conroy, M.J., Y. Cohen, F.C. James, Y.G. Matsino, and B.A. Maurer. 1995. Parameter
estimation, reliability, and model improvement for spatially explicit models of
animal populations. Ecological Applications 5: 17-19.
Cox, J.A. 1987. Status and distribution of the Florida Scrub Jay. Florida Omithol. Soc.
Spec. Pub. No. 3.
Christman, S.P. and W.S. Judd. 1990. Notes on plants endemic to Florida scrub. Florida
Scientist 53: 52-73.
Dawkins, R. 1976. The Selfish Gene. Oxford: Oxford University Press. 224 pp.
DeGange, A.R., J.W. Fitzpatrick, J.N. Layne, and G.E. Woolfenden. 1989. Acorn
harvesting by Florida Scrub Jays. Ecology 70: 348-356.
Doak, D.F., and L.S. Mills. 1994. A useful role for theory in conservation. Ecology 75:
615-626.
Duncan, B., D. Breininger, P. Schmalzer, and V. Larson. 1995. Validating a Florida
Scrub Jay habitat suitability model, using demography data on Kennedy Space
Center. Photogr. Eng. & Remote Sensing 61: 1361 1370.
Dunning, J.B., and D.J. Stewart. 1995. Spatially explicit population models: current
forms and future uses. Ecological Applications5: 3-11.
Ebenhard, T. 1991. Colonization in metapopulations: a review of theory and
observations. Pp. 105-121 in M.E. Gilpin and I. Hanski (eds.), Metapopulation
Dynamics: Empirical and Theoretical Investigations. Academic Press, London,
UK
E.S.R.I. 1990. Understanding GIS. Redlands, CA: Environmental Systems Research
Institute, Inc.
Fahrig, L. 1997. Relative effects of habitat loss and fragmentation on population
extinction. J. Wildlife Management 61: 603-610.

375
Fahrig, L. 1992. Relative importance of spatial and temporal scales in a patchy
environment. Theor. Pop. Biol. 41: 300-314.
Ferson, S. 1991. RAMAS/stage: Generalized Stage-based Modeling for Population
Dynamics. Setauket, New York: Applied Biomathematics. 108 pp.
Fishwick, P.A. 1995. Simulation model design and execution: building digital worlds.
New Jersey: Prentice-Hall. 448 pp.
Fitzpatrick, J.W., R. Bowman, D.R. Breininger, M.A. O'Connell, B. Stith, J. Thaxton, B.
Toland, and G.E. Woolfenden. In prep. Habitat Conservation Plans for the
Florida Scrub Jay: A Biological Framework.
Fitzpatrick, J.W., B. Pranty, and B. Stith. 1994. Florida Scrub Jay statewide map. U.S.
Fish and Wildlife Service Report, Cooperative Agreement no. 14-16-0004-91 -
950.
Fitzpatrick, J.W., and G.E. Woolfenden. 1986. Demographic routes to cooperative
breeding in some New World jays. Pp. 137-160 in M.H. Nitecki and J.A. Kitchell,
(eds.), Evolution of Behavior. New York: Oxford University Press.
Fitzpatrick, J.W., G.E. Woolfenden, and M.T. Kopeny. 1991. Ecology and development-
related habitat guidelines of the Florida Scrub Jay (Aphelocoma coerulescens
coerulescens). Florida Nongame Wildlife Program Technical Report, No. 8. 49
pp.
Florida Atlas and Gazeteer. 1997. (4th ed.). DeLorme, Freeport, Maine.
Forman, R.T.T. and M. Godron. 1986. Landscape Ecology. New York: Wiley.
Greenwood, P.J. 1980. Mating systems, philopatry and dispersal in birds and mammals.
Animal Behavior 28: 1140-1162.
Hanski, I. 1994. A practical model of metapopulation dynamics. J. Animal Ecology 63:
151-162.
Hanski, I. and M.E. Gilpin 1991. Metapopulation dynamics: brief history and conceptual
domain. Pp. 89-103 in M.E. Gilpin and I. Hanski (eds.). Metapopulation
Dynamics: Empirical and Theoretical Investigations. London: Academic Press.
Harrison, S. 1991. Local extinction in a metapopulation context: an empirical evaluation.
Pp. 89-103 in M.E. Gilpin and I. Hanski (eds.), Metapopulation Dynamics:
Empirical and Theoretical Investigations. London: Academic Press.

376
Harrison, S. 1994. Metapopulations and conservation. Pp. 111-128 in P.J. Edwards, R.M.
May and N.R. Webb (eds.), Large-scale Ecology and Conservation biology.
Blackwell Scientific Publication, Oxford, UK
Harrison, S., A. Stahl, and D. Doak 1993. Spatial models and Spotted Owls: exploring
some biological issues behind recent events. Conservation Biology 7: 950-953.
Hooper, R., J.C. Watson, and R.E.F. Escao. 1990. Hurricane Hugos initial effects on
red-cockaded woodpeckers in the Francis Marion National Forest. Trans. 55th
N.A. Wildlife and Natural Resources Conference: 220-224.
Howe, R.W. and G. Davis. 1991. The demographic significance ofsink populations.
Biological Conservation 57: 239-255.
Huston, M., D. DeAngelis, and W. Post. 1988. New computer models unify ecological
theory. BioScience 38: 682-691.
Ims, R.A. and N.G. Yoccoz. Studying transfer processes in metapopulations: emigration,
migration, and colonization. Pages 247 265 in I. Hanski and M.E. Gilpin (eds.),
Metapopulation Biology. New York: Academic Press.
Johnson, M.L. and M.S. Gaines. 1990. Evolution of dispersal: theoretical models and
empirical tests using birds and mammals. Ann. Rev. Ecol. Svst. 21: 449-480.
Judson, O.P., 1994. The rise of the individual-based model in ecology. Trends in Ecology
and Evolution 9: 9-14.
Kale, H.W. II, B. Pranty, B. Stith, and C.W. Biggs. 1992. The Atlas of the Breeding Birds
of Florida. Final report to Fla. Game & Fresh Water Fish Comm., Tallahassee,
FL.
Karieva, P., D. Skelly, and M. Ruckelshaus. 1997. Reevaluating the use of models to
predict the consequences of habitat loss and fragmentation. Pages 156-166 in
S.T.A. Pickett, R. Ostfels, M. Shachak, and G. Likens (eds.). The Ecological
Basis of Conservation. New York: Chapman and Hall.
Kautz, R.S., D.T. Gilbert, and G.M. Mauldin. 1993. Vegetative cover in Florida based on
1985-1989 Landsat Thematic Mapper imagery. Florida Scientist 56: 135-154.
Koenig, W.D., F.A. Pitelka, W.J. Carmen, R.L. Mumme, and M.T. Stanback. 1992. The
evolution of delayed dispersal in cooperative breeders. Quarterly Review of
Biology 67: 111-150.
Koenig, W.D., D.Van Vuren, and P. Hooge. 1996. Detectability, philopatry, and the
distribution of dispersal distances in vertebrates. Trends in Ecology and Evolution
11:514-517.

377
Komdeur, J. 1992. Importance of habitat saturation and territory quality for the evolution
of cooperative breeding in the Seychelles Warbler. Nature. 358: 493-495.
Krebs, J.R. and N.B. Davies. 1993. An Introduction to Behavioural Ecology (3rd ed.).
Blackwell Science, Inc. Cambridge, MA.
Lamberson, R.H., R. McKelvey, B.R. Noon, and C. Voss. 1992. A dynamic analysis of
Northern Spotted Owl viability in a fragmented forest landscape. Conservation
Biology 6: 505-512.
Lacy, R.C. 1993. VORTEX: a computer simulation model for Population Viability
Analysis. Wildlife Research 20: 45-65.
Lande, R. 1988a. Demographic models of the northern spotted owl (Strix occidentals
caurina). Oecologia 75: 601-607.
Lande, R. 1988b. Genetics and demography in biological conservation. Science 241:
1455-1459.
Lesica, P., and F. Allendorf. 1995. When are peripheral populations valuable for
conservation? Conservation Biology 9: 753-760.
Letcher, B.H., J.A. Priddy, J.R. Walters, and L.B. Crowded 1998. An individual-based,
spatially-explicit simulation model of the population dynamics of the endangered
red-cockaded woodpecker, Picoides borealis. Biological Conservation 86: 1-14.
Levins, R. 1969. Some demographic and genetic consequences of environmental
heterogeneity for biological control. Bull. Entomol. Soc. Amer. 15: 237-240.
Levins, R. 1970. Extinction. Lectures on Mathematics in the Life Sciences 2: 75-107.
Lima, S. and P. Zollner. 1996. Towards a behavioral ecology of ecological landscapes.
Trends in Ecology and Evolution 11: 131-135.
MacArthur, R. H. and E. O. Wilson. 1967. The Theory of Island Biogeography.
Princeton, NJ: Princeton University Press. 203 pp.
Mace, G. M., and R. Lande. 1991. Assessing extinction threats: toward a reevaluation of
IUCN threatened species categories. Conservation Biology 2: 148-157.
Maley, C. and H. Caswell. 1993. Implementing I-state configuration models for
population dynamics: an object-oriented programming approach. Ecological
Modelling 68: 75-89.

378
McCarthy, M.A. 1996. Extinction dynamics of the helmeted honeyeater: effects of
demography, stochasticity, inbreeding and spatial structure. Ecological Modelling
85:151-163.
McCarthy, M.A., D.C. Franklin, and M.A. Burgman. 1994. The importance of
demographic uncertainty: an example from the helmeted honeyeater. Biological
Conservation 67: 135-142.
McDonald, D.B. and H. Caswell. 1993. Matrix methods for avian demography. Current
Ornithology 10: 139-184.
McGarigal, K. and B. Marks. 1994. FRAGSTATS: spatial pattern analysis program for
quantifying landscape structure. Corvallis, OR: Forest Science Dept., Oregon
State University.
McGowan, K.J. and G.E. Woolfenden. 1989. A sentinel system in the Florida Scrub Jay.
Animal Behavior 37: 1000-1006.
McKelvey, K., B. Noon, and R. Lamberson. 1993. Conservation planning for species
occupying fragmented landscapes: the case of the northern spotted owl. Pages
424-452 in Kingsolver, J., P. Karieva, and R. Huey (eds.), Biotic interactions and
global change. Sunderland, MA: Sinauer.
Meyers, R.L. and J.J. Ewel (eds.). 1990. Ecosystems of Florida. University of Central
Florida Press, Orlando, Florida.
Miller, G.L. and B.W. Carroll. 1989. Modeling vertebrate dispersal distances: alternatives
to the geometric distribution. Ecology 70: 977-986.
Mills, L. S., S. Hayes, C. Baldwin, M. Wisdom, J. Citta, D. Mattson, and K. Murphy.
1996. Factors leading to different viability predictions for a grizzly bear data set.
Conservation Biology 10: 863-873.
Mooij, W and D. DeAngelis. 1999. Error propagation in spatially explicit dispersal
models. Conservation Biology 11: 1298-1306.
Morrison, M.L., B.G. Marcot, and R.W. Mannan. 1992. Wildlife-habitat Relationships:
Concepts and applications. Madison, WI: University of Wisconsin Press.
Mumme, R. and T. Below. In press. Evaluation of translocation for the threatened Florida
Scrub-Jay. Wildlife Management.
Mumme, R.J., S.J. Schoech, G.E. Woolfenden, and J.W. Fitzpatrick. In revision. Life
and death in the fast lane: demographic consequences of road mortality in the
Florida Scrub-Jay. Conservation Biology.

379
Murphy, D.D. and B.R. Noon. 1992. Integrating scientific methods with habitat
conservation planning: reserve design for Northern Spotted Owls. Ecological
Applications 2: 3-17.
Noon, B.R. and D.S. McKelvey. 1992. Stability properties of the Spotted Owl
metapopulation in southern California. Pages 187-206 in J. Vemer et al. (eds.),
The California Spotted Owl: A Technical Assessment of Its Current Status.
General Technical Report PSW-GTR-133. Albany, California: USDA Forest
Service.
Noss, R., m.A. OConnell, and D.D. Murphy. 1998. The science of Conservation
Planning: Habitat Conservation Under the Endangered Species Act. Washington,
DC: Island Press. 246 pp.
O'Connell, M. 1994. Preserving scrub habitat. Florida Naturalist 67: 9-12.
Papi, F. (ed.). 1992. Animal Homing. London: Chapman and Hall. 390 pp.
Porter, J.H. and J.L. Dooley. 1993. Animal dispersal patterns: a reassessment of simple
mathematical models. Ecology 74: 2436-2443.
Price, M. and M. Gilpin. 1996. Modelers, mammalogists, and metapopulations: designing
Stephens Kangaroo Rat reserves. Pages 217-240 in D. McCullough (ed.),
Metapopulations and Wildlife Conservation. Washington, DC: Island Press. 429
pp.
Pulliam, H.R. 1988. Sources, sinks, and population regulation. American Naturalist 132:
652-61.
Pulliam, H.R. 1992. Population dynamics in complex landscapes: a case study.
Ecological Applications 2: 165-177.
Pusey, A. and C. Packer. 1987. Dispersal and philopatry. Pages 250 266 in B. Smuts,
D. Cheney, R. Seyfarth, R. Wrangham, and T. Struhsaker (eds.), Primate
Societies. Chicago, IL: U. Chicago Press. 578 pp.
Rolstad, J. 1991. Consequences of forest fragmentation for the dynamics of bird
populations: conceptual issues and the evidence. Pages 149-163 in M.E. Gilpin
and I. Hanski (eds.), Metapopulation Dynamics: Empirical and Theoretical
Investigations. London: Academic Press.
Root, K. 1996. Population Viability Analysis for the Florida Scrub-Jay (Aphelocoma
coerulescens coerulescens) in Brevard County, Florida. Ph. D. dissertation,
Florida Institute of Technology, Melbourne. 199 pp.

380
Root, K. 1998. Evaluating the effects of habitat quality, connectivity, and catastrophes on
a threatened species. Ecological Applications 8: 854-865.
Ruckelhaus, M., C. Hartway, and P. Karieva. 1997. Assessing the data requirements of
spatially explicit dispersal models. Conservation Biology 11: 1298-1306.
Rykiel, E. 1996. Testing ecological models: the meaning of validation. Ecological
Modeling 90: 229-244.
Shaffer, M.L. 1981. Minimum population sizes for species conservation. BioScience 31:
131-134.
Simberloff, D. 1988. The contribution of population and community biology to
conservation science. Ann. Rev. Ecol. Svst. 19: 473-511.
Sjogren, P. 1991. Extinction and isolation gradients in metapopulations: the case of the
pool frog (Rana lessonae). Pp. 135-147 in M.E. Gilpin and I. Hanski (eds.),
Metapopulation Dynamics: Empirical and Theoretical Investigations. London:
Academic Press.
Smith, A. 1980. Temporal changes in insular populations of the Pika (Ochotona
princeps). Ecology 61: 8-13.
Smith, A. and M. Peacock. 1990. Conspecific attraction and the determination of
metapopulation colonization rates. Conservation Biology 4: 320-323.
Soule, M. (ed.). 1987. Viable populations for conservation. Cambridge, UK: Cambridge
U. Press.
South, A. 1999. Dispersal in spatially explicit population models. Conservation Biology
4: 1039- 1046.
Stenseth, N. and W. Lidicker (eds.). 1992. Animal Dispersal: Small Mammals as a
Model. New York: Chapman and Hall. 366 pp.
Stith, B.M., J.W. Fitzpatrick, G.E. Woolfenden, and B. Pranty. 1996. Classifying and
conserving metapopulations: a case study of the Florida Scrub Jay. Pages 187-216
in D. McCullough (ed.), Metapopulations and Wildlife Conservation.
Washington, DC: Island Press. 429 pp.
Swain, H.M., P.A. Schmalzer, D.R. Breininger, K.V. Root, S.A. Bergen, S.R. Boyle, and
S.MacCaffree. 1995. Scrub Conservation and Development Plan: Brevard
County. Appendix B: Biological Consultants Report to the Natural Resources
Management Division of Brevard County, Florida. August 1995.

381
Taylor, B. 1995. The reliability of using population viability analysis for risk
classification of species. Conservation Biology 9: 551-558.
Thaxton, J.E. and T.M. Hingtgen. 1996. Effects of suburbanization and habitat
fragmentation on Florida Scrub-Jay dispersal. Florida Field Naturalist 24: 25-60.
Thomas, C.D. 1990. What do real population dynamics tell us about minimum viable
population sizes? Conservation Biology 4: 324-327.
Thomas, C.D. 1994. Difficulties in deducing dynamics from static distributions. Trends
in Ecology and Evolution 9: 300.
Thomas, J.W., E.D. Forsman, J.B. Lint, E.C. Meslow, B.R. Noon, and J. Vemer. 1990. A
Conservation Strategy for the Northern Spotted Owl. Report of the Interagency
Scientific Committee to Address the Conservation of the Northern Spotted Owl.
Portland, Oregon: U.S. Fish and Wildlife Service. 427 pp.
Turner, M.G., G.J. Arthaud, R.T. Engstrom, S.J. Hejl, J. Liu, S. Loeb, and K. McKelvey.
1995. Usefulness of spatially explicit population models in land management..
Ecological Applications 5: 12-16.
Turchin, P. 1998. Quantitative Analysis of Movement. Sunderland, MA: Sinaeur.
U.S. Fish and Wildlife Service. 1991. Floridians advised of Scrub Jay conservation
measures. June 18. News Release. Jacksonville, FL.
Waser, P.W. 1985. Does competition drive dispersal? Ecology 66: 1170-1175.
Waser, P.W., S.R. Creel, and J.R. Lucas. 1994. Death and disappearance: Estimating
mortality risks associated with philopatry and dispersal. Behavioral Ecology 5:
135-141.
With, K., R. Gardner, and M. Turner. 1997. Landscape connectivity and population
distributions in heterogeneous environments. Oikos 78: 151-169.
Wolff, W.F. 1994. An individual-oriented model of a wading bird nest colony. Ecological
Modelling 72: 75-114.
Woolfenden, G.E. and J.W. Fitzpatrick. 1984. The Florida Scrub Jay: Demography of a
Cooperative-breeding Bird. Monogr. Pop. Biol. no. 20. Princeton, NJ: Princeton
University Press.
Woolfenden, G.E. and J.W. Fitzpatrick. 1986. Sexual asymmetries in the life history of
the Florida Scrub Jay. Pp. 87-107. In D. Rubenstein and R.W. Wrangham (eds.),
Ecological Aspects of Social Evolution: Birds and Mammals. Princeton, NJ:
Princeton University Press.

382
Woolfenden, G.E., and J.W. Fitzpatrick. 1991. Florida Scrub Jay ecology and
conservation. Pp. 542-565 in C.M. Perrins, J.D. Lebreton, and G.J.M. Hirons
(eds.), Bird Population Studies: Relevance to Conservation and Management.
Oxford, UK: Oxford University Press.
Woolfenden, G.E. and J.W. Fitzpatrick. 1996. Florida Scrub-Jay (Aphelocoma
coerulescens). In A. Poole and F. Gill (eds.), The Birds of North America, No.
228. Philadelphia: The Academy of Natural Sciences, and Washington, DC: the
American Ornithologists Union.
Wooton, T.J. and D.A. Bell. 1992. A metapopulation model of the Peregrine Falcon in
California: viability and management strategies. Ecological Applications 2: 307-
321.
Zack, S. 1990. Coupling delayed breeding with short-distance dispersal in cooperatively
breeding birds. Ethology 86: 265-286.
Zar, J.H. 1996. Biostatistical Analysis (3rd ed.). New Jersey: Prentice Hall. 662 pp.
Zollner, P.A. and S.L. Lima. 1999. Search strategies for landscape-level interpatch
movements. Ecology 80: 1019- 1030.

BIOGRAPHICAL SKETCH
Bradley M. Stith was bom in Indianapolis, Indiana, but moved to Sarasota,
Florida, at the age of three. He spent most of his youth there, collecting reptiles, scuba
diving, and exploring the many wild areas around Sarasota. With his growing love of
nature came the sad realization that humans were rapidly destroying the playground of
his youth. His interest in conservation issues were put on hold for awhile as he obtained a
B.S. in geology in 1980 from the University of Arizona in Tucson. He then moved to
Houston, Texas and worked for a small consulting firm that developed and marketed
computer mapping software (now known as GIS), and built digital databases used
primarily by the oil industry. His job as programmer/software installer/trainer allowed
him to travel around the U.S. and overseas. By 1986 the economy in Texas became badly
depressed, and he decided to seek a graduate degree. He entered the graduate program in
the Department of Wildlife and Range Sciences at the University of Florida in 1987. He
immediately began work on the Florida Breeding Bird Atlas under Dr. Stephen R.
Humphrey. In 1990 he completed his thesis entitled Satellites, Landscapes, and GIS: A
Case Study in the Atlantic Forest of Brazil and obtained a Master of Science degree. He
then re-enrolled as a Ph.D. student in the same program, now known as the Department
of Wildlife Ecology and Conservation and will receive his Ph.D. in 1999.
383

I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy-
Professor of Wildlife Ecology and
Conservation
1 certify- that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and qualitv.
as a dissertation for the degree of Doctor of Philosophy
rofessor c
Conservation
Cochairman
Ecologv and
1 certify- that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and qualitv.
as a dissertation for the degree of Doctor of Philosophy
Associate Professor of Wildlife Ecology
and Conservation
1 certify- that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy
CD.
Jon v Allen
^Profassor of Entomology and
Jematologv

I certifV that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality.
This dissertation was submitted to the Graduate Faculty of the College of
Agriculture and Life Sciences and to the Graduate School and was accepted as partial
fulfillment of the requirements for the degree of Doctor^of Philosophy
December. 1999
Dean. College of Agriculture and Life
Sciences
Dean. Graduate School



5-15d. Martin and N. Palm Beach county quasi-extinction graphs. Top) no
acquisition, Bottom) maximum acquisition 284
Fig 5-16a. Central Palm Beach county map 1992 1993 jay and habitat
distribution 288
5-16b. Central Palm Beach county acquisition map 289
5-16c. South Palm Beach county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition 291
5-16d. South Palm Beach county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition 292
5-17a. Ocala National Forest county map 1992 1993 jay and habitat
distribution 296
5-17b. Ocala National Forest county acquisition map 297
5-17c. Ocala National Forest county trajectory graphs. No acquisition 299
5-17d. Ocala National Forest county quasi-extinction graphs. No acquisition 300
5-18a. N.E. Lake county map 1992 1993 jay and habitat distribution 304
5-18b. N.E. Lake county acquisition map 305
5-18c. N.E. Lake county trajectory' graphs. Top) no acquisition, Bottom)
maximum acquisition 307
5-18d. N.E. Lake county quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition 308
5-19a. S.W. Volusia county map 1992 1993 jay and habitat distribution 312
5-19b. S.W. Volusia county acquisition map 313
5-19c. S.W. Volusia county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition 315
5-19d. S.W. Volusia county quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition 316
5-20a. Central Lake county map 1992 1993 jay and habitat distribution 320
5-20b. Central Lake county acquisition map 321
xxii


Prop. Cover
68
GARO
--RADI
CURV
-K CTRE
*YYYY
ARDS
ITOWR
FUN
XOVR
O BUDD
aSOAR
6NTRL
-H-CLVT
*TRGT
*DTCH
)FRST
TWNP
DUMP
Buffer Distance
Fig. 3-9. Percent tree cover for individual territories for 4 zones (inside territories 100
200, 400 m buffer) North territories.


328
< HTAUisney^iOOT^sf-r y-^
rr \% P^\K
*>\
' rrL>
]ay Territory Locations
(after restoration) Protection Status Water Bodies Lake Wales Ridge
Scrub Polygons /\/ Protected .
BC Polygon ID .-V Unprotected V ^ ' *
Jays outside of labeled, bold polygons are considered to be Suburban jays. -| 720 000
30 Kilometers
Fig. 5-2la. Lake Wales Ridge map overview.


345
occur in these patches and in unmapped habitat. Additional surveys are urgently needed
in this area.
Western Polk
A few jays are known to occur in small scrub patches in southern Polk county to
the west of the Lake Wales Ridge, several of which have been preserved for endemic
plant species (e.g. Lake Blue and Lake McLeod), but are lacking scrub-jays. The western
portion of Polk county has been heavily modified by the phosphate industry, and partially
successful attempts have been made by industry to restore scrub or physically move scrub
soils and vegetation. A few scrub-jays occur on these sites, and other jays undoubtedly
occur on areas unsurveyed by the SMP.
Bright Hour Ranch
A substantial and apparently viable population of jays, estimated at 21 pairs
during the SMP, exists in western DeSoto county at the Bright Hour Ranch. Now a
conservation easement, this property is only 20 miles west of Archbold Biological
Station. The jays at Bright Hour Ranch have a distinctive hiccup call (J. Fitzpatrick,
pers. comm.; B. Stith, pers. obs.) which suggests that they may be highly isolated from
the nearby jays on the Lake Wales Ridge.


154
Results
These results summarize the output of different simulations performed for each of
the 21 metapopulations. Results for each metapopulation are reported in separate
sections. Each section begins with a general description, lists the protected areas,
discusses restoration potential, summarizes the simulation results, and provides
recommendations. Maps are provided showing the distribution of jays and habitat during
the 1992-1993 SMP, and the distribution of jays in relation to the simulations and
acquisition possibilities. The acquisition maps depict what jay populations might look
like if all habitat were restored and fully occupied by jays. Tables are provided that
summarize the patch statistics (number of jays in each patch) for different reserve
configurations. The results from all simulations are presented in tables, and quasi
extinction and trajectory graphs are provided for at least 2 reserve design configurations.
The simulation results also include the statewide rankings developed and explained in the
last section (see Recommendations).


Table 3-1. Demographic and habitat parameters for North and South Sandy Hill (1994 1995).
Location
Territory
Year Group
Size
Fledgling
Production
Yearling
Production
Breeder
Surv.
Helper
Surv.
Fledgl.
Surv.
Yearl.
Surv.
Bare
sand
Tree cover
(inside
territory)
Tree cover
(100 m
buffer)
NRIDGE
ARDS
1994
4
2
2
1
1
1
0
0.02
0.02
0.02
NRIDGE
BUDD
1994
2
2
2
0.5
1
0
0.15
0.01
0.01
NRIDGE
CLVT
1994
4
3
2
1
1
0.66
0
0.28
0.01
0.07
NRIDGE
CTRE
1994
4
2
0
0.5
0.5
0
0
0.04
0.03
0.02
NRIDGE
CURV
1994
2
4
1
0.5
0.25
0.25
0.16
0.03
0.03
NRIDGE
DTCH
1994
2
0
0
1
0.2
0.02
0.06
NRIDGE
DUMP
1994
3
0
0
1
1
0.23
0
0.06
NRIDGE
FLIN
1994
5
1
0
1
0.66
0
0
0.05
0
0.02
NRIDGE
FRST
1994
4
3
3
0.5
0.5
1
0.66
0.34
0
0
NRIDGE
GARD
1994
3
4
1
1
1
0.25
0
0.26
0.02
0.05
NRIDGE
NTRL
1994
2
2
1
1
0.5
0.5
0.19
0.02
0.02
NRIDGE
RADI
1994
6
0
0
0.5
0.33
0.16
0.03
0.03
NRIDGE
SQAR
1994
8
3
2
1
0.66
0.66
0
0.12
0.03
0.02
NRIDGE
TOWR
1994
5
4
3
0
0.66
0.75
0.25
0.15
0.04
0.06
NRIDGE
TRGT
1994
2
2
1
1
0.5
0
0.11
0.01
0.02
NRIDGE
TWNP
1994
4
3
3
1
0.5
1
0
0.25
0
0.02
NRIDGE
YYYY
1994
4
1
1
0.5
1
1
0
0.08
0.04
0.04
SRIDGE
BRIK
1994
2
4
3
0.5
0.75
0.25
0.3
0.05
0.07
SRIDGE
DEAD
1994
3
3
2
1
0
0.66
0
0.16
0.16
0.24
SRIDGE
ECHO
1994
3
0
0
1
0
0.25
0.16
0.26
SRIDGE
FARS
1994
4
5
3
1
0
0.6
0.2
0.21
0.28
0.28
SRIDGE
GPSS
1994
5
3
0
1
1
0
0
0.21
0.08
0.1
SRIDGE
HTOP
1994
2
3
2
0.5
0.66
0
0.15
0.33
0.53
SRIDGE
JSUS
1994
2
1
0
0.5
0
0
0.25
0.31
0.43


275
0>
N
(/)
C
o
76
3
Q.
o
CL
*-
i-
10 0~
r
p
5 0 ^
L
20

40
60
Year
-
L
20
40
Year
60
Fig. 5-14c. St. Lucie N. Martin county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.
j


Table 5-5a. Sarasota and W. Charlotte county patch statistics (number of jay territories for different configurations)
Patch
Status
1992-1993#
No
30%
30%
70%
70%
Maximum
id
jay territories
acquisition
preserved
preserved
preserved
preserved
acquisition
(restored)
by contiguity
by connectivity
by contiguity
by connectivity
Chari
Charlotte Harbor
S.B.P.
3
3
3
3
3
3
3
Char2
13
9
1
13
7
13
Char3
2
2
2
2
2
Char4
5
1
5
3
5
Char5
2
2
2
Char6
4
1
3
4
Sari
1
1
2
Sar2
1
1
1
1
Sar3
1
1
1
1
1
1
Sar4
Lemon Bay Scrub
Cty. Pk.
2
2
2
2
2
2
2
Sar5
1
1
1
3
2
3
Sar6
1
1
3
3
Sar7
Casperson Beach
Cty. Pk./Brohard
Pk.
7
13
13
13
13
13
13
Sar8
Oscar Sherer S.P.
19
30
30
30
30
30
30
Sar9
Private reserve
1
1
1
1
1
1
1
Sano
2
1
3
2
3
Sari 4
Myakka S.F.
1
1
1
1
1
Totals
64
50
61
61
77
77
89
o
o


Probability
259
ro ;
I*
L
0
>.
S L
15
n g-r
£ h
0- |.
0 4*f
0.2*1
"
(
2 4 6
Threshold Pop Size
Fig. 5-12d. Central Brevard county quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition.


277
Table 5-14b. St. Lucie county simulation statistics
Data type
No
acquisition
Maximum
acquisition
starting
population size
23
33
x end pop. size
9.8
19.4
s.d.
5.9
7.1
percent
decline
57.4
41.2
extinction
risk
0.20
0.03
quasi-extinction
risk (10 pairs)
0.73
0.27


272
]ay Territory Locations 1.75 km dispersal buffer
Scrub Polygons
Lo Disturbance Protection Status A /Interstates
2] Lo Density Housing L......1 Protected /\/ State highways
Hi Density Housing I Pr0Psed County roads
M Ranch/Ag Water Bodies County lines
Ml 4 St. Lucie
0 4 8 12
1 : 210.000
Kilometers
+
Fig. 5-14a. St. Lucie N. Martin county map 1992 1993 jay and habitat distribution.


177
extinction risk (p=0.30; Table 5-3b). The maximum acquisition configuration, estimated
to support only 6 additional jay pairs (total of 69), had no extinction risk and a reduced
quasi-extinction risk (p=0.233).
Recommendations: The Pasco-Hemando metapopulation should be considered
poorly surveyed; fire suppression has forced jays to occupy atypical, unsurveyed habitat,
as evidenced by the recent discovery of jays at Pas2 (Pranty et al., manuscript). This
small population connects the 12 km dispersal buffer between Cross-bar and Serenova,
and may be an important acquisition. In the absence of new survey data, further
acquisition options appear very limited.
The potential for restoration of protected habitat in this metapopulation is large.
The Weeki Watchee State Park (Herl in Fig. 5-3b) is a large sand pine forest with a
dense oak understory that has had a single resident jay family residing in a small bum for
many years. This forest has the potential to support 17 or more pairs of jays, but the one
resident family may have recently disappeared (Pranty et al., manuscript). Portions of
this forest should be restored to scrub as soon as possible. The largest population of jays
in this metapopulation, and perhaps the 2nd largest jay population along the Gulf Coast,
occurs on the Cross-bar/Al-bar Wellfields (Pas3), which is currently being restored by
Pasco county (B. Pranty, pers. comm.). Habitat restoration is urgently needed at the
Starkey and Serenova properties (Pasl), as jays are nearly extirpated at this site.


5-2 la. Lake Wales Ridge patch statistics (number of jay territories for different
configurations) 339
5-2lb. Lake Wales Ridge simulation statistics 343
5-22. Metapopulation viability statistics 350
5-23. Metapopulation vulnerability ranking no acquisition (sorted by
decreasing quasi-extinction probability) 351
5-23a. Metapopulation vulnerability ranking maximum acquisition (sorted by
increasing percent protection) 352
5-24. Percent protected ranking (sorted by increasing percent protection) 353
5-25. Metapopulation priority ranking (sorted by decreasing priority) 354
5-26. Summary of recommendations (highest priority first) 355
xiv


284
z
1
£
1 O
O 8
O 6
O 4
02
/ i i
/
20
40
Threshold Pop. Size
60
-|
80
1 0
0 8
* 06
0 4
02
I
I
I
i
50 100
Threshold Pop Size
Fig. 5-15d. Martin and N. Palm Beach county quasi-extinction graphs. Top) no
acquisition, Bottom) maximum acquisition.


5-7b. Central Charlotte county acquisition map 215
5-7c. Central Charlotte county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition 217
5-7d. Central Charlotte county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition 218
5-8a. Lee and N. Collier county map 1992 1993 jay and habitat distribution 222
5-8b. Lee and N. Collier county acquisition map 223
5-8c. Lee and N. Collier county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition 225
5-8d. Lee and N. Collier county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition 226
5-9a. Flagler and N.E. Volusia county map 1992 1993 jay and habitat
distribution 230
5-9b. Flagler and N.E. Volusia county acquisition map 231
5-9c. Flagler and N.E. Volusia county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition 233
5-9d. Flagler and N.E. Volusia county quasi-extinction graphs. Top) no
acquisition, Bottom) maximum acquisition 234
5-10a. Merritt Island and S.E. Volusia county map 1992 1993 jay and habitat
distribution 238
5-1 Ob. Merritt Island and S.E. Volusia county acquisition map 239
5-10c. S.E. Volusia and Merritt Island county trajectory graphs. Top) no
acquisition, Bottom) maximum acquisition 241
5-1 Od. S.E. Volusia and Merritt Island county quasi-extinction graphs. Top) no
acquisition, Bottom) maximum acquisition 242
5-1 la. N. Brevard county map 1992 1993 jay and habitat distribution 247
5-1 lb. N. Brevard county acquisition map 248
xx


229
vulnerability and low potential for improvement. Simulations of the currently protected
configuration of 5 territories produced a high extinction (p=0.933) and quasi-extinction
risk (p=1.0) for the small population (Table 5-9b). The maximum acquisition option
produced a substantially reduced extinction risk (p=0.57; Table 5-9b), but the quasi
extinction risk remained high (p=l .0).
Recommendations: The habitat map for this metapopulation developed for the
SMP is based on old soil maps and is very outdated along the coast; it does not reflect the
extensive habitat destruction that has occurred subsequent to the production of the soil
maps. Recent aerial photographs should be used to update this habitat information.
Acquisition options are very limited along the coast in this area. The best
opportunity may be several small tracts of land near Marineland and Washington Oaks
(Flagl, Flag2). The prognosis for this small population of jays is not good, as they
probably face problems similar to those described by Breininger (1999) for the urban jays
on the the south Brevard county barrier island. Some large tracts of apparently
unoccupied scrub may still exist a few kilometers inland. Given the high risk faced by the
coastal jays and their potentially unique genetic traits, the possibility of acquiring and
restoring these unoccupied patches and translocating jays from nearby coastal areas
should be considered.


228
Flaeler-N.E. Volusia (M9)
General description: The Flagler-N.E. Volusia county metapopulation is the most
north-eastern population of jays occurring along the Atlantic Coast. This metapopulation
is isolated from the Volusia-Merritt Island metapopulation (mlO) to the south by the city
of Daytona Beach. All of the jays in this metapopulation occur near or along the beach;
consequently habitat loss due to oceanfront development has greatly reduced this
population. The SMP documented about 12 jay territories, excluding suburban jays, in
this metapopulation. Estimated potential population size after habitat restoration and full
occupancy is 5 pairs in currently protected areas, and 12 pairs maximum.
Protected areas: The only protected jays in this metapopulation occur in the N.
Peninsula State Recreation Area (Voll in Fig. 5-9b). Three protected areas had jays
prior to the SMP: Gamble Rogers Memorial State Recreation Area, Flagler Beach State
Recreation Area, and Washington Oaks State Gardens (jays still occasionally seen). Jays
occur near these parks, as well as near Marineland, but appear to occupy territories in
suburban or urban settings.
Restoration potential: Improved management of the unoccupied, protected areas
(Gamble Rogers, Flagler Beach, and Washington Oaks) might attract jays from nearby
suburban settings. The potential also exists to re-introduce jays to these parks, but for
modeling purposes jays were excluded from these areas. The population of jays at N.
Peninsula State Recreation Area as measured for the SMP was assumed to be close to
carrying capacity.
Simulation results: This metapopulation ranked 7th in vulnerability (table 5-23),
10th in percent protected (41.7%; table 5-24), and 16th in priority (table 5-25), with high


39
Fig. 2-10. Portion of the largest mainland-midland-island metapopulation in the interior,
consisting of the Lake Wales Ridge and associated smaller sand deposits. The large
central subpopulation (enclosed by the thin black line) contains nearly 800 pairs of jays.
Small subpopulations to the south and east are within known dispersal distance of the
large, central mainland. A small metapopulation to the west (in DeSoto County) contains
a single subpopulation of 21 territories. This small system qualifies as a patchy
metapopulation, since jays occur in two or more patches but the patches are so close
together that they function as a single demographic unit.


Background Landscape Image 141
Map Production 143
Statewide metapopulation map 143
1992-1993 SMP maps 143
Acquisition maps 144
GIS Database Preparation 146
Estimation of jay populations after restoration 146
Identification of protected areas 147
Assessment of unprotected areas 147
Suburban jays 148
Simulation runs 149
Repetitions and duration of simulations 149
Reserve design configurations 150
Output statistics 151
Model Validation/Calibration 152
Interpreting Simulation Results 152
Results 154
Levy (Cedar Key) (Ml) 155
Citrus-S.W. Marion (M2) 164
Pasco-Hemando (M3) 176
Manatee-S. Hillsborough (M4) 186
Sarasota-W. Charlotte (M5) 196
N. W. Charlotte (M6) 204
Central Charlotte (M7) 212
Lee and N. Collier (M8) 220
Flagler-N.E. Volusia (M9) 228
Merritt Island-S.E. Volusia and (M10) 236
N. Brevard (Mil) 244
Central Brevard (Ml2) 253
S. Brevard-Indian River-N. St. Lucie (Ml3) 261
St. Lucie N. Martin (Ml4) 270
Martin and N. Palm Beach (Ml5) 278
South Palm Beach (Ml6) 286
Ocala National Forest (Ml 7) 294
N.E. Lake (Ml 8) 302
S.W. Volusia (M19) 310
Central Lake (M20) 318
Lake Wales Ridge (M21) 326
Other Metapopulations 344
Brevard barrier island 344
Clay county 344
Osceola 344
Western Polk 345
Bright Hour Ranch 345
Recommendations 346
Ranking Metapopulation Vulnerability 346
IX


235
Table 5-9b. Flagler and N.E. Volusia county simulation statistics
Data type
Original
1992-1993
scenario
No
acquisition
Maximum
acquisition
Starting
population size
12
5
12
Mean ending
population size
2.4
0.27
2.4
s.d.
2.91
0.995
2.91
Percent
population
decline
75.8
80.1
75.8
Extinction
Risk
0.57
0.933
0.57
Quasi
extinction
Risk (10 pairs)
1.00
1.00
1.00


4-2. Kaplan-meier survival curves of daily survival rates for 10 jays released at 3
sites in Highlands county, Florida. Upper curve: maximum possible
survival; middle curve: best guess survival; lower curve: minimum
possible survival 126
4-3. Distribution of daily distances moved by released jays (solid line), and
inverse function fitted to observed movements (dashed line) 127
4-4. Comparison of dispersal data from Archbold Biological Station and
simulated dispersal distances for male jays 128
4-5. Comparison of dispersal data from Archbold Biological Station and
simulated dispersal distances for female jays 129
4-6. Comparison of stage-age data from Archbold Biological Station and
simulated stage-age data for breeders 130
4-7. Comparison of stage-age data from Archbold Biological Station and
simulated stage-age data for helpers 131
5-0. Delineations of 21 Florida Scrub-Jay metapopulations based on 1992 1993
statewide survey 145
5-la. Levy county maps 1992 1993 jay and habitat distribution 158
5-lb. Levy county acquisition map 159
5-lc. Levy county trajectory graphs. Top) no acquisition, Bottom) maximum
acquisition 161
5-ld. Levy county quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition 162
5-2a. Citrus county map 1992-1993 jay and habitat distribution 168
5-2d. S.W. Marion county acquisition map 171
5-2e. Citrus and S. Marion county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition 173
5-2f. Citrus and S. Marion county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition 174
5-3a. W. Pasco and Hernando county map 1992 1993 jay and habitat
distribution 178
XVlll


92
Florida Scrub-Jay is to simulate both types of dispersal behavior (i.e., delay-and-foray
and float).
Methods
General Approach
I simulated dispersal in the Florida Scrub-Jay with two distinct algorithms, one
for jays that engage in short forays away from their natal territory (philopatric
algorithm), another for jays that become floaters and move long distances from their natal
territory (floater algorithm). The philopatric dispersal algorithm is intended, 1) to
simulate the prevalent mode of dispersal observed in real jay populations, 2) to produce
the majority of dispersals within a given simulation, and 3) to produce the modal
distribution of dispersal distances in simulated populations. The philopatric algorithm
produces no dispersals beyond a specified radius referred to as the assessment sphere.
The floater algorithm completely determines the tail of the distribution, as philopatric
jays settle only within the radius of the assessment sphere. Together, the two algorithms
produce the combined dispersal curve; the philopatric algorithm produces dispersals
ranging from the natal territory (i.e. inheritance) out to the radius of the assessment
sphere, the floater algorithm produces dispersals beyond the assessment sphere.
Considerable information from long term, color band studies (e.g. Woolfenden
and Fitzpatrick 1984) is available to aid in the simulation of philopatric dispersal. In
contrast, much less is known about jays that disperse as floaters. Although some long
distance dispersals have been documented, the number of observed movements is small


240
Table 5-10a. S.E. Volusia and Merritt Island county patch statistics (number of jay
territories for different configurations)
Patch id
Status
1992-1993# jay
territories
No
acquisition
(restored)
Maximum
acquisition
Vol2
3
3
Vol3
4
4
Vol4
7
7
Vol5
1
1
Vol6
6
6
Vol7
4
4
Vol8
1
1
Vol9
2
2
Brev19
4
4
Brev20
3
3
Brev21
3
3
Brev22
1
1
Brev23
1
1
Brev24
1
1
Brev25
Merritt Island N.W.R. & Kennedy
377
377
377
Space Center
Brev26
Cape Canaveral Air Station
118
118
118
Totals
536
495
536


168
Scrub Polygons
^ Lo Disturbance
2¡ Lo Density Housing
Hi Density Housing
¡ggj Ranch/Ag
Protection Status
Protected
Proposed
Water Bodies
A/ Interstates
State highways
County roads
County lines
M2 Citrus
1 250,000
Kilometers
Fig. 5-2a. Citrus county map 1992-1993 jay and habitat distribution.


312
Jay Territory Locations
1.75 km dispersal buffer
Scrub Polygons
[ I Lo Disturbance
K 7 -1 Lo Density Housing IsHlIi ^rotectec*
IfV'/j Hi Density Housing Proposed
Water Bodies
Protection Status
/\/ Interstates
v State highways
County roads
County lines
Ml 9 W. Volusia
1 : 140,000
9 Kilometers
Fig. 5-19a. S.W. Volusia county map 1992 1993 jay and habitat distribution.


313
Water Bodies Ml9 W. Volusia
]ay Territory Locations
(after restoration) Protection Status
Scrub Polygons /\/ Protected
CCC## Polygon ID /''/ Unprotected
Jays outside of labeled, bold polygons are considered to be Suburban jays.
County lines
9 Kilometers
1 : 140,000
Fig. 5-19b. S.W. Volusia county acquisition map.


19
Extinction vulnerability was estimated using a single-population viability model
(Fitzpatrick et al. 1991). We then chose thresholds to delineate islands, midlands, and
mainlands, much as Mace and Lande (1991) used extinction probabilities to propose
IUCN threatened species categories.
Dispersal Distances
Between 1970 and 1993 we documented 233 successful natal dispersals from the
marked population under long-term study at Archbold Biological Station (Figure 2-4; see
also Woolfenden and Fitzpatrick 1984, 1986). Unlike the situation for most field studies
of birds (e.g., Barrowclough 1978), many characteristics of our study and the behavior of
jays themselves enhance our ability to locate dispersers that leave the main study area.
Once established as breeders, for example, Florida Scrub-Jays are long-lived and
completely sedentary. Furthermore, we have mapped in detail all scrub habitat within the
local range of the species, and we census these tracts periodically in search of dispersed
jays. (Such censuses reveal remarkably few banded dispersers among the many hundreds
of jays encountered.) Because banded Florida Scrub-Jays from our study usually are
tame to humans, both our own searches and casual encounters by local homeowners have
high likelihood of exposing any off-site dispersers to us once they become paired on a
territory. Indeed, if we assume that immigration and emigration rates are about equal in
our study area, evidence suggests that we have succeeded in locating all but a low
percentage of the jays that have departed over the 25-year period of our study. Therefore,
although some dispersers do escape our detection, our observed dispersal curve (Fig. 2-4)
can be only marginally biased toward the shorter distances.


25
expanding the distances among patches, rescue (on midlands) may be at least as
important as turnover (on islands) in Florida Scrub-Jay metapopulation dynamics.
Use of empirically derived dispersal-buffers and extinction probabilities provides
an explicit method for quantitatively describing metapopulation structure. Application of
this technique to the Florida Scrub-Jay demonstrates that a species can exhibit a variety
of metapopulation patterns across its range. Patterns of aggregation and isolation do not
conform to a single metapopulation class in the Florida Scrub-Jay. Such complex spatial
structure is probably common in nature, particularly among species with large and widely
dispersed populations restricted to a patchy habitat. Such patterns may be further
complicated by perturbations of the natural system caused by humans.
Caveats
We offer several caveats as to the generality of dispersal-buffer methodology in
conservation. (1) The technique is best suited for organisms occupying discrete
territories, home ranges, or habitat patches amenable to mapping. (2) The technique is
predicated on having a comprehensive survey. Missing data can lead to misleading
results, especially as regards connections among metapopulations or subpopulations. (3)
The technique presents a static, snapshot view of metapopulations. It does not easily
reveal important dynamics among subpopulations, such as those obtainable from an
SEPM. The viability of different configurations is best determined from SEPMs rather
than single population PVAs. (4) Populations in decline or in sinks can present an
overly optimistic picture (Thomas 1994). Indeed, we suspect that many of the island
and midland subpopulations of Florida Scrub-Jays currently are failing to replace
themselves demographically, as a result of habitat degradation from fire suppression.


227
Table 5-8b. Lee and N. Collier county simulation statistics
Data type
Original
1992-1993
scenario
No
acquisition
30% acquisition
by area
70% acquisition
by area
Maximum
acquisition
starting
population size
47
15
25
48
62
x end pop. size
0.9
1.1
2.0
4.5
5.6
s.d.
2.4
2.2
3.2
6.1
4.7
percent
decline
98.0
92.7
92.0
90.1
91.0
extinction
risk
0.87
0.73
0.67
0.43
0.40
quasi-extinction
risk (10 pairs)
1.0
1.0
1.0
0.90
0.90


58
of supporting high quality habitat are present, but high quality habitat occurs only in
small areas that were recently burned or cleared experimentally. Five of 6 experimental
plots were classified as relatively large patches of high quality habitat (see Fig. 3-21).
Only plot 6 was not classified as high quality; for unknown reasons it had a dark
reflectance when the area was photographed in March 1994. No jays have become
established in the three isolated experimental plots (2, 3, and 4), despite their appearance
as high quality. We suspect that conspecific attraction is extremely important to this
species, greatly reducing the likelihood that solitary jays will become established in
unoccupied, isolated patches.
The most important relationships I found for individual territories among the
habitat-demography variables is a decline in group size with increasing tree cover, both
within territories and within the 100 m buffer zones (Figs. 3-14 and 3-15). Large family
sizes only occur in territories with low tree cover and adjacent to low tree cover. Variance
in group size is high in the North because some groups accrue large size here, but not in
the South. Territories in or adjacent to habitat with moderate to high tree cover have
predictably small group sizes, presumably a result of successive years with poor
productivity and low survival. A similar pattern between group size and tree cover can be
seen in Figs. 3-16 and 3-17, where group size is lumped into 3 categories of roughly
equal size.
Demographic parameters other than group size showed no clear patterns with the
habitat variables at the individual territory level. Group size may be the least noisy
measure of demographic success, since it integrates past and current demographic
performance. Helper survival may be especially sensitive to habitat quality, since helpers


185
Table 5-3b. Pasco county simulation statistics
Data type
1992-1993
No
acquisition
Maximum
acquisition
starting
population size
29
63
69
x end pop. size
6.2
20.2
22.0
s.d.
4.5
10.5
percent
decline
78.6
67.9
68.2
extinction
risk
0.23
0.03
0.0
quasi-extinction
risk (10 pairs)
0.97
0.30
0.233


104
transported to their release sites in cages covered with heavy material to prevent the jays
from seeing the passing landscape and thereby potentially developing a homing direction.
All jays were released in the late afternoon.
Radio-transmitters were mounted with a technique already in use by another
researcher (Keith Tarvin) on Blue Jays at Archbold Biological Station. A 2-g transmitter
(manufactured by Wildlife Materials of Carbondale, IL ) was mounted on the back of
each jay and secured with elastic cord. The transmitters had small tubes on the front and
back that the cord was run through to create a loose, independent loop of equal size
around each wing. A small loop of cord was run between the wing loops and across the
belly to make the harness snug. The cord was kept loose enough to avoid cutting into the
skin or restrict movement or breathing, but tight enough to eliminate slack cord that feet
or branches might catch on. Newly outfitted jays were observed in an outdoor aviary for
at least half a day, and sometimes overnight. When initially rigged, jays were
preoccupied with trying to remove the outfit, suggesting that initially they might be
especially vulnerable to predators. Properly fitting harnesses were ignored after a few
hours and more normal behave resumed including taking peanuts and drinking water.
My failure to tie the knots on the harness tightly occasionally allowed jays to untie and
remove the transmitter, but only while in the aviary. I used forceps to cinch knots tightly
to reduce their size; knots were placed away from areas where they might rub against
pressure points. The elastic cord, obtainable from many fabric stores, degrades and falls
apart within a year or so; one jay that returned to Archbold wore her transmitter for a year
before it finally fell off. Her antennae also broke off after several months (reinforcing
each antennae base with epoxy might have been beneficial). Since most electronics fail


Population Size
283
40 0
u
F
20 0
20 40 60
Year
h
t
20
:
40
Year
J
60
Fig. 5-15c. Martin and N. Palm Beach county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition.


66
SNdS
dOlH
SOIS
SOOT
OVNS
0H03
NdM
0V3Q
sim
VldX
xiua
SSdO
idon
1S01
uo(podoid
Fig. 3-7. Measurements of sand, tree cover, and mixed vegetation for Spring 1994
territories on S. Sandy Hill.
Territories


374
Burgman, M., S. Ferson, and H.R. Akcakaya. 1993. Risk Assessment in Conservation
Biology. New York: Chapman & Hall. 314 pp.
Carmen, W. 1989. Behavioral ecology of the California Scrub Jay: a noncooperative
breeder with close cooperative relatives. Ph.D. dissertation. University of
California, Berkeley, CA. 134 pp.
Caughley, G. 1994. Directions in conservation biology. J. Animal Ecology 63: 215-244.
Conroy, M.J., Y. Cohen, F.C. James, Y.G. Matsino, and B.A. Maurer. 1995. Parameter
estimation, reliability, and model improvement for spatially explicit models of
animal populations. Ecological Applications 5: 17-19.
Cox, J.A. 1987. Status and distribution of the Florida Scrub Jay. Florida Omithol. Soc.
Spec. Pub. No. 3.
Christman, S.P. and W.S. Judd. 1990. Notes on plants endemic to Florida scrub. Florida
Scientist 53: 52-73.
Dawkins, R. 1976. The Selfish Gene. Oxford: Oxford University Press. 224 pp.
DeGange, A.R., J.W. Fitzpatrick, J.N. Layne, and G.E. Woolfenden. 1989. Acorn
harvesting by Florida Scrub Jays. Ecology 70: 348-356.
Doak, D.F., and L.S. Mills. 1994. A useful role for theory in conservation. Ecology 75:
615-626.
Duncan, B., D. Breininger, P. Schmalzer, and V. Larson. 1995. Validating a Florida
Scrub Jay habitat suitability model, using demography data on Kennedy Space
Center. Photogr. Eng. & Remote Sensing 61: 1361 1370.
Dunning, J.B., and D.J. Stewart. 1995. Spatially explicit population models: current
forms and future uses. Ecological Applications5: 3-11.
Ebenhard, T. 1991. Colonization in metapopulations: a review of theory and
observations. Pp. 105-121 in M.E. Gilpin and I. Hanski (eds.), Metapopulation
Dynamics: Empirical and Theoretical Investigations. Academic Press, London,
UK
E.S.R.I. 1990. Understanding GIS. Redlands, CA: Environmental Systems Research
Institute, Inc.
Fahrig, L. 1997. Relative effects of habitat loss and fragmentation on population
extinction. J. Wildlife Management 61: 603-610.


302
N.E. Lake (Ml8)
General description: The N.E. Lake metapopulation is separated from the W.
Volusia metapopulation (Ml9) by the heavily wooded St. Johns riverine system to the
west. The ONF metapopulation (Ml7) to the northwest is separated from the N.E. Lake
metapopulation by more than 30 km, with an intervening matrix of dense forest stands.
The SMP documented about 109 jay territories, excluding suburban jays, in this
metapopulation. Estimated potential population size after habitat restoration and full
occupancy is 67 pairs in currently protected areas, and 161 pairs maximum.
Protected areas: Ocala N.F.(Lakel 1), Seminole S. F. (Lake7, Lake8), Rock
Springs Run S.R. (Ora2, Lake6), Wekiwa Springs S.P. (Semi, Oral), Wekiva R.
Buffers C.A. (Semi), Yankee Lake Waste Water (Sem3).
Restoration potential: Habitat patches in this metapopulation are heavily
overgrown, and have an enormous potential for restoration. The largest patches of
occupied habitat currently are unprotected (Lake 10, Lake9), and could support many
more jays than were found during the SMP (compare the second and last data columns in
Table 5-18a).
Simulation results: This metapopulation ranked 17th in vulnerability (table 5-23),
11th in percent protected (41.6%; table 5-24), and 17th in priority (table 5-25), with low
vulnerability and low potential for improvement. The no acquisition option had a low
risk of quasi-extinction (p=0.03), no extinction risk, and a 33.7% mean percent
population decline (Table 5-18b). The maximum acquisition option has no risk of
extinction or quasi-extinction, and a 10.9% mean percent population decline (Table 5-
18b).


213
extremely vulnerable to extinction and quasi-extinction (Table 5-7b). The 70%
acquisition by area configuration is considerably improved, but still has a substantial
quasi-extinction risk (p=0.33). The maximum acquisition has a much lower quasi
extinction risk (p=0.07).
Recommendations: This metapopulation ranks 2nd in vulnerability due to the near
absence of jays on protected lands. It has a priority ranking of 3, with low protection and
high potential for improvement.
The private reserve (Charl) would benefit considerably by the acquisition of
nearby jay habitat, especially Chari 5. Substantial tracts of largely unsurveyed,
unprotected scrub occur along both sides of Prairie Creek (Chari 7, Chari 8), and jays
were documented in the western portions of these patches for the SMP. Acquisition and
restoration of these patches would greatly bolster this metapopulation. The large,
unsurveyed patch along Shell Creek also should be investigated. Consideration should be
given to adding the isolated scrub patch (Lee 5) to the proposed CARL addition to the
Babcock-Webb W.M.A.


37
t
Fig. 2-8. Examples of non-equilibrium metapopulations from North Gulf coast. Each of
the six metapopulations contains fewer than 10 pairs of jays, except for the centrally
located system that contains a single, midland-size subpopulation.
i


304
Jay Territory Locations
Scrub Polygons
1.75 km dispersal buffer
Lo Disturbance
3 Lo Density Housing
Hi Density Housing
Ranch/Ag
Protection Status
mm Protected
Proposed
Water Bodies
/\/ Interstates
State highways
County roads
. County lines
Ml 8 N.E. Lake
1 : 200,000
12
Kilometers
+
Fig. 5-18a. N.E. Lake county map 1992 1993 jay and habitat distribution.


373
Best, L.B., and D.F. Stauffer. 1986. Factors confounding the evaluation of bird-habitat
relationships. Pages 209-217 in J. Vemer, M.L. Morrison, and C.J. Ralph (eds.),
Wildlife 2000: Modeling habitat relationships of terrestrial vertebrates. Madison:
University of Wisconsin Press.
Blanchard, J., S. Jue, and A. Crook. 1998. Florida Conservation Lands 1998.
Bowman, R. and L. Averill. 1993. Demography of a suburban population of Florida
Scrub Jays. Annual Progress Report to the U.S. Fish and Wildlife Service.
Boyce, M.S. 1992. Population viability analysis. Ann. Rev. Ecol. Syst. 26: 481-506.
Breininger, D.R. 1999. Florida Scrub-Jay demography and dispersal in a fragmented
urban landscape. Auk 116: 520-527.
Breininger, D.R. 1997. Florida Scrub-Jay demography in an urban metapopulation. 1997
Final Report to Endangered Species Office, U.S. Fish and Wildlife Service,
Jacksonville, FL.
Breininger, D.R., V.L. Larson, D.M. Oddy, R.B. Smith, M.J. Barkaszi. 1996.
Demography of Florida scrub-jays in different landscapes. Auk: 113: 617-625.
Breininger, D.R., V.L. Larson, R.W. Duncan. R.B. Smith, D.M. Oddy, M.F. Goodchild.
1995. Landscape patterns of Florida Scrub Jay habitat use and demographic
success. Conservation Biology 9: 1442 1453.
Breininger, D.M., M.J. Provancha, and R.B. Smith. 1991. Mapping Florida Scrub Jay
habitat for purposes of land-use management. Photogr. Eng, and Remote Sensing
57: 1467-1474.
Breininger, D.M., M.A. Burgman, and B.M. Stith. In press. Influence of habitat quality,
catastrophes, and population size on extinction risk of the Florida Scrub-Jay.
Wildlife Society Bulletin
Brook, B.W., L. Lim, R. Harden, and R. Frankham. 1997. Does population viability
analysis software predict the behaviour of real populations? A retrospective study
on the Lord Howe Island woodhen Tricholimnas sylvestris (Sclater). Biological
Conservation 82: 119-128.
Brown, J.H., and A. Kodric-Brown. 1977. Turnover rates in insular biogeography: effect
of immigration on extinction. Ecology 58: 445-449.
Buechner, M. 1987. A geometric model of vertebrate dispersal: tests and implication.
Ecology 68: 310-318.


8
A series of simulations were run for each metapopulation based on different
reserve design scenarios. These scenarios ranged from a minimal configuration consisting
of only currently protected patches (no acquisition option), to a maximal configuration
consisting of all significant patches (complete acquisition option). For all simulations, the
assumption was made that all protected areas were restored and properly managed, and
that jays had demographic performance and densities typical of high quality habitat.
These assumptions should be viewed as optimistic. Jays outside of protected areas were
assumed to have poor demographic performance typical of suburban areas.
The output from the simulation runs included estimates of extinction, quasi
extinction (probability of falling below 10 pairs), and percent population decline.
Comparisons of these results provided the basis for ranking the vulnerability of different
metapopulations around the state. Metapopulations were ranked in terms of vulnerability
assuming no further acquisition, and in terms of potential for improvement through
acquiring all unprotected habitat. The proper uses and limitations of population modeling
are discussed.
Chapter 6 synthesizes previous chapters, focusing on some of the limitations of
metapopulation theory. The chapter closes by presenting a set of landscape rules that
provide guidelines for developing a statewide Habitat Conservation Plan for the Florida
Scrub-Jay. Adherence to these landscape rules would likely maintain the viability of
different jay populations across the state, while allowing for further loss of jays to human
development in some areas.


Population Size
193
Fig. 5-4c. Manatee and S. Hillsborough county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition.


Table 5-4a. Manatee and S. Hillsborough county patch statistics (number of jay territories for different configurations)
Patch
Status
1992-
No
30%
30%
70%
70%
Maximum
id
1993# jay
acquisition
preserved
preserved
preserved
preserved
acquisition
territories
(restored)
by contiguity
by connectivity
by contiguity
by connectivity
Har1
8
3
8
8
8
Har 2
1
Har 3
1
Har 4
4
2
3
HilH
2
3
2
5
HII2
Little Manatee R.
2
2
2
2
2
2
2
HHI3
Little Manatee R. S.
Rec. Area
1
2
2
2
2
2
2
HII4
8
3
8
8
8
HHI5
1
1
1
1
1
HII6
1
HHI7
2
2
HII8
Balm-Boyette Scrub
Pr.
1
6
6
6
6
6
6
HII9
Golden Aster Scrub
Nature Pr.
2
2
2
2
2
2
2
HHI10
4
Him 1
2
2
2
2
HilH 2
1
HilH 3
3
Man1
4
5
3
5
5
5
Man2
4
4
2
4
4
4
Man3
3
5
Man4
3
2
5
5
5
Man5
1
6
5
10
Man6
2
5
3
5
3
5
Man7
2


316
-
o.2~r
5 10 15
Threshold Pop. Size
Fig. 5-19d. S.W. Volusia county quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition


309
Table 5-18b. N.E. Lake county simulation statistics
Data type
No
acquisition
Maximum
acquisition
starting
population size
67
161
x end pop. size
44.4
143.8
s.d.
12.9
14.5
percent
decline
33.7
10.9
extinction
risk
0.0
0.0
quasi-extinction
risk (10 pairs)
0.03
0.0


Table 5-16b. South Palm Beach county simulation statistics
Data type
No
acquisition
Maximum
acquisition
starting
population size
9
16
x end pop. size
0.47
1.20
s.d.
1.34
2.40
percent
decline
94.8
91.0
extinction
risk
0.90
0.77
quasi-extinction
risk (10 pairs)
1.0
1.0


217
20 40 60
Year
Fig. 5-7c. Central Charlotte county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.


310
S.W. Volusia (Ml91
General description: The S.W. Volusia metapopulation is separated from the N.E.
Lake metapopulation (Ml 8) to the west by the St. Johns riverine system. The SMP
documented about 54 jay territories, excluding suburban jays, in this metapopulation.
Estimated potential population size after habitat restoration and full occupancy is 17 pairs
in currently protected areas, and 70 pairs maximum.
Protected areas: The only protected jays occur on the Blue Springs State Park
(VollO in Fig. 5-19b); a single family was found during the SMP.
Restoration potential: For modeling purposes, Blue Springs State Park was
estimated to support 17 families of jays after restoration (probably an overly optimistic
estimate). The largest population of jays occurs on the unprotected Stewart Ranch
(Voll9), and this population likely could support more jays, but additional information
is needed. Other patches in this metapopulation occur in the rapidly developing Deltona
area south of Deland. The scrub in this area is heavily overgrown and the restoration
potential is unknown.
Simulation results: This metapopulation ranked 11th in vulnerability (table 5-23),
15th in percent protected (24.3%; table 5-24), and 4th in priority (table 5-25), with high
vulnerability and high potential for improvement. The no acquisition option had a high
risk of extinction (p=0.33) and quasi-extinction risk (p=0.90), and a 54.9% mean
population decline (Table 5-19b). The maximum acquisition option had no risk of
extinction or quasi-extinction, and a 20.2% mean population decline (Table 5-19b).


CHAPTER 5
METAPOPULATION VIABILITY ANALYSIS OF THE FLORIDA SCRUB-JAY
Introduction and Objectives
The Florida Scrub-Jay is a Federally threatened bird species that occurs only in
Florida. Formerly found in 39 of 40 counties of peninsular Florida, by 1981 the Florida
Scrub-Jay was known to be extirpated from 7 counties (Broward, Dade, Duval, Gilchrist,
Hendry, Pinellas, and St. Johns; Cox 1987). A 1992-1993 statewide mapping project
(SMP) added Alachua and Clay to the list of extirpations, and estimated that less than 10
breeding pairs remained in 6 other counties (Flagler, Hardee, Hernando, Levy, Orange,
and Putnam; Fitzpatrick et al. 1994). In nearly all other counties jay populations have
declined drastically. Since 1993, a severe decline to near extinction was documented on
the barrier island south of Patrick Air Force Base in Brevard county (Breininger 1999),
and a huge decline exceeding 50% was documented on mainland S. Brevard county
during the same period (Breininger 1998). Undoubtedly, similar drastic declines are
occurring throughout the state. Humans are directly responsible for this dramatic
population reduction, primarily through suppression of natural fires which are necessary
to maintain high rates of reproduction and survival in scrub-jay populations, and through
outright destruction or degradation of jay habitat due to roads, housing developments, and
agriculture (e.g. citrus groves).
132


141
territories were given parameter values for either high quality habitat or suburbs (see
parameter settings in Table 5-1).
Background Landscape Image
Bit-mapped GIS files provided the landscape setting upon which the population
dynamics and dispersal movements were simulated. These files were created by
overlaying the scrub patches in the 1992-1993 SMP database onto a statewide habitat
classification map produced by the Florida Game and Freshwater Fish Commission
(FGFWFC) based on 1985-1989 Landsat Thematic Mapper data (Kautz et al. 1993). All
GIS files had a spatial resolution of 30 m. The original landcover types coded in the
FGFWFC classification are shown in Table 4-2 (chapter 4), along with the associated
attractiveness values that affected the movement of floaters. In the simulations
completed for this chapter, the landscape was assumed to be static through time.


Table 5-22. Metapopulation viability statistics.
Metapopulation
Protected
population
size
Maximum
population
size
Extinction
prob. (no
acquisition)
Extinction
prob
(maximum
acquisition)
Quasi-ext.
prob
(no
acquisition)
Quasi-ext.
prob.
(maximum
acquisition)
% decline
(no
acquisition)
% decline
(maximum
acquisition)
M1 Levy
17
75
1.0
0.0
1.0
0.0
100.0
1.3
M2 Citrus
47
125
0.17
0 0
047
0.33
70.6
55.0
M3 Pasco
63
69
0.03
00
0.30
0.233
67.9
68.2
M4 Manatee
36
145
0.97
0 30
1.0
0 90
95.3
96.6
M5 Sarasota
50
89
0 03
0 0
0.10
0.0
47.4
47 3
M6 N.W. Charlotte
28
56
067
0.07
1.0
0.30
91.8
60.7
M7 Cen. Charlotte
5
61
1.0
0.07
1.0
0 07
100.0
65.4
M8 Lee
15
62
0.73
040
1.0
0 90
92.7
90.9
M9 Flagler
5
12
0.93
0.57
1.0
1.0
80.1
75.8
M10 Merritt Island
495
536
0.0
0.0
00
0.0
0.70
6.5
M11 N. Brevard
4
110
1.0
0.0
1.0
0.0
100.0
14.5
M12 Cen. Brevard
5
40
1.0
0.0
1.0
0.10
100.0
36.3
M13 S. Brevard
62
165
0.07
0.0
0.20
0.0
54 0
24.8
M14 St. Lucie
23
37
0.20
0.03
0.73
0.27
57.4
19.4
M15 Martin
85
120
0.0
00
00
0.0
17.3
7.5
M16 Palm Beach
9
13
0.90
0.77
1.0
1.0
94.8
91.0
M17- Ocala N.F.
448
???
0 0
0.0
roo
0.0
25.0
???
M18-N.E. Lake
67
161
0.0
0.0
003
0.0
33.7
10.9
M19-W. Volusia
17
70
0.33
0.0
0.90
0.0
54 9
20.2
M20 Cen. Lake
0
20
1.0
0.70
1.0
1.0
100.0
90.4
M21 Lake Wales
535
858
0.0
0.0
0.0
0.0
18.5
17.4


75
>
O
O
0)
0)
Q.
O
U-
Q-
0.6
0.5
0.4
0.3
0.2
0.1
0.0
Large Me(F
Group size
Small
Fig. 3-16. Group size (small = 2, medium = 3, large = 4 7 jays) vs. percent tree cover
within 100 m buffer (North and South populations pooled). Trend towards smaller group
size with higher tree cover is not significant.


138
Field studies have shown that in the Florida Scrub-Jay, environmental stochasticity of
fecundity and mortality are positively correlated (Woolfenden and Fitzpatrick 1984).
Annual mortality. Breeder and helper annual mortality rates are based on the
current territory configuration (see Table 5-1). All breeder vacancies available to
dispersers are created in this step. Helpers do not actually die in this step, but are
considered to have disappeared. The floater frequency parameter later determines
whether they die or become floaters (see dispersal section below). Juveniles are not
subjected to mortality in this step since their annual mortality is already reflected in the
fecundity rate. Epidemics occur with an annual probability of 0.05, and increase the
mortality rate of juveniles by 100% and adult mortality by 20%. These percentages are
conservative; the actual mortality rates may be considerably higher, although the long
term values are unknown. Fitzpatrick and Woolfenden (1991) reduced adult survival to
0.55 and juvenile survival to 0.0 in their population model.
Promotion to next stage. Survivors of the mortality step are promoted to the
appropriate experienced stage: novice breeders to experienced breeders; 1-year helpers
to experienced helpers.
Dispersal. Two types of dispersal are modeled: philopatric dispersal forays
around the natal territory within an assessment sphere, and floater dispersal long
distance search in which a disperser permanently leaves its area of intimate knowledge
and moves through the landscape searching for breeder vacancies or empty territories.
All helpers that survive the mortality step engage in philopatric dispersal. The
order in which philopatric dispersal events occur mirrors the dominance hierarchy of
jays: males dominate females, older jays dominates younger, jays closer to their natal


I certify that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy-
Professor of Wildlife Ecology and
Conservation
1 certify- that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and qualitv.
as a dissertation for the degree of Doctor of Philosophy
rofessor c
Conservation
Cochairman
Ecologv and
1 certify- that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and qualitv.
as a dissertation for the degree of Doctor of Philosophy
Associate Professor of Wildlife Ecology
and Conservation
1 certify- that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality,
as a dissertation for the degree of Doctor of Philosophy
CD.
Jon v Allen
^Profassor of Entomology and
Jematologv


246
properties (especially around Tico and Seminole Ranch), and the acquisition and
restoration of some of the southern habitat patches (Brevl5, Brevl6, Brevl7,
Brevl8).


226
20
40
Threshold Pop Size
60
Fig. 5-8d. Lee and N. Collier county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition.


88
Individual-based models are appealing because they allow the inclusion of almost
any biological detail, giving them unrivaled realism (Huston et al. 1988; Judson 1994).
Because they often incorporate behavioral mechanisms, the parameters in an individual-
based model tend to have clear biological meaning and typically are directly measurable
in the field. The increased level of detail inherent in an individual-based model also
provides more opportunities to compare model output with different types of field data
for validation purposes. Finally, the increased realism of individual-based models makes
them more likely to reveal dynamics that would otherwise be missed in a less detailed
model.
Individual-based models have been developed for a few avian species, including
Bachman Sparrow (Pulliam et al. 1992), Northern Spotted Owl (McKelvey et al. 1993),
American Wood Stork (Wolff 1994), Helmeted Honey Eater (McCarthy 1996), and Red-
cockaded Woodpecker (Letcher et al. 1998). The excellent field data and well-known
behavioral characteristics of the Florida Scrub-Jay make this species an excellent
candidate for an individual-based modeling approach.
The overall objectives of this chapter are as follows:
Develop a set of algorithms and parameters specifically for the Florida Scrub-
Jay to simulate dispersal in an individual-based model.
Calibrate the dispersal module using long term field data from Archbold
Biological Station data, and radiotelemetry data acquired from a displace-and-
release experiment.


270
St. Lucie N. Martin (Ml4)
General description: The St. Lucie-N. Martin metapopulation is separated from
the S. Brevard-Indian River-N. St. Lucie metapopulation (Ml 3) by the city of Fort Pierce
to the north, and the Martin-N. Palm Beach metapopulation (Ml5) by the St. Lucie Inlet
to the south (see map in Fig. 5-14a,b). The SMP documented about 28 jay territories,
excluding suburban jays, in this metapopulation. Estimated potential population size after
habitat restoration and full occupancy is 23 pairs in currently protected areas, and 33
pairs maximum.
Protected areas: Savannas State Park (Stl4), and portions of the S. Savannas
CARL site (Marl).
Restoration potential: The densities of jays measured by the SMP probably were
close to maximum, even though habitat conditions were not optimal. For modeling
purposes, the only population that was increased over the SMP was at Savannas State
Park (15 pairs increased to 20).
Simulation results: This metapopulation ranked 12th in vulnerability (table 5-23),
7th in percent protected (62.2%; table 5-24), and 7th in priority (table 5-25), with high
vulnerability and high potential for improvement. Quasi-extinction and extinction risk
was substantially higher for the no acquisition option (p=0.73 and 0.20 respectively)
compared to the the maximum acquisition option (p=0.27 and 0.03 respectively), even
though the difference in population size was small (14 territories; Table 5-14b).
Recommendations: Habitat restoration and proper management of the Savannas
State Park is crucial to this metapopulation. Acquisition of jay habitat within and south of
the S. Savannas CARL site (Marl, Mar2, Mar3, Mar4) will substantially


32
/
Fig. 2-3. Schematic depiction of different kinds of metapopulations, illustrating use of
dispersal-distance buffers to predict recolonization rates among subpopulations. Dotted
lines separate functional subpopulations, based on frequency of dispersal beyond them.
Solid lines separate metapopulations, based on poor likelihood of dispersal among them,
A. Patchy metapopulation. B. Classical metapopulation. C. Nonequilibrium
metapopulations. D. Mainland-island metapopulation.


152
ending population size from the starting population size, dividing this difference by the
starting population size, and multiplying this result by 100.
Model Validation/Calibration
Efforts to validate this model using long-term data from Archbold Biological
Station (ABS) (Woolfenden and Fitzpatrick 1984) are described in chapter 4. The
demographic parameters measured at ABS were used to parameterize jays in optimal
habitat (table 5-1). A constraint analysis and a small radiotelemetry study (chapter 4)
were used to develop parameters for the dispersal algorithm.
Interpreting Simulation Results
Two key assumptions have a large influence on the simulation results reported in
this chapter. First, the assumption has been made that the density of jays in all occupied
habitat is the maximum expected if the habitat were fully restored. The second
assumption is that the demographic performance of jays is maximal, corresponding to
measurements made in optimal habitat in the long-term study at Archbold Biological
Station (Woolfenden and Fitzpatrick 1984). Both of these assumptions are likely to be
very optimistic for most metapopulations around the state.
Many habitat patches, including those in public ownership, are not currently
managed properly for scrub-jays. In the absence of aggressive management, jay
demographic success decreases; small changes of 10% can produce dramatic declines in
population size and rapid extinction (Fitzpatrick and Woolfenden 1986; Breininger 1998;
Root 1998). The simulations results presented in this chapter assume continuous, optimal
habitat conditions. Unfortunately, even proper management of habitat may not guarantee


Table 5-23a. Metapopulation vulnerability ranking maximum acquisition (sorted by increasing percent protection).
Rank
Metapopulation
Quasi-ext.
prob (maximum
acquisition)
Extinction
prob.
(maximum
acquisition)
Protected
population
size
Maximum
population
size
1
M16 Palm Beach
1.0
0.77
9
13
2
M20 Cen. Lake
1.0
0.70
0
20
3
M9 Flagler
1.0
0.57
5
12
4
M8 Lee
0.90
0.40
15
62
5
M4 Manatee
0 90
0.30
36
145
6
M2 Citrus
0.33
0.0
47
125
7
M6-N.W. Charlotte
0.30
0.07
28
56
8
M14 St. Lucie
0.27
0.03
23
37
9
M3 Pasco
0.23
0.0
63
69
10
M12 Cen. Brevard
0.10
0.0
5
40
11
M7 Cen. Charlotte
0.07
0.07
5
61
12
M19-W. Volusia
0.0
0.0
17
70
13
M1 Levy
0.0
0.0
17
75
14
M5 Sarasota
0.0
0.0
50
89
15
M11 N. Brevard
0.0
0.0
4
110
16
M15 Martin
0.0
0.0
85
120
17
M13 S. Brevard
0.0
0.0
62
165
18
M18-N E. Lake
0.0
0.0
67
161
19
M10 Merritt Island
0.0
0.0
495
536
20
M21 Lake Wales
0 0
0.0
535
858
21
M17-Ocala N.F.
0.0
00
448
???


Table 5-6a. N. W. Charlotte county patch statistics (number of jay territories for different configurations)
Patch id
Status
1992-1993#
No
70% preserved
70% preserved
Maximum
jay territories
acquisition
(restored)
by connectivity
by area
acquisition
Char7
11
8
11
11
Char8
4
3
4
4
Char9
Charlotte
Harbor State
Buffer Pr.
11
11
11
11
11
CharlO
3
2
1
3
Chari 1
2
2
3
3
Chari 1a
SWFMD?
2
5
5
5
5
Char12
Private Pr.
6
6
6
6
6
Sari 1
2
2
4
Sar12
2
2
4
Sar13
Myakka S.F
1
6
6
6
6
Totals
44
28
47
47
56
208


219
Table 5-7b. Central Charlotte county simulation statistics
Data type
Original
1992-1993
scenario
No
acquisition
70% acquisition
by connectivity
70% acquisition
by area
Maximum
acquisition
starting
population size
31
5
44
44
61
x end pop. size
11.7
0.0
21.3
35.5
19.4
s.d.
6.3
0.0
10.4
10.4
11.3
percent
decline
62.2
100.0
51.6
19.3
65.4
extinction
risk
0.13
1.0
0.17
0.10
0.07
quasi-extinction
risk (10 pairs)
0.70
1.0
0.30
0.33
0.07


Recommendations: This metapopulation has been the focus of intensive
acquisition efforts, and most major occupied habitat patches that are relatively
327
undeveloped appear to be acquired or in the process of being acquired. However, none of
the jays in Glades county are protected; many occur on the extensive landholdings of the
Lykes Brothers Corporation. In Highlands county, the Hendrie Ranch (High2) is an
important unprotected population that doesnt appear on most acquisition lists. The jay
population at Highlands Hammock State Park (Highl5) is very small and somewhat
isolated. Habitat restoration and additional acquisition is needed for this population. In
Polk county, unprotected jay habitat (Polk7) exists that would help connect Tiger
Creek (Polk6) and Catfish Creek (Polkl 1). The tiny population of jays at Lake
Kissimmee State Park (Polk 10) would benefit from the acquisition of jays and habitat
at Polk8 and Polk9. The most significant northerly population of jays occurs on the
northeast margin of Lake Marion (Polk 13) on unprotected habitat, and doesnt appear
on most acquisition lists.


256
Water Bodies
Jay Territory Locations
(after restoration)
_ Scrub Polygons
CCC## Polygon ID
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Protection Status
/V/ Protected .
s\/ Unprotected / V County lines
Ml2 Central Brevard
0 3 6 9 Kilometers
1 : 160,000 Hlh
Fig. 5-12b. Centra] Brevard county acquisition map.
Brev42
Brev43
Wickham
County
Brev44
Reg.


Scrub Polygons
I 1 Lo Disturbance
2] Lo Density Housing
IrWSj Hi Density Housing
Ranch/Ag
Protection Status
f i Protected
2 Proposed
Water Bodies
/\y Interstates
State highways
, County roads
County lines
M6 N.W. Charlotte
0 4 8 12
1 : 200,000
Kilometers
Fig. 5-6a. N. W. Charlotte county map 1992 1993 jay and habitat distribution.


67
o
>
O
O
*->
c
CD
O
u.
O
CL
M North
B South
Territory 100 m 200 m 400 m
Buffer Distance
Fig. 3-8. Percent tree cover (mean and standard deviation) for 4 zones (inside territories,
100, 200, 400 m buffer) North vs. South Sandy Hill. South population shows
significantly higher tree cover within all zones compared to North population.


38
Fig. 2-9. Example of a "classical" metapopulation from five counties in Central Florida.
Note the occurrence of jays in small islands of intermediate distance from one another.
L


182
Table 5-3a. Pasco and Hernando county patch statistics (number of jay territories for different
configurations)
Patch id
Status
1992-1993#
No
Maximum
jay territories
acquisition
(restored)
acquisition
Her1
Weeki Wachee
1
17
17
Pas1
Starkey Wellfield &
Serenova
3
13
13
Pas2
4
Pas3
Cross-Bar/AI-bar Wellfield
15
19
19
Pas4
2
2
Pas5
Alston Tract
2
2
2
Pas6
Green Swamp W.M.A.
6
12
12
Totals
29
63
69


52
Comparison of demographic performance of jays in North vs. South for two years
(1994, 1995) showed significant differences only for nonbreeder survival (Table 3-3;
Mann-Whitney U Test, Z = -2.396, P = 0.017). Group size was nearly significant (P =
0.060). Number of fledglings produced, fledgling survival, yearling survival, and breeder
survival were not significantly different.
Relationships between group size and percent tree cover within all territories (Fig.
3-13), and group size and percent tree cover within 100 m buffers (Fig. 3-14) showed
decreasing group size with increasing tree cover. I lumped group size into 3 categories of
roughly equal size and looked for differences in tree cover at the territory (Fig. 3-15) and
100 m buffer (Fig. 3-16). Large families (4 to 7 jays per group; n = 13) had lower median
and variance in tree cover within and adjacent to their territories compared to medium (3
jays per group; n = 7) and small (2 jays per group; n=15) families. The differences,
however, were not significant (Kruskal-Wallis one-way analysis of variance; tree cover P
= 0.205; 100 m buffer P = 0.186). I lumped the 3 group categores into 2 group size
categories (2-3 jays per group; 4-7 jays per group) and performed the same analysis, but
the results were not significant.
Figures 3-17, 3-18, and 3-19 show side-by-side views of the classified and raw
images for N. Sandy Hill, the N. portion of S. Sandy Hill, and the S. portion of S. Sandy
Hill respectively. Jay territories are outlined in black. The names of jay territories are
shown in Fig. 3-1. Three colors on the classified images correspond to tree cover (green),
bare sand (white) and mixed shrubby or grassy vegetation (brown).
Figures 3-20 and 3-21 show presumed habitat quality as computed from the three
HSI variables for the N. and S. portion of S. Sandy Hill respectively. High quality habitat


96
Estimating floater mortality and mobility
Long distance dispersers have two daily survival rates: one for floaters within
scrub, another for floaters outside of scrub. Within scrub, the daily survival rate is
assumed to be higher than outside of scrub, and similar to survival rates of nondispersing
jays of similar age and same sex. The survival rate for dispersers outside of scrub is
drawn from the best-guess Kaplan-Meier curve (Fig. 4-2) derived from the
displacement experiment described below. This curve was hard-coded into the model and
daily survival rates were drawn from the distribution and applied to floaters moving in
the matrix between scrub patches. An option to use a constant daily survival rate (as in
scrub habitat) was also included in the model and evaluated in the constraint analysis.
Each disperser moves until it exceeds a daily-distance-moved threshold value
selected for each jay from a function that approximates the observed distribution of daily
move distances. This distribution was derived from field data obtained from the
displacement experiment described below. The function that approximates the field
distribution was generated by the curve fitting procedure of SPSS (ver. 7.5). The
distribution of distances excluded 0 distances (i.e. days when jays did not move see later
discussion on displacement experiment). Once the daily-distance-moved threshold is
exceeded, each jays daily mortality rate is used to determine if the jay survives to the
next day. These steps are repeated until each jay dies, finds a mate or vacant territory, or
leaves the simulation area. The order in which dispersers move is randomized each time
all jays have taken a step. Jays that leave the area are considered dead (i.e. there is no
immigration from outside the simulation area). In contrast to short distant dispersers, long
distance dispersers do not return home.


LIST OF TABLES
Table page
2-1. Summary information for 42 Florida Scrub-Jay metapopulations, including
type of metapopulation, number of territories (pairs of jays), and number of
subpopulations 28
3-1. Demographic and habitat parameters for North and South Sandy Hill (1994
- 1995) 81
3-2. Kolmogorov-Smimov test for normality of demographic and habitat
variables (* significantly different from normal) 84
3-3. Mann-Whitney_U test for differences in demographic and habitat variables
between North and South jay populations (* significantly) 85
4-1. Landcover types from statewide habitat map (Kautz et al. 1993) used in
simulations and associated floater attractiveness values 120
43. Summary of philopatric dispersal rules showing sex differences and rules
used to implement the algorithm 122
4-4. Summary of jay movement data obtained from displacement experiment
(distances in km) 123
4-5. Summary of constraint analysis for 9 simulation scenarios (50 years x 30
repetitions) showing number of colonizations from Lake Wales Ridge to
Bright Hour Ranch, DeSoto county, Florida 124
5-1. Demographic and dispersal parameter settings for jays in optimal and
suburban conditions 142
5-la. Levy county patch statistics (number of jay territories for different
configurations) 160
5-lb. Levy county (Cedar Key) simulation statistics 163
5-2a. Citrus and S. Marion county patch statistics (number of jay territories for
different configurations) 172
xi


Table 5-13a. S. Brevard-Indian River-N. St. Lucie county patch statistics (number of jay territories for different configurations)
Patch id
Status
1992-
No
30%
30%
70%
70%
Maximum
1993# jay
acquisition
preserved
preserved
preserved
preserved
acquisition
territories
(restored)
by connectivity
by area
by connectivity
by area
Brev30
Malabar Scrub Sanct.
10
10
10
10
10
10
10
Brev31
Jordan (proposed)
23
4
5
11
16
23
Brev32a
Valkaria (partial)
7
7
7
7
7
7
7
Brev32
(Valkaria proposed)
17
9
17
17
Brev33
5
3
5
2
5
Brev34
8
5
7
8
Brev35
St. Sebastian River
St. Pk. & Micco Scrub
Sanct.
28
28
28
28
28
28
28
Brev36
11
3
11
6
11
11
Brev37
2
2
2
2
2
Brev38
(Babcock proposed)
6
2
6
5
6
6
InRil
St. Sebastian River
State Park
12
12
12
12
12
12
lnR¡2
Sebastian municipal
airport
8
5
8
8
8
8
lnRi3
Private HCP
3
3
3
3
3
3
3
lnRi4
Wabasso Scrub Pr.
2
2
2
2
2
2
2
lnRi5
4
2
4
lnRi6
5
2
5
lnR¡7
1
1
1
lnRi8
2
2
2
lnRi9
5
3
5
5
5
InRilO
7
5
7
7
StLu1
St. Lucie airport
3
2
3
StLu2
1
1
1
Totals
153
62
91
91
129
136
165
266


281
St Lucie
Martin
Seabranch St. Pk.
Jonathan Dickinson St. Pk.
Hills Natural Area
L. Universe Scrub Pr.
Station Preserve
Scrub Jay Preserve
]ay Territory Locations
(after restoration) Protection Status
3 Scrub Polygons /\/ Protected
CCC## Polygon ID / V UnProtected
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Water Bodies
County lines
M 1 5 Martin SC N.
0 5 10
1 : 250,000
Palm Beach
15 Kilometers
Fig. 5-15b. Martin and N. Palm Beach county acquisition map.


70
O)
>
O
U
O)
O)
c
O)
u
a>
CL
Jays
Available
u n 3;!L TLree cover within jay territories vs. total tree cover in North vs. South Sandy
Hill. Note that jays select habitat with lower tree cover in both areas.


34
V)
0)
JZ
o
*->
re
Q.
o
0>
'5.
3
CJ
O
O
c
o
'
o
O.
O
0.
1.00
o
o
o
O
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
o
O
o
o
o
o
o
o
o
o
o
o
(N
CD
CO
o
CN
CD
o
o
o
o
T
T
r
T
CM
CO
kO
Interpatch Distance (m)
Fig. 2-5. Proportion of suitable habitat patches occupied by Florida Scrub-Jays as a
function of their distance to the nearest separate patch of occupied habitat. Occupancy
rates are high (nearly 90 %) for patches up to 2 km apart and decline monotonically to 12
km. Note the scale change after 16 km.


97
Habitat attractiveness
Five attractiveness values are used in the model: 0 for a repulsive landcover that
jays do not enter, 1 for an unattractive landcover, 2 for a neutral landcover, 3 for a
somewhat attractive landcover, and 4 for a highly attractive landcover. The attractiveness
values assigned to different landcover classes in the GIS files (described in the previous
GIS section) are provided in table 4-1.
Floater detection radius
The floater detection radius is the maximum distance at which a disperser can be
expected to detect another jay or vacant territory. A starting point for estimating this
parameter is the distance between tape playback census points, as recommended by
Fitzpatrick et al. (1991b p. 13): Adequate spacing between transects can be estimated
roughly as the distance at which a person listening to the tape directly in front of the
speaker perceives the bird to be no more than about 100 meters away. A distance of
100 to 200 meters between transects and between stations is generally adequate when
using a good-quality, hand-held cassette player broadcasting at full volume. Jays no
doubt see each other, especially during territorial display flights, at greater distances than
they can hear each other. A value of 450 meters was selected as the default value for the
detection radius.


207
jay Territory Locations
(after restoration)
Scrub Polygons
CCC## Polygon ID
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Protection Status
/\f Protected
A/Unprotected
Water Bodies
County lines
M6 N.W. Charlotte
1 : 200,000
12
Kilometers
Fig. 5-6b. N. W. Charlotte county acquisition map.


Table 5-lb. Levy county (Cedar Key) simulation statistics
Patch Name
Data type
Original
1992-1993
scenario
No
acquisition
30% acquisition
by area
70% acquisition
by area
Maximum
acquisition
Cedar Key
Scrub Reserve
starting
population size
4
17
34
54
75
x end pop. size
0
12
31
52
74
s.d.
0.8
5
6
5
5
percent
decline
100.0
29.4
8.8
3.7
1.3
extinction
risk
1.0
0.1
0
(8)
0
(26)
0
(45)
quasi-extinction
risk (10 pairs)

0.82
0.05
0
0


18
Harrisons metapopulation classes can be described using this island-midland-
mainland nomenclature as follows: a nonequilibrium metapopulation is a system of
one or more islands (i.e., extinction-prone subpopulations), with a total population size
too small to persist. A classical metapopulation is a system of island-size
subpopulations large enough and close enough together and of sufficient total size to
allow persistence. Any system containing a midland or mainland (by definition) cannot
be a nonequilibrium or classical metapopulation, as all subpopulations in the latter
systems are extinction-prone. A patchy metapopulation is a set of patches close enough
together to form a single subpopulation of sufficient size to persist (i.e. a midland or
mainland). Mainland-island metapopulations are self-explanatory.
Explicit reference to midlands-extinction-vulnerable patches of intermediate
population size-produces metapopulation types not described in Harrison (1991).
Systems with, for example, several midlands, or a mainland with several midlands, are
possible. We illustrate some of these configurations by applying our nomenclature,
quantitatively, to the Florida Scrub-Jay.
Metapopulation Structure of the Florida Scrub-Jav
Application of the above scheme to any species requires choosing two dispersal
buffer distances and two threshold values for extinction-vulnerability among single
populations. Here, for the Florida Scrub-Jay, we based each of these values on
empirically gathered biological data. Buffer distances were derived from long-term field
studies of marked individuals, and from information garnered on the statewide survey
regarding occupancy of habitat patches at various distances from source populations.


239
MIO
Jay Territory Locations
(after restoration) Protection Status
Scrub Polygons /\/ Protected
CCC## Polygon ID /V Unprotected
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Water Bodies
County lines
Merritt Island St
S.E. Volusia
12
1 : 310,000
18 Kilometers
Fig. 5-10b. Merritt Island and S.E. Volusia county acquisition map.


Table 4-4. Summary of jay movement data obtained from displacement experiment (distances in km).
Jay
color
band id
Territory
of
origin a
Release
c
site
Max. daily
movement4*
Active mean
daily
movement
Inactive mean
daily movement
Cumulative
distance
traveled
Start-
end
dist8
Days
h. i
moving
Days
alive*k
G-MR
MIDR
1
1.25
1.25
1.25
1.25
1.0
1
1
GZ-O
EERR
1
8.33
2.97
1.22
12.2
6.67
4 [6]
[10]
{ABS}
YB-GM
Disney*5
2
15-20
5.33-5.83
2.33-2.67
53.3-58.3
3.33
10
22
R-MO
BYRD
2
(4.17)
(4.17)
(2.17)
(4.17)
(4.17)
(1)
(1)
R-MF
TP68
2
(4.17)
(2.0)
(1.03)
(6.17)
(1.66)
(3)
(6)
R-MA
CWHP
1
(1.67)
(0.77)
(0.30)
(3.83)
19.2
(13)
{ABS}
LZ-A
DEAD
3
4.17
4.17
1.38
4.17
4.17
1
3
Y-MW
MARE
3
2.58
2.42
1.22
4.83
1.00
2
4
Y-CR
DRUM
3
0.42
0.38
0.38
0.75
0.75
2
2
Z-YY
TRVN
3
(4.67)
(2.38)
(2.38)
(21.7)
16.6
(1)
{ABS}
Mean
(4.67-5.17)
(2.60 2.63)
(1.37- 1.40)
(11.2- 11.7)
5.83
3.3
3.9
a Territory of origin from the south (experimental) tract of Archbold Biological Station.
b Jay obtained in mitigation deal from Disney property in Orange Co., Florida.
c Release sites (Highlands county, Florida): 1 = Large citrus grove with nearby sandhill on W. side of Grassy Lake; 2 =
MacArthur Agroecology Research Center improved pasture and numerous cabbage palm/oak hammocks; 3 =
Downtown Lake Placid.
if Numbers in parentheses indicate distances or days measured for birds lost for unknown reasons.
e Days when birds were not moving were excluded in distance calculations.
/All days were included in distance calculation, including days with no movement.
g Straight line distance between release site and last known location.
h Number of days when bird moved more than 0.33 km.
/ Numbers in brackets include days when jays were in scrub habitat.
j Number of days jays were known to be alive. See d and /' for explanation of numbers in parentheses or brackets.
k {ABS} indicates jay successfully returned to its territory of origin at Archbold Biological Station.


22
rare event beyond about 12 km from an occupied patch of habitat. We use this distance to
identify metapopulations that have become essentially demographically independent from
one another (i.e., the outer dispersal-buffer).
We selected the distance of 3.5 km (about 2 miles) as an inner dispersal-buffer to
delineate subpopulations. We choose this figure because: 1) behavioral information from
a variety of sources, including radiotracking data (B. Stith, unpubl.), shows that jays
begin to show reluctance to crossing habitat gaps at about this size (and at much smaller
gaps where open water or closed-canopy forest are involved); 2) known dispersals of
many banded jays included habitat gaps up to 3.5 km, but their frequency declines
dramatically thereafter; 3) the observed dispersal curve from Archbold (Fig. 2-4) shows
that in good habitat, more than 85% of dispersals by females, and fully 97% of those by
males, are shorter than 3.5 km; 4) patch occupancy data (Fig. 2-5) show significant
decline in colonization rates at distances above 3.5 km.
Population Viability Analysis
PVA based on a simulation model incorporating demographic (but not genetic)
stochasticity and periodic, catastrophic epidemics (Fitzpatrick et al. 1991; Woolfenden
and Fitzpatrick 1991) provided a quantitative method for defining boundaries along the
island (extinction prone), midland (vulnerable), and mainland (extinction resistant)
continuum (but see Taylor 1995). Among the several methods for expressing extinction
vulnerability (e.g., Burgman et al. 1993; Boyce 1992; Caughley 1994) we elect the
simple approach of specifying time-specific probability of persistence of populations of a
given size.


I certifV that I have read this study and that in my opinion it conforms to
acceptable standards of scholarly presentation and is fully adequate, in scope and quality.
This dissertation was submitted to the Graduate Faculty of the College of
Agriculture and Life Sciences and to the Graduate School and was accepted as partial
fulfillment of the requirements for the degree of Doctor^of Philosophy
December. 1999
Dean. College of Agriculture and Life
Sciences
Dean. Graduate School


370
however, additional rules of a non-spatial nature that should be added to the landscape
rules. Among the possible candidates, nothing is more important than the effects of
habitat quality. Florida Scrub-Jay populations will decline drastically and
deterministically as habitat quality declines (Woolfenden and Fitzpatrick 1991; Root
1997). It is unfortunate for the Florida Scrub-Jay that fire suppression by humans has
greatly reduced habitat quality in many areas. Thus, a rule to the effect that habitat
quality should be maintained at a high level is an obvious addition to the landscape rules.
Restoration efforts at Kennedy Space Center, Florida, suggest that the use of fire alone to
restore densely overgrown coastal scrub oaks may be ineffective (D. Breininger, pers.
comm.). Swain et al. (1995) suggest that unbumed scrub close to public roads or with
high pine canopy cover will be difficult to restore. The restoration experiment mentioned
in chapter 3 suggests that jays may be reluctant to colonize newly restored, vacant
habitat. Modeling results of Breininger et al. (in press) suggest that effective restoration
depends on having a surplus of local helpers. Although much remains to be learned about
habitat restoration, the selection of habitat patches to be restored clearly should take into
account spatial factors, especially the proximity of healthy jay populations that can
provide colonists to the habitat being restored.
The spatially explicit, individual-based modeling approach provides a powerful
framework for investigating conservation issues in a repeatable, quantifiable fashion.
The model I developed for this dissertation could be modified to incorporate features
such as dynamically changing landscapes, fire, road effects, genetics, etc. The
technology and programming tools are capable of this and much more. What is lacking is
the necessary field data to calibrate and validate the model. Perhaps the biggest problem


Probability
194
i oT
i
i-
0 8!
0 6~
r
04
l
0 2
r
50
100
Threshold Pop. Size
150
1 oT
|-
o.aT
i!
r!
611
0 4
0.2
50
100
Threshold Pop. Size
150
Fig. 5-4d. Manatee and S. Hillsborough county quasi-extinction graphs. Top) no
acquisition, Bottom) maximum acquisition.


Population Size
183
U
V)
L
20
Year
40
60
|-
i i 1 1 1 i 1 i ( i ¡ i |_
20 40 60
Year
Fig. 5-3e. Pasco and Hernando county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.


367
the information gleaned from modeling performed for this dissertation and by previous
modeling efforts.
Conserving Florida Scrub-Jay Metapopulations
Human-induced fragmentation and habitat loss already have split the Florida
Scrub-Jay into numerous metapopulations that are now effectively isolated from one
another. Further habitat loss will have the inevitable effect of driving each
metapopulation down an ever-steepening gradient of endangerment: mainland-midland
configurations will become midland-island ones, classical configurations will become
nonequilibrium ones, which in turn are headed inexorably to extinction. At different
stages of this process, conservation strategies should vary. For systems still containing
mainlands, preserving the mainlands usually overrides other concerns, as these large
subpopulations have the greatest role in persistence of the system. As mainlands are lost,
and subpopulations shift towards configurations of islands and midlands, conservation
emphasis should shift from maintaining area to maintaining connectivity. In this phase,
priority should be placed on keeping contiguous territories together, preserving centrally
located patches, and minimizing distances among patches, thereby facilitating philopatric
dispersal and maintaining opportunities for recolonization or rescue (Hanski 1994). As
patch size and connectivity both become problematic (i.e., approaching nonequilibrium
configuration), drastic measures are appropriate, such as intensive habitat restoration,
perhaps coupled with translocation and reintroduction as a substitute for natural dispersal.


13
Federally-owned lands were not surveyed for this project. Those with large
populations of jays include: Cape Canaveral Air Force Station, Merritt Island National
Wildlife Refuge, Canaveral National Seashore, and Ocala National Forest. Florida Scrub-
Jays at each of these areas are currently under study, so for the statewide summary we
used estimates of numbers and locations of jays provided by the respective biologists
conducting those studies.
To archive, map, and analyze the statewide data we developed a series of map
layers by means of a GIS at Archbold Biological Station. PC and Sun ARC/INFO were
used to input all GIS data (E.S.R.I. 1990). Habitat patches (both occupied and
unoccupied) and jay locations, originally hand-drawn on soil or topographic maps
(usually at 1:24000 scale), were digitized. Patch characteristics and jay family sizes were
entered into accompanying data files. Map layers included current and historic range of
jays, current distribution of suitable and potential habitat, and locations and numbers of
jay families encountered.
Statewide Survey: Results
We estimate that as of 1993 the total population of Florida Scrub-Jays consisted
of about 4,000 pairs (Fig. 2-1; Fitzpatrick et al. 1994). Both total numbers and overall
geographic range have decreased dramatically during this century (Cox 1987). In recent
decades the species has been extirpated from 10 of 39 formerly occupied counties, and it
is now reduced to fewer than 10 pairs in 5 additional counties (detailed tabulations in
Fitzpatrick et al., in prep.). Detailed, site-by site comparison of our survey with Cox's
(1987) suggests that the species may have declined as much as 25% to 50% during the
last decade alone.


122
Table 4-3. Summary of philopatric dispersal rules showing sex differences and rules used
to implement the algorithm.
Male
Female
Algorithmic inmlementation
Smaller assessment
sphere
Larger assessment
sphere
Only move to vacancies within
the specified assessment sphere.
Natal Inheritance
No natal inheritance
If both breeders die, resident
male helper has priority over
nonresident male helpers.
Extreme competitive
advantage near home
Nominal competitive
advantage near home
Sort territories within
assessment sphere by distance.
Allow each female to select
nearest vacancy within
assessment sphere. For males,
divide sphere into 3 nested
subspheres. Process each
subsphere separately, innermost
to outermost. Allow each male
to select nearest vacancy within
each subsphere.
Prefer unpaired breeders
over vacant territories
Prefer unpaired breeders
over vacant territories
Complete search for unpaired
breeders for all dispersers
before searching for vacant
territories.
Older helpers dominate
yearling helpers
Older helpers dominate
yearling helpers
During search for unpaired
breeders or vacant territories,
process all older helpers before
yearling helpers.


60
North Sandy Hill
dtch;
Florida Scrub Jay Territories
Spring 1994
South Sandy Hill
IN
A
2 Kilometers
Fig. 3-1. Map of Scrub-Jay Territories Spring 1994. Dividing line between North and
South populations is the Kissimmee Rd..


Population Size
233
Fig. 5-9c. Flagler and N.E. Volusia county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition.


42
objective of this chapter is to test whether remote sensing can accurately measure habitat
features that explain Florida Scrub-Jay demographic performance.
Methods
All image processing and GIS work was completed on a Sun workstation or an
Intel 486-based PC using Arc/Info (ver. 4.3D) and Erdas (ver. 7.5) software.
Image Source
Regular color, black and white, and color infrared photography were evaluated for
use in this project. Color infrared photography was flown for much of the APAFR
through the NAPP program of the U.S.G.S. in March of 1994.1 selected this photography
for three primary reasons: separation between sandy and grassy areas was more distinct,
wider coverage per frame made image mosaicing easier, and the photography is available
for study sites outside of the APAFR that are currently under investigation for Florida
Scrub-Jay conservation. Hence, results of this study could be extended to other areas.
Two photos covered the entire study area. Frame 6980-207 covered the area I
refer to as N. Sandy Hill, which is between Kissimmee and Smith roads, and part of S.
Sandy Hill south to Submarine Lake. Frame 6980-205 covered the remainder of S. Sandy
Hill south to the southern fence line.


115
behavioral flexibility. For example, in the Seychelles warbler (Acrocephalus
sechellensis), dispersers from high quality territories delay dispersal when high quality
territories are unavailable, while dispersers from low quality territories, where survival is
low, do not delay dispersal but instead settle for a low quality territory (Komdeur 1992).
Florida Scrub-Jays do exhibit some flexibility in dispersal, as suggested by the behavior
of suburban jays. For example, Bowman (1994), Thaxton and Hingten (1995), and
Breininger (1999) have documented traits in suburban jays that have much in common
with the floating behavior of the Western species, including reduced delays in dispersal
and much greater dispersal distances. Within good habitat floating behavior in Florida
Scrub-Jays is likely to be rare, but fire suppression may lower habitat quality sufficiently
to encourage floating and excessive numbers of helpers at a territory might encourage
subordinates to disperse early (so-called saturation dispersal). The possibility that
Florida Scrub-Jays may switch between these two dispersal strategies needs further
investigation.
The telemetry displacement study provided a number of useful findings that were
incorporated into the floater module. Jays clearly avoided open habitats lacking tree or
shrub cover, such as water bodies and pasture. In contrast, they moved freely through a
variety of landscapes with at least some tree or shrub cover, including citrus groves,
sandhill, pasture with small, scattered oak hammocks, and suburbs with trees. This
information was incorporated into the model by assigning different attractiveness values
to various landcover classes based on the amount cover each class is likely to provide.
Table 4-3 shows the original landcover classes encoded in the Florida Game and


80
Habitat Quality
S. Sandy Hill
Ex_plotr
/V Experimental Plots
Terr94
A/ 1994 Jay Territories
Roads
A. / Roads
0.5 0 0.5 1 Kilometers
Fig. 3-21. Habitat quality map of S. portion ofSouth Sandy Hill.


291
Fig. 5-16c. South Palm Beach county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.


378
McCarthy, M.A. 1996. Extinction dynamics of the helmeted honeyeater: effects of
demography, stochasticity, inbreeding and spatial structure. Ecological Modelling
85:151-163.
McCarthy, M.A., D.C. Franklin, and M.A. Burgman. 1994. The importance of
demographic uncertainty: an example from the helmeted honeyeater. Biological
Conservation 67: 135-142.
McDonald, D.B. and H. Caswell. 1993. Matrix methods for avian demography. Current
Ornithology 10: 139-184.
McGarigal, K. and B. Marks. 1994. FRAGSTATS: spatial pattern analysis program for
quantifying landscape structure. Corvallis, OR: Forest Science Dept., Oregon
State University.
McGowan, K.J. and G.E. Woolfenden. 1989. A sentinel system in the Florida Scrub Jay.
Animal Behavior 37: 1000-1006.
McKelvey, K., B. Noon, and R. Lamberson. 1993. Conservation planning for species
occupying fragmented landscapes: the case of the northern spotted owl. Pages
424-452 in Kingsolver, J., P. Karieva, and R. Huey (eds.), Biotic interactions and
global change. Sunderland, MA: Sinauer.
Meyers, R.L. and J.J. Ewel (eds.). 1990. Ecosystems of Florida. University of Central
Florida Press, Orlando, Florida.
Miller, G.L. and B.W. Carroll. 1989. Modeling vertebrate dispersal distances: alternatives
to the geometric distribution. Ecology 70: 977-986.
Mills, L. S., S. Hayes, C. Baldwin, M. Wisdom, J. Citta, D. Mattson, and K. Murphy.
1996. Factors leading to different viability predictions for a grizzly bear data set.
Conservation Biology 10: 863-873.
Mooij, W and D. DeAngelis. 1999. Error propagation in spatially explicit dispersal
models. Conservation Biology 11: 1298-1306.
Morrison, M.L., B.G. Marcot, and R.W. Mannan. 1992. Wildlife-habitat Relationships:
Concepts and applications. Madison, WI: University of Wisconsin Press.
Mumme, R. and T. Below. In press. Evaluation of translocation for the threatened Florida
Scrub-Jay. Wildlife Management.
Mumme, R.J., S.J. Schoech, G.E. Woolfenden, and J.W. Fitzpatrick. In revision. Life
and death in the fast lane: demographic consequences of road mortality in the
Florida Scrub-Jay. Conservation Biology.


16
patch, and for most terrestrial species it asymptotically approaches zero at some point
farther away than Harrisons single, discrete dispersal buffer. Therefore, we extend
Harrisons diagrammatic approach by adding a second buffer to delineate the distance
beyond which dispersal is effectively reduced to zero. We maintain that this second
buffer functionally identifies separate metapopulations. We acknowledge that
connectivity actually should be represented graphically as a continuous surface of
dispersal probabilities. However, discrete boundaries, placed at biologically meaningful
(and empirically determined) distances, greatly simplify the description of
metapopulations. They also provide explicit, repeatable methodology for comparative or
modeling purposes.
Harrisons metapopulation types may be characterized using these two buffers
(Fig. 2-3). In patchy systems (Fig. 2-3a), every patch belongs to the same
subpopulation, so they are all enclosed within a single, inner dispersal buffer. Classical
systems (Fig. 2-3b) have small subpopulations separately encircled, representing the fact
that each may go extinct temporarily, or may be rescued before going extinct (Brown
and Kodric-Brown 1977), both by way of colonization from another subpopulation
enclosed within the outer buffer. The simplest nonequilibrium systems (Fig. 2-3c) are
represented as bulls-eyes around small, isolated subpopulations. A mainland-island
metapopulation (Fig. 2-3d) has a large subpopulation and several small ones within a
single outer buffer.
The important point is that even more complicated patterns may be common in
nature, arising from combinations or intermediate cases, and many of these are not easily
fit into Harrisons (1991) four metapopulation classes. To deal with such complications,


102
that producing larger numbers of female floaters may inflate female survival rates
because many may survive if they find vacancies outside the assessment sphere, recall
that disappearance rates (Dtotai) for females are substantially higher than males, so more
female floaters must survive to achieve the equilibrium rate (Deq in Table 4-2).
Based on the above considerations, upper and lower bounds for Pfioater are listed in
Table 4-2, along with a best guess.. The sensitivity of the model to different Pfioater
settings likely will vary considerably with the configuration of the landscape, and will be
investigated at a later time. A constraint analysis was performed to assess model
sensitivity for the Archbold setting as described in a section below.
Floater algorithm development and calibration
Whereas a wealth of biological data is available to aid in the simulation of short
distance dispersal, much less is known about long distance dispersal. A few long distance
dispersals by Archbold birds have been documented during comprehensive off-station
surveys for color-banded birds, but these established birds provided little information
about the process by which they moved and became established. Radiotelemetry offers
the best hope for documenting the movement and interactions of long distance dispersers,
but such studies face serious logistic difficulties (Koenig et al. 1996). Foremost among
these is the low likelihood of tagging a reasonable number of jays that then become long
distance dispersers. Most birds will disperse a short distance, yielding little or no
information useful for modeling long distance dispersal. By knowing the sex, age, and
dominance of birds within a study area, the likelihood of tagging long distance dispersers
might be increased, but even then most of the prime candidates for dispersal will move
only a small distance. A further complication for the present study is that transmitters


Table 5-2a. Citrus and S. Marion county patch statistics (number of jay territories for different configurations)
Patch id
Status
1992-1993#
No
30%
30%
70%
70%
Maximum
jay territories
acquisition
preserved
preserved
preserved
preserved
acquisition
(restored)
by contiguity
by connectivity
by contiguity
by connectivity
Citrl
Crystal River
St. But. Pr.
6
12
12
12
12
12
12
Citr 2
1
2
1
2
2
2
Citr 3
3
6
7
7
Citr 4
4
2
4
4
4
Citr 5
4
2
2
4
Citr 6
Potts Preserve
1
6
6
6
6
6
6
Citr 7
3
3
3
Marl
Cross FI.
Greenway
1
1
1
1
1
1
Mar2
19
10
5
19
12
19
Mar3
6
1
4
4
6
Mar4
3
3
3
3
3
Mar5
Cross FI.
Greenway
8
8
8
8
8
8
8
Mar6
Cross FI.
Greenway
3
3
3
3
3
3
3
Mar7
14
3
6
11
14
Mar8
14
7
2
14
10
14
Mar9
2
2
2
2
Sumtl
Half Moon
W.M.A.
17
17
17
17
17
17
17
Totals
108
47
69
69
101
101
125


35
Fig. 2-6. Statewide jay distribution map with dispersal buffers. Shaded areas with thin,
solid lines depict subpopulations of jays within easy dispersal distance (3.5 km) of one
another. Thick lines delineate demograpnically independent metapopulations separated
from each other by at least 12 km.


258
Year
10 0
20 40 60
Year
Fig. 5-12c. Central Brevard county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.
I


339
Table 5-2 la. Lake Wales Ridge patch statistics (number of jay territories for different
configurations)
Patch id
Status
1992-1993#
iay territories
No acquisition
(restored)
Maximum
acquisition
Char22
2
2
Gladel
33
33
Glade2
2
2
Glade3
11
11
Glade4
41
41
Glade5
1
1
Highl
Platt Branch Mit. Pk.
14
14
14
High2
40
40
High3
4
4
High4
Archbold Biol. St. & Lake
141
134
141
Placid W.E.A.
High6
(proposed)
8
8
8
High7
Lake Wales Ridge W.E.A.
20
24
26
High8
Lake Wales Ridge W.E.A.
23
25
27
High9
8
8
HighIO
7
7
Highl 1
Lake June Scrub S.P.
18
19
21
High12
30
28
30
Highl 3
Lake Wales Ridge W.E.A.
10
8
10
High14
11
11
Highl 5
Highlands Hammock S.P.
8
9
12
High16
7
12
High17
Carter Creek
36
48
Highl 8
7
7
Highl 9
Sun 'n Lakes
10
11
Okeel
11
11
Osc2
2
2
2
Polkl
2
5
Polk3
Avon Park Air Force Range
114
152
153
Polk4
4
4
4
Polk5
2
6


380
Root, K. 1998. Evaluating the effects of habitat quality, connectivity, and catastrophes on
a threatened species. Ecological Applications 8: 854-865.
Ruckelhaus, M., C. Hartway, and P. Karieva. 1997. Assessing the data requirements of
spatially explicit dispersal models. Conservation Biology 11: 1298-1306.
Rykiel, E. 1996. Testing ecological models: the meaning of validation. Ecological
Modeling 90: 229-244.
Shaffer, M.L. 1981. Minimum population sizes for species conservation. BioScience 31:
131-134.
Simberloff, D. 1988. The contribution of population and community biology to
conservation science. Ann. Rev. Ecol. Svst. 19: 473-511.
Sjogren, P. 1991. Extinction and isolation gradients in metapopulations: the case of the
pool frog (Rana lessonae). Pp. 135-147 in M.E. Gilpin and I. Hanski (eds.),
Metapopulation Dynamics: Empirical and Theoretical Investigations. London:
Academic Press.
Smith, A. 1980. Temporal changes in insular populations of the Pika (Ochotona
princeps). Ecology 61: 8-13.
Smith, A. and M. Peacock. 1990. Conspecific attraction and the determination of
metapopulation colonization rates. Conservation Biology 4: 320-323.
Soule, M. (ed.). 1987. Viable populations for conservation. Cambridge, UK: Cambridge
U. Press.
South, A. 1999. Dispersal in spatially explicit population models. Conservation Biology
4: 1039- 1046.
Stenseth, N. and W. Lidicker (eds.). 1992. Animal Dispersal: Small Mammals as a
Model. New York: Chapman and Hall. 366 pp.
Stith, B.M., J.W. Fitzpatrick, G.E. Woolfenden, and B. Pranty. 1996. Classifying and
conserving metapopulations: a case study of the Florida Scrub Jay. Pages 187-216
in D. McCullough (ed.), Metapopulations and Wildlife Conservation.
Washington, DC: Island Press. 429 pp.
Swain, H.M., P.A. Schmalzer, D.R. Breininger, K.V. Root, S.A. Bergen, S.R. Boyle, and
S.MacCaffree. 1995. Scrub Conservation and Development Plan: Brevard
County. Appendix B: Biological Consultants Report to the Natural Resources
Management Division of Brevard County, Florida. August 1995.


212
Central Charlotte (M7)
General description: The central Charlotte metapopulation is isolated from the
nearby northwest Charlotte metapopulation by the Peace River to the west. Most of the
habitat occurs along Prairie and Shell Creek, which drain into the Peace River. Two
somewhat isolated populations occur south of Punta Gorda (Char22) and into Lee
county (Lee5), the latter patch occurring near a proposed Carl addition to the Babcock-
Webb W.M.A. The SMP documented about 31 jay territories, excluding suburban jays, in
this metapopulation. Estimated potential population size after habitat restoration and full
occupancy is 5 pairs in currently protected areas, and 61 pairs maximum.
Protected areas: No jays are protected on public lands in this metapopulation; a
small population of jays occurs on a private reserve (Charl).
Restoration potential: The restoration potential for most patches in this
metapopulation probably is not significantly greater than the jay densities measured for
the SMP. However, two large incompletely surveyed patches occur along Prairie Creek
(Chari 7, Chari 8) and may have large restoration potential. For modeling purposes,
these patches were estimated to support considerably more jays than were found during
the SMP (Table 5-7a). Both patches probably have the potential to support substantially
more jays than the estimates used here. A large, unsurveyed patch east of Char20 along
Shell Creek may harbor jays, but no jays were included in any of the simulations.
Simulation results: This metapopulation ranked 2nd in vulnerability (table 5-23),
19th in percent protected (8.2%; table 5-24), and 3rd in priority (table 5-25), with high
vulnerability and high potential for improvement. The no acquisition option is


5-1 le. N. Brevard county trajectory graphs. Top) 30% acquisition, Bottom) 70%
acquisition 250
5-1 Id. N. Brevard county quasi-extinction graphs. Top) 30% acquisition, Bottom)
70% acquisition 251
5-12a. Central Brevard county map 1992 1993 jay and habitat distribution 255
5-12b. Central Brevard county acquisition map 256
5-12c. Central Brevard county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition 258
5-12d. Central Brevard county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition 259
5-13a. S. Brevard-Indian River-N. St. Lucie Metapopulation county map 1992 -
1993 jay and habitat distribution 264
5-13b. S. Brevard-Indian River-N. St. Lucie county acquisition map 265
5-13c. S. Brevard-Indian River-N. St. Lucie county trajectory graphs. Top) no
acquisition, Bottom) 30% acquisition by area 267
5-13d. S. Brevard-Indian River-N. St. Lucie county quasi-extinction graphs. Top)
no acquisition, Bottom) 30% acquisition by area 268
5-14a. St. Lucie N. Martin county map 1992 1993 jay and habitat
distribution 272
5-14b. St. Lucie N. Martin county acquisition map 273
5-14c. St. Lucie N. Martin county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition 275
5-14d. St. Lucie N. Martin county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition 276
5-15a. Martin and N. Palm Beach county map 1992 1993 jay and habitat
distribution 280
5-15b. Martin and N. Palm Beach county acquisition map 281
5-15c. Martin and N. Palm Beach county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition 283
xxi


Probability
324
1 OT
0 6
0 6
0 4
0.2
5 10
Threshold Pop Size
Fig. 5-20d. Central Lake county quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition.


Table 5-4b. Manatee and S. Hillsborough county simulation statistics
Data type
No
acquisition
70% acquisition
by connectivity
70% acquisition
by area
Maximum
acquisition
starting
population size
36
112
112
145
x end pop. size
0.2
5.5
2.4
4.9
s.d.
0.7
8.2
4.0
4.6
percent
decline
95.3
96.6
extinction
risk
0.97
0.43
0.60
0.30
quasi-extinction
risk (10 pairs)
1.0
0.90
0.97
0.90


3-10. Percent tree cover for individual territories for 4 zones (inside territories,
100,200,400 m buffer) South territories 69
3-11. Tree cover within jay territories vs. total tree cover in North vs. South
Sandy Hill. Note that jays select habitat with lower tree cover in both areas 70
3-12. Percent bare sand (mean and standard deviation) for 4 zones (inside
territories, 100, 200, 400 m buffer) North vs. South Sandy Hill.
Differences between two areas are not significant 71
3-13. Group size vs. percent tree cover within all territories (North and South
populations pooled). Trend towards smaller group size with higher tree
cover is not significant 72
3-14. Group size vs. percent tree cover within 100 m buffer for all territories
(North and South populations pooled). Trend towards smaller group size
with higher tree cover is not significant 73
3-15. Group size (small = 2, medium = 3, large = 4-7 jays) vs. percent tree cover
within all territories (North and South populations pooled). Trend towards
smaller group size with higher tree cover is not significant 74
3-16. Group size (small = 2, medium = 3, large = 4-7 jays) vs. percent tree cover
within 100 m buffer (North and South populations pooled). Trend towards
smaller group size with higher tree cover is not significant 75
3-17. Images and territories (black polygons) of North Sandy Hill. Right: color-
infrared image. Left: classified image (white = bare sand; green = trees;
brown = shrubs/grass; black = water) 76
3-18. Images and territories (black polygons) of N. portion of South Sandy Hill.
Right: color-infrared image. Left: classified image (white = bare sand;
green = trees; brown = shrubs/grass; black = water) 77
3-19. Images and territories (black polygons) of S. portion of South Sandy Hill.
Right: color-infrared image. Left: classified image (white = bare sand;
green = trees; brown = shrubs/grass; black = water) 78
3-20. Habitat quality map of N. portion of South Sandy Hill 79
3-21. Habitat quality map of S. portion of South Sandy Hill 80
4-1. Daily distances moved and number of days movements were tracked for 10
jays released at 3 sites in Highlands county, Florida 125
xvii


61
Fig. 3-2. Habitat Suitability Index graphs for percent bare sand, percent tree cover, and
distance to forest (modified from Duncan et al. 1995).


297
]ay Territory Locations
(after restoration)
Scrub Polygons
CCC## Polygon ID / V
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Protection Status
/\/ Protected
A/ Unprotected
Water Bodies
County lines
Ml 7 Ocala National Forest
0 6 12 18 Kilometers
1 : 320,000
Fig. 5-17b. Ocala National Forest county acquisition map.


Population Size
299
Fig. 5-17c. Ocala National Forest county trajectory graphs. No acquisition.


263
1998). This population decline is not predicted by the model, and illustrates clearly the
influence of the model parameter settings on the simulation results, which assume
optimal habitat conditions. Similar declines likely are occurring in many other parts of
the state, and highlight the importance of habitat restoration and management; land
acquisition alone is insufficient to preserve jay populations.


TABLE OF CONTENTS
Eige
ACKNOWLEDGMENTS iii
LIST OF TABLES xii
LIST OF FIGURES xv
ABSTRACT xxiw
CHAPTERS
1 INTRODUCTION 1
Historical Background 1
Biological Background 3
Objectives 4
2 CLASSIFYING FLORIDA SCRUB-JAY METAPOPULATIONS 9
Introduction 9
Statewide Survey of the Florida Scrub-Jay 11
Statewide Survey: Methods 11
Statewide Survey: Results 13
A Method for Classifying Metapopulations 14
Metapopulation Structure of the Florida Scrub-Jay 18
Dispersal Distances 19
Patch Occupancy 20
Population Viability Analysis 22
Metapopulation Structure 23
Caveats 25
3 REMOTE SENSING OF FLORIDA SCRUB-JAY HABITAT 40
Introduction 40
Methods 42
Image Source 42
Image Scanning and Conversion 43
Image Rectification and Mosaicing 43
vii


262
extinction (Table 5-13b and Fig. 5-13f; quasi-extinction risk = 0.167) and show a
substantial population decline (Table 5-13b and Fig. 5-13e; mean ending population size
= 32.5; percent decline = 50.0). Intermediate configurations all show a substantial
reduction in quasi-extinction, but only the 70% preserved by area and maximum
acquisition have no quasi-extinction risk (Table 5-13b).
Recommendations: Comparison of simulations with equal population size but
different spatial configuration (area vs. connectivity) indicates that maintaining contiguity
of territories is more important than maintaining connectivity. Given this criteria,
unprotected patches such as Jordan (Brev31), Valkaria (Brev32), Babcock
(Brev38), and Brev36 should be high priority acquisition sites. South of Sebastian
along the coast the jay populations are in small, isolated populations that are extinction-
prone. Acquisition of habitat (e.g. InRi5) near the Wabasso Scrub Preserve (InRi4)
may bolster the long-term viability of that population. Significant numbers of unprotected
jays occur along the Ten Mile Ridge in Indian River county (InRi9 and InRilO), and
other populations likely occur nearby (Breininger 1998), making this an important area
for future acquisition. Habitat management plans are needed at the 3 airports known to
have jays within this metapopulation (St. Lucie County, Sebastian Municipal, and
Valkaria).
Recent surveys and color-band studies of Brev30, Brev31, Brev32,
Brev35, Brev36, Brev37, and Brev38 by Breininger (1998) documented an
alarming population decline exceeding 50% since 1993. This decline is due primarily to
habitat degradation resulting from fire suppression (Breininger 1998). An epidemic in
late 1997-early 1998 also may have had a significant effect in this region (Breininger


98
Estimating floater frequency
For most bird species, all surviving young depart from their natal territory as
floaters upon reaching sexual maturity, often during the first year of life. From a
modeling standpoint, the proportion of young that become floaters is simply the
proportion that survive to dispersal age. Upon reaching dispersal age, all young
disappear permanently from their natal territory and become floaters searching for
breeder vacancies or unoccupied habitat. For Florida Scrub-Jays and other species that
delay dispersal, the situation is more difficult to model, as some dispersal age young may
return to their natal territory after making unsuccessful searches for nearby breeder
vacancies, while other young may depart permanently from their natal territory.
The objective of this section is to estimate the annual proportion of helpers that
become floaters (Dfioater)- The dispersal model uses Dfloater to establish the proportion of
jays that become floaters from the pool of jays that disappear. The starting point for
estimating Dfioater is D,otai, the total proportion of helpers that disappear annually. All
floaters must come from this pool of disappearing jays, so Dfioater must be less than or
equal to Dtotai. D,otai is calculated from field data as the number of helpers disappearing
during a year divided by the original number present at the start of the year. Dt0tai for
helpers is shown in Table 4-2 (taken from appendix M of Woolfenden and Fitzpatrick
1984).
From the modeling standpoint, Dtotai has two components: jays that disappear by
dying locally (i.e. within their assessment sphere), and jays that disappear by becoming
floaters who move beyond their assessment sphere and either die or become breeders.
This can be represented as equation 4-1:


287
vulnerability and low potential for improvement. Simulations of the no acquisition and
maximum acquisition option both show a high quasi-extinction risk (p=l .0 for both)
and a high extinction risk (p=0.90 and 0.77 respectively).
Recommendations: Because of the small size of this metapopulation and its
individual patches, and the heavily urbanized landscape which subjects these jays to
additional sources of mortality, the long-term prognosis for this metapopulation is poor.
An experimental program involving intensive human intervention might be necessary to
maintain this metapopulation. Such a program likely would involve intensive habitat
management, food supplementation, predator control, control of vehicular speed, and
translocation of jays to supplement local population declines. No such program has been
attempted for scrub-jays, but because of the huge human population in this area which
could support and benefit from such a progam, this metapopulation might be the best
candidate for such an experiment.
The two most significant habitat patches that remain unprotected include the
Overlook Scrub (PB14), and the Tradewind / Winchester Site (PB13). Both of these
patches occur near the already protected Rolling Green Scrub Preserve (PB11) and
Galaxy School Scrub Preserve (PB12). Acquisition and restoration of both of these
sites would benefit the two nearby protected areas.


Probability
242
02 !
L
h
I l
100
l|j. I i__J 1 I ; ( 1 L
I I
200 300 400
Threshold Pop Size
i.o7
8"t
i-
f
6
l
J-
04[
-
0.27
i i i i iii
100
r
i
j
j
i
r
J
J
J
r¡
200 300 400 500
Threshold Pop Size
Fig. 5-1 Od. S.E. Volusia and Merritt Island county quasi-extinction graphs. Top) no
acquisition, Bottom) maximum acquisition.


181
Water Bodies
County lines
E. Pasco
jay Territory Locations
(after restoration) Protection Status
Scrub Polygons /\/ Protected
CCC Polygon ID A/ Unprotected
Jays outside of labeled, bold polygons are considered to be Suburban jays.
10
15
1 : 250,000
Kllometei
Fig. 5-3d. E. Pasco and Hernando county acquisition map.


71
30
25
JD
a
Territory 100 m 200 m 400 m
Buffer Distance
Fig. 3-12. Percent bare sand (mean and standard deviation) for 4 zones (inside territories,
100, 200,400 m buffer) North vs. South Sandy Hill. Differences between two areas are
not significant.


Table 5-la. Levy county patch statistics (number of jay territories for different configurations)
Patch id
Status
1992-1993#
jay territories
No
acquisition
(restored)
30%
preserved
by contiguity
70%
preserved
by contiguity
Maximum
acquisition
Levyl
Cedar Key
4
17
17
17
17
Scrub Reserve
Levy2
4
17
37
58
Totals
8
17
34
54
75


130
Fig. 4-6. Comparison of stage-age data from Archbold Biological Station and simulated
stage-age data for breeders.


214
Jay Territory Locations
Scrub Polygons
2 Lo Disturbance
2 Lo Density Housing
l/VYj Hi Density Housing
gggj Ranch/Ag
1.75 km dispersal buffer
Protection Status
Hifll- Protected
~ Proposed
Water Bodies
/V'interstates M7 Central Charlotte
A
State highways
County roads
County lines
9 Kilometers
1 : 250,000
Fig. 5-7a. Central Charlotte county map 1992 1993 jay and habitat distribution..


329
)ay Territory Locations
1.75 km dispersal buffer
Scrub Polygons
^ Lo Disturbance
3 Lo Density Housing
IMftj Hi Density Housing
Ranch/Ag
Protection Status
: ) Protected
J Proposed
Water Bodies
/\/ Interstates
State highways
County roads
County lines
M21 Glades
10
1 : 250,000
15 Kilometers
-+
Fig. 5-2lb. Lake Wales Ridge map 1992 1993 jay and habitat distribution, Glades
County.


62
Fig. 3-3. Ilius.ra.ive map of 100,200. and 400 m buffer zones around LOST territory.


dwna
dNMl
ISdd
Hoia
10H1
1A10
IdlN
dVOS
aans
HAOX
NHd
dM 01
sad*
AAJkA
3H10
Aano
iavd
OdVO
a
6 o
to
6
to
6
to n -
6 o' 6 o'
o
uo|)iodaid
Fig. 3-6. Measurements of sand, tree cover, and mixed vegetation for Spring 1994
territories on N. Sandy Hill.
Tcrrltoriei


186
Manatee-S. Hillsborough (M41
General description: This metapopulation occurs predominantly in Manatee and
S. Hillsborough counties, with a few jays occurring in west Hardee and northeast
Sarasota county. The configuration of habitat in this region as mapped by the SMP is
very unusual, occurring as small patches isolated from each other by small to moderate
distances (1-10 km). During the 1992-1993 survey, a significant number of patches
were on private lands that could not be accessed. The SMP documented about 65 jay
territories, excluding suburban jays, in this metapopulation. Estimated potential
population size after habitat restoration and full occupancy is 36 pairs in currently
protected areas, and 145 pairs maximum.
Protected areas: Golden Aster Scrub Nature Preserve (Hill9), Balm-Boyette
Scrub (H118), Little Manatee River (H112), Little Manatee River State Recreation
Area (H113), Duette Park (Manl5), Lake Manatee Lower Watershed (ManlO),
Beker (Man 12, Man 16), Lake Manatee State Recreation Area (Man 17), Myakka
River State Park (Sari5), Verna Wellfield (Sarl9).
Restoration potential: Many of the patches in this metapopulation are heavily
overgrown, but the restoration potential of the numerous small patches may not be much
greater than the SMP population estimates. Nevertheless, habitat restoration is urgently
needed in many of the protected areas, since local populations of jays are small and very
vulnerable to local extinction.
Simulation results: This metapopulation ranked 6th in vulnerability (table 5-23),
14th in percent protected (24.8%; table 5-24), and 10th in priority (table 5-25), with high
vulnerability and moderate potential for improvement.


231
]ay Territory Locations
(after restoration)
Scrub Polygons
Protection Status
/\/ Protected
/\J Unprotected
CCC## Polygon ID
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Water Bodies M9 Flagler sc N.E. Volusia
' County lines 0 4 8 12 Kilomete!
_ N
1 : 250,000
Fig. 5-9b. Flagler and N.E. Volusia county acquisition map.


158
Jay Territory Locations ^ 1.75 km dispersal buffer
Scrub Polygons
Lo Disturbance Protection Status
3 L Density Housing Protected
|WM Hi Density Housing Dr rmA
I Ranch/Ag
H Proposed
Water Bodies
/\/ Interstates
*\/ State highway
County roads
County lines
Ml Cedar Key
0 5 10 15 Kilometers
1 : 250,000 ^
Fig. 5-la. Levy county map 1992 1993 jay and habitat distribution.


116
Freshwater Fish Commissions statewide classification, and each classs respective
attractiveness value as assigned by me following the telemetry study.
The animation graphics displayed during a simulation run allows the movement
of long distance dispersers to be visually checked. Dispersers behaved as expected,
completely avoiding landcover types with highly repulsive values, moving in an
essentially straight line in habitat with neutral attractiveness, and preferentially turning
towards habitat with high attractiveness values. Simulated floaters could be observed
moving long distances, and the dispersal curves generated by the program showed that
they occasionally settled after moving long distances.
Quantitative comparison of movement patterns generated by the model with those
recorded for radiotagged jays was not feasible because the sample size, resolution of data
recorded, and movements of several jays all were too small for meaningful statistical
tests. Simple circular statistics could be used with a better data set to test jay movements
for directional bias (Batschelet 1981; Zar 1996). Turchin (1998) outlines several more
sophisticated analyses, but besides requiring better data sets, Turchins analyses are
designed to test for correlated random walk, which is more typical of insects than the
landscape-directed movements of vertebrates. Statistical tests appropriate for this model
probably do not exist, and would likely involve boot-strap tests based on simulations
developed for specific landscape settings.
Although Florida Scrub-Jays are fairly weak fliers (Woolfenden and Fitzpatrick
1996), jays that were experimentally displaced were capable of moving fairly long
distances, up to a maximum of 20 km per day, with an average daily movement of about
2.67 km. This information was used to parameterize the daily movement parameter in the


73
Fig. 3-14. Group size vs. percent tree cover within 100 m buffer for all territories (North
and South populations pooled). Trend towards smaller group size with higher tree cover
is not significant.


Fig. 3-17. Images and territories (black polygons) of North Sandy Hill. Right: color-infrared image. Left: classified image (white -
bare sand; green = trees; brown = shrubs/grass; black = water).
Os


369
occurs with numerous narrowly adapted, range-restricted scrub endemics (e.g.,
Christman and Judd 1990). Many of these species will automatically be preserved if the
full jay distribution is maintained; others, however, will require habitat preservation in
areas deemed nonviable for the jay.
3) Favor preservation of contiguous territories. Jays that exist in clusters of
contiguous territories are less extinction-prone than populations of equivalent size that
occur as noncontiguous territories. The risks associated with floater dispersal are high
(chapter 4), and breeder vacancies that arise within contiguous territories can be found
using the much less risky philopatric dispersal strategy. Currently, I can offer no
minimum population size or distance thresholds as quantitative guidelines for what
constitutes a stable cluster of contiguous territories. I suspect that epidemics are the
critical factor for determining this threshold (assuming habitat quality is high), but further
modeling is needed to come up with appropriate guidelines.
4) Prohibit the splitting of metapopulations. Habitat gaps larger than 12 km
represent barriers to natural dispersal and recolonization (chapter 2). To maintain all
existing metapopulations, therefore, all habitat gaps must be kept well below this 12 km
threshold. Failure to do so would effectively split the system apart and create two
smaller, hence less viable, systems. Because coastal populations of Florida Scrub-Jays are
distributed in narrow strips parallel to the coast line (dune and shoreline deposits), they
are especially vulnerable to being split as a result of elimination of small habitat patches.
The emphasis of this dissertation has been on the influence of spatial factors on
metapopulation viability. Landscape rules integrate much of the information derived from
previous chapters and prior research, in a form useful for conservation. There are,


153
demographic success, especially for patches that are juxtaposed against landscapes that
subject jays to detrimental edge effects. These edge effects include human factors such as
road mortality (Mumme et al., in press), and predation by domestic cats, and natural
predators that occur at artificially high densities (e.g. racoons, grackles; Breininger 1999).
More subtle edge effects may occur in natural landscapes where jay habitat is located
next to forests or other habitats favorable to jay predators and competitors (Breininger et
al. 1998).
It is likely that these negative factors, which depress jay densities and
demographic performance, are the norm throughout much of the state. Fire management
programs are being developed and implemented on many public lands, but as
development continues in Florida many jays on these properties will have reduced
reproductive success simply because they are surrounded by human landscapes. A recent
study in well-managed jay habitat at Archbold Biological Station by Mumme et al. (in
press) documented substantial negative effects of road mortality in a rural setting. The
situation in suburban and urban settings is likely to be even worse (Breininger 1999;
Bowman 1993). Because the simulation model does not consider such edge effects, the
model results should be viewed as optimistic.
Given these caveats, the recommended use of the estimates reported below is to
compare the relative viability of jay metapopulations around the state as a guide for land
acquisition and to rank areas in terms of vulnerability. Probability and trajectory
estimates produced by the model should not be taken literally; they are best used for
comparative purposes.


Ill
censored in the figure because they successfully returned to their home territory. The
remaining jays either perished or were lost due to transmitter failure.
A Kaplan-Meier survival curve (Fig. 4-2) shows the daily probability of survival
for the pooled sample of jays. Three survival curves were generated by making different
assumptions about unknown disappearances. The upper curve assumes that all unknown
disappearances were caused by transmitter failure, the bottom all due to mortality, and
the middle curve makes an educated-guess about each disappearance based on
circumstantial evidence. The 3 jays that successfully returned to their home territory were
right-censored. All three curves show an early rapid decrease in survivorship caused by
known mortality of several jays shortly after release. The constant daily survivorship out
to day 21 reflects a single jay released at the ranch which was depredated at day 22.
Floater Parameter Estimation
Floater mobility was parameterized by fitting a function generated by the curve
fitting procedure of SPSS (ver. 7.5) to the distribution of daily distances (Fig. 4-3) moved
obtained from radiotelemetry field data (excluding days with no movement see
explanation above). An inverse function of the form x = bl / (y bO) produced a good fit
with the observed data (r2 = 0.851; F = 91.43; d.f. = 16; p = 0.000). Values for the two
coefficients were: bO = -0.48; bl = 5.57.
Daily survival of floaters within scrub was assumed to be only slightly less than
survival of non-dispersing stages, as measured at Archbold Biological Station. The
default daily survival value was set at 0.9988, which was computed from the estimated
annual survival rate of 2nd year female helpers (0.66365). Daily survival of floaters


Probability
10
Threshold Pop. Size
20
30
Fig. 5-14d. St. Lucie N. Martin county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition.


127
Observed
Inverse
Fig. 4-3. Distribution of daily distances moved by released jays (solid line), and inverse
function fitted to observed movements (dashed line).


164
Citrus-S.W. Marion (M2)
General description: The Citrus and S.W. Marion county metapopulation consists
of small, scattered groups of jays living within about ten miles of the Gulf Coast, and
larger populations of jays concentrated mostly in the Big Scrub area of southwest Marion
and northwest Sumter counties. Most jays occur in small, somewhat isolated clusters; the
largest population is an unprotected group of 19 pairs at Mar2 (fig. 5-2b). Extensive
mosaics of scrub, scrubby flatwoods, sand pine, and sandhill occur throughout this
region. The habitat as mapped by the SMP makes no distinction among habitat types and
should be treated as very incomplete. Because of the severely overgrown habitat
conditions, many jays occur in marginal habitat, and the small, isolated populations along
the Gulf Coast are especially vulnerable to blinking out. The connection between the Gulf
Coast and Big Scrub area may be very poor; the 12 km dispersal buffer (fig. 5-1) shows
the tenuous connection occurring at the Twisted Oak golf course (Citr5 in fig. 5-2b).
The SMP documented about 108 jay territories, excluding suburban jays, in this
metapopulation. Estimated potential population size after habitat restoration and full
occupancy is 47 pairs in currently protected areas, and 145 pairs maximum.
Protected areas: The only protected jays along the coastal portion of this
metapopulation occur at the Crystal River State Buffer Preserve (Citrl). Protected jays
inland occur at Potts Preserve (Citr6), Cross Florida Greenway (Marl, Mar5,
Mar6), and Half Moon Wildlife Management Area (Sumtl). CARL sites with jays
include Mar8.
Restoration potential: Nearly all of the habitat in this metapopulation is heavily
overgrown and occurs as a complex mosaic of scrub, scrubby flatwoods, sandhill and


Probability
162
5 10 15 20
Threshold Pop Size
20 40 60
Threshold Pop Size
Fig. 5-Id. Levy county quasi-extinction graphs. Top) no acquisition, Bottom) maximum
acquisition.


211
Table 5-6b. N. W. Charlotte county simulation statistics
Data type
1992-1993
configuration
No
acquisition
70%
acquisition by
connectivity
70%
acquisition
by area
Maximum
acquisition
Starting
population size
44
28
47
47
56
Mean ending
population size
19.4
2.3
12.8
14.5
22.1
s.d.
11.3
3.9
9.1
8.6
11.3
% population
decline
55.9
91.8
72.8
69.1
60.7
Extinction
Risk
0.10
0.67
0.17
0.17
0.07
Quasi
extinction
Risk (10 pairs)
0.37
1.0
0.63
0.57
0.30


131
Fig. 4-7. Comparison of stage-age data from Archbold Biological Station and simulated
stage-age data for helpers.


14
Degraded quality of many currently occupied habitat patches suggests that
further, substantial declines in the jay population are inevitable. Specifically, those jays
occupying suburban areas (approximately 30% of all territories) are unlikely to persist as
these suburbs continue to build out, given the rapid rate at which Floridas human
population continues to expand. Furthermore, jays living in fire-suppressed, overgrown
habitat (at least 2,100 families, or 64% of all occupied scrub patches by area) already are
likely to be experiencing poor demographic performance (Fitzpatrick and Woolfenden
1986). These can be expected to decline further unless widespread restoration of habitat
is begun soon.
A Method for Classifying Metapopulations
The patchy distribution and variable clustering of territories throughout the range
of the Florida Scrub-Jay (Fig. 2-1) challenges us to expand upon traditional
metapopulation concepts in order to describe the spatial structure of this species. In this
section we describe our conceptual approach, and in the next we apply it to the Florida
Scrub-Jay data.
Harrisons (1991) four classes of metapopulations can be presented graphically
(Fig. 2-2) as different regions on a plot of degree of isolation against patch size
distribution. Thus, Harrisons "non-equilibrium" metapopulation is that set of small
patches in which each has a high probability of extinction, and among which little or no
migration occurs. Local extinctions are not offset by recolonization, resulting in overall
decline toward regional extinction. The classical model developed by Levins (1969,
1970) is a set of small patches which are individually prone to extinction, but which are


232
Table 5-9a. Flagler and N.E. Volusia county patch statistics (number of jay territories for
different configurations)
Patch id
Status
1992-
1993# jay
territories
No
acquisition
(restored)
Maximum
acquisition
Flagl
3
3
Flag2
2
2
Flag3
2
2
Vol1
N. Peninsula St. Rec. Area
5
5
5
Totals
12 5 12


243
Table 5-1 Ob. S.E. Volusia and Merritt Island county simulation statistics
Data type
No
acquisition
Maximum
acquisition
starting
population size
495
536
x end pop. size
491.5
501.3
s.d.
5.7
14.5
percent
decline
0.7
6.5
extinction
risk
0.0
0.0
quasi-extinction
risk (10 pairs)
0.0
0.0


72
Prop.
Tree Cover
Fig. 3-13. Group size vs. percent tree cover within all territories (North and South
populations pooled). Trend towards smaller group size with higher tree cover is not
significant.


155
Levy (Cedar Key) (Mil
General description: The Levy county metapopulation is the most northerly
population of jays occurring along the Gulf Coast, and is highly isolated from other
metapopulations. The SMP delineated a single large scrub patch in this area, and found 8
groups of jays, 4 of which occurred in the Cedar Key Scrub State Reserve (see Fig. 5-la;
Table 5-la). The SMP found the condition of the scrub to be severely overgrown, and
noted that the number of jays present was only one-third the number found by Cox in
1980 (Pranty et al. manuscript). A 1997-1998 study of jays at Cedar Key (T. Webber in
F.D.E.P. 1998) found only 1 pair on the reserve, 2 pairs in a nearby junkyard, and 4
groups in the town of Rosewood 7-8 miles to the east. Estimated potential population size
after habitat restoration and full occupancy is 17 pairs in currently protected areas, and 75
pairs maximum.
Protected areas: The only protected jays in this metapopulation occur in the Cedar
Key Scrub State Reserve (Levyl).
Restoration potential: Some restoration has taken place at the Cedar Key Scrub
State Reserve, but many areas remain heavily overgrown. A recent report (F.D.E.P. 1998)
noted that a shortage of staff and difficulties associated with burning sand pine forests
have delayed restoration needed at this reserve. For modeling purposes, the currently
protected area is estimated to support about 17 jay families after restoration and full
occupancy (Fig. 5-lb; Table 5-la). A large, contiguous patch of unprotected habitat
(Levy2) was mapped by the SMP, but much of this habitat is not suitable for jays.
Restorable patches of scrub and scrubby flatwoods occur within a complex matrix of


5-12a. Central Brevard county patch statistics (number of jay territories for
different configurations) 257
5-12b. Central Brevard county simulation statistics 260
5-13a. S. Brevard-Indian River-N. St. Lucie county patch statistics (number of jay
territories for different configurations) 266
5-13b. S. Brevard-Indian River-N. St. Lucie county simulation statistics 269
5-14a. St. Lucie N. Martin county patch statistics (number of jay territories for
different configurations) 274
5-14b. St. Lucie county simulation statistics 277
5-15a. Martin and N. Palm Beach county patch statistics (number of jay territories
for different configurations) 282
5-15b. Martin and N. Palm Beach county simulation statistics 285
5-16a. South Palm Beach county patch statistics (number of jay territories for
different configurations) 290
5-16b. South Palm Beach county simulation statistics 293
5-17a. Ocala National Forest county patch statistics (number of jay territories for
different configurations) 298
5-17b. Ocala National Forest county simulation statistics 301
5-18a. N.E. Lake county patch statistics (number of jay territories for different
configurations) 306
5-18b. N.E. Lake county simulation statistics 309
5-19a. S.W. Volusia county patch statistics (number of jay territories for different
configurations) 314
5-19b. S.W. Volusia county simulation statistics 317
5-20a. Central Lake county patch statistics (number of jay territories for different
configurations) 322
5-20b. Central Lake county simulation statistics 325
xiii


288
Scrub Polygons
Lo Disturbance Prtectc>n Status A/ Interstates
3 Lo Density Housing i.l.'lD Protectec* /\/ Sute highways
If-V/j Hi Density Housing i ProPse^ County roads
Ranch/Ag Water Bodes ' ' County lines
Ml 6 S. Palm Beach
0 2 4 6
1 : 120,000
Kilometers
Fig 5-16a. Central Palm Beach county map 1992 1993 jay and habitat distribution.


CHAPTER 6
SYNTHESIS
The technique developed in chapter 2 to classify the metapopulation structure of
the Florida Scrub-Jay provided qualitative expectations about the viability of different
types of metapopulations. For example, systems composed only of islands are more
vulnerable than systems with midlands, which in turn are more vulnerable than systems
with mainlands. However, the technique provides no quantitative estimates of the
viability of different configurations. The individual-based model described in chapter 5
permits such viability estimates, and allowed an assessment to be made of the viability of
the major scrub-jay metapopulations around the state.
The reserve design scenarios simulated in chapter 5 did not allow the landscape to
change over time. In a theoretical paper, Fahrig (1992) argues that temporal changes in
habitat (patch lifespan) are likely to be much more important than spatial factors. She
found that the temporal scale of dispersal (dispersal frequency) far outweighed the spatial
scale (dispersal distance) in affecting population recovery from patch disturbance. The
most applicable implication of this finding for Florida Scrub-Jays today is that given the
relatively short life span of scrub patches under the current human-dominated regime of
fire suppression, large numbers of dispersing jays exploring many areas are needed to
find the few recently burned, unoccupied patches. The ability to move long distances is
much less unimportant than the ability to send out many dispersers to canvass a large
365


336
Jay Territory Locations
(after restoration) Protection Status
Scrub Polygons /\/ Protected
CCC Polygon ID A/1Unprotected
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Water Bodies
County lines
M 21 N. Highlands SC S. Polk
0 5 10 15 Kilometers
1 : 250,000
Fig. 5-2 li. Lake Wales Ridge acquisition map, N. Highlands and S. Polk county.


Table 5-19a. S.W. Volusia county patch statistics (number of jay territories for different configurations)
Patch id
Status
1992-1993#
No
30%
30%
70%
70%
Maximum
jay territories
acquisition
preserved
preserved
preserved
preserved
acquisition
(restored)
by contiguity
by connectivity
by contiguity
by connectivity
VoMO
Blue Springs
1
17
17
17
17
17
17
S.P.
Vol11
7
3
3
7
5
7
Vol12
2
1
1
2
2
2
Vol13
2
1
1
1
2
2
Vol14
5
2
2
4
5
Vol15
4
1
1
3
4
Vol16
2
1
1
1
2
Vol17
4
1
1
3
4
Vol18
3
2
2
3
2
3
Vol19
Stewart Ranch
(proposed)
24
4
4
24
15
24
Totals
54
17
33
33
54
54
70


Table 2-1. Summary information for 42 Florida Scrub-Jay metapopulations, including
type of metapopulation, number of territories (pairs of jays), and number of
subpopulations.
28
Metapopulation type
(after Harrison 1991)
Mainland-Island (?)
Unknown
Metapopulation Type Size Number of
(Mainland, Midland, (pairs) subpopulations
and/or Island)a
Mn
lOMd
391
1247
50
Mn
Md
51
1036
7
Mn
Md
51
466
7
Mn
2Md
61
237
9
Mn
Md
51
179
7
Mn
Md
126
2
4Md
111
120
15
2Md
I
103
3
Md
51
94
6
Md
I
58
2
2Md
31
55
5
Md
I
50
2
Md
21
29
3
Md
31
22
4
Md
21
18
3
Md
26
1
Md
22
1
Patchy


257
Table 5-12a. Central Brevard county patch statistics (number of jay territories for
different configurations)
Patch id
Status
1992-
No
70%
Maximum
1993# jay
acquisition
preserved
acquisition
territories
(restored)
by area
Brev40
Rockledge Scrub Pr.
3
3
3
3
Brev41
(DRI/EELS -
proposed)
15
16
19
Brev42
(Wickham Rd. CARL
site)
9
9
9
Brev43
Wickham County Pk.
2
2
2
2
Brev44
Melbourne regional
airport
7
7
Totals 36 5 30 40


286
South Palm Beach (Ml6)
General description: The South Palm Beach metapopulation is the most southerly
metapopulation on the Atlantic coast. It is isolated from the Martin county
metapopulation (Ml5) to the north by heavy urbanization associated with West Palm
Beach, Palm Springs, and Lake Worth. All of the scrub patches are small and occur in
suburban or urban settings. The SMP found 8 groups of jays, and characterized the
condition of the scrub to be severely overgrown. The number of jays present during the
SMP was only one-third the number found by Cox in 1980 (Pranty et al. manuscript).
Estimated potential population size after habitat restoration and full occupancy is 9 pairs
in currently protected areas, and 16 pairs maximum. Grace Iverson, who has studied jays
in Palm Beach county for a number of years, provided invaluable information on this
metapopulation.
Protected areas: Rolling Green Scrub Preserve (PB11), Galaxy School Scrub
Preserve (PB12), Yamato Scrub NAP (PB14). A number of small scrub preserves
that did not have jays during the SMP were excluded from all simulations (Osborne
Scrub NAP, Gopher Tortoise Scrub NAP, Rosemary Ridge Scrub NAP, Leon Weeks
Scrub Preserve NAP, Seacrest Scrub NAP, Rosemary Scrub NAP).
Restoration potential: The Yamato Scrub NAP (PB14) had only 1 pair of jays
during the SMP, but is estimated to potentially support about 6 pairs of jays if fully
restored and managed.
Simulation results: This metapopulation ranked 8th in vulnerability (table 5-23),
6th in percent protected (69.2%; table 5-24), and 14th in priority (table 5-25), with high


6
this area. The relationship between demographic variables and the remotely-sensed
habitat variables were examined. At issue is the possibility of measuring habitat quality,
and potential demographic success remotely, and across large areas.
Chapter 4 examines the difficult subject of dispersal and describes an approach
used to simulate dispersal in an individual-based model (IBM) developed for this
dissertation. Two types of dispersal are simulated by the IBM. A close-distance dispersal
module mimics a stay-home-and-foray strategy that results in most dispersing jays
settling close to their natal territory. This module incorporates many details of Florida
Scrub-Jay biology documented by long-term color-band studies (Woolfenden and
Fitzpatrick 1984), including sex and age dominance relations. A long-distance dispersal
module simulates a floater strategy, which accounts for the infrequent, though
potentially important, tendency of some jays to abandon their natal territory and move
long distances, often between habitat patches and through hostile landscape matrices.
Empirical data on long-distance dispersal are poor, and a simple field experiment was
conducted with radiotelemetry to obtain information useful for modeling purposes. To
induce behavior that might be similar to long distance dispersal, radio-collared jays were
experimentally displaced kilometers away from their natal territories. Habitat
preferences, movement abilities, and mortality rates were recorded and incorporated into
the long-distance dispersal module. In combination, the close- and long-distance dispersal
modules produced a dispersal and stage-age curve that closely resembled results of long
term data from Archbold Biological Station. A constraint analysis was used to place
plausible bounds on several of the poorly known long distance dispersal parameters. This


48
Habitat Quality Model
A habitat quality model was developed from the habitat variables using a habitat
suitability index (HIS) approach similar to Duncan et al. (1995). The HSI model
combines three HSI values for percent bare sand within territories (BS), percent tree
cover within territories (TC), and distance to nearest forest (DF), to calculate a single
habitat quality value (HQ) for each territory. The equation used for this model is
hq = Mbs *tc*df
The HSI values for each of the three habitat variables was obtained from step
functions relating the habitat variable to an estimate of habitat quality (see Fig. 3-2
modified from Duncan et al. 1995). The shapes of these step functions were developed
subjectively by D. Breininger. The habitat quality values were mapped automatically
across the entire study area using the Spatial Modeler in the Imagine software package.
The BS variable was mapped using the focal sum operator to count the number of bare
sand pixels within 10 m of every point in the study area, and computing an HSI value
using the function in Fig. 3-2a. The TC variable was similarly mapped by measuring tree
pixels within a 60 m radius, and computing an HSI value using the function in Fig. 3-2b.
To compute the DF variable, the TC layer was use ... and pixels surrounded by greater
than 30% tree cover were coded as forest. The search operator was used to measure the
distance to nearest forest for all pixels. The DF v.aue was then computed using the
function in Fig. 3-2c. The Imagine summary" function was used to output HSI values


178
Scrub Polygons
Lo Disturbance
\ Lo Density Housing
jgg Hi Density Housing
I R.anch/Ag
Protection Status
Protected
[j Proposed
Water Bodies
Interstates
State highways
County roads
County lines
M3 W. Pasco
1 : 250,000
Kilometers
Fig. 5-3a. W. Pasco and Hernando county map 1992 1993 jay and habitat distribution.


340
Table 5-21a continued.
Patch id
Status
1992-1993#
jay territories
No acquisition
(restored)
Maximum
acquisition
Polk6
Tiger Creek & Lake Wales
11
17
17
S.F.
Polk7
9
13
Polk8
6
8
Polk9
6
6
PolklO
Lk. Kissimmee S.P.
6
6
6
Polk11
Catfish Creek
34
41
41
Polk12
Disney Wilderness Pr.
39
39
39
Polk13
17
20
Totals
655
535
858


Probability
234
roT
0 8
%
CL
0.6~f
04
0 2
2 4
Threshold Pop. Size
Fig. 5-9d. Flagler and N.E. Volusia county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition.


Table 3-1 continued.
Location Territory Year Group Fledgling Yearling Breeder Helper
Size Production Production Surv. Surv.
NRIDGE XOVR
1995
2
4
0
0
NRIDGE
YYYY
1995
5
0
0
0.5
0.66
SRIDGE
DEAD
1995
4
1
0
1
1
SRIDGE
ECHO
1995
2
0
0
1
SRIDGE
EXP5
1995
3
0
0
0.5
1
SRIDGE
FARS
1995
4
4
3
1
0
SRIDGE
GPSS
1995
5
0
0
0
0.33
SRIDGE
HTOP
1995
4
0
0
0
0
SRIDGE
JSUS
1995
2
4
2
1
SRIDGE
LOGG
1995
2
0
0
0
0
SRIDGE
LOPI
1995
4
0
0
0.5
0.5
SRIDGE
LOST
1995
3
0
0
1
0
SRIDGE
NORW
1995
3
1
0
0.5
1
SRIDGE
SNAG
1995
4
1
1
0
0.5
SRIDGE
SRNG
1995
3
3
0
1
1
SRIDGE
TRIS
1995
3
1
1
1
1
SRIDGE
WPND
1995
3
3
2
0.5
0
Fledgl. Yearl. Bare Tree cover Tree cover
Surv. Surv. sand (inside (100m
territory) buffer)
0
0
0.75
0.5
0
1
0
1
0.66
0
0
0.25
0
0
0
0
0
0
0.13
0.07
0.11
0.28
0.24
0.22
0.32
0.13
0.09
0.16
0.26
0.17
0.21
0.14
0.25
0.21
0.13
0.02
0.06
0.26
0.17
0.1
0.26
0.01
0.36
0.53
0.18
0.07
0.14
0.03
0.48
0.33
0.08
0.23
0.01
0.05
0.24
0.18
0.26
0.3
0.05
0.48
0.55
0.21
0.1
0.17
0.03
0.45
0.37
0.12
0.29
oo


377
Komdeur, J. 1992. Importance of habitat saturation and territory quality for the evolution
of cooperative breeding in the Seychelles Warbler. Nature. 358: 493-495.
Krebs, J.R. and N.B. Davies. 1993. An Introduction to Behavioural Ecology (3rd ed.).
Blackwell Science, Inc. Cambridge, MA.
Lamberson, R.H., R. McKelvey, B.R. Noon, and C. Voss. 1992. A dynamic analysis of
Northern Spotted Owl viability in a fragmented forest landscape. Conservation
Biology 6: 505-512.
Lacy, R.C. 1993. VORTEX: a computer simulation model for Population Viability
Analysis. Wildlife Research 20: 45-65.
Lande, R. 1988a. Demographic models of the northern spotted owl (Strix occidentals
caurina). Oecologia 75: 601-607.
Lande, R. 1988b. Genetics and demography in biological conservation. Science 241:
1455-1459.
Lesica, P., and F. Allendorf. 1995. When are peripheral populations valuable for
conservation? Conservation Biology 9: 753-760.
Letcher, B.H., J.A. Priddy, J.R. Walters, and L.B. Crowded 1998. An individual-based,
spatially-explicit simulation model of the population dynamics of the endangered
red-cockaded woodpecker, Picoides borealis. Biological Conservation 86: 1-14.
Levins, R. 1969. Some demographic and genetic consequences of environmental
heterogeneity for biological control. Bull. Entomol. Soc. Amer. 15: 237-240.
Levins, R. 1970. Extinction. Lectures on Mathematics in the Life Sciences 2: 75-107.
Lima, S. and P. Zollner. 1996. Towards a behavioral ecology of ecological landscapes.
Trends in Ecology and Evolution 11: 131-135.
MacArthur, R. H. and E. O. Wilson. 1967. The Theory of Island Biogeography.
Princeton, NJ: Princeton University Press. 203 pp.
Mace, G. M., and R. Lande. 1991. Assessing extinction threats: toward a reevaluation of
IUCN threatened species categories. Conservation Biology 2: 148-157.
Maley, C. and H. Caswell. 1993. Implementing I-state configuration models for
population dynamics: an object-oriented programming approach. Ecological
Modelling 68: 75-89.


2
delta during the mid-Pleistocene. As conditions became more mesic, xeric habitat became
reduced and isolated into desert-like islands within which a remarkable assemblage of
organisms evolved. Among these organisms that diverged from their western kin is the
Florida-Scrub Jay, the subject of this dissertation.
Not long after the most recent Wisconsin glacial retreat, a mere 12,000 years ago,
nomadic people known as the Clovis entered North America from Siberia and
encountered a continent largely or entirely devoid of humans. At that time the continent
teemed with giant animals such as mammoths, mastodons, ostrich-size flightless birds,
huge ground sloths, horses, camels, saber-toothed cats, and giant tortoises. This
megafaunal scene rivaled anything seen today in Africa. Within 1,000 to 2,000 years all
of these species and many more became extinct. Dozens of large vertebrates appear to
have made their last stand in Florida, their demise apparently coinciding with the arrival
of the Clovis. It remains uncertain whether humans were largely to blame for these
extinctions, as the megafauna also faced great changes in climate and landscape. Yet, it is
highly likely that humans contributed significantly to these megafaunal extinctions.
The Florida Scrub-Jay managed to survive this period of massive extinctions. But
by the mid-20th century, a new threat to the fauna and flora of Florida appeared.
Discoveries in applied sciences and engineering paved the way for the demise of the
Florida Scrub-Jay and its scrub habitat. Among these were the discovery that citrus could
be grown on the formerly worthless, sandy, infertile soils upon which Florida Scrub-Jay
habitat grew. The invention of air conditioning made tolerable the Florida summers,
ushering in an era of massive suburban sprawl, much of it devouring Florida Scrub-Jay
habitat.


Probability
08
(-
o 6~r
04
-
i-
i i i i i j
100
I
;
200 300
Threshold Pop Size
400
Fig. 5-17d. Ocala National Forest county quasi-extinction graphs. No acquisition.


Table 5-4a continued.
Patch id
Status
1992-1993#
No
30%
jay territories
acquisition
preserved
(restored)
by contiguity
Man8
Man9
Man10
Lake Manatee
Lower
Watershed
3
3
Man11
Man12
Beker
1
2
2
Man 13
1
Man14
1
Man15
Duette Park
6
11
11
Man16
Beker
6
8
Man17
Lake Manatee
St. Rec. Area
1
2
Man18
1
Sar15
Myakka River
St. Pk.
3
3
Sar16
3
Sar17
2
2
Totals
65
36
69
30%
preserved
by connectivity
70%
preserved
by contiguity
70% Maximum
preserved acquisition
by connectivity
3
3
3
1
1
3
1
2
1
1
11
6
2
1
2
4
11
15
2
1
2
4
1
12
15
2
2
2
4
3
12
15
2
3
3
3
3
2
2
4 4 4
2 2 2
69 112 112 145
vO
K)


26
Similarly, abnormally high densities may exist due to the crowding effect (Lamberson
et al. 1992) following recent habitat losses. (5) The technique relies on numerous
simplifying assumptions about dispersal behavior in defining connectivity among
patches. Most important, it assumes random movement between patches, equal
traversibility of interpatch habitats, absence of dispersal biases owing to habitat quality
differences at the origin or the destination, and absence of density-dependence in
behavior. More elaborate applications, of course, could incorporate alternative
assumptions about these and other factors.
Another important consideration are the kinds of data to buffer. To create
biologically meaningful--but very differentdescriptions for the Florida Scrub-Jay we
could have buffered around jay territories (our choice), occupied patches, suitable habitat
patches both occupied and unoccupied, or all scrub habitat patches regardless of current
suitability. Organisms such as Florida Scrub-Jays that are reluctant to become established
in unoccupied, suitable habitat (e.g., Ebenhard 1991), or have high conspecific attraction
or an allee effect (Smith and Peacock 1990) are best buffered around actual territories
or occupied patches. This is because unoccupied sites have a low probability of becoming
occupied regardless of their degree of isolation, hence contribute little to the current
metapopulation dynamics of the species. On the other hand, excellent colonizers of empty
habitat or species adept at long-distance dispersals via unoccupied stepping stones
probably should be buffered around all habitat patches.
In summary, this method of classifying metapopulations provides a compact
means of describing both connectivity and local population size through the use of simple
terminology. Separate metapopulations are easily delineated using the maximum


371
facing these types of models is ascertaining the reliability of model predictions. Rykiel
(1996) provides a general review of various means of model validation, and concludes
that model validation has many different meanings and no standard methods. The use of
postdiction or retrospective testing to compare historic data with model predictions has
been used only occasionally for PVA models. Brook et al. (1997) documented the 10-
year population trajectory of the Lord Howe Island wood hen following the release of 86
captive-bred individuals. Model predictions were unreliable unless accurate post facto
estimates of carrying capacity were used. Intuitively, the strongest form of validation
occurs when model predictions are not falsified by future events. This strong form of
model validation, the statistical comparison of predicted and actual trajectories, requires
long term data replicated in different areas for different landscapes. Obtaining such data
may be impossible for most species, but substantial data sets for at least 6 different color-
banded populations of Florida Scrub-Jays (Archbold Biological Station, Lake Placid,
Avon Park Air Force Range, Sarasota county, S. Brevard county, and Kennedy Space
Center) are already available, and coarser survey data for other parts of the state also are
available. Stable and declining population trajectories have been documented in these
areas, and offer an excellent opportunity to further test and refine this model. It is my
hope to continue working on this model and intriguing species, the Florida Scrub-Jay, for
which so much is known, but so much more remains to be discovered.


209
J ¡L
20 40 60
Year
Fig. 5-6c. N. W. Charlotte county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.


101
age or experience who are not making frequent forays away from their natal territory (i.e.
breeders or older fledglings). As an example, consider yearling helpers that disappeared
at Archbold Biological Station (Dtotai female = 0.42, male = 0.22; table 4-2). The local
death rate of yearling helpers is almost certainly greater than that of breeders (0.18;
Woolfenden and Fitzpatrick 1984, table 9.2, p. 265), but may be lower than the death rate
of older fledglings (0.30; Woolfenden and Fitzpatrick 1984, fig. 9.1, p. 255). If we
assume that P|0Cai is less than 0.30, suppose 0.25, then the maximum value for Pfioater
would be 0.75 (Pfioater must be less than (1 Pi0Cai); eqn. 4-2).
In the real world, the proportion of helpers that become floaters may vary greatly,
depending on factors such as habitat quality or number of neighboring territories. Some
of these factors are reviewed below in the discussion. If we consider the Archbold
setting, for yearling male helpers, it is likely that most disappearances can be attributed to
local mortality within their assessment sphere, while the proportion of females dying
within their assessment sphere may be less because more females disperse as floaters.
However, since the female assessment sphere is larger than the male, the proportion of
females dying within their assessment sphere may be substantial. Furthermore, 2nd year
males are estimated to have the largest proportion of disappearances due to emigration
(Table 4-2; column En0niocai/ Dtotai), suggesting that Pfioater also might be considerable for
older male helpers.
Since the disappearance rates (Dtotai) of female helpers are considerably greater
than males, setting Pfioater equivalent for both sexes would produce substantially more
female than male floaters. The values for Pfioater likely are larger for females than males,
which further increases this bias in the number of female floaters. Although it may seem


56
georeferencing section). Since there is no reason to expect errors in georeferencing to
have the same bias as errors in GPS readings, the difference between the two could
compound to exceed 15 meters. An error of this magnitude probably exceeds the
expected error from a field person using a high resolution photo map to position a
transect. Unfortunately, this leaves us with no way of quantifying and correcting
positional inaccuracies. Recent advances in GPS technology may solve these problems as
sub-meter accuracy becomes increasingly affordable and practical.
Quadrat samples and visual comparison of the classified image with the aerial
photographs plus field knowledge suggest that the classification accurately reflected
biologically important differences among habitats. The image processing techniques,
combined with GIS files of the locations of jay territories and buffer zones, provided
quantitative measurements of jay habitat in a quick and efficient manner. The quantitative
results show dramatic differences in habitat structure between the North and South areas
corresponding well to impressions reported by field researchers (R. Bowman pers.
comm.). N. Sandy Hill jays have far fewer trees within and adjacent to their territories
compared to S. Sandy Hill jays.
The image processing results confirmed our general field impressions about
which jay families were living in good and poor quality habitat on Sandy Hill. Jays that
occupied the poorest habitat were in the southern part of S. Sandy Hill. Here, several
families were living in low quality habitat near a recent bum that was occupied by two or
three other families. The presence of jays in poor habitat resulted from conspecific
attraction (Smith and Peacock 1990) or "queueing" behavior, where individuals stay near
high quality habitat to wait for breeding vacancies. Jays living in such poor habitat,


Table 5-26. Summary of recommendations (highest priority first).
Metapopulation
Primary Acquisition
Targets
Secondary
Acquisition
Targets
Restoration (protected areas)
Other actions
N. Brevard
(Mil)
Brev4 (Buck Lake),
Brev7 (addition to South
Lake), Brev8 (Seminole
Ranch), Brev9, BrevlO,
Brevl 1, Brevl2, Brevl5,
Brev 16, Brev 17, Brev 18
Brev6, Brev 13
New jay surveys needed
at most acquisition sites
Cedar Key (Ml)
Levy2 (as much as
possible)
Levyl (Cedar Key Scrub Preserve)
Central
Charlotte (M7)
Chari 7, Chari 8, Chari 5,
Chari4, Chari9
Lee5
Charl
Surveys needed along
Prairie and Shell Creek
(Char 17,18,20)
Central Brevard
(Ml 2)
Brev41, Brev42
Brev40 (Rockledge Scrub Pr.), Brev43
(Wickham County Pk.), Brev44
(Melbourne regional airport)
W. Volusia
(Ml 9)
Voll9, Voll8, Volll
VollO (Blue Springs)
Additions to VollO
(Blue Springs)
N.W. Charlotte
(M6)
Char8, Char7, Sari 3
(expand), Sari 1
CharlO, Chari 1,
Sari 2
Char9 (Charlotte Harbor S.B. Pr.)
St. Lucie-N.
Martin (Ml4)
Marl (additions), Mar2,
Mar3, Mar4
Stl4 (Savannas S.P.), Marl
Investigate status of Stl3


343
Table 5-2lb. Lake Wales Ridge simulation statistics
Data type
No
acquisition
Maximum
acquisition
starting
population size
535
858
x end pop. size
435.7
708.5
s.d.
56.4
65.4
percent
decline
18.5
17.4
extinction
risk
0.0
0.0
quasi-extinction
risk (10 pairs)
0.0
0.0


349
Summary of Recommendations
Table 5-26 lists each metapopulation in the order given by the priority ranking
(table 5-25), and summarizes the recommendations provided earlier in each
metapopulation section. Unprotected habitat patches with a high priority for acquisition
are listed in the Primary Acquisition Target column; patches that may be of lower
priority are listed in the Secondary Acquisition Target column. I decided which patches
to place into the two acquisition categories subjectively, based on the results of different
reserve design simulations and my overall impression from the modeling results that
maintaining habitat contiguity is much more important than maintaining habitat
connectivity. A systematic analysis of this contiguity vs. connectivity issue is needed.
Habitat patches that are already protected but are in immediate need of restoration and
management are listed in the Restoration column. The last column, Other actions,
lists miscellaneous activities that are recommended, such as additional surveys.


Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
METAPOPULATION DYNAMICS AND LANDSCAPE ECOLOGY OF THE
FLORIDA SCRUB-JAY, APHELOCOMA COERULESCENS
By
Bradley M. Stith
December 1999
Chairman: Dr. Stephen R. Humphrey
Major Department: Wildlife Ecology and Conservation
Floridas only endemic bird species, the Florida Scrub-Jay (Aphelocoma
coerulescens), is rapidly disappearing throughout much of its range. A 1992-1993
statewide survey shows that it has effectively gone extinct in 10 of 39 formerly occupied
counties in less than two decades. To characterize the spatial structure and vulnerability
of the Florida Scrub-Jay throughout the state, I developed and applied a new method to
describe the species metapopulation structure. This method uses GIS-generated buffers
based on documented dispersal distances to identify separate metapopulations and highly
connected subpopulations called mainlands (extinction resistant), islands (extinction
prone), or midlands (vulnerable to extinction). Of the 42 jay metapopulations identified,
only five include mainlands; 21 consist only of extinction-prone islands. The resulting
xxiv


126
Days Alive
Fig. 4-2. Kaplan-meier survival curves of daily survival rates for 10 jays released at 3
sites in Highlands county, Florida. Upper curve: maximum possible survival; middle
curve: best guess survival; lower curve: minimum possible survival.


298
Table 5-17a. Ocala National Forest county patch statistics (number of jay territories for
different configurations)
Patch id Status 1992-1993#
jay territories
Marl 5
Ocala National
448
Forest
Mar16
9
Mar17
6
Lake5
7
Totals
470


114
The biological consequence of these philopatric dispersal rules is that they
increase the probability that dispersers engaging in delay-and-foray behavior will acquire
a nearby vacancy. A fascinating controversy relating to this feature exists in the literature
on the evolution of delayed dispersal. Experts on the Florida Scrub-Jay attribute the
origins of this behavior to ecological constraints: under normal conditions high quality
jay habitat is fully occupied (saturated). Survival and fecundity of dispersers outside of
high quality habitat is so low that it is better to remain at the natal territory and maximize
opportunities for acquiring nearby vacancies in the future. While acting as resident
helpers, jays can engage in behavior that increases their likelihood of acquiring a
territory. Such behavior includes engaging in frequent short forays to investigate potential
vacancies due to a breeder death or illness, enlisting the help of parents and siblings to
bud a territory, establishing dominance over nearby helpers (potential competitors),
developing furtive relationships with neighboring opposite sex breeders, and learning
territory boundaries, refugia from predators, and roost sites. Besides these direct
benefits to staying home, certain indirect benefits accrue by helping relatives
(Woolfenden and Fitzpatrick 1984). Zack (1990), Stacey and Lign (1991), Brown
(1989) and others present these factors as benefits of philopatry, and argue that they are
more important than ecological constraints. Krebs and Davies (1998, p. 304) suggest
that the two views are not so much alternative hypotheses as different sides of the same
equation. They suggest that the pay-offs depend on the quality of breeding vacancies
available and how much competition exists.
Carmen (1989) showed that Western Scrub-Jays do not engage in delay-and-foray
dispersal behavior. In some cooperative breeders dispersers show considerable


3
In 1969, Glen E. Woolfenden turned his ornithological focus on the Florida
Scrub-Jay at the Archbold Biological Station. Thus began a continuous 31 year scientific
study of this single organism, making it one of the most thoroughly studied wild bird
species in the world. By 1975 his research on jays became a classic example of altruism
cited prominently in E.O. Wilsons (1975) influential Sociobioloev. In the mid-1970s
Woolfenden teamed up with John W. Fitzpatrick, an ornithologist with a strong
background in population modeling. In 1984, they produced a highly-acclaimed
Princeton monograph (Woolfenden and Fitzpatrick, 1984) describing the demography
and cooperative behavior of this intriguing bird. Today, the number of publications on
this species approaches one hundred, and the Florida Scrub-Jay continues to be widely
cited as a classic example of altruism (e.g. Krebs and Davies 1998). No brief summary
can do justice to this large body of work. Nonetheless, a short review of the basic natural
history of the Florida Scrub-Jay follows (consult Woolfenden and Fitzpatrick 1996 for an
extensive list of references).
Biological Background
The Florida Scrub-Jay, Florida's only endemic bird species, is a disjunct, relict
taxon separated by more than 1600 km from its closest western relatives (Woolfenden
and Fitzpatrick 1984). This habitat specialist is restricted to a patchily distributed scrub
community found on sandy, infertile soilsmostly pre-Pleistocene and Pleistocene
shoreline deposits. The vegetation is dominated by several species of low-stature scrub
oaks (Quercus spp.). Jays rely heavily on acorns for food, especially during the winter,
when they retrieve thousands of acorns cached in open, sandy areas during the fall
(DeGange et al. 1989). Florida Scrub-Jays show a strong preference for low, open


169
]ay Territory Locations
Scrub Polygons
^ Lo Disturbance
Lo Density Housing
tV/J Hi Density Housing
gggj Ranch/Ag
1.75 km dispersal buffer
Protection Status
i j Protected
~ Proposed
Water Bodies
Interstates
/\/ State highways
County roads
' County lines
M2 SW Mahon
0 5 10
1 : 250,000
15 Kilometers
Fig. 5-2b. S.W. Marion county map 1992-1993 jay and habitat distribution.


204
N. W. Charlotte M61
General description: The N.W. Charlotte metapopulation is isolated from the
Sarasota metapopulation to the west by the Myakka River, and is isolated from the
Central Charlotte metapopulation to the east by the Peace River. The SMP documented
about 44 jay territories, excluding suburban jays, in this metapopulation. Estimated
potential population size after habitat restoration and full occupancy is 28 pairs in
currently protected areas, and 56 pairs maximum.
Protected areas: Protected jays occur on the Charlotte Harbor State Buffer
Preserve (Char9; formerly known as Tippecanoe Scrub), and on the Myakka River
State Forest (Sari3), a private reserve (Chari2) and near 1-75 (Chari la).
Restoration potential: Most patches in this metapopulation are small, and jay
densities measured during the SMP were probably close to maximal. The Myakka River
State Forest (Sari 3), which had one pair of jays during the SMP, may have sufficient
habitat for 6 pairs.
Simulation results: This metapopulation ranked 10th in vulnerability (table 5-23),
9th in percent protected (50.0%; table 5-24), and 6th in priority (table 5-25), with high
vulnerability and high potential for improvement. All simulations had significant
extinction and quasi-extinction risk, and large percent population declines (see Table 5-
6b).
Recommendations: The risk estimates for the maximum acquisition
configuration are greatly improved compared to the no acquisition option. Acquisition
and restoration of as much habitat as possible is recommended. The most important
population of protected jays in this metapopulation probably occurs at the Charlotte


Table 5-2b. Citrus and S. Marion county simulation statistics
Data type
No
acquisition
30%
preserved by
connectivity
30%
preserved by
area
70%
preserved by
connectivity
70%
preserved by
area
Maximum
acquisition
starting
population
size
47
69
69
101
101
125
x end pop.
13.8
15.0
24.4
32.7
48.3
56.2
size
s.d.
8.5
11.4
16.1
20.1
20.1
17.7
percent
decline
70.6
78.3
65.2
67.6
52.2
55.0
extinction
risk
0.17
0.10
0.03
0.03
0.05
0.04
quasi
extinction
risk (10 pairs)
0.47
0.57
0.43
0.17
0.27
0.0


320
]ay Territory Locations ) 75 km dispersal buffer
Scrub Polygons
^ Lo Disturbance
3 Lo Density Housing
IfMj Hi Density Housing
Ranch/Ag
Protection Status
IlSagj Proteaed
]] Proposed
Water Bodies
/\f Interstates
State highways
County roads
' County lines
M20 Central Lake
0 5 [0 15
1 : 250,000
Kilometers
Fig. 5-20a. Central Lake county map 1992 1993 jay and habitat distribution.


50
Results
The locations and names of Florida Scrub-Jay territories throughout the study area
are shown in Fig. 3-1. The division between the North and South population occurs at
Kissimmee Rd. Note the absence of jays within the South population in the central part
of S. Sandy Hill. This area has high densities of pine trees and little bare sand, and has
three experimental plots where habitat restoration is underway.
A best-fit regression line, constrained to pass through the origin, showed a
significant relationship between percent bare sand measured from quadrats vs. imagery
(Fig. 3-4; r-squared = 0.60). The difference between the quadrat and image measurements
was not significant (Paired T-test; mean difference = 0.9372, n=12, p=0.256), indicating
no systematic bias in the measurements. Figure 3-5 shows the relationship between
percent tree cover measured from transect data vs. imagery. A best-fit logarithmic
regression line was drawn through the points (r-squared = 0.25). The differences between
the transect and image measurements was not significant (Paired T-test; mean difference
= 0.9372, n=40, p=0.824), indicating no systematic bias in the measurements. However,
the large scatter of points deviated considerably from the expected distribution which
would fall on a regression line intercept! 4 the origin and having a slope of one. The
logarithmic trend line suggested that the nage measurements give higher than expected
measurements at low tree cover values, id low-: nan expected measurements at high
tree cover values.
Percent tree cover ranged from 4% to ,% on S. Sandy Hill (Fig. 3-7), and only
0% to 4% on N. Sandy Hill (Fig. 3-6). The difference in tree cover between North and
South was highly significant (Table 3-3; Mann-Whitney U Test, Z = -6.291, P <


100
For example, Waser et al. (1994) describe a flexible approach for estimating from
census data the proportion of unobserved emigrants that survive, based on several types
of data such as the known number of successful immigrants, and the known survival rates
of non-dispersing sex and age classes. They cite Woolfenden and Fitzpatrick (1984) as
the earliest example of such an approach. Woolfenden and Fitzpatrick (1984 appendix M)
estimated the number of dispersers expected to successfully emigrate off their study area
and subtracted these estimates fr<&n the known disappearances to calculate more accurate
helper mortality rates. Their approach assumes that immigration and emigration are at
equilibrium. Table 4-2 shows their equilibrium mortality rates (column labeled Deq)
next to the total disappearance rates (Dtotai). The difference between D,0tai and Deq
(Enoniocai in table 4-2) is the proportion of disappearing helpers expected to become
breeders off the study area. If we assume that this emigration rate (En0niocai) is the
proportion of disappearing helpers that successfully became breeders by dispersing as
floaters, we can use En0niocai as a lower bound for Pfioater- That is, the proportion of
disappearing helpers becoming ffbaters must be at least as big as the proportion of helpers
that successfully emigrate off the study area. Thus, En0nioca]/ Dtotai, which is the proportion
of disappearing jays estimated to became breeders off Archbold Biological Station,
establishes a lower bound for Pnoater and because the proportion of floaters that die is
likely to be very high, Pfloaier is likely to be much greater than En0niocai/ Dtotai.
We can develop an upper bound for Pnoater by using the fact that the sum of Pfioater
and Pioci must equal 1 (equation 4-2) and estimating P|0cai, the proportion of
disappearances due to local death rather than floating. We begin with the assumption that
jays engaged in local forays are likely to die at the same or higher rate than jays of similar


375
Fahrig, L. 1992. Relative importance of spatial and temporal scales in a patchy
environment. Theor. Pop. Biol. 41: 300-314.
Ferson, S. 1991. RAMAS/stage: Generalized Stage-based Modeling for Population
Dynamics. Setauket, New York: Applied Biomathematics. 108 pp.
Fishwick, P.A. 1995. Simulation model design and execution: building digital worlds.
New Jersey: Prentice-Hall. 448 pp.
Fitzpatrick, J.W., R. Bowman, D.R. Breininger, M.A. O'Connell, B. Stith, J. Thaxton, B.
Toland, and G.E. Woolfenden. In prep. Habitat Conservation Plans for the
Florida Scrub Jay: A Biological Framework.
Fitzpatrick, J.W., B. Pranty, and B. Stith. 1994. Florida Scrub Jay statewide map. U.S.
Fish and Wildlife Service Report, Cooperative Agreement no. 14-16-0004-91 -
950.
Fitzpatrick, J.W., and G.E. Woolfenden. 1986. Demographic routes to cooperative
breeding in some New World jays. Pp. 137-160 in M.H. Nitecki and J.A. Kitchell,
(eds.), Evolution of Behavior. New York: Oxford University Press.
Fitzpatrick, J.W., G.E. Woolfenden, and M.T. Kopeny. 1991. Ecology and development-
related habitat guidelines of the Florida Scrub Jay (Aphelocoma coerulescens
coerulescens). Florida Nongame Wildlife Program Technical Report, No. 8. 49
pp.
Florida Atlas and Gazeteer. 1997. (4th ed.). DeLorme, Freeport, Maine.
Forman, R.T.T. and M. Godron. 1986. Landscape Ecology. New York: Wiley.
Greenwood, P.J. 1980. Mating systems, philopatry and dispersal in birds and mammals.
Animal Behavior 28: 1140-1162.
Hanski, I. 1994. A practical model of metapopulation dynamics. J. Animal Ecology 63:
151-162.
Hanski, I. and M.E. Gilpin 1991. Metapopulation dynamics: brief history and conceptual
domain. Pp. 89-103 in M.E. Gilpin and I. Hanski (eds.). Metapopulation
Dynamics: Empirical and Theoretical Investigations. London: Academic Press.
Harrison, S. 1991. Local extinction in a metapopulation context: an empirical evaluation.
Pp. 89-103 in M.E. Gilpin and I. Hanski (eds.), Metapopulation Dynamics:
Empirical and Theoretical Investigations. London: Academic Press.


53
is shown in red (HSI = 0.81 1.0), medium quality habitat is shown in blue (HSI = 0.61 -
0.80), and low quality habitat is shown as white (HSI < 0.61). Jay territories (outlined in
black) generally included substantial areas of low quality habitat in both the North and
South areas.
I searched for correlations between demographic (fledglings, independent young,
and yearlings produced, survival of fledglings, yearlings and breeders, and group size)
versus the habitat variables (percent sand, percent tree cover within territories, percent
tree cover within 100 m buffer, HSI for sand, HSI for distance to forest, HSI for trees
within habitat, and combined HSI). The clearest bivariate patterns were for group size
versus percent tree cover within territories (Fig. 3-13) and percent tree cover within the
100 m buffer (Fig. 3-14), but Kruskal-Wallis one-way analysis of variance results showed
no significant differences between different group size comparisons and percent tree
cover. Large group size variance existed in territories with low tree cover or adjacent to
low tree cover, but there was a strong, nonsignificant trend towards smaller groups as tree
cover increases. Kolmogorov-Smimov tests for normality showed that bare sand was the
only normally distributed habitat variable. Normality plots indicated that deviations from
normality could not be corrected by commonly used (e.g. arcsine, inverse, log, square
root) transformations. Nonparametric spearman rank correlation coefficients were low for
all pairings of demographic and habitat variables. Multiple regression models never
explained more than about 22% of the variation in demographic parameters using all
combinations of habitat variables. Similarly, no habitat variables in several logistic
regression models were significant.


29
Table 2-1 cont.
Metapopulation type
(after Harrison 1991)
Metapopulation Type
(Mainland, Midland,
and/or Island)a
Size
(pairs)
Number of
subpopulations
Md
15
1
Classical
161
49
16
101
24
10
61
21
6
Nonequilibrium
31
5
3
31
3
3
21
7
2
21
3
2
21
3
2
21
2
2
21
2
2
21
2
2
I
6
1
I
2
lb
I
1
lc
a Numerical prefix indicates number of Mainlands (Mn), Midlands (Md), and Islands (I).
See text for nomenclature.
b There was a total of 4 single Island systems composed of 2 pairs in one subpopulation.
c There was a total of 8 single Island systems composed of a subpopulation of one pair.


342
1 O
I*
-Q
n
a
o
£1
0 6
F
0 A~~
02
L
L
100
200 300
Threshold Pop Size
400
500
1 0
A
1
c
£
04
rJ
[
L
500
Threshold Pop Size
1000
Fig. 5-2 lm. Lake Wales Ridge quasi-extinction graphs. Top) no acquisition, Bottom)
maximum acquisition.


Probability
202
r
0 2~[
i *
50 100 150
Threshold Pop Size
roT
0 8~
j-
(
os'!;
04
p
0 2~T
it
50 100
Threshold Pop. Size
150
Fig. 5-5d. Sarasota and W. Charlotte quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition.


238
)ay Territory Locations 1 75 km dispersal buffer
Scrub Polygons
^ Lo Disturbance
\7~7~\ Lo Density Housing
Hi Density Housing
B Ranch/Ag
Protection Status
[ I Protected
1 Proposed
Water Bodies
/\/ Interstates
State highways
County roads
. County lines
M10 Merritt Island St
S.E. Volusia
0 6 12 18
1 : 310,000
Kilometers
Fig. 5-10a. Merritt Island and S.E. Volusia county map 1992 1993 jay and habitat
distribution.


54
Discussion
Accuracy of my remotely sensed habitat measurements is difficult to assess. Most
remote sensing studies are conducted at a much coarser scale, and they attempt to identify
discrete data classes (e.g. vegetation types). Such studies typically use a simple error
matrix analysis where percent correct classification is given. I have little precedence to
follow, since the goal of this classification was to provide continuous measurements (i.e.
percent cover) from structural classes (e.g. tree cover or bare sand) rather than discrete
vegetation classes. To assess the accuracy of these measurements quantitatively, I
compared them to ground based measurements using paired T-Tests and simple
correlation analysis. The bare sand measurements obtained from quadrats showed a fairly
good correlation with the image measurements (Fig. 3-4). In contrast, the transect
measurements of tree cover showed a weak correlation (Fig. 3-5). Some of these
differences resulted from classification errors noticeable in comparisons of the
photography with the classified image. Underestimates of tree cover were apparent in
some of the young or very dense pine plantations, which tended to form a uniform
canopy with few shadows. Overestimates were noted in areas with large scrub oaks, such
as in some of the long unbumed scrub patches on the W. side of N. Sandy Hill. Larger
oaks spectrally may look very similar to pine trees. From a Florida Scrub-Jay standpoint,
stands of large oaks may be as unusable as pine forests, so for modeling jay habitat it may
be unnecessary to distinguish these tree cover types.
I suspect that many of the differences in tree cover estimates resulted from
differences in the locations of transects measured on the ground vs. the image. Because
transects are sampling vegetation intercepting a thin vertical plane, relatively small


280
St.Luoie
Martin \
> \
Seabranch St Pk.
Dickinso
Preserve
Jay Preserve
County Pk.
Ridge Natural Area
Hills Natural Area
Universe Scrub Pr.
]ay Territory Locations
Scrub Polygons
Lo Disturbance
^ Lo Density Housing
Hi Density Housing
Ranch/Ag
1.75 km dispersal buffer
Protection Status
|||1-1 Protected
H Proposed
Water Bodies
/\J Interstates
State highways
County roads
County lines
Ml 5 Martin St N. Palm Beach
0 5 10 15 Kilometers
1 : 250,000
Fig. 5-15a. Martin and N. Palm Beach county map 1992 1993 jay and habitat
distribution.


112
outside scrub was parameterized with the best-guess Kaplan-Meier curve (Fig. 4-2)
generated from the radiotelemetry field data.
Other floater parameter values were estimated indirectly, and are listed in table 4-
5. The selection of these values is described in the Methods section.
Constraint Analysis
The constraint analysis provided a valuable upper bound for several of the floater
dispersal parameters (table 4-5). An unrealistically high number of dispersals from the
Lake Wales Ridge to the Bright Hour Ranch occured for parameter settings that are not
substantially higher than the default settings (table 4-5).
Calibration and Validation
Comparisons between model output and dispersal data obtained from long term
observations of marked jays (unpublished data provided by J.W. Fitzpatrick and G.E.
Woolfenden) are shown for male and female dispersal in Figs. 4-4 and 4-5. The general
shapes of the dispersal curves produced by the model closely resembled the field data.
Model output did not differ significantly for females and was marginally different for
males compared to the field data (Kolmogorov-Smimov: male Z = 1.373, p = 0.046;
female Z= 0.784, p = 0. 570). Comparisons between model and field data for age-stage
structure of helpers and breeders are shown in Figs. 4-6 and 4-7 (obtained from tables 9.8
and 9.9 in Woolfenden and Fitzpatrick 1984). Model and field data comparisons are not
significantly different for helpers or breeders (Kolmogorov-Smimov: breeder Z = 0.447,
p = 0.988; helpers Z= 0.267, p = 1.00).


93
(< 20 birds from Archbold Biological Station; John W. Fitzpatrick, pers. comm.), and the
movement and behavior of such jays remains essentially unobserved. To acquire some
empirical data on floaters, a small radiotelemetry study was conducted for this
dissertation.
GIS Files
The GIS files used in the simulations for this chapter and chapter 5 were created
by overlaying the scrub patches obtained during the 1991-1992 statewide Florida Scrub-
Jay survey (see chapter 2) onto a statewide habitat classification map produced by the
Florida Game and Freshwater Fish Commission (FGFWFC) in 1992 (Kautz et al. 1993).
Spatial resolution of the GIS file was 30 m. The original landcover types coded in the
FGFWFC classification are shown in Table 4-1.
Simulating Philopatric Dispersal
I modeled the behavior of helpers searching for breeding vacancies near their
natal territory using a small set of behavioral rules. Helpers engaged in philopatric
dispersal are assumed to have perfect knowledge of the status of each territory within
their assessment sphere. Helpers compete for vacancies; older helpers out compete
younger helpers, and closer helpers out compete more distant helpers. Dispersers that
find no vacancies during this search return to their natal territory and remain as helpers
until the following year.
Philopatric disperser survival rate is not directly specified in the model, but is
related to the floater frequency rate described below.


382
Woolfenden, G.E., and J.W. Fitzpatrick. 1991. Florida Scrub Jay ecology and
conservation. Pp. 542-565 in C.M. Perrins, J.D. Lebreton, and G.J.M. Hirons
(eds.), Bird Population Studies: Relevance to Conservation and Management.
Oxford, UK: Oxford University Press.
Woolfenden, G.E. and J.W. Fitzpatrick. 1996. Florida Scrub-Jay (Aphelocoma
coerulescens). In A. Poole and F. Gill (eds.), The Birds of North America, No.
228. Philadelphia: The Academy of Natural Sciences, and Washington, DC: the
American Ornithologists Union.
Wooton, T.J. and D.A. Bell. 1992. A metapopulation model of the Peregrine Falcon in
California: viability and management strategies. Ecological Applications 2: 307-
321.
Zack, S. 1990. Coupling delayed breeding with short-distance dispersal in cooperatively
breeding birds. Ethology 86: 265-286.
Zar, J.H. 1996. Biostatistical Analysis (3rd ed.). New Jersey: Prentice Hall. 662 pp.
Zollner, P.A. and S.L. Lima. 1999. Search strategies for landscape-level interpatch
movements. Ecology 80: 1019- 1030.


140
attractiveness. They move away from habitat with low attractiveness, and towards habitat
with high attractiveness. Dispersers remember previously visited locations, which makes
it less likely that they will backtrack unless alternative directions are very unattractive.
Long distance dispersers have two mortality rates: one for dispersers within scrub,
another for dispersers outside of scrub. Within scrub, the disperser survival rate is higher
than outside of scrub (Table 5-1). Each disperser moves until it exceeds a random daily-
distance-moved threshold selected for each jay from a distribution of daily move
distances. Once the latter distance is exceeded, a daily mortality rate is used to determine
if the jay survives to the next day. These steps are repeated for each jay until it dies, finds
a mate or leaves the simulation area. Jays that leave the area are considered dead (i.e.
there is no immigration from outside the simulation area). In contrast to short distant
dispersers, long distance dispersers do not return home.
After all floaters have settled, died, or left the simulation area, the annual cycle is
repeated with the reproduction step. If all jays are extinct or the last year of the last
repetition is reached, the simulation terminates.
Territories
The model tracks individual territories, maintains a list of jays occupying each
territory, and graphically displays the occupancy status of each territory. In the
simulations completed for this project, all territories were assumed to be 9 hectares and
shaped as squares. Territory locations were read in from an ASCII file exported from
Arcview. Position, number, and habitat quality of territories did not change over time. All


Population Size
174
h
i
20 40 60
Year
20 40 60
Year
Fig. 5-2f. Citrus and S. Marion county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition


programming resources, and John Dixon for statistical help. Dr. Ken Portier also
provided valuable statistical advice.
Last, but no least, I thank my family and especially my wife, Ellen Thoms.
vi


xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EY4E8414J_QJXCFL INGEST_TIME 2015-02-24T19:58:49Z PACKAGE AA00028807_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES


268
1 o
O 8
r
60
10
2
g
2
o
£
0 8
0 6
04 _
0 2
20
40
Threshold Pop. Size
60
80
Fig. 5-13d. S. Brevard-Indian River-N. St. Lucie county quasi-extinction graphs. Top) no
acquisition, Bottom) 30% acquisition by area.


59
in poor quality habitat may have a greater tendency to emigrate than helpers in high
quality habitat.
The remote sensing techniques described above show significant potential for
evaluating Florida Scrub-Jay habitat. Tree cover and bare sand are important habitat
variables that are relatively easy to measure with these techniques. Oak cover is likely to
be important to scrub-jays, but the techniques I investigated could not discriminate oaks
from other low-lying vegetation such as palmettos. Because of low mast failure and high
acorn production, oak cover may not be a limiting factor for many scrub-jay populations.
Further investigation of the importance of oak cover is needed.
The results of this study suggest that tree cover exceeding 20% 30% within or
adjacent to territories is associated with reduced demographic performance and may
create sink populations of Florida Scrub-Jays. Although I lack direct evidence, much
indirect evidence suggests that forest-dwelling predators and competitors explain the
negative relationship between tree cover and demographic success. Sink populations can
constitute a major proportion of a species population and may contribute to
metapopulation longevity (Howe and Davis 1991), but the loss of a single critical source
population may result in the extinction of all dependent sink populations (Pulliam 1988).
Thus, management practices should seek to convert sink populations to self-sustaining
source populations. The results of this study suggest that for the Florida Scrub-Jay, this
entails keeping tree cover within and adjacent to jay territories at relatively low levels.


253
Central Brevard (Ml2)
General description: The Central Brevard metapopulation is separated from the N.
Brevard metapopulation (Ml 1) by the city of Cocoa, from the S. Brevard- Indian River-
N. St. Lucie metapopulation (Ml 3) by the city of Melbourne, and from the Merritt Island
metapopulation (M10) to the east by the Indian River (see maps in Fig. 5-12a, b). The
SMP documented about 36 jay territories, excluding suburban jays, in this
metapopulation. Estimated potential population size after habitat restoration and full
occupancy is 5 pairs in currently protected areas, and 40 pairs maximum.
Protected areas: Rockledge Scrub Preserve (Brev40); Wickham County Park
(Brev43). Portions of the large, contiguous habitat patch (Brev41) just south of
Rockledge are targeted for acquisition (1999 Carl project), as is habitat (Brev42 -
CARL 1999 site) just north of Wickham County Park. The Melbourne regional airport
(Brev29), which lacks a habitat management plan, was not included as a protected area.
Restoration potential: At the time of the SMP, jay densities in most areas probably
were close to maximum (compare first and last data columns in Table 5-12a).
Simulation results: This metapopulation ranked 4th in vulnerability (table 5-23),
18th in percent protected (12.5%; table 5-24), and 4th in priority (table 5-25), with high
vulnerability and high potential for improvement. The no acquisition option has an
extremely high probability of extinction (p=1.0) and quasi-extinction (p=1.0). The 70%
acquisition by area has a high quasi-extinction risk (p=0.43) and a moderate extinction
risk (p=0.10). Risk estimates for the maximum acquisition option are substantially
reduced for both quasi-extinction (p=0.0) and extinction (p=0.10), as is percent
population decline (see Table 5-12b).


113
Discussion
Each of the behavioral rules (table 4-3) implemented in the software to generate
dispersal and stage-age curves similar to the Archbold field data has clear biological
meaning and is readily matched to features of known jay biology. Early versions of the
model, incorporating simpler rules, produced dispersal curves quite different from the
Archbold data. Interestingly, some of these curves were more similar to dispersal curves
for non-cooperative species. The final set of rules seems to capture key elements of
dispersal behavior that may partly account for differences between cooperative and non-
cooperative species. For example, after reviewing bird studies with adequate dispersal
data, Zack (1990) found that cooperative species dispersal distances were strongly
skewed towards the natal territory compared to non-cooperative breeding species.
Dispersal curves generated by early versions of my model resembled non-cooperative
species in that the mode was several territories away from the natal territory. This pattern
arose when the probability of settling on a vacancy was nearly the same for any territory
within an individuals assessment sphere. To produce the strongly right-skewed pattern of
a cooperative breeder, rules must be added to increase the probability of acquiring nearby
vacancies rather than those that are more distant within the assessment sphere.
Algorithmically, I simulated this behavior by keeping a distance-sorted list of nearby
territories within each dispersers assessment sphere. When a breeder death creates a
vacancy, the helper closest to that territory was given first choice at the opening (ties
were broken by random draw). This dominance hierarchy is further modified to account
for age by processing older helpers before younger. These rules are summarized in table
4-3.


323
20 40 60
Year
Fig. 5-20c. Central Lake county trajectory graphs. Top) no acquisition, Bottom)
maximum acquisition.
J


31
Highly
Connected
A
Patch
Isolation

Highly
Isolated
Patchy
Classical
Mainland- Mainland-
Island Mainland
Nonequilibrium Disjunct
All
Small
Mixture of All
Small & Large
Large
Patch Size
Fig. 2-2. Classification scheme showing different types of metapopulations based on
patch size distribution (patches all small in size, mixture of small and large, and all large
in size) along the horizontal axis, and degree of patch isolation (highly connected to
highly isolated) on the vertical axis. Nonequilibrium, classical, mainland-island, and
patchy classes are named according to Harrison (1991).


55
difference in transect position can result in large differences in measurements. Quadrat
measurements are probably less sensitive to positional inaccuracy than transects.
It was surprisingly difficult to estimate the magnitude of the positional
inaccuracies of the transect locations. The locations of the transects were predetermined
by a program that generated random locations for transect endpoints. These transect
locations were plotted on high resolution photo maps which were taken into the field and
used to stake out the transects. Thus, the correspondence between the GIS-based location
and the actual ground location depended on the field persons ability to find the exact
location from the photo map. My qualitative impression was that accuracy of positioning
the transects depended on whether features visible on the photo map could be located on
the ground. In sparsely forested areas, individual trees and bare sand patches were
identifiable on the photo and ground, and served as good reference points. Under these
conditions, a transect could probably be placed within several meters of its position on
the photo. In heavily forested areas, good reference points were absent, and positional
accuracy probably decreased, to errors of 10 meters or more. It might seem that
differential GPS could solve this potential problem. Unfortunately, several factors make
this approach more problematic than anticipated. First, the stated accuracy of the
differential GPS approach available to us is 2 5 meters, which can misplace the ends of
a transect by 3 pixels in any direction. Also, I have occasionally encountered averaged,
differentially corrected points that are off by considerably more than 5 meters. Second,
current GPS units often are unable to pick up the necessary signals within forests;
precisely where they are most needed for this study. Third, the positional accuracy of the
imagery itself is unknown, but is probably on the order of 2 to 10 meters (see


49
for each territory. Because summary requires integer values for input, the HSI values
were resampled to 5 equal intervals.
Collection of Demographic Data
Demographic data for jays on Sandy Hill were collected by a team of field
researchers (Brad Stith, Reed Bowman, Doug Stotz, Larry Riopelle, Natalie Hamel, and
Mike McMillan) during 1994 1995, as part of a larger, on-going project that now
monitors jays on the entire APAFR. All jays on Sandy Hill were captured, color-banded,
and monitoried quarterly using techniques similar to Woolfenden and Fitzpatrick (1984).
Nests were found and monitored during the spring and early summer. The raw
demographic data are presented in Table 3-1.
Habitat-Demographic Analysis
I compared the demographic performance and habitat characteristics between the
North and South Sandy Hill populations of jays. Owing to lack of normality for nearly all
parameters (Table 3-2; Kolmorov-Smimov test for normality), the nonparametric Mann-
Whitney U statistic was used to test for demographic and habitat differences between the
North and South study areas (Table 3-3).
I searched for habitat-demographic relationships by performing multiple linear
regression (maximum R2 improvement technique for all combinations of variables) and
logistic regression (SAS Institute). Demography parameters served as dependent
variables, and habitat measurements as indepe- ,.ent variables.


Table 5-13b. S. Brevard-Indian River-N. St. Lucie county simulation statistics.
Data type
No
acquisition
30% preserved
by connectivity
30% preserved
by area
70% preserved
by connectivity
70% preserved
by area
Maximum
acquisition
Starting
population size
62
91
91
136
136
165
Mean ending
population size
28.5
44.3
71.9
91.1
107.3
124.1
s.d.
13.5
17.9
13.9
21.4
15.3
15.0
Percent
population
decline
54.0
51.3
21.0
33.0
21.1
24 8
Extinction
risk
0.07
0.0
0.0
0.0
0.0
0.0
Quasi
extinction
Risk (10 pairs)
0.20
0.07
0.03
0.03
0.0
0.0


Table 5-26 (cont.)
Metapopulation
Primary Acquisition
Targets
Secondary
Acquisition
Targets
Restoration (protected areas)
Other actions
Ocala National
Forest (Ml7)
Creation of scrub-jay preserve in
Marl 5
Lake Wales
(M21)
High2 (Hendrie Ranch),
High 15 (additions to
Highlands Hammock
S.P.), Polk7, Polk8,
Polk9, Polk 13
all patches in
Glades county?
High 15 (Highlands Hammock S.P.),
High 17 (Carter Creek), High 19 (Sun
n Lakes)


Population Size
201
a
w
a.
o
CL
I
40
10
20
l
30
Year
50
20 0
20
40
Year
60
Fig. 5-5c. Sarasota and W. Charlotte county trajectory graphs. Top) no acquisition,
Bottom) maximum acquisition.


120
Table 4-1. Landcover types from statewide habitat map (Kautz et al. 1993) used in
simulations and associated floater attractiveness values.
Landcover Type
Attractiveness Value0
Coastal Strand 3
Dry Prairies 1
Pinelands 2
Sand Pine Scrub 2
Sandhill 2
Xeric Oak Scrub 4
Mixed Hardwood Pine Forest 1
Hardwood Hammocks and Forests 1
Tropical Hardwood Hammock 1
Coastal Salt Marshes 1
Freshwater Marsh and Wet Prairie 2
Cypress Swamp 1
Hardwood Swamp 1
Bottomland Hardwoods 1
Bay Swamp 1
Shrub Swamp 2
Mangrove Swamp 2
Aquatic 0
Grassland 1
Shrub and Brushland 2
Exotic Plant Communities 2
Barren 0
a Attractiveness values range from 0 to 4 (0 = repulsive; 1 = unattractive; 2 = neutral; 3
= attractive; 4 = highly attractive).


250
Fig. 5-1 lc. N. Brevard county trajectory graphs. Top) 30% acquisition, Bottom) 70%
acquisition.


348
Percent population protected. Table 5-24 shows the percent of each population
that is currently protected (assuming all habitat is restored and fully occupied). Eleven of
the 21 metapopulations have less than 50% of their potential population protected.
Priority ranking. Table 5-25 provides a priority ranking of metapopulations based
on a simple classification scheme I devised. The highest priority ranking is given to those
metapopulations that are most vulnerable and have the highest potential for improvement.
The lowest ranking is given to metapopulations with the lowest vulnerability and least
potential for improvement. I arbitrarily defined three vulnerability categories based on
the quasi-extinction estimate for the no acquisition scenarios (low vulnerability: p =
0.0 0.05, moderate vulnerability : p = 0.5 0. 20, and high vulnerability: p > 0.20). I
arbitrarily defined three potential for improvement categories based on the difference
between the quasi-extinction estimates for the maximum acquisition option and the no
acquisition option (low improvement: p = 0.0 0.05, moderate improvement: p = 0.5 -
0.20, and high improvement: p > 0.20).
This classification scheme indicates that there are 13 metapopulations of moderate
or high vulnerability that also have moderate or high potential for improvement. These 13
metapopulations (N. Brevard, Levy, Central Charlotte, Central Brevard. W. Volusia,
N.W. Charlotte, St. Lucie, Citrus, Lee, Manatee, Pasco, S. Brevard, and Sarasota) score
highest on the priority list. Three metapopulations (Palm Beach, Central Lake, and
Flagler) have high vulnerability, but low potential for improvement. The remaining 5
metapopulations have low vulnerability and low potential for improvement, at least as
measured by quasi-extinction estimates. As discussed earlier, other statistics such as
percent population decline should be used to evaluate these large metapopulations.


27
dispersal buffer. The internal structure of each metapopulation is easily described using
the inner dispersal buffer to delineate islands, midlands, and mainlands. Enumerating all
metapopulations and describing their internal structure (table 2-1) reveals much about the
distribution and viability of Florida Scrub-Jays. The differing internal configurations of
metapopulations present different conservation problems and require different
management approaches. Discussion of these matters is deferred until chapter 6,
following the presentation of the modeling results (chapter 5) which analyze the viability
of metapopulations around the state.
Note: This chapter has been published as Stith et al., 1996, and is reproduced with some
modifications with the permission of Island Press.


CHAPTER 3
REMOTE SENSING OF FLORIDA SCRUB-JAY HABITAT
Introduction
The importance of understanding relationships between wildlife and habitat has
been recognized for many decades (review in Morrison et al. 1992). The ultimate success
of wildlife management and conservation efforts depends to a large degree on our ability
to understand these relationships. Unfortunately, measuring habitat variables over large
areas, and with sufficient spatial resolution to capture the essential habitat features a
particular species responds to, is a difficult and time consuming task. In this chapter I
apply a technique that greatly assists in the measurement and analysis of wildlife-habitat
relationships across large areas. This technique, which relies heavily on recent advances
in computer hardware and software technology, uses image processing and GIS software
to measure habitat variables directly from scanned aerial photography.
Habitat requirements differ widely among most species, and choosing the
appropriate habitat variables to measure is greatly facilitated for species whose habitat
requirements are well understood. Long-term studies of the Florida Scrub-Jay have
revealed much about the habitat requirements of this species (Woolfenden and Fitzpatrick
1984; Breininger et al. 1991, 1995, 1996). Research indicates that scrub-jays have a fairly
simple "habitat template" consisting of low vegetation dominated by scrub oaks, open
40


Table 5-26 (cont.)
Metapopulation
Primary Acquisition
Targets
Secondary
Acquisition
Targets
Restoration (protected areas)
Other actions
Citrus-S.W.
Marion (M2)
Mar2, Mar8, Citr2
(Re-evaluate after new
survey completed)
Citr5, Mar7
(Re-evaluate after
new survey
completed)
Citrl (Crystal R. St. Buffer Pr.), Citr6,
Mar5, Sumtl
Improved survey.
Purchase/restore
unoccupied patches e.g.
N. side of Citr3 & Mar3
Lee-Collier
(M8)
Colli, Coll4, Coll5,
Coll7,
Lee3 (Estero Bay Aquatic Pr.), Coll2
(Rookery Bay N. Estuarine Research
R.), Coll3 (Immokalee airport)
Purchase/restore
unoccupied patches
south of Lee3
Manatee (M4)
Sari6, Man, Mani,
Man 15, Man9, Man 10,
Manl 1, Man5, Harl
(Re-evaluate after new
survey completed)
(Re-evaluate after
new survey
completed)
Sari 5 (Myakka S.P.), Sari 9 (Verna
Wellfield), Manl5 (Duette),
Manl2&16 (Beker), Manl7-18 (Lake
Manatee), H112-3 (Little Manatee),
H118 (Balm-Boyette Scrub Pr.), HH19
(Golden Aster Scrub Nature Pr.)
New survey (emphasize
atypical habitat)
Pasco (M3)
Pas2
(Re-evaluate after new
survey completed)
(Re-evaluate after
new survey
completed)
Herl (Weeki Wachee), Pasl
(Starkey/Serenova), Pas3 (Cross-bar/
Al-bar), Pas5 (Alston Tract), Pas
(Green Swamp W.M.A.)
New survey (emphasize
atypical habitat)
S. Brevard
(Ml 3)
Brev31, Brev32, Brev38
Brev36, InRi9, InRilO,
InRi5, InRi6
Brev30 (Malabar Scrub Sanct.),
Brev31 (Valkaria), Brev35 (St.
Sebastian River S.P.& Micco Scrub
Sanct.), InRil (Sebastian airport),
lnRi2 (private), InRi4 (Wabasso Scrub
Pr.), StLul (St. Lucie airport)
Survey for additional
jays on south end of Ten
Mile Ridge


XXV
classification reveals key subpopulations requiring special attention to maintain the long
term viability of the existing metapopulations.
I developed and applied a technique for measuring habitat features and estimating
habitat quality over large areas using image processing and GIS methods. The technique
showed that jays in central Florida had a strong preference for open sandy areas, and few
or no pine trees. A proximity analysis showed that demographic performance decreased
near forests. Measurement of habitat variables using this technique will be a valuable
technique for habitat management and conservation.
I developed a spatially explicit, individual-based model to simulate the
metapopulation dynamics of Florida Scrub-Jays. Special emphasis was placed on
realistically modeling dispersal. I conducted a small radio-tracking study and used data
from long-term studies to parameterize and validate the model. Stage-age structure and
dispersal distances generated by the model showed good fit to field data.
I used this simulation model to investigate the viability of 21 major Florida Scrub-
Jay metapopulations across the state. For each metapopulation I simulated 2 or more
hypothetical reserve designs, ranging from a minimal design with only currently
protected jays, to a maximal design containing all significant populations as of 1993. All
habitat was assumed to be restored and fully occupied. Model results indicated that only
3 of 21 metapopulations would be adequately protected without further habitat
acquisition. At least 4 metapopulations appear to be at great risk of extinction.


335
Jay Territory Locations
(after restoration)
| Scrub Polygons
CCC## Polygon ID
Jays outside of labeled, bold polygons are considered to be Suburban jays.
Protection Status
/\/ Protected
/\/ Unprotected
Water Bodies
County lines
M 21 S. Highlands
O
1 : 250,000
i 5 Kilometers
V
4*
Fig. 5-2lh. Lake Wales Ridge acquisition map, S. Highlands county.


143
Map Production
A statewide metapopulation map was produced to depict the 21 metapopulations
that were analyzed for this chapter (Fig. 5-1). For each of the 21 metapopulations, two
types of detailed maps were produced to depict the status of jays in 1992-1993 as
determined by the SMP, and to depict what jay populations might look like if all habitat
were restored and fully occupied by jays. The maps showing restored populations of jays
provided the basis for the simulations, as explained below.
Statewide metapopulation map
Metapopulations were delineated initially using a dispersal buffer approach
discussed in chapter 2. A GIS was used to generate a 12 km buffer around all jay
territories to enclose populations that are likely to be connected by dispersal. In certain
areas the buffers joined populations that probably are not connected due to physical
barriers to movement, such as a large river systems or cities. These physical barriers are
identified in the written accounts for each metapopulation (see Metapopulation Viability
Analysis section). Figure 5 -1 shows a map of the 21 metapopulations identified for the
entire state. Each of these 21 metapopulations were modeled as demographically
independent units as described below.
1992-1993 SMP maps
The status of jays and habitat as determined by the 1992-1993 SMP were
portrayed for each metapopulation on one or more maps. The jay data included some


117
floater algorithm, and the constraint analysis suggested that actual movements should not
be much larger than the measured distribution.
The constraint analysis showed that simulated jays originating from Archbold or
nearby patches rarely colonized the Bright Hour Ranch for certain parameter settings,
even though the dispersal curve showed that some Archbold dispersers did succeed in
settling much greater distances within the Lake Wales Ridge than the distance to Bright
Hour. This bias in dispersal success is largely due to the floater mortality rates set in the
model for different habitat types; jays that moved the longest distances were those that
managed to stay in scrub habitat the longest.
The displacement experiment suggested that jays suffered much higher mortality
rates while moving through non-scrub landscapes than they would likely experience in
scrub habitat. Predators took several jays, and it is possible that several other jays that
vanished inexplicably were also predated. These losses suggest that dispersal outside
scrub habitat is a costly activity for Florida Scrub-Jays, which is consistent with other
findings and a suite of behavioral traits described earlier, including the sentinel system,
delayed dispersal, and unusually short dispersal distances. The Kaplan-Meier curve
developed from the radiotelemetry data provided strong evidence for the vulnerability of
jays to predation in unfamiliar, non-scrub habitat. The Kaplan-Meier curve also provided
a useful means of parameterizing the daily survival of floaters (fig. 4-2), and
demonstrated graphically the potentially high cost of long distance dispersal through
unfamiliar terrain. The mortality rates measured for displaced jays are extremely high,
and likely are higher than would be experienced by naturally dispersing, untagged jays
due to negative effects of the backpack harness and radio transmitter. Nevertheless, the


Table 3-1 continued.
Location Territory Year Group Fledgling Yearling Breeder Helper
Size Production Production Surv. Surv.
SR1DGE
LOGG
1994
4
0
0
0.5
0.5
SRIDGE
LOPI
1994
2
3
1
0
SRIDGE
LOST
1994
3
2
2
1
0
SRIDGE
NORE
1994
2
1
0
0.5
SRIDGE
SLOG
1994
3
0
0
0.5
0
SRIDGE
SNAG
1994
2
4
1
1
SRIDGE
SRNG
1994
2
4
1
1
SRIDGE
TRIS
1994
2
3
1
1
SRIDGE
WPND
1994
2
3
1
1
NR1DGE
ARDS
1995
5
3
0
0.5
0.33
NRIDGE
BUDD
1995
6
2
0
0.5
0.66
NR1DGE
DTCH
1995
2
3
0
1
NRIDGE
DUMP
1995
2
4
1
1
NRIDGE
FLIN
1995
2
1
1
0.5
NRIDGE
FRST
1995
5
0
0
1
0.66
NRIDGE
GARD
1995
4
2
0
0
0.5
NRIDGE
JUVI
1995
2
1
0
1
NRIDGE
NTRL
1995
3
3
1
0.5
1
NRIDGE
PURP
1995
2
3
0
0
NRIDGE
SQAR
1995
5
3
1
1
0.66
NRIDGE
TOWR
1995
5
3
1
0.5
0.66
NRIDGE
TRGT
1995
3
1
1
0.5
1
NRIDGE
TWNP
1995
6
2
2
1
0.33
Fledgl. Yearl. Bare Tree cover Tree cover
Surv. Surv. sand (inside (100m
territory) buffer)
0.33
1
0
0.25
0.25
0.33
0.33
0
0
0
0.25
1
0
0
0.33
0
0.33
0.33
1
1
0.33
0
0
0
0.25
0.33
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0.14
0.28
0.21
0.24
0.21
0.11
0.29
0.17
0.1
0.01
0.19
0.23
0.29
0.09
0.24
0.32
0.08
0.26
0.1
0.15
0.21
0.21
0.16
0.22
0.04
0.08
0.07
0.13
0.46
0.31
0.14
0.28
0.02
0.07
0.01
0.01
0.03
0.04
0.09
0.02
0.01
0.02
0.03
0.01
0.06
0
0.28
0.08
0.01
0.12
0.25
0.44
0.35
0.19
0.24
0.02
0.01
0.01
0.01
0.01
0.02
0.06
0.02
0.01
0.02
0.02
0.04
0.02
0.02
oo
UJ


87
insensitive to errors in dispersal parameters except under very limited conditions. The
debate over the reliability of SEPMs is likely to continue, but as South (1999) points out,
model sensitivity to dispersal parameters may be greatly reduced by increasing model
realism.
The Florida Scrub-Jay offers an excellent opportunity to develop an extremely
realistic SEPM. Extensive dispersal information is available for this intensively studied
species (e.g. Woolfenden and Fitzpatrick 1984). Furthermore, there is great concern for
this species, which is listed as a threatened species Federally and by the State of Florida.
Much is known about the spatial distribution of the species (e.g. chapter 2), making it
feasible to model the entire population. Root (1996,1998) used RAMAS GIS to develop
the first SEPM for scrub-jays. Roots research was focused on four somewhat isolated
populations within Brevard county, Florida. She modeled dispersal among these four
populations for female jays only, using the interpatch distances and the ABS dispersal
curve to estimate migration rates. Her results suggested that interpatch dispersal was
important for offsetting the deleterious effects of epidemics. Root (1998) stated that her
estimates of dispersal likely were optimistic and suggested that a better approach would
account for differential dispersal rates based on the interpatch matrix.
In this chapter I describe a SEPM I developed to account for the influence of
interpatch matrix on dispersal, as well as a host of other biological details documented in
the extensive scrub-jay literature (see Woolfenden and Fitzpatrick 1996 for a recent
literature review). The SEPM is an individual-based model; it tracks all individuals of
both sexes from birth to death, and simulates the daily movement of individuals during
dispersal within and between habitat patches.


147
al. manuscript). Most of these estimates differed little from the densities determined by
the SMP.
Identification of protected areas
Boundaries were digitized around occupied, protected jay habitat using Arcview.
These boundaries appear on the acquisition maps for each metapopulation as heavy solid
lines. Protected areas were identified through the use of the 1998 F.N.A.I. publication
(Blanchard et al. 1998), annual C.A.R.L. reports (especially Anonymous 1999), Arc/Info
coverages obtained from water management districts, and from verbal updates provided
by individuals familiar with specific sites. The source data for the F.N.A.I. publication
and the Arc/Info coverages date from late 1997. A significant number of acquisitions
have been made since that time, some of which may not be identified as protected in this
document. In a few cases, local experts suggested that a particular tract of land be treated
as protected even though it was in private hands or the property had been only partially
purchased. In these cases, it is possible that areas designated as protected may be less
than shown on the maps. The ongoing process of land acquisition ensures that any map
will be obsolete as soon as it is published, and such errors will affect the outcome of
some simulations.
Assessment of unprotected areas
Unprotected, occupied patches of jay habitat delineated by the 1992-1993 SMP
were grouped into two categories: patches having sufficient potential to be considered for
acquisition, and patches with little or no acquisition value due to excessive human


362
which may create source-sink dynamics along roads. The absence of these factors from
my model provides further reason to view' the model predictions as underestimating
extinction risk.
Thus, the results presented in this chapter must be viewed with these biases and
assumptions in mind. One of the important steps in the modeling process is to present the
structure of a model and its assumptions explicitly, so that others can decide whether the
results are useful (Rykiel 1994). Models are assumption analyzers (Bart 1995) they
provide a means of integrating empirical data, hypotheses, theories, and intuition into a
formal framework that reveals the consequences of the underlying assumptions. In my
estimation, the assumptions I made for the simulations presented in this chapter should be
viewed as optimistic with regard to Florida Scrub-Jay viability. Therefore, the simulation
results which are summarized below should be viewed as optimistic scenarios for Florida
Scrub-Jay metapopulations.
Assuming that no additional scrub-jay habitat is protected, 11 of 21
metapopulations were estimated to be highly vulnerable to quasi-extinction (N. Brevard,
Levy, Central Charlotte, Central Brevard, W. Volusia, N.W. Charlotte, St. Lucie, Citrus,
Lee, Manatee, Pasco). Of these 11 metapopulations, the risk of quasi-extinction could be
greatly reduced for 7 metapopulations by acquiring all or major portions of the remaining
jay habitat (N. Brevard, Levy, Central Charlotte, Central Brevard, W. Volusia, N.W.
Charlotte, St. Lucie). However, even after total acquisition Central Charlotte and N.W.
Charlotte showed large mean population declines (65% and 61% respectively). The other
4 metapopulations (Citrus, Lee, Manatee, and Pasco) showed high quasiextinction
vulnerability, and moderate potential for improvement through acquisition. However,


7
analysis relied on data that suggested where successful dispersal from Archbold
Biological Station could and could not take place.
Chapter 5 describes the complete individual-based, spatially explicit population
model, which incorporates the dispersal algorithms described in chapter 4. The model
provides a framework for integrating much of what is currently known about the Florida
Scrub-Jay. Simulations take place on a landscape provided by a geographic information
system (GIS) file. Non-dispersing jays occupy discrete territories. Both sexes are
modeled, and individual jays progress through 5 stages (juvenile, 1-year helper, older
helper, inexperienced and experienced breeder). Each territory has a separate set of
demographic parameters assigned to each sex and stage. Breeder experience and presence
of helpers may affect fecundity. Helpers monitor neighboring territories within their
assessment sphere and vie for breeder openings; the outcome of such competition is
determined by simple dominance rules. Helpers may leave on long distance dispersals,
during which time mortality and movement varies depending on landcover type.
The statewide population of jays was divided into 21 metapopulations thought to
be demographically isolated from each other (fig. 5-0). Two series of maps were
developed for each metapopulation. One map type depicts jays and habitat as mapped in
1992-1993. A second map type, referred to as an acquisition map, depicts jays as they
might exist if all habitat patches were restored to optimal conditions, and distinguishes
among jays within protected areas, unprotected habitat patches, and suburban areas. Key
habitat patches are labeled on the acquisition maps, and are cross-referenced in the text
descriptions, tables, and recommendations.


Table 5-8a. Lee and N. Collier county patch statistics (number of jay territories for different configurations)
Patch id
Status
1992-1993#
No
30%
70%
Maximum
jay territories
acquisition
acquisition
acquisition
acquisition
(restored)
by area
by area
Lee1
8
6
8
Lee2
15
8
15
Lee3
Estero Bay Aquatic Pr.
2
9
9
9
9
Lee4
1
1
Colli
2
6
6
7
Coll2
Rookery Bay Nat. Estuarine
Research Reserve
3
6
6
6
6
Coll3
Immokalee Airport
4
4
4
4
Coll4
2
2
2
Coll5
5
5
5
C0II6
3
3
Coll7
2
2
2
Totals
47
15
25
48
62
K>
K>


184
20 40 60 80
Threshold Pop. Size
20 40 60 80
Threshold Pop. Size
Fig. 5-3f. Pasco and Hernando county quasi-extinction graphs. Top) no acquisition,
Bottom) maximum acquisition.


Recommendations: Although afforded little protection, the viability of this
metapopulation could be greatly increased through acquisition of the few remaining
254
habitat patches. The long term viability of this metapopulation and the small Rockledge
Scrub Preserve (Brev40) depends critically on substantial acquisition and restoration of
habitat at Brev41 (EELS/CARL 1999 site). The small population at Wickham County
Park (Brev43) would benefit greatly from proposed acquisition of habitat just to the
north (Brev42 1999 CARL). A habitat management plan is needed for the jays at
Melbourne Regional Airport (iBrev44).


137
Starting Population Stage Structure
At the start of each repetition of each simulation run, all territories are initialized
with a pair of inexperienced breeders (both 2 years old) and one inexperienced helper (1
year old; randomly selected sex). The location of each territory is obtained from an
ASCII file provided to the model (see GIS section below).
Annual Life Cycle
The scrub-jay annual life cycle is simulated by a series of events scheduled in an
event queue; each event is completed for the entire metapopulation before the next event
begins. The following is a summary of the major events in the annual cycle, which begins
with reproduction.
Reproduction. Each territory produces a poisson distribution of juveniles (see
Burgman et al. 1993) with 3 different means (see Table 5-1) for: 1) at least one
experienced breeder with at least one helper, 2) at least one experienced breeder but no
helpers, or 3) both novice breeders. The fecundity parameter values are set to the number
of one-year old offspring produced, rather than fledglings produced. From a software
efficiency standpoint, this greatly reduces the number of jays that must be created and
then destroyed in the mortality step that immediately follows (i.e. the model does not
subject juveniles to mortality during their first year the fecundity rate accounts for this
mortality). Demographic stochasticity of fecundity is implemented by randomly
selecting the sex of offspring. Environmental stochasticity of fecundity is not
implemented, but would be expected to have a negative effect on population persistence.


45
reflectance or very high reflectance in all three spectral bands. Extremely low reflectance
values corresponded to tree crowns, the shadows cast by tree crowns, or standing water.
Although appearing in the same spectral class, water was easily distinguished from tree
crowns and shadows by pattern and texture. High reflectance values corresponded to bare
sand patches, and human disturbances such as dirt roads and excavations. Naturally
occurring bare sand patches were nearly always associated with xeric habitats, and had a
distinctive, fine-grained pattern and texture compared to ground features created by
humans. Spectral classes with intermediate reflectance were much more difficult to
associate with ground features. In general, grass-dominated prairies, such as occur
between scrub patches on N. Sandy Hill, had high reflectances that were only slightly less
than bare sand. Areas dominated by oak shrubs were spectrally similar to areas
containing various proportions of palmetto and wire grass. Discriminating among mixed
shrub classes was difficult and also was believed unlikely to affect jay dispersion at
APAFR. The dominant and most recognizable spectral classes corresponded to tree
canopies and the shadows they cast, and bare sand patches. All spectral classes were
recoded to the following landcover classes: 1 = tree cover, 2 = bare sand, 3 = mixed
grass/shrub, 4 = wetlands/seeps. These classes were ;ntended to reflect key structural
components of the habitat rather than vegetative types.
Manual Editing of Classification
Manual editing of the final classification was necessary in some areas with very
dark signatures that were confused with tree shadows. These areas of confusion were all
wetlands or poorly drained areas with temporary standing water at the time the imagery


Table 5-5b. Sarasota and W. Charlotte county simulation statistics
Data type
No
acquisition
30% acquisition
by connectivity
30% acquisition
by area
70% acquisition
by connectivity
70% acquisition
by area
Maximum
acquisition
starting
population size
50
61
61
77
77
89
x end pop. size
26.3
31.3
33.6
43.9
45.7
46.9
s.d.
10.7
96
8.0
11.1
13.2
11.6
percent
decline
47.4
48.7
44.9
43.0
40.7
47.3
extinction
risk
0.03
0.0
0.0
00
0.0
0.0
quasi-extinction
risk (10 pairs)
0.10
0.07
0.0
0.0
0.02
0.0
K>
o
u>


23
Model results indicated that a population of jays with fewer than 10 breeding
pairs has about a 50% probability of extinction within 100 years, while a population with
100 pairs has a 2% to 3% probability of extinction in the same time period. These two
population sizes--10 and 100 pairsprovide convenient and biologically meaningful
values by which to classify subpopulations as islands (< 10 pairs), midlands (10-99
pairs), and mainlands (> 99 pairs). Although subjectively chosen, these values
effectively separate population sizes having fundamentally different levels of protection.
These values also receive empirical support from several long-term bird studies
(reviewed by Thomas 1990; Thomas et al. 1990; Boyce 1992).
Metapopulation Structure
We used a GIS buffering procedure (E.S.R.I. 1990) to generate dispersal-buffers
around groups of jays occurring within 3.5 km (for subpopulations) and 12 km (for
metapopulations) of each other (Fig. 2-6). We buffered jay territories rather than habitat
patches because we strongly suspect that dispersing Florida Scrub-Jays cue on the
presence of other, resident jays even more strongly than on habitat, so the functional
boundaries of occupied patches may be determined by where actual jay families exist.
We modified the resulting buffers in the following areas to reflect the presence of hard
barriers to dispersal in the form of open water with forested margins: Myakka River,
Peace River, St. Johns River, St. Lucie River, and Indian River Lagoon.
Using a 3.5 km dispersal-buffer we delineated 191 separate Florida Scrub-Jay
subpopulations (Fig. 2-6). Over 80% (N=152 islands) are smaller than 10 pairs (Fig. 2-
7), and 70 of these consist of only a single pair or family of jays. Only 6 subpopulations
contain at least 100 pairs (mainlands), leaving 32 midlands (10-99 pairs).



PAGE 1

METAPOPULATION DYNAMICS AND LANDSCAPE ECOLOGY OF THE FLORIDA SCRUB-JAY, APHELOCOMA COERULESCENS By BRADLEY M. STITH A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 1999

PAGE 2

Dedicated to my wife and best friend, Ellen Mary Thorns

PAGE 3

ACKNOWLEDGMENTS I wish to acknowledge each of my committee members for their unique contributions to this dissertation. My chairman, Dr. Stephen R. Humphrey, stimulated my interests in issues relating to philosophy, policy, and management. His broad interests and ability to take on multiple careers were remarkable. In my times of need he provided unfailing support while allowing me great freedom to pursue my own course of study. As director of Archbold Biological Station, co-chairman Dr. John W. Fitzpatrick provided me with a wonderful opportunity to study the Florida Scrub-Jay at this world-class research facility. His ability to perform brilliantly as director, researcher, and conservationist is inspirational. I am grateful to him for sending a generous U.S.F.W.S. grant my way to support me towards the end of my program. I was also privileged to work under the tutelage of the legendary pioneer of Florida Scrub-Jay research, Dr. Glen E. Woolfenden, who showed me the world from a Florida Scrub-Jay's perspective. My collaborative research with Dr. Lyn Branch on the Vizcacha and the Florida scrub lizard, was a joy. She forced me to think about landscape ecology and the metapopulation dynamics of "lesser" organisms. Her support, financial and otherwise, was extraordinary, and a nicer person I have never met. Dr. Jon Allen exposed me to population modeling from an entomologists perspective. His enthusiasm for teaching and helping graduate students solve technical problems was exceptional. The staff at Archbold Biological Station made my stay there rewarding beyond words. Working with Dr. Reed Bowman at the Avon Park Bombing Range, and learning iii

PAGE 4

about his suburban Florida ScrubJays, was most stimulating. His invitation to participate in the Brevard county Habitat Conservation Plan scientific committee was most appreciated. Special thanks go to Steve Friedman and Roberta Pickert for their help with GIS problems. Dan Childs, former manager of the affiliated MacArthur Agricultural Research Center, and his staff provided unfailing assistance in many ways. Current Archbold director Dr. Hilary Swain provided generous travel support for the 1997 AOU meeting. For collecting the bulk of the demographic data used in chapter 3, 1 am indebted to Reed Bowman, Doug Stotz, Larry Riopelle, Natalie Hamel, and Mike McMillan. I thank the staff at the Natural Resources Office at the Avon Park Air Force Range, especially Bob Progulske, Paul Ebersbach, and Pat Walsh for their support. Dave McDonald was instrumental in recruiting me to work on the statewide jay survey and exposing me to the opportunities at Archbold. Dr. Keith Tarvin and Dr. Curt Atkinson kindly lent me radiotelemetry equipment and provided much valuable advice. Special thanks go to Dr. Ron Mumme who kindly allowed me to use color banded jays from the south tract of Archbold which he has been monitoring for many years. His prior work in banding, taming, and sexing these jays was critical to identifying candidates for the displacement experiment. Bill Pranty worked tirelessly with me to digitize major portions of the statewide survey. Steve Schoech provided occasional field support and many hours of entertainment on the Archbold tennis court. For stimulating discussions about jays and landscape rules, I am indebted to many colleagues engaged in the study of Florida Scrub-Jays and their conservation. Participants in the Habitat Conservation Planning group whom I have not yet mentioned include iv

PAGE 5

David Breininger, Grace Iverson, Michael O'Connell, Parks Small, Jon Thaxton, and Brian Toland. The following people contributed data for the statewide survey: Reed Bowman, Dave Breininger, Jack Dozier, Florida Game and Fresh Water Fish Commission personnel, Grace Iverson, David McDonald, Ron Mumme, Ocala National Forest personnel, Bill Pranty, Hilary Swain, Jon Thaxton, and Brian Toland. I thank David Wesley and Dawn Zattau, of the U.S. Fish and Wildlife Service, who provided lead funding for the statewide survey and helped stimulate our discussions of habitat conservation planning. For special assistance with site-specific questions related to the maps in chapter 5: Mary Barnwell, Jim Beever, Reed Bowman, Dave Breininger, Mike Eagen, Mary Huffman, Grace Iverson, Mike Jennings, Laura Lowry, Dan Pearson, Gary Popotnik, Bill Pranty, Park Smalls, Hank Smith, Jon Thaxton, Brian Toland, Jane Tutin. For assistance with GIS data: Reed Bowman, Dave Breininger, Kathy Bronson, Beth Needham, Bill Pranty, Roberta Pickert, Katy NeSmith. For advice regarding modeling: Reed Bowman, Dave Breininger, John Fitzpatrick, Glen Woolfenden. Bill Pranty deserves special thanks for the exceptionally detailed information he has collated on scrub-jay locations around the state (Pranty et al. manuscript). Input from members of the Recovery Team was most helpful. Thanks go to the U.S.F.W.S. for funding the research in chapter 5, and especially to Dawn Zattau for her support and patience. I thank the support staff at the University of Florida Department of Wildlife Ecology and Conservation. Joe Gasper provided extraordinary computer support. Leonard Pearlstein and the entire USFWS coop unit helped with innumerable computer problems. I thank support staff at Circa computing, especially Jiannong Xin for v

PAGE 6

programming resources, and John Dixon for statistical help. Dr. Ken Portier also provided valuable statistical advice. Last, but no least, I thank my family and especially my wife, Ellen Thorns. vi

PAGE 7

TABLE OF CONTENTS page ACKNOWLEDGMENTS iii LIST OF TABLES xii LIST OF FIGURES xv ABSTRACT CHAPTERS 1 INTRODUCTION ! Historical Background 2 Biological Background 3 Objectives 4 2 CLASSIFYING FLORIDA SCRUBJAY METAPOPULATIONS 9 Introduction g Statewide Survey of the Florida ScrubJay 1 1 Statewide Survey: Methods Statewide Survey: Results 13 A Method for Classifying Metapopulations 14 Metapopulation Structure of the Florida ScrubJay " 18 Dispersal Distances 19 Patch Occupancy 20 Population Viability Analysis 22 Metapopulation Structure 23 Caveats 25 3 REMOTE SENSING OF FLORIDA SCRUB-JAY HABITAT 40 Introduction 4 q Methods 42 Image Source 42 Image Scanning and Conversion 43 Image Rectification and Mosaicing 43 vii

PAGE 8

Image Classification 44 Manual Editing of Classification 45 Assessment of Classification Accuracy 46 Digitization of Territories and Background Features 47 Tree Cover Buffering Procedure 47 Habitat Quality Model 4g Collection of Demographic Data 49 Habitat-Demographic Analysis 49 Results " 50 Discussion 54 4 MODELING DISPERSAL IN THE FLORIDA SCRUB-JAY 86 Introduction g^ Dispersal Strategies g9 Dispersal Traits of the Florida and Western ScrubJay ' 90 Methods " ^ General Approach 92 GIS Files 92 Simulating Philopatric Dispersal 93 Short distance dispersal algorithm development and calibration 94 Simulating Long Distance Dispersal 95 Estimating floater mortality and mobility 96 Habitat attractiveness 9-7 Floater detection radius 97 Estimating floater frequency 9g Floater algorithm development and calibration 102 J ay displacement experiment 1 03 Constraint Analysis 1 06 Model Validation I0g Results Radiotelemetry Displacement Experiment \ 08 Floater Parameter Estimation ] 2 j Constraint Analysis 112 Calibration and Validation l\2 Discussion j j ^ 5 METAPOPULATION VIABILITY ANALYSIS OF THE FLORIDA SCRUBJAY 132 Introduction and Objectives j 32 Methods j„ Simulation Model Description 135 Life Stages 136 Starting Population Stage Structure ZZ....... 137 Annual Life Cycle " ' 137 Territories , , A 140 viii

PAGE 9

Background Landscape Image 141 Map Production 143 Statewide metapopulation map 143 1992-1993 SMP maps ZZZZZZZZ 143 Acquisition maps ..144 GIS Database Preparation 146 Estimation of jay populations after restoration 146 Identification of protected areas 147 Assessment of unprotected areas 147 Suburban jays 14g Simulation runs I49 Repetitions and duration of simulations 149 Reserve design configurations 150 Output statistics 151 Model Validation/Calibration 1 52 Interpreting Simulation Results 152 Results Levy (Cedar Key) (Ml) ZZZZZZZZ 155 Citrus-S.W. Marion (M2) ZZZZZZ! " 164 Pasco-Hernando (M3) 1 76 Manatee-S. Hillsborough (M4) ' 1 g6 SarasotaW. Charlotte (M5) 196 N. W. Charlotte (M6) ZZZZZZZZZZ! 204 Central Charlotte (M7) 212 Lee and N. Collier (M8) 220 Flagler-N.E. Volusia (M9) ZZZZZZZZZZZZZZ 228 Merritt Island-S.E. Volusia and (M 1 0) 236 N. Brevard (Ml 1) ZZZZZZZ 244 Central Brevard (M 1 2) 253 S. Brevard-Indian River-N. St. Lucie (M13) ZZ! Z 261 St. Lucie -N. Martin (Ml 4) 270 Martin and N. Palm Beach (Ml 5) 278 South Palm Beach (Ml 6) ZZZZZZ ZZZZ! 286 Ocala National Forest (M 1 7) 294 N.E. Lake (Ml 8) ZZZZ 302 S.W. Volusia (Ml 9) ZZZZZZZZZ 310 Central Lake (M20) I!!!!!!!!!!!! 318 Lake Wales Ridge (M21) ZZZZZZZZZZZZZZZ" 326 Other Metapopulations 344 Brevard barrier island 344 Clay county 344 Osceola "" 344 Western Polk ^45 Bright Hour Ranch 345 Recommendations 346 Ranking Metapopulation Vulnerability 346 ix

PAGE 10

Summary of Recommendations 349 Discussion ..359 6 SYNTHESIS 365 REFERENCES 372 BIOGRAPHICAL SKETCH 383 x

PAGE 11

LIST OF TABLES MS ^ 21. Summary information for 42 Florida Scrub-Jay metapopulations, including type of metapopulation, number of territories (pairs of jays), and number of subpopulations 28 31. Demographic and habitat parameters for North and South Sandy Hill (1994 " 19 *5) 81 3-2. Kolmogorov-Smirnov test for normality of demographic and habitat variables (* significantly different from normal) 84 33. Mann-Whitney U test for differences in demographic and habitat variables between North and South jay populations (* significantly) 85 41. Landcover types from statewide habitat map (Kautz et al. 1993) used in simulations and associated floater attractiveness values 120 4-3. Summary of philopatric dispersal rules showing sex differences and rules used to implement the algorithm 122 A-A. Summary of jay movement data obtained from displacement experiment (distances in km) 123 45. Summary of constraint analysis for 9 simulation scenarios (50 years x 30 repetitions) showing number of colonizations from Lake Wales Ridge to Bright Hour Ranch, DeSoto county, Florida 124 51. Demographic and dispersal parameter settings for jays in optimal and suburban conditions 142 5la. Levy county patch statistics (number of jay territories for different configurations) j 5lb. Levy county (Cedar Key) simulation statistics 163 5-2a. Citrus and S. Marion county patch statistics (number of jay territories for different configurations) 1 7 2 xi

PAGE 12

5-2b. Citrus and S. Marion county simulation statistics 175 5-3a. Pasco and Hernando county patch statistics (number of jay territories for different configurations) Ig2 5-3b. Pasco county simulation statistics 1 35 5-4a. Manatee and S. Hillsborough county patch statistics (number of jay territories for different configurations) 191 5-4b. Manatee and S. Hillsborough county simulation statistics 195 5-5a. Sarasota and W. Charlotte county patch statistics (number of jay territories for different configurations) 200 5-5b. Sarasota and W. Charlotte county simulation statistics 203 5-6a. N. W. Charlotte county patch statistics (number of jay territories for different configurations) 208 5-6b. N. W. Charlotte county simulation statistics 21 1 5-7a. Central Charlotte county patch statistics (number of jay territories for different configurations) 216 5-7b. Central Charlotte county simulation statistics 219 5-8a. Lee and N. Collier county patch statistics (number of jay territories for different configurations) 224 5-8b. Lee and N. Collier county simulation statistics 227 5-9a. Flagler and N.E. Volusia county patch statistics (number of jay territories for different configurations) 232 5-9b. Flagler and N.E. Volusia county simulation statistics 235 510a. S.E. Volusia and Merritt Island county patch statistics (number of jay territories for different configurations) 240 5-10b. S.E. Volusia and Merritt Island county simulation statistics 243 5-1 la. N. Brevard county patch statistics (number of jay territories for different configurations) 249 51 1 b. N. Brevard county simulation statistics 252 xii

PAGE 13

5-12a. Central Brevard county patch statistics (number of jay territories for different configurations) 257 512b. Central Brevard county simulation statistics 260 5-13a. S. Brevard-Indian River-N. St. Lucie county patch statistics (number of jay territories for different configurations) 266 5-1 3b. S. Brevard-Indian River-N. St. Lucie county simulation statistics 269 514a. St. Lucie N. Martin county patch statistics (number of jay territories for different configurations) 274 514b. St. Lucie county simulation statistics 277 515a. Martin and N. Palm Beach county patch statistics (number of jay territories for different configurations) 282 5-1 5b. Martin and N. Palm Beach county simulation statistics 285 5-1 6a. South Palm Beach county patch statistics (number of jay territories for different configurations) 290 516b. South Palm Beach county simulation statistics 293 517a. Ocala National Forest county patch statistics (number of jay territories for different configurations) 298 517b. Ocala National Forest county simulation statistics 301 518a. N.E. Lake county patch statistics (number of jay territories for different configurations) 30^ 51 8b. N.E. Lake county simulation statistics 309 519a. S.W. Volusia county patch statistics (number of jay territories for different configurations) 3 j 4 519b. S.W. Volusia county simulation statistics 317 5-20a. Central Lake county patch statistics (number of jay territories for different configurations) 322 5-20b. Central Lake county simulation statistics 325 xiii

PAGE 14

5-2 la. Lake Wales Ridge patch statistics (number of jay territories for different configurations) 339 5-2 lb. Lake Wales Ridge simulation statistics 343 5-22. Metapopulation viability statistics 359 5-23. Metapopulation vulnerability ranking "no acquisition" (sorted by decreasing quasi-extinction probability) 352 5-23a. Metapopulation vulnerability ranking "maximum acquisition" (sorted by increasing percent protection) 352 5-24. Percent protected ranking (sorted by increasing percent protection) 353 5-25. Metapopulation priority ranking (sorted by decreasing priority) 354 5-26. Summary of recommendations (highest priority first) 355 xiv

PAGE 15

LIST OF FIGURES Bgure Qggg 2-1. 1993 distribution of Florida Scrub Jay groups (small black circles). Note the discontinuous distribution and variability in patterns of aggregation 30 2-2. Classification scheme showing different types of metapopulations based on patch size distribution (patches all small in size, mixture of small and large, and all large in size) along the horizontal axis, and degree of patch isolation (highly connected to highly isolated) on the vertical axis. Nonequilibrium, classical, mainland-island, and patchy classes are named according to Harrison (1991) 31 2-3. Schematic depiction of different kinds of metapopulations, illustrating use of dispersal-distance buffers to predict recolonization rates among subpopulations. Dotted lines separate functional subpopulations, based on frequency of dispersal beyond them. Solid lines separate metapopulations, based on poor likelihood of dispersal among them. A. Patchy metapopulation. B. Classical metapopulation. C. Nonequilibrium metapopulations. D. Mainland-island metapopulation 32 2-4. Dispersal frequency curve. Dispersal distances from natal to breeding territories for color-banded jays at Archbold Biological Station, 1970-1993. About 85% of documented dispersals were within 3.5 km, and 99% within 8.3 km. The longest documented dispersal was 35 km 33 2-5. Proportion of suitable habitat patches occupied by Florida ScrubJays as a function of their distance to the nearest separate patch of occupied habitat. Occupancy rates are high (nearly 90 %) for patches up to 2 km apart and decline monotonically to 12 km. Note the scale change after 16 km 34 2-6. Statewide jay distribution map with dispersal buffers. Shaded areas with thin, solid lines depict subpopulations of jays within easy dispersal distance (3.5 km) of one another. Thick lines delineate demographically independent metapopulations separated from each other by at least 12 km 35 2-7. Frequency of Florida Scrub-Jay metapopulation sizes. Note that 21 metapopulations have 10 pairs or less of jays. These represent nonequilibrium metapopulations 35 xv

PAGE 16

2-8. Examples of non-equilibrium metapopulations from North Gulf coast. Each of the six metapopulations contains fewer than 10 pairs of jays, except for the centrally located system that contains a single, midland-size subpopulation 37 2-9. Example of a "classical" metapopulation from five counties in Central Florida. Note the occurrence of jays in small islands of intermediate distance from one another 3g 210. Portion of the largest mainland-midland-island metapopulation in the interior, consisting of the Lake Wales Ridge and associated smaller sand deposits. The large central subpopulation (enclosed by the thin black line) contains nearly 800 pairs of jays. Small subpopulations to the south and east are within known dispersal distance of the large, central mainland. A small metapopulation to the west (in DeSoto County) contains a single subpopulation of 21 territories. This small system qualifies as a patchy metapopulation, since jays occur in two or more patches but the patches are so close together that they function as a single demographic unit 39 31. Map of Scrub-Jay Territories Spring 1994. Dividing line between "North" and "South" populations is the Kissimmee Rd 60 3-2. Habitat Suitability Index graphs for percent bare sand, percent tree cover, and distance to forest (modified from Duncan et al. 1995) 61 3-3. Illustrative map of 100, 200, and 400 m buffer zones around LOST territory 62 3-4. Accuracy assessment correlation graph for bare sand based on quadrat vs. classified image measurements (r-squared = 0.60) 63 3-5. Accuracy assessment correlation graph for tree cover based on transect vs. classified image measurements (r-squared * 0.25) 64 3-6. Measurements of sand, tree cover, and mixed vegetation for Spring 1994 territories on N. Sandy Hill 65 3-7. Measurements of sand, tree cover, and mixed vegetation for Spring 1994 territories on S. Sandy Hill 66 3-8. Percent tree cover (mean and standard deviation) for 4 zones (inside territories, 100, 200, 400 m buffer) North vs. South Sandy Hill South population shows significantly higher tree cover within all zones compared to North population 0/ 3 " 9 ' onf II! 66 ? °L er f ° r individual territories for 4 zones (inside territories, 1 00, 200, 400 m buffer) North territories 68 xvi

PAGE 17

3-10. Percent tree cover for individual territories for 4 zones (inside territories 100, 200, 400 m buffer) South territories ' 69 3-11. Tree cover within jay territories vs. total tree cover in North vs. South Sandy Hill. Note that jays select habitat with lower tree cover in both areas 70 3-12. Percent bare sand (mean and standard deviation) for 4 zones (inside territories, 100, 200, 400 m buffer) North vs. South Sandy Hill. Differences between two areas are not significant ,.71 3-13. Group size vs. percent tree cover within all territories (North and South populations pooled). Trend towards smaller group size with higher tree cover is not significant 72 3-14. Group size vs. percent tree cover within 100 m buffer for all territories (North and South populations pooled). Trend towards smaller group size with higher tree cover is not significant 73 3-15. Group size (small = 2, medium = 3, large = 47 jays) vs. percent tree cover within all territories (North and South populations pooled). Trend towards smaller group size with higher tree cover is not significant 74 3-16. Group size (small = 2, medium = 3, large = 4 7 jays) vs. percent tree cover within 100 m buffer (North and South populations pooled). Trend towards smaller group size with higher tree cover is not significant 75 3-17. Images and territories (black polygons) of North Sandy Hill. Right: colorinfrared image. Left: classified image (white = bare sand; green = trees; brown = shrubs/grass; black = water) 76 3-18. Images and territories (black polygons) of N. portion of South Sandy Hill Right: color-infrared image. Left: classified image (white = bare sand; green trees; brown = shrubs/grass; black = water) .... 77 3-19. Images and territories (black polygons) of S. portion of South Sandy Hill Right: color-infrared image. Left: classified image (white = bare sandgreen trees; brown = shrubs/grass; black = water) .... 78 3-20. Habitat quality map of N. portion of South Sandy Hill 79 321. Habitat quality map of S. portion of South Sandy Hill g0 41. Daily distances moved and number of days movements were tracked for 10 jays released at 3 sites in Highlands county, Florida 125 xvii

PAGE 18

4-2. Kaplan-meier survival curves of daily survival rates for 10 jays released at 3 sites in Highlands county, Florida. Upper curve: maximum possible survival; middle curve: "best guess" survival; lower curve: minimum possible survival 126 4-3. Distribution of daily distances moved by released jays (solid line), and inverse function fitted to observed movements (dashed line) 127 4-4. Comparison of dispersal data from Archbold Biological Station and simulated dispersal distances for male jays 128 4-5. Comparison of dispersal data from Archbold Biological Station and simulated dispersal distances for female jays 129 4-6. Comparison of stage-age data from Archbold Biological Station and simulated stage-age data for breeders 130 47. Comparison of stage-age data from Archbold Biological Station and simulated stage-age data for helpers 131 50. Delineations of 21 Florida ScrubJay metapopulations based on 1992 1993 statewide survey 145 5la. Levy county maps 1992 1993 jay and habitat distribution 158 5lb. Levy county acquisition map 159 5-lc. Levy county trajectory graphs. Top) no acquisition, Bottom) maximum acquisition ^\ 5Id. Levy county quasi-extinction graphs. Top) no acquisition, Bottom) maximum acquisition 1 $y 5-2a. Citrus county map 1992-1993 jay and habitat distribution 168 5-2d. S. W. Marion county acquisition map 1 7 j 5-2e. Citrus and S. Marion county trajectory graphs. Top) no acquisition, Bottom) maximum acquisition 173 5-2f. Citrus and S. Marion county quasi-extinction graphs. Top) no acquisition, Bottom) maximum acquisition 174 5-3a. W. Pasco and Hernando county map 1992 1993 jay and habitat distribution , _„ xviii

PAGE 19

5-3b. E. Pasco and Hernando county map 1992 1993 jay and habitat distribution ..179 5-3c. W. Pasco and Hernando county acquisition map 1 80 5-3d. E. Pasco and Hernando county acquisition map 181 5-3e. Pasco and Hernando county trajectory graphs. Top) no acquisition, Bottom) maximum acquisition Ig3 5-3f. Pasco and Hernando county quasi-extinction graphs. Top) no acquisition, Bottom) maximum acquisition 184 5-4a. Manatee and S. Hillsborough county map 1992 1993 jay and habitat distribution j g^ 5-4b. Manatee and S. Hillsborough county acquisition map 190 5