Citation
Hurricane Impacts on Coastal Dunes and Spatial Distribution of Santa Rosa Beach Mice (Peromyscus polionotus leucocephalus) in Dune Habitats

Material Information

Title:
Hurricane Impacts on Coastal Dunes and Spatial Distribution of Santa Rosa Beach Mice (Peromyscus polionotus leucocephalus) in Dune Habitats
Creator:
PRIES, ALEXANDER JAMES
Copyright Date:
2008

Subjects

Subjects / Keywords:
Beaches ( jstor )
Dunes ( jstor )
Habitat conservation ( jstor )
Habitat loss ( jstor )
Hurricanes ( jstor )
Mice ( jstor )
Scrub vegetation ( jstor )
Storm surges ( jstor )
Storms ( jstor )
Vegetation ( jstor )
Santa Rosa Island ( local )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Alexander James Pries. Permission granted to University of Florida to digitize and display this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Embargo Date:
7/24/2006
Resource Identifier:
496181279 ( OCLC )

Downloads

This item has the following downloads:


Full Text











HURRICANE IMPACTS ON COASTAL DUNES AND SPATIAL DISTRIBUTION
OF SANTA ROSA BEACH MICE (Peromyscus polionotus leucocephalus) IN DUNE
HABITATS













By

ALEXANDER JAMES PRIES


A THESIS PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA INT PARTIAL FUFILLMENT OF THE
REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE

UNIVERSITY OF FLORIDA


2006

































Copyright 2006

by

Alexander James Pries
















ACKNOWLEDGMENTS

I would like to acknowledge the support received from numerous organizations,

groups, and people during the development of this project. Funding and logistic support

was provided by the National Park Service and Eglin Air Force Base. The University of

Florida, Milton campus and UF Department of Wildlife Ecology and Conservation

provided equipment and additional funding.

My committee members (Dr. Lyn C. Branch, Dr. Debbie L. Miller, and Dr.

George W. Tanner) went above and beyond the call of duty in helping with development

and implementation of this research. Dr. Miller provided housing after my trailer broke

and continued to graciously welcome me into her home during the endless process of data

collection. I have enjoyed thoroughly our conversations on restoration ecology and

ethics of being a scientist. Dr. Tanner was most insightful in steering me towards proper

techniques for assessment of vegetation and other habitat features. Dr. Branch is a sound

editor who masterfully checked and rechecked my thesis for clarity and scope.

Thanks go to Riley Hoggard at Gulf Islands National Seashore and Bruce

Hagedorn, Bob Miller, and Dennis Teague at Jackson Guard, Eglin AFB. These

individuals served as important contacts and sources of support when materials were

lacking or I had questions about the accessibility of certain sites. I am most thankful for

their support. Many individuals assisted me during various phases of this project and I

will attempt to name them all here. I apologize for those who I may miss but your deeds

are not forgotten. Tanya Alvarez endured the hot conditions and was covered in black









carbon powder for weeks. Mica Schneider and Lisa Yager created the initial cover layer

of dunes (with over 750 polygons) on Santa Rosa Island that was massively important.

Jonathan Shore and Cathy Hardin were fantastic as temporary field technicians during the

difficult conditions. I also offer appreciation to Bob Schooley and Arpat Ozgul for

statistical assistance during initial data analysis. Conversations with many graduate

students including Jason Martin, Elizabeth Swiman, Ann George, Dan Thornton, and

Traci Darnell helped to craft and refine appropriate research questions.

Finally, I am thankful to my immediate family and friends outside of the scientific

or graduate school community. Your ability to listen and be a sounding board when

things became difficult is a maj or reason why I have been able to complete this research.

Your support and patience are massively important to me and I dedicate this work to you.




















TABLE OF CONTENTS


IM Le

ACKNOWLEDGMENT S ............ ..... .__ .............. iii...


LIST OF TABLES ............ ..... ._ ..............vii...


LIST OF FIGURES .............. .................... ix


AB S TRAC T ..... ._ ................. ............_........x


CHAPTER


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


Beach Mice and Threats to Survival ................. ...............1...............
Habitat Use by Beach M ice ................. .......... ...............1 .....
Coastal Dunes, Development, and Erosion .............. ...............2.....
Dune Restoration and Protection for Beach Mice .............. ...............3.....


2 INFLUENCE OF DUNE STRUCTURE ON STORM-RELATED EROSION
FOR FOREDUNES AND SECONDARY DUNES ON SANTA ROSA ISLAND,
FLORIDA. .............. ...............5.....


Introducti on ............ ..... ._ ...............5....
M ethods .............. ...............7.....

Study Area .............. ......__ ...............7....
Characteristics of Hurricane Ivan ....._ .....___ .........__ ...........8
Dune M apping ............ ...... ...............9...
Statistical Analyses............... ...............10
Re sults........................._ .....__ .............1
Conditions before Hurricane Ivan ............. .. ...._ .......__ ............1
Hurricane Ivan' s Impact on Foredunes and Secondary Dunes. ........................12
Regression Trees .............. ...............13....
Discussion ............ ..... .._ ...............15...


3 INFLUENCE OF HABITAT AND LANDSCAPE FEATURES ON SPATIAL
DISTRIBUTION OF SANTA ROSA BEACH MICE INT TWO DUNE
HABITATS BEFORE AND AFTER A HIURRICANE.............__ .........___.......24


Introducti on ................. ...............24........_ .....











M ethods ............ ............. .... ...............2
Study Area and Habitat Mapping ................. ...............26........... ...
Dune Occupancy .................. ........... ... .... ............2
Predictor Variables: Vegetation Cover and Landscape Structure .......................29
Occupancy M odels .............. ...............30....
R e sults.........._.... .. ...._.__ ....... .._._.. .. ...... ... .............3
Hurricane Impacts on Habitat Availability at EAFB .............. .....................3
Dune Occupancy .............. ...............32....
Habitat M odels .............. ...............33....
Discussion ................. ...............34.................


4 CONCLUSIONS AND CONSERVATION IMPLICATIONS ................ ...............42


Dune Erosion and Loss of Beach Mouse Habitat ................. ......... ................42
Habitat Restoration for Beach Mice .............. ...............44....


APPENDIX

A DELINEATION OF DUNES INT THE FIELD ................. .............................46


B CORRELATION MATRICES FOR VARIABLES BY HABITAT ................... .......47

C COMPARISON OF FRONTAL DUNES AT EGLIN AIR FORCE BASE AND
GULF ISLANDS NATIONAL SEASHORE ................. ..............................49

D PREDICTORS OF CHANGE INT OCCUPANCY OF FRONTAL DUNES
AFTER HIURRICANE IVAN .............. ...............50....

E CORRELATION MATRIX FOR STRUCTURAL FEATURES OF FRONTAL
DUNES ON EGLIN AIR FORCE BASE ................. ...............51...............


F CORRELATION MATRIX FOR STRUCTURAL FEATURES OF
SECONDARY DUNES ON EGLIN AIR FORCE BASE ................. ................ ...52

LITERATURE CITED .............. ...............53....


BIOGRAPHICAL SKETCH .............. ...............59....

















LIST OF TABLES


Table pg

2-1 Means and standard errors for structural variables measured to explain dune
erosion in foredunes and secondary dunes on Santa Rosa Island. ...........................17

2-2 Statistics for evaluation of dune characteristics as predictors of dune loss as a
result of Hurricane Ivan ................. ...............19................

3-1 Means and standard errors for structural and vegetation variables measured for
modeling occupancy of frontal and scrub habitat by Santa Rosa beach mice on
Eglin Air Force Base (EAFB) and Gulf Islands National Seashore (GINS) on
Santa Rosa Island, FL............... ...............38...

3-2 AIC-based selection of site occupancy models of dune occupancy for Santa Rosa
beach mice in frontal and scrub dune habitat ................. .............................40

3-3 Relative importance (wsum), model-averaged parameter estimates, and
unconditional standard errors for variables used to model occupancy for beach
mice in frontal and scrub habitat before and after Hurricane Ivan. ................... .......41

B-1 Correlations for variables measured on 61 scrub dunes surveyed for beach mice
before Hurricane Ivan (Jun. 2004 Sep. 2004). ............. ...............47.....

B-2 Correlations for variables measured on 61 scrub dunes surveyed for beach mice
after Hurricane Ivan (Oct. 2004 Jan 2005) ................. .............................48

B-3 Correlations for variables measured on foredunes (Eglin Air Force Base, n = 11,
and Gulf Islands National Seashore, n = 15) surveyed for beach mice after
Hurricane Ivan. (Oct. 2004 Feb. 2005). ............. ...............48.....

C-1 Results of t-tests comparing vegetation, structure and landscape context for
frontal dunes on Eglin Air Force Base and Gulf Islands National Seashore
measured after Hurricane Ivan. ............. ...............49.....

D-1 Mean values, standard errors, and t-test results for habitat variables on frontal
dunes on EAFB that became unoccupied and for dunes that remained occupied
after Hurricane Ivan. ............. ...............50.....

E-1 Correlations for structural and landscape context variables measured on frontal
dunes (N = 93) on Santa Rosa Island prior to Hurricane Ivan. ............... ...............51










F-1 Correlations for structural and landscape context variables measured on
secondary dunes on Santa Rosa Island prior to Hurricane Ivan............_.._ .............52

















LIST OF FIGURES


Finure pg

2-1 Map of Santa Rosa Island, FL. The study area encompasses the section
between Navarre and Fort Walton Beach............... ...............20.

2-2 Cross validation relative error for regression trees for (a) foredunes and (b)
secondary dunes to explain dune loss from Hurricane Ivan in relation to
measured predictor variables ................. ...............21........... ....

2-3 Regression trees relating percentage of dune lost from Hurricane Ivan for (a)
foredunes (N = 93) and (b) secondary dunes (N = 52) to physical features of
dunes, spatial location of dunes with respect to where Hurricane Ivan made
landfall, and width of island ......................._ ...............22. ..

2-4 Regression trees relating percentage of dune lost from Hurricane Ivan for (a)
secondary dunes > 0.25 ha (N = 61) and (b) secondary dunes <0.25 ha (N = 34)
to dune features, spatial location, and island width ................. .......................23
















Abstract of Thesis Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Master of Science

HURRICANE IMPACTS ON COASTAL DUNES AND SPATIAL DISTRIBUTION
OF SANTA ROSA BEACH MICE (Peromyscus polionotus leucocephahts) IN DUNE
HABITATS

By

Alexander James Pries

May 2006

Chair: Lyn C. Branch
Cochair: Deborah L. Miller
Maj or Department: Wildlife Ecology and Conservation

I examined the impact of Hurricane Ivan on dune erosion and changes in spatial

distribution of beach mice (Peromyscus polionotus leucocephahts) in two dune habitats

on Santa Rosa Island, FL. Foredunes (i.e., frontal dunes) and secondary (i.e., scrub)

dunes were mapped and surveyed for the presence of beach mice before and after the

hurricane using polyvinyl chloride (PVC) tracking tubes. I also collected data on

physical structure, vegetation, and landscape context for each dune during these two time

periods. Regression trees were used to evaluate structural features of dunes that

explained patterns in dune erosion as a result of Hurricane Ivan for the two dune types. I

used site-occupancy models and an information-theoretic approach to evaluate predictors

of occupancy for beach mice in frontal and secondary dune habitats before and after the

hurricane. Hurricane Ivan removed 68.2% of frontal dune area surveyed for beach mice

(Chapter 3) and 76.8% of all frontal dune area mapped (Chapter 2). Secondary dunes









surveyed for beach mice only lost 14.8% of their total area (Chapter 3) and all secondary

dunes surveyed lost 19.3% of their area (Chapter 2). Dune erosion for frontal dunes was

related inversely to distance from where the eye of Hurricane Ivan passed over the island,

dune height, and dune width. Dune erosion for large secondary dunes was reduced when

dunes were located behind foredunes. Erosion of small secondary dunes increased with

distance of the dune from where the eye of the hurricane passed over the island. The

reason for this pattern is unknown, but it may be related to spatial distribution of storm

surge in the Santa Rosa Sound. Dune erosion decreased with increasing length and area

of small secondary dunes. Beach mice occupied 100% of frontal dunes before the

hurricane and 59% of these dunes after the hurricane. Occupancy of scrub habitat by

beach mice was not statistically different before and after the hurricane, but was higher

(~70% of sites) than frontal dune occupancy after the storm. Frontal dune occupancy

was influenced largely by percent cover of woody vegetation and distance to nearest

occupied dune. Probability of occupancy by beach mice in scrub habitat increased with

an increase in dune area and amount of dune habitat surrounding the dune within 200 m.

This study indicates that scrub habitat, which currently is not protected for beach mice, is

important for mice because of the stochastic and severe impacts of storms on frontal

habitat. With further removal of frontal dune habitat, scrub could become essential for

long-term persistence of beach mice. The study suggests that restoration programs for

frontal dunes set targets for the construction of dunes that are tall and wide. Dunes with

these two features likely will mitigate storm surge associated with hurricanes and protect

coastal dunes, mouse habitat, or infrastructure located further inland.















CHAPTER 1
INTTRODUCTION

Beach Mice and Threats to Survival

Beach mice (Peromyscus polionotus spp.) are a complex of eight subspecies of the

oldfield mouse, which occupy the coastal dunes of Alabama and Florida (Holler 1992).

Two subspecies along the Atlantic coast of Florida are federally protected and one is

extinct. Gulf Coast populations of beach mice comprise five subspecies, four of which

are listed as threatened or endangered (Potter 1985; Milio 1998). The remaining

subspecies along the Gulf coast, the Santa Rosa beach mouse (Peromyscus polionotus

leucocephahts), is not yet listed because its geographic range includes several federally

managed lands (Gore and Schaffer 1993). All populations of beach mice suffer from

severe habitat loss from coastal development and habitat disturbance from hurricanes

(Gore and Schaffer 1993; Swilling et al. 1998). Additional threats to beach mice include

predation by feral cats (Felis silvestris), competition with house mice (M~us muscuhts) in

dunes around coastal development, and low population levels in late summer when

hurricane activity is most prevalent (Humphrey and Barbour 1981; Oli et al. 2001).

Habitat Use by Beach Mice

Beach mice are considered habitat specialists on dune habitats with burrow

locations strongly correlated to these habitat types (Blair 1951; Humphrey and Barbour

1981). Foredunes or frontal dunes, located immediately adj acent to the Gulf of Mexico,

are believed to be optimal habitat for beach mice (USFWS 1987). Mice also occur in

secondary dunes (also known as scrub), which are located farther from the shoreline and










are characterized by increased dominance of woody vegetation. Although densities of

beach mice generally are highest in frontal habitat (Swilling et al. 1998), abundance of

beach mice in scrub can increase after hurricanes. Prior to Hurricane Opal, population

abundance of Alabama beach mice (P. p. amnmobates) on trapping grids in scrub habitat

was less than frontal dune habitat (Swilling et al. 1998). Four months after the storm,

abundance in scrub habitat was almost twice that of frontal dune habitat. Despite use of

scrub by beach mice, this habitat is not designated as critical habitat under USWFS

recovery plans for Gulf coast beach mouse populations (USFWS 1987). Additionally,

little is known about what features define suitable scrub habitat for beach mice or how

much of this habitat is utilized by mice.

Although hurricanes are a natural feature of coastal disturbance, their impacts work

in concert with coastal development and anthropogenic habitat loss to impact habitat

availability for beach mice (Holler 1992). Population models suggest that hurricane

impacts pose a significant threat to all subspecies of beach mice (Oli et al. 2001).

Hurricanes, in addition to directly destroying dunes, fragment remaining dunes and

potentially change features of dunes that make them suitable for beach mice.

Fragmentation of dunes from hurricanes may require beach mice to travel more

frequently between remaining dune patches, exposing them to predators. Habitat loss and

fragmentation also may reduce landscape connectivity for beach mice, limiting their

ability to recolonize frontal dunes after storm impacts or force them to utilize more

marginal habitats.

Coastal Dunes, Development, and Erosion

The coastal dunes that beach mice occupy are valued for their fauna and flora,

natural beauty, and ability to protect human-made infrastructure (Nordstrom et al. 2000;










Martinez et al. 2005). Despite this, many coastal dune ecosystems have been changed

irreversibly as a result of exploitation of natural resources and anthropogenic

development (Martinez et al. 2005). Increases in storm severity and frequency are

predicted and storm impacts will continue to alter dunes, reduce infrastructure protection

and disturb habitat for wildlife. As a result, interest has increased in the creation or

restoration of dunes that will withstand storm impacts.

Coastal dunes are formed from aeolian processes with dune development occurring

where sediment is trapped by existing vegetation (Hesp 2004; Psuty 2005). Foredunes,

located closest to the shoreline, are dynamic structures that are influenced greatly by the

flow of water, wind, and sediment with normal environmental fluctuations and periodic

storm events (Psuty 2005). Secondary dunes are created by sediment flows from existing

foredunes or they may be old foredunes. These dunes are no longer maintained by the

processes that drive foredune morphology.

Hurricanes alter dune ecosystems by burying native vegetation under centimeters

of deposited sand and also change the configuration or presence of dune structures

(Ehrenfeld 1990). Dune erosion occurs as a result of storm surge and waves repeatedly

narrowing the dune face to cause an eventual breach or when storm surge overwashes a

dune and pushes sediment landward (Hesp 2002; Judge et al. 2003). Dune erosion is

influenced by duration and intensity of a storm event; however, structural features of the

dune also alter the risk of erosion.

Dune Restoration and Protection for Beach Mice

Increasing the amount of protected beach mouse habitat generally is not an option

as most coastal dunes in Florida already are in public lands or have been developed (Bird

2002). Therefore, other approaches such as habitat restoration may be important for









long-term maintenance of beach mouse populations. Although habitat restoration is cited

as critically important for the recovery of beach mouse populations (USFWS 1987; Oli et

al. 2001), little work has been conducted to identify habitat and landscape features that

influence use of frontal or secondary dune habitat by beach mice. Restoration techniques

have been developed to promote regeneration of physical structure to dunes after storms

(Miller et al. 2001; 2003). These techniques can be used to create dunes with particular

structural features (e.g., tall and wide), but identification of structural features of dunes

that confer resistance against storm-related erosion is limited.

Restoration techniques that promote creation of dunes and dunes with key habitat

requirements of beach mice may aid in management of existing protected habitats for

these mice. My study contributes to this effort in the following ways:

*Examining the relationship between dune erosion as a result of Hurricane Ivan and

the physical structure of frontal and secondary dunes (Chapter 2)

*Assessing the impact of Hurricane Ivan on the overall occupancy of frontal and

secondary dune habitat by beach mice (Chapter 3)

*Identifying habitat variables at the patch and landscape scale that influence

occupancy of frontal and secondary dunes by beach mice (Chapter 3).

Chapter 2 and 3 are written as stand-alone papers for publication. Therefore, some

background material is repeated in each chapter.















CHAPTER 2
INFLUENCE OF DUNE STRUCTURE ON STORM-RELATED EROSION FOR
FOREDUNES AND SECONDARY DUNES ON SANTA ROSA ISLAND, FLORIDA

Introduction

Coastal dunes are valued for their aesthetic beauty and their ability to protect

human-made structures during storms (Nordstrom et al. 2000; Nordstrom and Mitteager

2001). Dunes absorb wave energy, block storm surge, and reduce damage to

infrastructure. Coastal dunes also are important wildlife habitat (Martinez et al. 2005).

Hurricanes and tropical storms have altered coastal dunes on barrier islands along the

northern portion of the Gulf of Mexico in the last decade (Stone et al. 2004). Increases in

the severity and frequency of tropical cyclones are predicted and will further modify dune

configuration, reduce infrastructure protection, and disturb wildlife habitat (Emmanuel

2005). As a consequence, creation and restoration of dunes has become an important

issue in coastal management strategies (Nordstrom et al. 2000). Strategies for dune

protection and restoration could benefit from information on physical and spatial factors

that influence storm impacts on dunes.

Impacts of storms on dune erosion are a function of storm characteristics and

structural features of dunes. Dune erosion occurs when storm surge and waves

repeatedly narrow a dune face, causing irregular slumping of sediment and an eventual

breach, or when overtopping by storm surge completely overwashes a dune and pushes

sediment landward (Hesp 2002; Judge et al. 2003). Although severity and length of a

storm influence dune erosion (Kriebel et al. 1997; Sallenger 2000), key structural features









of dunes (e.g., height, width) also provide protection against dune erosion (Judge et al.

2003). Laboratory research and numerical models of dune erosion are extensive

(Vellinga 1982), but few studies have evaluated importance of dune structure in storm-

related erosion in the field (but see Judge et al. 2003). Additionally, past evaluations of

dune erosion often have been limited to foredune structures (i.e., dunes nearest to the

high tide line).

Coastal foredunes are formed from aeolian processes with dune development

occurring where sediment is trapped by vegetation (Hesp 2004; Psuty 2005). Secondary

dunes generally are found landward of foredunes and develop from sediment originating

on foredunes or they may be relict foredunes that are no longer controlled by aeolian

processes (Hesp 2004). Foredunes are differentiated as either incipient or established.

Incipient foredunes are low-lying developing dunes associated with pioneer plant

communities. Established foredunes evolve from incipient dunes and are distinguished

by presence of an intermediate plant community, including woody species. These dunes

have greater height and width than incipient dunes (Hesp 2002). Although the location

and development of incipient dunes may change annually, development of large

established foredunes takes decades, and these dunes remain in a relatively fixed position

unless removed by storms or anthropogenic disturbance. Evolution and maintenance of

established foredunes is not determined solely by sediment flows but rather by a suite of

additional factors like vegetation density and the frequency of wave and wind forces

(Hesp 2004).

Established foredunes and secondary dunes, by way of their size, should provide

greater resistance to increased tide levels and storm events than incipient dunes.










However, storm surge and waves associated with hurricanes of category 3 or above on

the Saffir-Sampson scale can cause even large (> 3 m tall) established foredunes to return

to a more erosional form or to be destroyed (Hesp 2002). Effects of strong hurricanes on

secondary dunes are less well documented. Impact of storm surge on secondary dunes

may be less severe as these structures are no longer governed by sand exchange, storm

tides or wave activity associated with foredune development (Hesp 2004). Additionally,

as a result of their spatial location behind wave-absorbing foredunes, dune erosion from

storm events may be lower for secondary dunes.

I assessed dune erosion along a barrier island ecosystem in the Gulf of Mexico after

Hurricane Ivan. The objectives of this study were to examine impacts of Hurricane Ivan

on established foredune and secondary dunes and to evaluate structural features of dunes

as predictors of dune vulnerability for these two dune types. I also examined dune

erosion as a function of the landscape context of the dune, including island width at the

location of the dune, distance to neighboring dunes and distance of the dune from the

position where Hurricane Ivan passed over the island. Identification of structural features

that allow dunes to resist storm-related erosion and evaluation of landscape attributes that

influence erosion are important for future manipulation of coastal dunes in a restoration

context.

Methods

Study Area

The study was conducted on Santa Rosa Island, which is a barrier island

approximately 60 km long and 0.5 km wide, in the Gulf of Mexico. The study site is

located on property owned and managed by Eglin Air Force Base (30024' N, 81037' W).

This portion of the island is approximately 20 km long and includes the island' s entire









width (Fig. 2-1). This area contains several military structures and a paved road for

military traffic but otherwise is undeveloped.

A thorough description of Santa Rosa Island's geomorphology can be found in

Stone et al. 2004. Foredunes are found near the high tide line and, in the absence of

hurricane activity, can run continuously the length of the island. Prior to Hurricane Opal

(1995), mean dune height was 3.8 m (Stone et al. 2004). These dunes are dominated by

sea oats (Uniola paniculata),) cakile (Calkile spp.), beach morning glory (Ipomoea

imperati), and seashore elder (Iva imbricata) but various woody species can be present on

foredunes in the absence of frequent disturbance. Secondary dunes are located behind

foredunes on the bayside of the island. Woody species dominate these dunes, including

fal se rosemary (Ceratiola ericodes), woody goldenrod (Chrysoma pauciflosculosa),

scrubby oaks (Quercus geminata) and sand pine (Pinus clausa). Between these two types

of dunes is grassland dominated by maritime bluestem (Schizachrium maritimum) and

bitter panic grass (Pan2icum ama~rum), interspersed with densely vegetated ephemeral

wetlands.

Characteristics of Hurricane Ivan

Hurricane Ivan made landfall as a category 3 hurricane on 16 September 2004,

west of Gulf Shores, Alabama, and approximately 100 km west of our study site. Storm

surge from the hurricane was estimated at 3 4.5 m from Mobile, AL to Destin, FL

(Stewart 2005), which encompassed all of Santa Rosa Island. Ivan was the most

destructive hurricane to make landfall along the Gulf coast in 100 years with a maj ority

of damage resulting from wave action associated with unusually high storm surge

(Stewart 2005).









Dune Mapping

Established dunes (foredunes, N = 93, and secondary dunes, N = 484) were

delineated in the field after Hurricane Opal (1995). Because established dunes change

very slowly over time, except when they are impacted by storms, these data could be

used as a baseline for dune structure prior to Hurricane Ivan. Dunes were mapped again

after Hurricane Ivan (2004). Geographic location of dune perimeters were recorded with

a TRIMBLE GPS unit in UTMs (Universal Tranverse Mercator) and differentially

corrected for < 1 m accuracy. Dunes were included if they were greater than 1.0-m high

with woody vegetation or greater than 1.5-m high with grasses or other herbaceous

vegetation. Dunes were considered distinct if they were separated by more than 3.0 m of

sand. Dune height (m) was measured every 15 m along the long axis of each dune using

a telescoping pole. Dune perimeters were incorporated into ArcView 3.2 (ESRI 1996)

and the following variables were calculated: dune area (ha), dune width (perpendicular to

the shoreline), length (parallel to the shoreline), and distance of each dune from the

position where Hurricane Ivan made landfall. Coordinates for the position where

Hurricane Ivan made landfall were obtained from the National Oceanic and Atmospheric

Association (Stewart 2005). Aerial photographs taken in 1995 were overlaid on dune

location in ArcView 3.2 to calculate island width at each dune location. I also recorded

presence or absence of foredunes located seaward of secondary dunes before Hurricane

Ivan. Gap distance for each dune was calculated as the average of the distance between

the closest dunes located immediately to the west and east of the target dune.

After Hurricane Ivan all remaining foredunes (N = 26) were remapped or recorded

as completely destroyed (100% loss) if not found during remapping (N = 67). A random

subset of small secondary dunes (< 0.25 ha, N = 34) were remapped after Hurricane Ivan.









All large secondary dunes (> 0.25 ha, N = 61) were remapped. The percentage of each

foredune or secondary dune lost from Hurricane Ivan was calculated by subtracting the

dune's area after Hurricane Ivan from the post-Opal dune area and by dividing this value

by the post-Opal dune area.

Statistical Analyses

For statistical analysis, I used data from all foredunes and all secondary dunes

prior to Hurricane Ivan, and I used all foredunes and a subset of secondary dunes sampled

after the hurricane. Because all small secondary dunes were not remapped after

Hurricane Ivan, I determined the proportion of the landscape occupied by large dunes (>

0.25 ha) and small dunes (< 0.25 ha) prior to Hurricane Ivan. I used these proportions to

determine the sample size for large and small dunes in analyses. The total area of scrub

dunes prior to Hurricane Ivan was 131.55 ha with large dunes making up 109.99 ha

(83.6%) of this total. To maintain the proportional area of the two dune types, I used the

34 small dunes randomly chosen for remapping after Hurricane Ivan and I randomly

selected 18 large dunes from the larger pool we mapped.

I used Pearson correlation coefficients to examine relationships among structural

variables for dunes, spatial location, and island width for all frontal dunes and secondary

dunes. Variables were examined for normality prior to examining correlations between

variables. For frontal and secondary dunes, data on percentage of dune loss were

transformed using arcsine transformation, and dune area and dune height prior to

Hurricane Ivan and dune area after Hurricane Ivan were transformed using log-

transformation (Zar 1998). I used t-tests to examine differences in dune area and dune

height between dune types before Ivan. Univariate linear regression initially was used to

identify structural or spatial variables that were important predictors of the percentage of









dune erosion after Hurricane Ivan for foredunes and secondary dunes, and I used logistic

regression to evaluate the importance of presence of foredunes on dune erosion in

secondary dunes. Changes in dune area of frontal and secondary dunes with the impacts

of Hurricane Ivan were examined with paired t-tests. All univariate tests were

conducted in SPSS version 13.0 (SPSS Inc., 2004) and I rej ected null hypotheses of no

influence on dune loss when p < 0.05.

Traditional multiple regression techniques may not work well when variables do

not meet parametric assumptions or when relationships between variables are complex or

non-linear (Bourg et al. 2005). I wanted to simultaneously assess the influence of all

predictor variables on dune erosion from Hurricane Ivan and examine relationships

between physical features of dunes and spatial location on dune erosion. I used

classification trees to assess how multiple predictor variables explained the impacts of

Hurricane Ivan on dune structure. This non-parametric approach splits the dataset into

smaller groupings with relatively homogeneous values of response variables (Breiman et

al. 1984). An advantage of classification trees is that they are simple to create, provide

intuitive descriptions of complex relationships, and explain variance in a dataset in a

manner similar to multiple regression or analysis of variance procedures (De'ath and

Fabricius 2000).

I used the RPART package in R (R Development Core Team, 2003) to build and

evaluate classification trees. Trees for foredunes were constructed with the percentage of

dune lost as the response variable and the following predictor variables: dune area (ha),

dune width (m), dune height (m), island width (km), gap distance (m), distance from

where the eye of Hurricane Ivan made landfall (km). All data except distance from the










eye of the hurricane were from measurements made prior to the hurricane. Classification

trees for secondary dunes used the same response and predictor variables as trees for

foredunes, but also included presence or absence of a foredune before Ivan. I used a

cross-validation procedure to evaluate the rate of misclassification as a function of tree

size (e.g., number of groupings) to select trees that were not over-fit (Breiman et al.

1984).

Results

Conditions before Hurricane Ivan

Before Hurricane Ivan, small dunes (< 0.25 ha) were numerous comprising 80 of

the 93 (86.1%) foredunes and 403 of the 484 (83.3%) secondary dunes. Secondary dunes

were larger in area but not taller than foredunes prior to Hurricane Ivan (dune area t = -

2.265, df = 575, p = 0.03; dune height t = 1.020, df = 575, p > 0. 10; Table 2-1).

Foredune area was correlated highly with dune height (r = 0.63, p < 0.01), but was not

correlated with west-east location as might be expected because Hurricane Opal made

landfall west of the study site (r = 0.06, p > 0.5). Correlations only considering foredunes

0.25 ha or larger also indicated no relationship between dune area and position on the

island's landscape (r = -0.05, p > 0.5). Similarly, dune area for secondary dunes was

correlated with dune height (r = 0.65, p < 0.01) and not correlated with west-east location

on the landscape (r = 0.02, p = 0.69).

Hurricane Ivan's Impact on Foredunes and Secondary Dunes

Dune area for foredunes and secondary dunes was reduced significantly by

Hurricane Ivan (foredunes, t = 5.160, df = 92, p < 0.01; secondary dunes, t = 3.267, df

= 51, p < 0.01, Table 2-1). Hurricane Ivan's storm surge physically removed 76.8% of

the foredune area. Of the original 93 foredunes measured, 67 were destroyed completely.









The 34 small secondary dunes sampled lost 42.1% of their total area. The 61 large dunes

lost 14.8% of total area. Based on the proportion of the area occupied by small and large

dunes on the pre-Ivan landscape, the total estimated loss of secondary dune area with

Hurricane Ivan is 19.3%. Reduction in dune area to these dunes was significantly less

than to foredunes (t = -9.953, df = 143, p < 0.01).

Univariate analyses of the relationship between dune structure, dune location,

island width, and dune loss for foredunes indicated that all variables except gap distance

were related significantly to dune loss (Table 2-2). However, many of these variables

were highly correlated making these tests difficult to interpret (Appendix E). In contrast,

for secondary dunes only dune height, dune width, dune length, and the presence of a

foredune were related to dune loss. Distance from the eye of Hurricane Ivan was an

important predictor of dune loss in foredunes, but not in secondary dunes (Table 2-2).

Regression Trees

Cross validation indicated the smallest classification trees to fit data from

foredunes and secondary dunes without an increase in misclassification error rate each

had 5 branches (Fig. 2-2). Regression trees for foredunes and secondary dunes indicated

that a different set of structural features were linked to dune erosion for dunes on the

oceanfront and bayside of the island (Fig. 2-3).

Percent of dune lost in foredunes after Ivan was related to dune structure (e.g.,

height and width) and the distance from where the eye of Ivan made landfall (Fig. 2-3a).

The amount of variance (R2) in dune erosion explained by the classification tree was

78.9%. The regression tree for foredunes indicated that structural features influencing

dune erosion in foredunes changed with distance of the dune from the location where

Ivan passed over the island (Fig. 2-3a).










The regression tree for secondary dunes sampled to represent proportional area of

small and large dunes on the landscape indicated that dune erosion of secondary dunes

was related to structural features of the dune, their position on the landscape, and the

presence of foredunes (Fig. 2-3b). This tree explained 76.3% of the variance in dune

erosion for secondary dunes. The tree first divided dunes by the width of a dune. For

wider dunes, the presence or absence of a foredune was important in determining dune

erosion. Dune erosion was lowest where foredunes were present. Where foredunes were

absent, dune erosion increased as distance from where Hurricane Ivan made landfall

increased. For narrow secondary dunes, island width was the only important factor

influencing dune erosion. Dune erosion was greater where the island was wide. Island

width and distance from where Hurricane Ivan made landfall are correlated (r = 0.46)

and, thus, may provide some of the same information (Appendix F).

When large (N = 61) and small (N = 34) secondary dunes were analyzed separately

with regression trees, results were easier to interpret. The presence of foredunes reduced

dune erosion for large secondary dunes and this was the only important variable (Fig. 2-

4a). However, this tree explained only 19.7% of the variance in dune erosion for large

secondary dunes. Erosion of small secondary dunes was lowest for dunes nearer to

where Hurricane Ivan made landfall, and no other variable appeared to be important in

predicting dune erosion for these dunes (Fig. 2-4b). Dune length and area were related

negatively to dune erosion for dunes at greater distances from where Hurricane Ivan

made landfall. This tree explained 76.6% of the variance in dune erosion for small

secondary dunes.









Discussion

My field study and mechanistic research in the laboratory (Vellinga 1982)

indicate that dune structure plays an important role in resistance of dunes to storm

damage. In addition, this study clearly demonstrates the influence of landscape context

of dunes on their vulnerability to dune erosion, including spatial location relative to a

hurricane's eye and presence of other dune structures. Identification of features that

promote resistance to storm-related erosion can aid agencies in the classification of

coastal areas that are especially vulnerable to future storm events. This information also

can assist in defining targets for coastal restoration.

Larger dunes on Santa Rosa Island experienced less erosion than smaller dunes

from Hurricane Ivan. However, the importance of location of the dune on the landscape,

and the specific structural features of dunes important in describing amount of erosion

were different for foredunes and secondary dunes, suggesting the processes that act upon

dunes during storms are different depending on distance from the shoreline. Much of the

erosion for foredunes probably is a result of storm surge. Foredunes that remained after

Hurricane Ivan showed signs of sediment slumping, dead or uprooted vegetation, and

blowouts; all of which are common effects of storm surge, wave action, and overwash.

Under these conditions height of a dune is likely to play a key role in resistance of dunes

to erosion, as demonstrated by the importance of this variable in our regression trees for

foredunes. Along exposed oceanfront beaches, the magnitude of storm surge and wave

action decreases with distance from the edge of the hurricane' s eyewall and damage to

foredunes follows a similar spatial pattern.

Secondary dunes on the island that experienced erosion lost sediment along dune

edges from passing storm surge and not from continual wave action. Dune erosion for









secondary dunes likely is influenced by storm surge from the Gulf of Mexico merging

with rising water levels in the Santa Rosa Sound, located behind the island. The presence

of foredunes substantially reduces erosion of secondary dunes. This observation

reinforces the importance of foredunes as buffers of storm surge for coastal features

located further landward and for protection of human-made structures.

One non-intuitive result of this research is that small secondary dunes that were

father from the eye of the hurricane and that were on the widest part of the island were

subject to more erosion that small secondary dunes nearer to the eye of Hurricane Ivan

and on the narrower part of the island. Storm damage on small secondary dunes

increased from west to east. The island widened from west to east, and the Santa Rosa

Sound narrowed as the island widened. When the flow of storm surge is confined and

water is shallow, high penetration distances have been observed for washover (Morton

and Sallenger 2003). I hypothesize that as the sound became narrower, the magnitude of

storm surge on the bayside of the island increased and smaller secondary dunes were

impacted more strongly, resulting in an inverse relationship between storm damage and

distance from the hurricane and an inverse relationship between storm damage and island

width.

Previous research on coastal dune systems has suggested that dune systems exist in

two opposing states: one where dune structure and vegetation communities are arranged

by environmental gradients generated from normal wind and wave activity and one

dominated by periodic but high levels of disturbance (e.g., hurricanes and tropical storms;

Synder and Boss 2002; Stallins and Parker 2003). I believe that coastal dunes on Santa

Rosa Island are beginning to shift towards the latter state, though assessment of dune









erosion after additional storms is needed to evaluate this statement. Prior to hurricane

Opal in 1995, the island's shoreline contained continuous foredunes (Stone et al. 2004).

Repeated hurricane activity has eroded or destroyed many foredune structures. My

analysis indicates that foredunes are important in protecting secondary dunes and thus,

further storm impacts may begin to affect secondary dunes more severely. The

consequences of this changing coastal landscape are large for maintenance of human-

developed infrastructure, success of restoration proj ects, and conservation of wildlife

species that depend on coastal dune habitat.







18


Table 2-1. Means and standard errors for structural variables measured to explain dune
erosion in foredunes and secondary dunes on Santa Rosa Island from
Hurricane Ivan. Differences between means before and after Hurricane Ivan
were compared using paired-t tests adjusted with Bonferroni's correction for
multiple tests ** p < 0.05.
Variable Foredunes Secondary Dunes'
Before Ivan After Ivan Before Ivan After Ivan
(N = 9 3) (N = 9 3) (N = 5 2) (N = 5 2)
Dune area (ha) 0.14 (0.03)** 0.05 (0.01) 0.37 (0.09)** 0.26 (0.07)
Dune height (m) 2.84 (0.15)** 0.89 (0.16) 3.28 (0.18)** 2.81 (0.15)
Dune length (m) 42.2 (5.8)** 12.1 (2.9) 80.6 (14.5)** 64.3 (11.3)
Dune width (m) 29.9 (2.6)** 10.7 (2.2) 56.6 (6.2)** 44.3 (6.4)
Gap distance (m) 95.6 (25.9) NA 50.8 (6.8) NA
Dunes were sub-sampled to represent proportional area of large (83.6%) and small (16.4%) dunes on the
landscape (N = 52 with 34 small dunes and 18 large dunes).










Table 2-2. Statistics for evaluation of dune characteristics as predictors of dune loss as a
result of Hurricane Ivan. Variables were assessed along a 20-km stretch of
Santa Rosa Island, FL prior to the hurricane. The significance of a variable's
role as a predictor of dune loss was evaluated using univariate linear
regression or a two sample t-test when examining the presence of absence of a
foredune before a secondary dune. Hypotheses of no association were rej ected
at p < 0.05.
Variable t p-value R2
Foredunes (N~ = 93)
Dune area (ha) 5.21 <0.01 0.21
Dune height (m) 5.43 <0.01 0.25
Dune length (m) -2.08 0.04 0.05
Dune width (m) -5.03 <0.01 0.22
Gap distance (m) 0.40 0.68 0.01
Island width (km) 3.08 <0.01 0.09
Distance from Ivan's eve (km) 6.09 <0.01 0.25

Secondary Dunesl (N = 52)
Dune area (ha) -1.65 0.10 0.05
Dune height (m) -2.39 0.02 0.10
Dune length (m) -2.35 0.02 0.10
Dune width (m) -3.50 0.01 0.20
Gap distance (m) 0.78 0.40 0.02
Island width (km) 1.05 0.30 0.02
Distance from Ivan's eve (km) -0.24 0.81 0.01
Presence / absence of foredune 2.47 0.02 NA
Dunes sub-sampled to represent proportional area of large (83.6%) and small (16.4%)
dunes on the landscape (N = 52 with 34 small dunes and 18 large dunes).








































Fig. 2-1. Map of Santa Rosa Island, FL. The study area encompasses the section
between Navarre and Fort Walton Beach.














1 2 3 4 5 7 9


size of tree


Ini 0.28 0.18 0.075 0.021 0.012 0.01
cp


size of tree
1234567891011


Ini 0.22 0.088 0.039 0.019 0.012



Fig 2-2. Cross validation relative error for regression trees for (a) foredunes and (b)
secondary dunes to explain dune loss from Hurricane Ivan in relation to
measured predictor variables. I used the 1-SE rule (Breiman et al. 1984) to
identify regression trees that had the smallest number of branches but were
closest to the overall minimum misclassification error (dotted line). Arrows
point to the best sized regression tree for each dune type. The complexity
parameter, cp, represents a balance between the complexity of a tree (i.e.,
more branches) and the costs of utilizing a simpler tree.












a)

Distance from Ivan Distance from Iv-an
> 114 km < 114 km




Dune height IDune height Dune width Dune width
123 m <23m 245S5m <45.5 m



Dune height Dune height
>41m < 41m
30.6% 91.4%m 1 98.9%
N= 11 N= 5 N=64




53.6%/ 801%
N= 6 N= 7


b)

Dune width Dune width
2 47.8 m <47.8 m





Presence of Absence of Island width Island width
foredmne foredmne < 062 kmn > 0.62 kmn





Distance from Distance from
Ian < 104.1n km Ivan 2 104.1 lan

113% 38.8%/ 64.6%
N=14 N=7 N=20





14.1% 46.99
N= 4 N= 7


Fig. 2-3. Regression trees relating percentage of dune lost from Hurricane Ivan for (a)
foredunes (N = 93) and (b) secondary dunes (N = 52) to physical features of
dunes, spatial location of dunes with respect to where Hurricane Ivan made
landfall, and width of island. Data for secondary dunes are based on sampling
of small (<0.25 ha) and large dunes (10.25 ha) according to their proportional
area on the landscape. Numbers at the ends of terminal nodes are the average
percentage of dune lost for all observations in that group. N is the number of
observations within that group.


























































I


Presence of
foredane


Absence of
foredine


19.4%
N= 44


36.6%
N=17


Distance from Ivan
< 104.8 lan


Distance from Ivan
2 104.8 lnn


Dune length
> 54.6 m


Dune length
< 54.6 m


21.9%/
N= 8


29.9%
N= 4


Dune area
10.07 ha










N= 9


Fig. 2-4. Regression trees relating percentage of dune lost from Hurricane Ivan for (a)
secondary dunes > 0.25 ha (N = 61) and (b) secondary dunes <0.25 ha (N =

34) to dune features, spatial location, and island width. Numbers at the ends
of terminal nodes are the average percentage of dune lost for all observations
in that group. N is the number of observations within that group.


Dune area
<0.07 ha









71.1%m
N= 13














CHAPTER 3
INFLUENCE OF HABITAT AND LANDSCAPE FEATURES ON SPATIAL
DISTRIBUTION OF SANTA ROSA BEACH MICE IN TWO DUNE HABITATS
BEFORE AND AFTER A HURRICANE

Introduction

Identification and protection of habitat are critical for species conservation and an

integral part of many conservation programs. GAP analysis, for example, utilizes

information on land cover and predicted species distributions to identify habitats that are

poorly represented in reserves (Flather et al. 1997). Habitat suitability index (HSI)

models establish relationships between a species' distribution and habitat variables to

create an index of suitable habitat that can be used to evaluate suitability of other areas

for habitat management or protection (UWFWS 1981). To aid in species recovery, the

Endangered Species Act provides for designation of critical habitat, which is the

geographic area that contains physical and biological features necessary for conservation

of the species (16 U.S.C. @ 153 1 et seq.). However, the effectiveness of critical habitat

designation is controversial as critical habitat often is defined using limited data or only

anecdotes, or found to be not determinable (Hoekstra et al. 2002; Taylor et al. 2005).

Habitat models are a common tool used to examine the role of habitat features in

explaining the spatial distribution, density, or diversity of species occurring on a

landscape and these models aid in the development of conservation strategies (Segurado

and Arauj o 2004; Guisan and Thuiller 2005). Incorporation of spatial structure of habitat

(e.g., size, shape and spatial distribution of habitat patches) along with traditional

assessments of habitat quality has improved the functionality of models (Cox and










Engstrom 2001). However, one problem with application of habitat models to

conservation problems is that their conclusions are based upon a limited range of

conditions because they generally are developed over short time periods (Pearce and

Ferrier 2000). Habitat availability and quality can shift rapidly with stochastic events

(VanHorne et al. 1997; Carlsson and Kindvall 2001), and key features that determine

spatial distribution or population density could change. Designation of protected

habitats often does not consider impacts of environmental stochasticity on distribution

and persistence of species even though disturbance may influence habitat turnover and

ultimately impact population persistence through impacts on habitat (Oli et al. 2001;

Jonzen et al. 2004; Frank 2005; Schrott et al. 2005). We demonstrate this issue with an

analysis of beach mouse habitat along the Gulf Coast of Florida before and after

Hurricane Ivan which made landfall in September 2004.

Coastal dunes are among the most dynamic and threatened habitats world-wide

(Martinez et al. 2005). Gulf Coast populations of beach mice comprise 5 subspecies, all

of which are subj ect to extreme stochastic events in the form of tropical storms and

hurricanes (e.g., Hurricanes Opal, 1995; Ivan, 2004; Dennis, 2005). Four of these

subspecies are listed as threatened or endangered (Potter 1985; Milio 1998). The

remaining subspecies, the Santa Rosa beach mouse (Peromyscus polionotus

leucocephalus), is not yet listed because its geographic range includes several federally

managed lands (Gore and Schaffer 1993). All subspecies suffer from severe habitat loss

from destruction of coastal sand dunes by development, and this habitat loss is

exacerbated greatly by hurricanes (Swilling et al. 1998). Optimal habitat for beach mice

is believed to be frontal dune habitat with sparse vegetative cover of sea oats (Uniola









paniculata) adjacent to the high tide line (USFWS 1987). Mice also occur in scrub

dunes, which are located farther from the beach and are characterized by increased

dominance of woody vegetation. These dunes provide refugia for mice during and

immediately after storms but are viewed as marginal habitat because of lower population

density in this habitat (Swilling et al. 1998). Three subspecies of beach mice on the Gulf

Coast are covered under the same federal recovery plan. This plan calls for protection of

dune communities within 152 m (500ft) of the high tide line and includes all frontal

dunes but excludes scrub habitat in most areas (Potter 1985; Swilling et al. 1998). The

St. Andrews beach mouse (P. p. peninsularis) is protected under its own recovery plan,

which states designation of critical habitat is not necessary for conservation of this

species (Milio 1998).

I examined impacts of Hurricane Ivan on the structure of frontal and scrub dunes,

compared occupancy patterns of beach mice in these two habitats, and determined how

these occupancy patterns changed after the hurricane. I also developed habitat models for

predicting dune occupancy by Santa Rosa beach mice and evaluated whether the factors

that influenced patterns of habitat occupancy were similar for frontal (optimal) and scrub

(marginal) habitats. I examined whether predictors of habitat occupancy changed after

the hurricane. My study demonstrates problems associated with narrowly defining

critical habitat as optimal habitat, particularly in systems characterized by high

stochasticity.

Methods

Study Area and Habitat Mapping

The study was conducted on Santa Rosa Island, a barrier island approximately 46-

km long and 0.5-km wide, located in the Gulf of Mexico near Fort Walton Beach, FL










(30024' N, 81037' W). My study area incorporated a 15-km section of the island on Eglin

Air Force Base (EAFB) and a 10-km section of the island on Gulf Island National Sea

Shore ginsS). Dune habitat was similar in these two areas and both sections contain a

single paved road and only a few structures. Frontal dunes were oriented parallel to high

tide line and were dominated by sea oats (Uniola paniculata),) cakile (Calkile spp.), beach

morning glory (Ipomoea imperati), and beach elder (Iva imbricate), and various woody

species in the absence of frequent disturbance. Scrub dunes were located on the bayside

of the island and woody species dominate scrub habitat, including false rosemary

(Ceratiola ericodes), woody goldenrod (Chrysoma pauciflosculosa), scrubby oaks

(Quercus geminate) and sand pine (Pinus clause). The area between frontal and scrub

dunes consisted of gently rolling grasslands interspersed with densely vegetated

wetlands.

EAFB dunes were mapped in the field before and after Hurricane Ivan by recording

their perimeters using a TRIMVBLE GPS unit and then differentially corrected for < 1 m

accuracy. GINS dunes were mapped only after the hurricane. Data were incorporated

into a cover layer in ArcView 3.2 (ESRI 1996).

Dune Occupancy

I surveyed for presence of beach mice in all frontal (N = 15) and scrub (N = 61)

dunes equal to or larger than 0.25 ha on EAFB before Hurricane Ivan (June September

2004) and after the hurricane (October 2004 December 2004). Frontal dunes (N = 15)

on GINS also were surveyed for beach mice after Hurricane Ivan (December 2004 -

February 2005). Presence of beach mice in each dune was determined with tracking

tubes that register footprints of mice that enter the tube. Tracking with tubes is less

dependent upon weather and less labor intensive than live trapping and, therefore,









particularly useful for large scale surveys of distribution (Mabee 1998; Glennon et al.

2002).

Tracking tubes were constructed with PVC pipe (33-cm long x 5-cm diameter) and

elevated 5-7 cm off the ground to prevent access by ghost crabs (Ocypode quadrata).

Dowels placed at either end of the tube allowed mice, but not crabs, to climb to the tube.

A paper liner was inserted into the bottom of each tube, and the tube was baited in the

middle with rolled oats. Felt inkpads located at each end of the paper liner were coated

with a 2: 1 mineral oil and carbon power solution (Mabee 1998). Hispid cotton rats

(Sigmodon hispidus) leave footprints that are substantially larger than footprints of Santa

Rosa beach mice. No other small rodents occur on undeveloped portions of the island

(Gore and Schaffer 1993).

Dunes less than 0.50 ha received eight tubes; dunes > 0.50 ha < 2.00 ha received

16 tubes; and dunes greater than > 2.00 ha received 32 tubes. Tracking tubes were placed

at 15-m intervals along transects that began and ended at the dune's boundary and ran

parallel to the long axis of the dune. The starting point for the first transect was selected

randomly and, when more than one transect was needed, parallel transects were

established 15 m apart. During each tracking session, tracking tubes remained in a dune

for five nights and were checked after each night.

For many species, probability of detection during presence/absence surveys is less

than one resulting in underestimates of occupancy, biased parameter estimates for habitat

models, and incorrect estimates of population persistence (Gu and Swihart 2004; Kery

2004). Therefore, I used recent statistical approaches for analysis of site occupancy that

build on traditional capture-recapture methods and used repeated censuses to calculate









detection probability (p) and to estimate the proportion of sites that are occupied ('P) after

accounting for detectability (MacKenzie et al. 2002). To estimate detection probability

within each habitat type, I re-sampled a random subset of scrub dunes (N = 30) with

tracking tubes three times after initial pre-storm surveys, and I resurveyed another

random subset of scrub dunes (N = 30) and all frontal dunes on EAFB three times after

initial post-storm surveys. Each repeat survey was conducted over 5 nights following the

sampling protocol described above.

Predictor Variables: Vegetation Cover and Landscape Structure

I measured vegetation cover and dune height on scrub dunes before and after

Hurricane Ivan. Surveys for these variables were not completed on frontal dunes before

the hurricane hit and, therefore, data on these variables were analyzed for frontal dunes

only post-hurricane. Vegetation cover was quantified using the line-intercept method

(Bonham 1989) along three 50-m transects placed 20 m apart and perpendicular to the

long axis of each dune. I recorded distances (cm) that sea oats, other herbaceous

vegetation, woody vegetation, and open sand occupied along each transect and divided

the distance for each cover class by total length of the transect to obtain percent cover for

each cover class. I averaged data for the three transects prior to analysis. Sea oats and

many herbaceous species are important food sources for beach mice (Moyers 1996).

Amount of open sand may be important for burrow construction and woody vegetation

may stabilize dunes during storms and provides food and cover for foraging.

I recorded dune height (m) by measuring height every 15 m along the long axis of

each dune using a telescoping pole and then averaged all values for each dune. Height of

a dune may influence perception of dune habitat by beach mice moving through the

landscape or influence the impact of storm surge on dunes. I calculated dune area and









amount of dune habitat surrounding each dune in ArcView 3.2 from the GIS database

created from field mapping of dunes. I also calculated the distance to the nearest

occupied dune as a measure of isolation. Habitat area may influence size of local

populations (Hanski 1994). We used the BUFFER function in ArcView 3.2 to estimate

the total area of dune habitat surrounding each dune at the foraging (200 m) and dispersal

(1 km) scales of beach mice (Bird 2002; Swilling & Wooten 2002). The east-west

coordinate (UTM) at the center of each dune was included in habitat models to examine

how spatial location relative to the eye of Hurricane Ivan influenced dune occupancy by

mice. The eye of the hurricane passed approximately 75 km west of the western end of

GINTS and 100 km west of the western end of EAFB.

Occupancy Models

I created and ranked a series of models with the program PRESENCE to identify

variables that influenced distribution of beach mice in frontal and scrub habitats

(MacKenzie et al. 2003). Correlations among variables were examined and correlated

variables ( r > 0.60) were not included in the same model (Welch and MacMahon

2005), or if correlated variables were used in the same model, a regression was conducted

for the two variables and residuals were included in the model as an independent measure

of one of the variables (Cooper & Walters 2002). Correlated variables requiring this

approach were pre-hurricane dune habitat within 1 km and pre-hurricane east-west

coordinate for scrub dunes (r = 0.69), post-hurricane dune habitat within km and post-

hurricane east-west coordinate for scrub dunes (r = 0.70), post-hurricane dune habitat

within 200 m and post-hurricane distance to nearest occupied dune for frontal dunes (r = -

0.62), and post-hurricane dune habitat within 1 km and post-hurricane east-west

coordinate for frontal dunes (r = 0.69).










Fifty-six candidate models were evaluated for scrub dunes using a combination of

variables measured before Hurricane Ivan and similar models were created with post-

hurricane data. The first eight base models included a combination of patch-level features

(e.g., dune area, % cover of woody vegetation, % cover of herbaceous vegetation, dune

height). An additional 24 models were created by adding distance to nearest occupied

dune, the 200-m habitat buffer or 1-km habitat buffer to the original base models.

Finally, I created another 24 models by including the east-west coordinate in the "base

model + landscape context" models.

I developed 40 candidate models for frontal dunes on EAFB and GINS after

Hurricane Ivan. All frontal dunes were occupied before Hurricane Ivan, so no model was

created for this period. To reduce risk of an over-parameterized model, I restricted the

total number of variables in a model to three. The first eight base models were the same

as in scrub habitat (i.e., patch-level features). An additional 32 models were created by

including distance to nearest occupied dune, 200-m habitat buffer, 1-km habitat buffer, or

spatial coordinate to base models. I also modeled post-hurricane occupancy of frontal

and scrub dunes on EAFB with pre-hurricane conditions to assess the role of pre-

hurricane conditions on post-hurricane occupancy. Models were created using the same

procedure as described above.

I used an Akaike Information Criterion (AICe) corrected for small sample bias to

select the best model and rank the remainder. I present AIC differences (Ai = AICci -

minimum AIC,), so that the best model has Ai = 0 (Burnham & Anderson 2002). Models

with Ai < 2 are considered competitive models. I also include Akaike weight (wi), which

indicates relative likelihood that model i is the best model. The relative importance of









each habitat variable (ws,;;,) was obtained by summing w, for all models that contained

this variable (Burnham & Anderson 2002). I performed model averaging to obtain

parameter estimates and unconditional standard errors for each habitat variable of interest

to reduce the bias of estimating parameter effects from a single model (Burnham and

Anderson 2002). When the confidence interval around a model-averaged parameter

estimate is > 0, an increase in the variable significantly increases the probability of

occupancy, and a value < 0 indicates that an increase in the variable decreases the

probability of occupancy (Buskirk 2005; Mazerolle et al. 2005). Estimated probability of

detection (p) and overall occupancy rate (WI) also were obtained using this approach.

Results

Hurricane Impacts on Habitat Availability at EAFB

Hurricane Ivan significantly reduced mean area of both types of dunes (Table 3-1),

but frontal dunes lost a much greater proportion of area. Storm damage resulted in a loss

of 68.2% of the total area of frontal dunes surveyed for beach mice, including complete

destruction of four dunes. No scrub dunes were destroyed entirely but the total area of

scrub dunes surveyed for beach mice was reduced by 14.8%. Dune height also was

reduced significantly for both dune types (Table 3-1). The amount of habitat within 200

m and 1km of a dune was reduced significantly for scrub dunes but not frontal dunes.

However, frontal dunes already had little habitat within 200 m prior to the hurricane

(Table 3-1).

Dune Occupancy

Beach mice were detected in 100% of frontal dunes and 72. 1% of scrub dunes prior

to the hurricane, and in 51.8% of frontal dunes and 73.8% of scrub dunes after the

hurricane. Probability of detection was high in all surveys. Site-occupancy models










suggest that before the hurricane 75.1 & 5.5% (model-averaged estimate & unconditional

SE) of scrub dunes were occupied, with a detection rate of 88.6 & 5.6%, and after the

hurricane 78.6 & 4.9% of sites were occupied, with a detection rate of 90. 1 & 3.1%.

Differences in occupancy of scrub dunes before and after the hurricane were not

significant (t = 0.5, df = 60, p > 0. 10). Occupancy in frontal dunes dropped to 59.7 &

5.1%, with a detection rate of 89.8 & 5.5% after the hurricane, and occupancy in frontal

dunes was significantly lower than occupancy of scrub dunes (t = 1.8, df = 42, p < 0.05).

Habitat Models

A combination of patch-level and landscape-level features ranked high in models of

occupancy of scrub dunes before and after the hurricane (Table 3-2 and 3-3). The

strongest model for scrub dunes before the hurricane included dune area, percent woody

vegetation cover, and amount of dune habitat within 200 m. No other models were

competitive. After the hurricane, the same model was the strongest; however, the Akaike

weight was much lower and several additional models were competitive (Table 3-3). All

models with Ai < 2 contained some combination of the variables in the best model except

dune height and total herbaceous cover were included in several models. Ranking of

variables based on the sum of their Akakie weights revealed that amount of dune habitat

within 200 m of scrub dunes was the most important variable in explaining probability of

occupancy of scrub dunes by mice before and after the hurricane (Table 3-3), followed

closely by dune area, and percent woody vegetation cover. Before and after the

hurricane, probability of beach mice occupying a scrub dune increased as amount of

habitat surrounding the dune within 200 m increased and dune area increased (Table 3-3).

Occupancy of scrub habitat by beach mice also appeared to increase with increasing

cover of woody vegetation before and after the hurricane, but this relationship was not









statistically significant. Top models of post-hurricane occupancy in scrub dunes using

pre-hurricane conditions retained the same suite of predictor variables (Table 3-2).

The strongest model for occupancy of frontal dunes after Hurricane Ivan included

percent woody vegetation cover and distance to nearest occupied dune (Table 3-2). In

contrast to scrub habitat, the amount of habitat surrounding a dune within 200 m was not

a factor in any competitive models. The likelihood of occupancy increased with

increasing cover of woody vegetation and this variable was the top ranked variable in

models of occupancy (Table 3-3). Increasing distance to the nearest occupied dune also

appeared to reduce the probability of occupancy after the hurricane but this relationship

was not statistically significant (Table 3-3). Dune height was the third ranked variable

and an increase in dune height appears to increase occupancy by beach mice in frontal

habitats but also was not statistically significant (Table 3-3). When post-hurricane

occupancy of frontal dunes on EAFB was modeled with variables related to the structural

and landscape context of dunes prior to the hurricane, dune height and distance to the

nearest occupied dune were the most important predictors of occupancy (Tables 3-2 and

3-3). The likelihood of occupancy of frontal dunes by beach mice after the hurricane

increased with a greater dune height prior to the hurricane, and a greater distance to the

nearest occupied dune prior to the hurricane decreased the likelihood of occupancy after

the hurricane (Table 3-3). Models for occupancy of frontal dunes, with pre and post-

hurricane habitat data, had a better fit than models of scrub dunes (Table 3-2).

Discussion

Frontal dunes near the high tide line are subj ected to major impacts during

hurricanes. Prior to Hurricane Opal (1995), frontal dunes ran relatively continuously

along the entire length of Santa Rosa Island (Stone et al. 2004). This hurricane and









subsequent tropical storms fragmented frontal dunes. Storm surge from Hurricane Ivan

removed close to 70% of the remaining frontal dunes. In contrast, no scrub dunes, which

are located on the bay side of the island, were completely lost with Hurricane Ivan and

reduction in area of scrub dunes occurred along dune edges from passing storm surge.

Distance from the eye of the hurricane influenced dune lost for frontal and scrub dunes

along this portion of Santa Rosa Island (Chapter 2). Tropical storms and hurricanes are

predicted to be increasing in number and severity (Emanuel 2005). Frontal habitat for

beach mice will continue to be fragmented and removed if the interval between

hurricanes and other tropical storms remains shorter than the time required for dunes to

develop. In contrast, my results suggest that the amount and configuration of scrub dunes

on this barrier island may remain relatively consistent. However, as buffering capacity

provided by frontal dunes is lost, scrub dunes may suffer more impacts. Also, Hurricane

Ivan was a category 3 hurricane; stronger hurricanes could have greater impacts.

Predictors of occupancy for beach mice in frontal habitat after the hurricane were

closely tied to local habitat features (e.g., percent cover of woody vegetation and dune

height) and proximity to other occupied dunes. Optimal beach mouse habitat generally is

described as tall frontal dunes vegetated by sea oats and other herbaceous plants (Holler

1992). My habitat model indicates that woody vegetation cover also is important to mice,

at least during hurricane cycles. Foraging experiments demonstrate that mice consume

more seeds under vegetation cover than in the open (Bird 2002). Woody plants provide

cover for foraging, serve as a food source for mice, and also may promote dune stability

during storms (Moyers 1996; Musila et al. 2001). Similarly, dune height may be an

important factor in dune stability, particularly in preventing overwash by storm surge.










Beach mice also are semi-fossorial and an increase in dune height may facilitate

conditions appropriate for burrow construction. For frontal dunes on EAFB where the

impact was severe, dune height prior to the hurricane was a significant predictor of post-

hurricane occupancy, but after the hurricane the importance of this variable was not as

clear.

Isolation explained post-hurricane occupancy of frontal habitat, whether modeled

with pre- or post-hurricane habitat conditions. This observation likely reflects the history

of disturbance and loss of frontal habitat on this island. Beach mice occupying frontal

dunes prior to Hurricane Opal experienced a fairly continuous habitat where habitat

quality might have been determined largely by resource availability or appropriate

burrow conditions. The current fragmented frontal dunes are too small to support

separate populations of beach mice, but rather may serve as resource patches for mice

moving among dunes.

The most important predictors of occupancy before and after the hurricane for

beach mice in scrub habitat were landscape features related to habitat amount (i.e., dune

area and amount of surrounding habitat). Predictors of occupancy in scrub habitat were

similar before and after the hurricane, presumably because the impact of the hurricane on

the structure of these dunes was minimal. Woody vegetation also may play a role in

occupancy of scrub dunes by beach mice, but this relationship is not as clear as in frontal

dunes. The amount of dune habitat surrounding scrub dunes is greater than for frontal

dunes, which may indicate less isolation for these dunes. The importance of surrounding

dune habitat for occupancy of scrub may reflect reduced habitat quality in scrub dunes.

Alabama beach mice travel further distances to forage in scrub habitat than in frontal









habitat during the winter and spring (Sneckenberger 2002). If habitat quality and

population density are lower in scrub dunes, larger areas may be required to maintain

mouse populations.

Dune restoration after hurricanes primarily has focused on re-establishment of sea

oats, which produces a lattice of rhizomes that accumulate sand and also is an important

food plant for beach mice. My results suggest that restoration programs for frontal dunes

also should include re-establishment of woody plants and promote increases in dune

height. Beach mice also should benefit from restoration programs that reduce isolation of

frontal dunes.

The results of my study suggest that optimal habitat for beach mice differs under

different environmental conditions. Lower occupancy of scrub habitat than frontal

habitat by beach mice prior to the hurricane, and documentation of lower density in scrub

habitat from other studies, suggest that scrub habitat could be lower quality than frontal

dunes under pre-hurricane conditions, though density is not always a good indicator of

habitat quality (Van Horne 1983). However, persistence of scrub habitat and maintenance

of occupancy levels by beach mice through the hurricane in this habitat versus the severe

loss of habitat and significant reduction in occupancy of frontal habitat suggest scrub is

an essential habitat. Scrub was an important refugia habitat for Alabama beach mouse

populations during Hurricane Opal and a source of dispersing individuals after the

hurricane (Swilling et al. 1998). Re-colonization of frontal habitats by beach mice after

Hurricane Opal occurred within nine months (Swilling et al. 1998). I observed beach

mouse tracks on previously unoccupied frontal dunes in March 2005, approximately six

months after Hurricane Ivan. These mice may have dispersed from scrub or neighboring










frontal dunes. Given the inevitable loss of frontal dunes with hurricanes, incorporation of

scrub habitat into conservation efforts for Gulf Coast beach mice is warranted to ensure

long-term population persistence. Scrub habitat, even as marginal habitat, will improve

population persistence and lessen extinction risk as frontal habitat is further removed.

The role of stochasticity and uncertainty in management outcomes has been

explored extensively with respect to impacts on population size and persistence of species

of economic or conservation concern (Ellner and Fieberg 2003). Our study demonstrates

the need to incorporate these factors in habitat planning and protection. Habitat

availability for species in dynamic landscapes can change quickly and additional habitats

may become critically important after stochastic events (Carlsson and Kindvall 2001;

Biedermann 2004). When the dynamics of landscape and population are not understood,

protection of habitat should follow a conservative approach. Failure to consider and

protect habitats required under different environmental conditions may exacerbate the

impacts of habitat loss and change on extinction risk.











Table 3-1. Means and standard errors for structural and vegetation variables measured
for modeling occupancy of frontal and scrub habitat by Santa Rosa beach
mice on Eglin Air Force Base (EAFB) and Gulf Islands National Seashore
(GINS) on Santa Rosa Island, FL. Variables were assessed before and after
Hurricane Ivan made landfall on 18 September 2005. Differences between
means before and after Ivan were compared for dunes on EAFB using paired
t-tests adjusted with Bonfferoni's correction for multiple tests. ** p < 0.05.


GINS after
hurricane
(N= 15)
0.15 (0.03)
3.24 (0.20)
0.24 (0.05)

1.94 (0.29)

126.2 (39.4)

3.3 (2.1)

19.2 (2.3)

506117 (1709)


Scrub Dunes
Mean (a SE)


Dune perimeter was correlated highly (p <0.01l)with dune area and was omitted from analyses.
2 Distance to nearest occupied dune dropped after the hurricane because the four dunes that were destroyed were very
isolated (mean distance to nearest occupied dune for those dunes = 459.5 m). Pre-hurricane mean for distance to
nearest occupied dune for the 11 dunes to survive Ivan = 131.7 m.


Frontal Dunes
Mean (a SE)
EAFB after
hurricane
(N= 1 )
0.26 (0.06) **
3.13 (0.35) **
0.22 (0.08)

8.67 (1.44)


Variables


EAFB before
hurricane
(N = 6 1)
1.82 (0.38)
4.64 (0.3)
2.21 (0.28)

12.73 (1.19)

176.1 (38.3)

19.6 (1.6)

14.1 (1.4)

524519 (653)


EAFB after
hurricane
(N= 61)
1.55 (0.37) **
3.32 (0.17 ) **
1.71 (0.19) **

11.29 (1.03) **


EAFB before
hurricane
(N= 15)
0.59 (0.09)
4.01 (0.33)
0.29 (.09)

8.45 (1.48)


Dune area (ha)
Dune height (m)
Dune habitat within
200 m (ha)
Dune habitat within
1 km (ha)
Distance to nearest
occupied dune (m)
Percent woody
cover
Percent total
herbaceous cover
East-West
Coordinate (UTM)
(m)


174.2 (37.6) 219.2 (54.7) 161.9 (25.9)2


19.9 (1.6)

7.4 (0.8) **

524519 (653)


no data

no data

522959 (1337)


6.5 (1.4)

24.5 (2.9)

524208 (879)











Table 3-2. AIC-based selection of site occupancy models of dune occupancy for Santa
Rosa beach mice in frontal and scrub dune habitat. K = the number of
explanatory variables plus 1, Ai = AICci -minimum AICci, w, = Akaike
weights. Models with Ai < 2 are presented.


Habitat and
conditions
Scmub pre-
hurricane


Period of
occupancy
Pre-hurricane


a, w, Rza
0.00 0.42 0.242


Location
EAFB /
N =61


Model
Dune area, habitat within 200
m, percent woody cover


Scmub post- Post-
hurricane hurricane


EAFB / Dune area, habitat within 200
N= 61 m, percent woody cover
Dune area, habitat within 200 m
Dune height, percent woody
cover, habitat within 200 m
Percent woody cover, habitat
within 200 m
Percent total herbaceous cover,
habitat within 200 m
Dune height, habitat within 200

Dune area, percent total
herbaceous cover, habitat within
200 m

EAFB / Dune area, habitat within 200
N =61 m, percent woody cover
Dune habitat within 200 m
Percent woody cover, dune
habitat within 200 m
Dune area, dune habitat within
200 m
Dune height, dune habitat
within 200 m, percent woody
cover
Dune area, percent total
herbaceous cover, dune habitat
within 200 m


0.18

0.16

0.14


0.264


3 0.94 0.12

3 1.63 0.08

3 1.69 0.08


1.98 0.07


Scmub pre-
hurricane


Post-
hurricane


0.26

0.18

0.12


0.287


3 1.60 0.12


4 1.91 0.10


4 1.91 0.10


EAFB;'
GINS /
N=27


Frontal post-
hurricane


Post-
hurricane


Percent woody cover, distance
to nearest occupied dune


3 0.00 0.58 0.467



3 0.00 0.73 0.489


Frontal pre- Post-
hurricaneb hurricane


EAFB / Dune height, distance to nearest
N = 15 occupied dune


" There is no RL analogue for patch occupancy models, instead we used a niax-rescaled RL value as an approximate
measure of strength of association for the top model in each candidate set (Nagelk~erke 1991). All top models provided
a significantly better fit than a base model with no environmental predictors (p < 0.05).
b This model was developed with data on the structure and landscape context of dunes and does not include vegetation
variables that are in other models.





Table 3-3. Relative importance (wstmi), model-averaged parameter estimates, and
unconditional standard errors for variables used to model occupancy for
beach mice in frontal and scrub habitat before and after Hurricane Ivan. Wsum
was estimated by summing Akakie weights (wi) of all models with a variable
of interest. **Confidence intervals do not contain 0 and indicate variable

significantly influences occupancy.
Habitat W Parameter Estimate SE 9()% C.I.


Scrub pre-hurricane habitat
Pre-inovicane occupancy
Dume area (ha) **
Dume height (nt)
Percent cover woody vegetation
Percent cover herbaceous
Distance to nearest occupied dune (nl)
Dume habitat within 2()( n (ha) **
Dume habitat within 1 kn (ha)
East coordinate (na)

Scrub post-hurricane habitat
Post-inovicane occupancy
Dume area (ha) **
Dume height (nt)
Percent cover woody vegetation
Percent cover herbaceous
Distance to nearest occupied dune (nl)
Dume habitat within 2()( n (ha) **
Dume habitat within 1 kn (ha)
East coordinate (na)


().679
().165
().642
().128
).()46
().888
).()17
().()()




().536
().257
().498
().197
).()23
().864
).()65
().()()




().524
().162
().498
().163
().()()
().936
).()1()
().()()




().141
().652
().969
).()14
().769
).()32
).()48
).()71




).()67
().73()
().828
).()53
).()25
().()()


().7)9
).()29
1.717
).()(3
).()28
().719
).()(3


().418
).()39
1.221
().182
).()35
().351
).()(7


).()22
-).()35
-().292
-().296
-().)3
().142
-).()(9


1.396
).()93
3.726
().3()2
).()86
1.296
).()15


().548
).()81
1.)24
-().1)6
).()(9
1.)89
).()(6





().516
).()36
1.218
-().3)5

1.)22
().129


().319
().1)6
().943
().714
).()12
().574
).()(7





().298
).()45
1.)48
().377

().461
).()93





().629
().893
9.854
).()33
1.851
).()68
).()89
).()(9




).()98
().151
).()(8
().121
).()6()


).()23
-).()93
-().527
-1.281
-().()ll
().145
-).()(6


1.()73
().255
2.575
1.()69
).()29
2.()33
).()18


Scrub pre-hurricane habitat
Post-hunricane occupancy
Dume area (ha) **
Dume height (nt)
Percent cover woody vegetation
Percent cover herbaceous
Distance to nearest occupied dune (nl)
Dume habitat within 2()( n (ha) **
Dume habitat within 1 km (ha)
East coordinate (na)

Frontal post-hurricane habitat
Post-inovicane occupancy
Dume area (ha)
Dume height (nt)
Percent cover woody vegetation**
Percent cover herbaceous
Distance to nearest occupied dune (nl)
Dume habitat within 2()( n (ha)
Dume habitat within 1 kn (ha)
East coordinate (nt)

Frontal pre-hurricane habitat
Post-hunricane occupancy
Dume area (ha)
Dume height (nt) **
Distance to nearest occupied dune (nl) **
Dune habitat within 2()( n (ha)
Dume habitat within 1 km (ha)
East coordinate nx)


).()26 1.()(6
-).()38 ().11
-().5()6 2.942
-().925 ().315

().264 1.78()
-).()24 ().282


().383
1.)27
16.355
).()(9
-2.627
).()13
-).()55
).()(6




).()32
().485
-).()15
).()93
).()5()


-().652
-().353
().146 -
-).()45
-5.672
-().1)6
-().2)1
-).()(9




-().129
().237
-).()28 -
-().1)6
-).()48


- 1.418
- 2.4()7
32.564
- (.()63
- (.418
- (.132
- (.()91
- (.()21




- (.193
- (.733
-().()(2
- (.292
- (.148















CHAPTER 4
CONCLUSIONS AND CONSERVATION IMPLICATIONS

Habitat loss and fragmentation from coastal development and hurricanes are

believed to be maj or threats to the long-term population persistence of Gulf Coast beach

mice (Holler 1992; Oli et al. 2001). Frontal dunes are protected as critical habitat for

Gulf Coast beach mice, but they are disturbed greatly by hurricanes (Chapter 2). Scrub

dunes (also known as secondary dunes), which are not currently protected under federal

recovery plans, are impacted less by hurricanes and have been proposed to serve as

important refugia habitat for beach mice during hurricanes (USFWS 1987; Swilling et al.

1998). Although restoration techniques exist to promote regeneration of physical

structure of coastal dunes after storms (Miller et al 2001; 2003), understanding of the

features that confer resistance against storm erosion is limited. Also, prior to this study,

little quantitative information was available on: 1) how hurricanes impact habitat

availability for beach mice, 2) utilization of scrub habitat by beach mice, 3) habitat

features that predict occupancy of frontal and scrub dunes by beach mice, and 4) relative

impacts of hurricanes on beach mouse occupancy of frontal versus scrub dunes. My

study contributes to filling these gaps.

Dune Erosion and Loss of Beach Mouse Habitat

Frontal dunes received much greater impacts from Hurricane Ivan than scrub

dunes, and larger dunes in both frontal and scrub habitat experienced less erosion than

small dunes. Structural features that conferred resistance against storm erosion differed

for frontal and scrub dunes, suggesting that different processes act upon these two dune










types. For frontal dunes, tall and wide dunes experienced the least amount of erosion

from Hurricane Ivan's high storm surge. Dune erosion for secondary dunes was

influenced by storm surge from the Gulf and probably also by rising water levels in the

Santa Rosa Sound, located behind the island. Secondary dunes experience less erosion

when located behind a frontal dune. This observation highlights the importance of

maintaining frontal dunes as buffers of storm surge. Small secondary dunes located

farther from the eye of Hurricane Ivan and located on the widest parts of the island

experienced more erosion than small secondary dunes closer to the eye of Hurricane Ivan

and on narrow parts of the island. The reason for this pattern is unknown, but it may be

related to storm surge in the narrow parts of the Santa Rosa Sound. When the flow of

storm surge is confined and water is shallow, high penetration distances have been noted

for washover (Morton and Sallenger 2003).

Hurricanes will continue to fragment and reduce coastal dunes if the interval

between hurricanes and other tropical storms remains shorter than the time required to

redevelop dunes through natural processes or restoration. Tall and wide frontal dunes are

more resistant to storm erosion than smaller frontal dunes and may continue to provide

suitable habitat for beach mice if they maintain appropriate habitat conditions. However,

Hurricane Ivan alone reduced the frontal dune habitat of beach mice by 76.8% in our

study area. As dunes become smaller with subsequent storms, erosion of frontal dunes

may accelerate. In contrast, secondary dune habitat was reduced by only 19.3% by

Hurricane Ivan, indicating that, in periods of high hurricane activity, scrub dunes provide

more stable habitat for beach mice than frontal dunes. However, removal of frontal

dunes is likely to increase impacts of hurricanes on secondary dunes as the buffering










capacity of frontal dunes is lost. Given the inevitable loss of frontal dunes from

hurricanes, incorporation of scrub habitat into conservation efforts for Gulf Coast beach

mice is warranted. Although scrub habitat has been considered marginal for beach mice,

my data suggest that conservation of scrub habitat will promote population persistence

and lessen extinction risk as frontal dunes continue to be removed from the landscape.

Landscape-scale research is needed to understand the interdependency of subpopulations

of mice in frontal and scrub dunes, the conditions under which either of these habitats is

optimal or marginal, and the relative contributions of each of these habitats to long-term

persistence of beach mice populations.

Habitat Restoration for Beach Mice

Dune restoration for frontal dunes after storms typically has focused on the re-

establishment of sea oats, which produces a lattice of rhizomes that can quickly trap and

accumulate sand. My results indicate that cover of woody vegetation is important for

promoting occupancy of frontal dunes by beach mice and woody plants also may

influence occupancy of scrub dunes. This observation has important implications for

conservation and management of beach mouse habitat, as optimal habitat for beach mice

generally is believed to consist of tall frontal dunes vegetated by sea oats and other

herbaceous species (Holler 1992). My data suggest that restoration programs should

incorporate the re-establishment of woody plants on frontal dunes. Scrub dunes are

dominated by woody vegetation, and habitat management strategies for beach mice

should aim to maintain this vegetation. We do not know the exact mechanism by which

woody vegetation influences occupancy of dunes by mice, but woody species provide

cover and food for mice and may stabilize dunes during storms. More research will be

required to understand these mechanisms and to identify key woody species for mice.









Results of my study indicate that landscape context is important for enhancing

occupancy of dune habitats by beach mice regardless of dune type. Isolation restricts

occupancy of frontal dunes and amount of dune habitat surrounding scrub dunes

influences occupancy of these dunes. As dune systems are eroded by hurricanes, dune

fragments become more widely separated by open sand that does not provide resources

for beach mice (e.g., food and substrate for burrow construction) and mice are forced to

move over large open areas to obtain resources in different patches. Movement of mice

also is critical for recolonization of the landscape in areas where mice are extirpated

during hurricanes and for recolonization of restored habitat. These movements are likely

to entail considerable risk (e.g., increased risk of predation). Management efforts should

aim to minimize isolation of dunes. Restoration techniques that provide connectivity

(i.e., facilitate movement) between fragmented frontal dunes or between frontal and

secondary dunes also may benefit beach mice. Vegetation cover facilitates foraging of

beach mice (Bird et al. 2004) and, presumably, would enhance movement by reducing

risks associated with moving between fragments of habitat that remain after hurricanes.

Although my occupancy data provide general evidence that landscape connectivity is

important for beach mice, factors limiting mouse movement (e.g., the degree to which

large open sand gaps restrict movement) are unknown and this would be a fruitful area of

research for understanding the long-term persistence of beach mice in dynamic

landscapes. Finally, restoration techniques to promote increases in dune height for

frontal dunes also would be beneficial for beach mice as taller dunes are more resistant to

storm-related erosion and may facilitate conditions appropriate for burrow construction.















APPENDIX A
DELINEATION OF DUNES IN THE FIELD


Critical definition of dunes for delineation in the field:

*Dunes were mapped if greater than 1 m high with woody vegetation or greater than
1.5 m with grasses and other herbaceous vegetation.

*Dune spurs were considered part of a dune if the cleft between dunes was less than
1.5 m in height.

*Dunes were considered to be separate if they were separated by more than 3 m.

























Table B-1. Correlations for variables measured on 61 scrub dunes surveyed for beach
mice before Hurricane Ivan (Jun. 2004 Sep. 2004).
Dume Dume Dume Dume % w~Poody %6 total East-west Distance
araheight hab~itat habitat vegetation hembaeous Coordinnate to nearest
(h)(m) wi~thmn wpithin I cover cover (UTMu) occupied
200 me km (ha) dume (m)
(ha)

Dume area (ha): r 1 0.559 0.457 0.2L0 0.242 -0.204 0.278 -0.16i5
p <0.01 Dume height (m) :r 0.559 I 0.458 0.301r -0.183 -0060.451; -0.141
Sp <0.01 <00 0.017 0.159 0.613 <01 0.279
Dume hab~itat within 200 m (hai) :r 0.457 0.458 1 0.742 -0.111 0.090 0.488 -0.391

ip <0.01 <0.01 0.01 0.393 0.490 <0.01 0.02
Dume hab~ilat within 1 km(h) /r 0.250n 0.306; 0.742 1 -0.086; 0.002 0.1;89 -0.409)

/p 0.043 0.017 <0.01 -0.508 0.989 <0.01 0.01
%w~oodyvegeation cover :r 0.242 0.183 -0.111 -0.086 I -0.388 0.043 0.099

:p 0.06;0 0.159 0.393 0.508 -0.002 0.742 0.448

%/ totalhbembaeous cover r, -0.204 -O.Odd 0.090 0.002 -0.388 1 -0.047 0.050

!p 0.115 0.6i13 0.490 0.989 0.002 -0.718 0.700
East-west coordinate (UTM)f Ir 0.278 0.456; 0.488 0.1r89 0.043 -0.047 1 -0.166C

Ip 0.030 <0.01 <0.01 <0.01 0.742 0.718 0.200

Distance in nearest occupied dxute (m) r -0.165 -01.141 -0.391 -0.409 01.099 0.050 -0.166L I

:p 0.228 0.279 0.02 0.01 0.448 0.700 0.200


APPENDIX B
CORRELATION MATRICES FOR VARIABLES BY HABITAT









48




Table B-2. Correlations for variables measured on 61 scrub dunes surveyed for beach
mice after Hurricane Ivan (Oct. 2004 -Jan 2005).


Dumo Dime hitait %wnoody % total East-wasut )inao
area height within whi1 vegetation harbacous Coordinate ocpe
(ha) (m) 200 m m ha cover cover ([JTIV) dm m
(ha) a)due m

DL111 aroa~ha) ir I 0551 0.265 0.194 0973 -0.162 0.254 -0.154
ip 0.01 0.039 0.134 0.574 0.211 0.048 0.236
Duni hoigpt(m) ir 0-551 1 0958 0.155 0976 -0933 0.227 -0.104
p <0.01 0.657 0.234 0.563 0.798 0.078 0.426
Dunm habitatwithin2000m ha) r 0265 0958 I 0380 -0277 0909 0A77 -0383
p 0.039 0.657 -<0.01 0.031 0.945 <0.01 0.002
DLuIG habitat within l km (ha) r 0.194 0.155 0.780 1 -0.172 -0.124 0.702 -0A51
p 0.134 0.234 <0.01 .182 0.341 <0.01 <0.01
%woody vegetation cover Ir 0973 0976i -0277 -0.172 I -0974 -0078 0.171
ip 0.574 0.563 0.031 01.86 -0.540 0.549 0.187
% total harbacous cover ir -0.162 -033 0009 -0.124 -074 I -0.131 0929
p 0.211 0.798 0.945 0.341 0.540 -0.313 0.825
East-vest coordinate (UTMv) r 0254 0227 0A77 0.702 -0078 -0.131 1 -0A64

p 0.048 0.078 <0.01 <0.01 0.549 0.313 -<0.01

Distance to nearest occupied domeo (m) ir -0.154 -0.104 -0383 -0A51 0.171 OD29 -0A64

p 0.23(6 0.426 0.002 <0.01 0.187 0.825 <0.01



Table B-3. Correlations for variables measured on foredunes (Eglin Air Force Base,

N = 11, and Gulf Islands National Seashore, N = 15) surveyed for beach

mice after Hurricane Ivan. (Oct. 2004 Feb. 2005).


Drule ture D tat tat %woody % total EalSt-est Disancet
area height within itn1 vagotation harbacous Coordinateoccpe
(ha) (m) 200 m m ha cover cover (UTMv) dur m
(ha) a)due m

Dume aRen(Im) r I 0A33 0.110 0A69 0304 0322 0A31 -0D35
ip- 0.027 0.592 0.016; 0.132 0.109 0.028 0.862z
Dtun heiglt(m) r 0A33 I -0321 OD47 -0935 -0.118 OD36 0014
ip 0.027 -0.109 0.818 0.866; 0.565 0.861 0.946
Dume habitat within 200 m (ha) r 0.110 -0321 I 0.237 -0D74 -0.127 -0050 -0 21
IP 0.592 0.109 -0.244 0.718 0.536 0.809 <0.01
Doui habitatwnithinl krm Pa) ir 0469 OD47 0.237 I 0382 0A88 0088 -0978
ip 0.016 0.818 0.244 -0.049 0.011 <0.01 0.700
% woody vegetationl cover r 0304 -0035 -0974 0382 1 0605 0.740 0.139
ip 0.132 0,866 0.718 0.049 .001 <0,01 0.488
% total harbacous cover i r 0322 -0.118 -0.127 0A88 OLOB I 0304 -0231
: p 0.109 0.565 0.536 0.011 0.001 -0.123 0.247
East-west coordinate (UTMv) r 0431 0936 -0950 088 <091 0304 I 0099

Ip 0.028 0.861 0.809 <0.01 0.740 0.123 -0.624

Ditanco to noarestoucupied dto(m) i r -0935 0914 -0621 -0978 0.139 -0.231 0999 I

i p 0.862 0.946 <0.01 0.700 0.488 0.247 0.624
















APPENDIX C
COMPARISON OF FRONTAL DUNES AT EGLIN AIR FORCE BASE AND GULF
ISLANDS NATIONAL SEASHORE

Table C-1. Results of t-tests comparing vegetation, structure and landscape context for
frontal dunes on Eglin Air Force Base and Gulf Islands National Seashore
measured after Hurricane Ivan.
Variable t df p
Dune area (ha) 1.915 24 0.06
Dune height (m) -0.287 24 0.77
Dune habitat within 200 m (ha) -0.209 24 0.84
Dune habitat within 1 km (ha) 5.305 24 <0.01
% woody vegetation cover -2.260 24 0.03
% total herbaceous cover -1.060 24 0.30
Distance to nearest neighbor (m) 0.697 24 0.49
Distance to nearest scrub dune (m) -1.975 24 0.06
















APPENDIX D
PREDICTORS OF CHANGE IN OCCUPANCY OF FRONTAL DUNES AFTER
HURRICANE IVAN

Table D-1. Mean values, standard errors, and t-test results for habitat variables on frontal
dunes on EAFB that became unoccupied and for dunes that remained
occupied after Hurricane Ivan. Where variances were found to not be equal (p
< 0.05), a student t-test with the assumption of unequal variances was used.
For all other variables, t-statistics and p-value are for student t-tests with the
assumption of eaual variances.
Variable' Unoccupied Occupied
Mean (+SE) Mean (a SE) t df p
Before hurricane
Dune Area (ha) 0.56 (0.12) 0.62 (0.16) -0.31 13 0.76
East-west 522516 (1901) 523624 (1919) -0.39 13 0.70
coordinate (UTM)
Dune height (m) 3.44 (0.32) 4.87 (0.53) -2.53 13 0.03
Dune habitat within 0.25 (0.13) 0.37 (0.15) -0.61 13 0.55
200 m (ha)
Dune habitat within 6.59 (1.34) 11.24 (2.89) -1.63 13 0.13
1km (ha)
Distance to nearest 317.57 (87.41) 106.61 (28.51) 2.16 13 0.05
occupied dune (m)
After hurricane
Dune Area (ha) 0.11 (0.06) 0.28 (0.07) -1.96 13 0.07
East-west 522410 (2160) 525033 (793) -1.14 13 0.28
coordinate (UTM)
Dune height (m) 1.52 (0.59) 3.18 (0.53) -2.06 13 0.06
Dune habitat within 0.08 (0.03) 0.30 (0.12) -0.91 9 0.39
200 m (ha)
Dune habitat within 7.56 (1.04) 9.29 (2.21) -0.44 9 0.67
1 km (ha)
Distance to nearest 210.12 (25.46) 134.34 (35.07) 1.49 9 0.17
occupied dune (m)
Percent cover of 2.0 (1.4) 8.5 (4.8) -3.87 9 0.01
woody vegetation
Percent cover of 23.0 (4.0) 25.0 (5.0) -0.33 9 0.75
total herbaceous
SHabitat data are presented for variables measured before and after the hurricane. Occupancy
data presented are data taken after the hurricane.
SAnalysis of dune area, dune height, and east-west coordinate for frontal dunes post-Ivan
included the four dunes that were completely destroyed. These four dunes, however, were not
included when assessing vegetation, distance to nearest occupied dune, or amount of
surrounding habitat.

















APPENDIX E
CORRELATION MATRIX FOR STRUCTURAL FEATURES OF FRONTAL DUNES
ON EGLIN AIR FORCE BASE

Table E-1. Correlations for structural and landscape context variables measured on
frontal dunes (N = 93) on Santa Rosa Island prior to Hurricane Ivan.

Width
Dunce Dtme Dime 3 Distlano Onp D~une Dune
of
l ostLfrain hei grt ilsnlnc aren from Ivan Di ~lance lenglh width
rvan (96) (m (n) (k3 mr) (m) (m) (m


Dneln lost liom Ivan (96 i r 1 -0.50 025 -028 -0,50 -001 -021 -04C7
pv 0.01 0.02 0.01 <00 0.94 0.04 <0.01
Dunc hoi ILl~ln) r -050 I -0.17 047 0.12 -092 052 049
py <0.01 0.11 <0.01 0.24 0.92 <0.01 <0.01
Width~ of is~lanld km) I r OSS 0.17 I OSI 0.17 031 091 -0.01
Ip 0.02 0.11 091 0.1 I 0.94 0,96 0.97
Dneln aren (hn) I r *0.28 047 OB1I 1 0401 023 094 0.72
p 0.01 <0.01 0,91 -0.97 0 02 <0.01 <0,01
Distancc from Ivan Okm) I r *050 0.12 0.17 -001 1 0.4 005 0.10
p <0,01 0,24 0.1 0,97 -0.1 9 0,61 0,33
OnP distance (m) I r -0901 -042 081 033 0.14 1 0.15 020
: p 0.94 0.92 0.94 0.02 0.1 9 -0.14 0.06i
Dne Inclgth (In) r *021 052 091 094 .095 05 I 0 1

!y p 04 <0.01 0.96 <0.01 001 0.4 0.01
D~unewidth~(m) Ir *OA7 04Q9 .0.01 0.72 0.10 030 061 I

ip < 0.01 <0.01 0.97 <0.01 0.33 0.06 <0.01-




























dunes).

Wi dth
D3unc D~uame Dmam Di stance Gap Datea Duna
lostfromn heig TL isulanrl aro from Ivan Disalnce lenglh width
1 van (%) (m) (hr a) k m) (m) (m) (m)
(k m)

Dune lost from Ivan (%6) r 1 -037 0.19 -024 0.17 0.19 -031 -0AS
: -i 0.01 0.18 0.09 0.23 0.18 0.02 0.01
Otmro hoighl(mn) ; r -037 I -0.20 048 Oa5 0117 040 052
i p 0.01 *0.16 <0.01 0.72 0.61 <0.01 <0.01
Wi dt of: i slanld(km) R 0.19 .020 I Oh? OA6 0A6 01)3 008
ip 0.18 0.16, 61 0.01 ODI 0,86 0.56
Dllme aren (ha) jr 034 0 8 0907 I 0d2 *0.17 082 0 7
~p 0.09 <0.01 0.61 -0.12 0.23 <0.01 <0,01
Distance from Ivan (km) ir 0.17 095 0c16 062 1 *093 010 0.10
p 0.23 0,72 .01l 0,12 OS 0,16 0.50
Gap dlistanc (ml) r 0.19 097 046 -0.17 -093 I -0.13 -042
p 0.18 0.61 0.1 .23 0 6 0.37 0.08
DamecIcngth Om) r 1031 040 003 082 020 *0.13 I 035

p 0.02 <0.01 0.86 <0.01 0.16 0.37 0.01

Damew\pidthU IIm) : r 045 052 098 087 0.10 *0A2 035 I

jp 0.01 <0.01 0.56 <0.01 0.50 0.08 0.01


APPENDIX F
CORRELATION MATRIX FOR STRUCTURAL FEATURES OF SECONDARY
DUNES ON EGLIN AIR FORCE BASE

Table F-1. Correlations for structural and landscape context variables measured on
secondary dunes on Santa Rosa Island prior to Hurricane Ivan. Dunes were
sub-sampled to represent proportional area of large (83.6%) and small


(16.4%) dunes on the landscape (N


52 with 34 small dunes and 18 large
















LITERATURE CITED


Biedermann, R. 2004. Modeling the spatial dynamics and persistence of the leaf beetle
(Gonioctena olivacea) in dynamic habitats. Oikos 107:645-653.

Bird, B. L. 2002. Effects of predatory risk, vegetation structure, and artificial lighting on
the foraging behavior of beach mice. MS Thesis. University of Florida,
Gainesville, FL. 58 pp.

Bird, B. L., L. C. Branch, and D. L. Miller. 2004. Effects of coastal lighting on foraging
behavior of beach mice. Conservation Biology 18(5):1435-1439.

Blair, W. F. 1951. Population structure, social behavior and environmental relations
in a natural population of the beach mouse (Peromyscus polionotus
leucocephalus). Pages 1-47. Contributions from the Laboratory of Vertebrate
Biology, University of Michigan. Ann Arbor, MI.

Bonham, C. D. 1989. Measurements for terrestrial vegetation. Wiley. New York, NY.

Bourg, N.A., W.J. McShea, and D.E. Gill. 2005. Putting a cart before the search:
successful habitat prediction for a rare forest herb. Ecology 86:2793-2804.

Breiman, L., Friedman, J.H., Olshen, R.A., Stone, C.G. 1984. Classification and
regression trees. Wadsworth International Group. Belmont, CA.

Burnham, K. P. and D. R. Anderson 2002. Model selection and multimodel inference: a
practical information-theoretic approach. Springer. New York, NY.

Buskirk, J.V. 2005. Local and landscape influence on amphibian occurrence and
abundance. Ecology 86:1936-1947.

Carlsson, A. and O. Kindvall. 2001. Spatial dynamics in a metapopulation network:
recovery of a rare grasshopper (Stauroderus scalaris) from population
refuges. Ecography 24:452-460.

Cooper, C. B. and J. R. Walters. 2002. Independent effects of woodland loss and
fragmentation on Brown Treecreeper distribution. Biological Conservation
105:1-10.










Cox, J. and R. T. Engstrom. 2001. Influence of the spatial pattern of conserved lands
on the persistence of a large population of red-cockaded woodpeckers.
Biological Conservation 100:137-150.

Dahl, B.E. and Woodard, D.W. 1977. Construction of Texas coastal foredunes with
sea oats (Uniola paniculata) and bitter panicum (Panicum ama~rum).
International Joumnal of Biometeorology 21:267-275.

De'ath, G. and K.E. Fabricius. 2000. Classification and regression trees: a powerful
yet simple technique for ecological data analysis. Ecology 81:3178-3192.

Ehrenfeld, J.G.,1990. Dynamics and processes of barrier island vegetation. Critical
Reviews in Aquatic Science 2:437-480.

Ellner, S.P. and J. Fieberg. 2003. Using PVA for management despite uncertainty:
effects of habitat, hatcheries, and harvest on salmon. Ecology 84: 1359-1269.

Emanuel, K. 2005. Increasing destructiveness of tropical cyclones over the past 30 years.
Nature 436: 686-688.

Environmental Systems Research Insititue. 1996. ArcView GIS, version 3.2.
Environmental Systems Research Institute, Redlands, Califomnia, USA.

Flather, CH, K.R. Wilson, D.J. Dean, and W.C. McComb. 1997. Identifying gaps in
conservation networks: Of indicators and uncertainty in geographic-based
analyses. Ecological Applications 7:531-542.

Frank, K. 2005. Metapopulation persistence in heterogeneous landscapes: Lessons
about the effect of stochasticity. American Naturalist 165:3 74-3 88.

Glennon, M. J., W. F. Porter, and C. L. Demers. 2002. An alternative field technique for
estimating diversity of small-mammal populations. Journal of Mammalogy
83:734-742.

Gore, J. A. and T. L. Schafer. 1993. Distribution and conservation of the Santa Rosa
beach mouse. Proceedings Annual Conference Southeast Association of Fish
and Wildlife Agenices 47:378-385.

Gu, W. D. and R. K. Swihart. 2004. Absent or undetected? Effects of non-detection
of species occurrence on wildlife-habitat models. Biological Conservation
116:195-203.

Guisan, A. and W. Thuiller. 2005. Predicting species distribution: offering more than
simple habitat models. Ecology Letters 8:993-1009.










Hallermier, R.J. and P.E. Rhodes. 1988. Generic treatment of dune erosion for 100-year
event. Proceedings of the 20th International Conference on Coastal Engineering,
ASCE, New York, 1107-1115.

Hanski, I. 1994. A practical model of metapopulation dynamics. Journal of Animal
Ecology 63:151-162.

Hesp, P. 2002. Foredunes and blowouts: initiation, geomorphology, and dynamics.
Geomorphology 48:245-268.

Hesp, P. 2004. Coastal dunes in the tropics and temperate regions; Location,
formation, morphology and vegetation processes. Pages 29-49 in M. L.
Martinez, and N.P. Psuty, editor. Coastal Dunes, Ecology and Conservation.
Springer-Verlag. Berlin.

Hoekstra, J. M., W. F. Fagan, and J. E. Bradley. 2002. A critical role for critical habitat
in the recovery planning process? Not yet. Ecological Applications
12:701-707.


Holler, N.R. 1992. Perdido Key Beach Mouse. In Rare and Endangered Biota of
Florida, Volume 1. Mammals (S. R. Humphrey ed.). University Press of
Florida, Gainesville, FL 52pp.


Humphrey, S. R. and D. B. Barbour. 1981. Status and habitat of 3 subspecies of
Peromyscus polionotus in Florida. Journal of Mammalogy 62:840-844.

Jonzen, N., C. Wilcox, and H. P. Possingham. 2004. Habitat selection and population
regulation in temporally fluctuating environments. American Naturalist
164:E103-E114.

Judge, E.K., M.F. Overton, and J.S. Fisher. 2003. Vulnerability indicators for coastal
dunes. Journal of Waterway, Port, Coastal, and Ocean Engineering 129:270-278.

Kery, M. 2004. Extinction rate, estimates for plant populations in revisitation studies:
importance of detectability. Conservation Biology 18:570-574.

Kriebel, D.L., R. Dalrymple, A. Pratt, and V. Sakovich. 1997. A shoreline risk index
for northeasters. Proceedings of the ASCE International Conference on Natural
Disaster Reduction, ASCE, New York, 251-252.

Mabee, T. J. 1998. A weather-resistant tracking tube for small mammals. Wildlife
Society Bulletin 26:571-574.

MacKenzie, D. I., J. D. Nichols, G. B. Lachman, S. Droege, J. A. Royle, and C. A.
Langtimm. 2002. Estimating site occupancy rates when detection probabilities are
less than one. Ecology 83:2248-2255.











MacKenzie, D. I. J. D. Nichols, J. E. Hines, M. G. Knutson, and A. B. Franklin. 2003.
Estimating site occupancy, colonization, and local extinction when a species is
detected imperfectly. Ecology 84:2200-2207.

Martinez, M. L., N.P. Psuty., and R.A. Lubke. 2005. A perspective on coastal dunes.
Pages 1-10 in M. L. Martinez, and N.P. Psuty, editor. Coastal Dunes, Ecology and
Conservation. Springer-Verlag. Berlin.

Mazerolle, M.J., A. Desrochers, L. Rochefort. 2005. Landscape characteristics influence
pond occupancy by frogs after accounting for detectability. Ecological
Applications 15:824-834.

Mendelssohn, I.A., M.W. Hester, F.J. Montefemante, and F. Talbot. 1991. Experimental
dune building and vegetative stabilization in a sand-deficient barrier island setting
on the Louisiana Coast. Journal of Coastal Research 7: 137-149.

Milio, J.F. 1998. Endangered and Threatened Wildlife and Plants; Determination of
Endangered Status for the St. Andrew Beach Mouse. U.S. Fish and Wildlife
Service. Federal Register 63:70053-70062.

Miller, D.L. and M. Thetford. 2001. Evaluation of sand fence and vegetation for dune
building following overwash by Hurricane Opal on Santa Rosa Island, Florida.
Journal of Coastal Research 17:936-948.

Miller, D. L., L. Yager, M. Thetford, and M. Schneider. 2003. Potential use of Thriola
paniculata rhizome fragments for dune restoration. Restoration Ecology 11:359-
369.

Morton, R.A. and A.H. Sallenger. 2003. Morphological impacts of extreme storms on
sandy beaches and barriers. Journal of Coastal Research 19:560-573.

Moyers, J. E. 1996. Food habits of the Gulf Coast subspecies of beach mice
(Peromyscus polionotus spp.). MS Thesis. Auburn University, Auburn, AL. 54pp.

Musila, W.M., J.I. Kinyamario, and P.D. Jungerius. 2001. Vegetation dynamics of
coastal sand dunes near Malindi, Kenya. African Journal of Ecology 39: 170-
177.

Nagelkerke, N. J. D. 1991. A note on a general definition of the coefficient of
determination. Biometrika 78:691-692.

Nordstrom, K.F., R. Lampe., and L.M. Vandemark. 2000. Reestablishing naturally
functioning dunes on developed coasts. Environmental Management 25:37-51.










Nordstrom, K.F. and W.A. Mitteager. 2001. Perceptions of natural and restored beach
and dune characteristics by high school students in New Jersey, USA. Ocean
and Coastal Management 44: 545-559.

Oli, M. K., N. R. Holler, and M. C. Wooten. 2001. Viability analysis of endangered
Gulf Coast beach mice (Peromyscus polionotus) populations. Biological
Conservation 97: 107-118.

Pearce, J. and S. Ferrier. 2000. Evaluating the predictive performance of habitat models
developed using logistic regression. Ecological Modelling 133:225-245.

Potter, J. C. 1985. Endangered and Threatened Wildlife and Plants; Determination of
Endangered Status and Critical Habitat for Three Beach Mice. U.S. Fish and
Wildlife Service. Federal Register. 50:23872-23889.

Psuty, N.P. 2005. The coastal foredune: A morphological basis for regional coastal dune
development. Pages 11-27 in M. L. Martinez, and N.P. Psuty, editor. Coastal
Dunes, Ecology and Conservation. Springer-Verlag. Berlin.

R Development Core Team. 2003. R: A language and environment for statistical
computing. R Foundation for Statistical Computing. Vienna, Austria.

Sallenger, A.H., Jr. 2000. Storm impact scale for barrier islands. Journal of Coastal
Research 16:890-895.

Schrott, G. R., K. A. With, and A. T. W. King. 2005. On the importance of landscape
history for assessing extinction risk. Ecological Applications 15:493-506.

Segurado, P. and M.B. Arauj o. 2004. An evaluation of methods for modeling species
distributions. Journal of Biogeography 31:1555-1568.

Sneckenberger, S. I. 2001. Factors influencing habitat use by the Alabama beach mouse
(Peromyscus polionotus amnmobates). MS Thesis. Auburn University, Auburn,
AL. 101pp.

SPSS Inc. 2004. SPSS Base 13.0 for Windows User' s Guide. SPSS Inc., Chicago IL.

Stallins, J. A. and A. J. Parker. 2003. The influence of complex systems interactions
on barrier island dune vegetation pattern and process. Annals of the
Association of American Geographers 93:13-29.

Stewart, S.R. 2005. Tropical Cyclone Report: Hurricane Ivan 2-26 September 2004.
NOAA National Weather Service, NHC/TPC.










Stone, G.W., B. Liu, D.A. Pepper, and P. Wang. 2004. The importance of extratropical
and tropical cyclones on the short-term evolution of barrier islands along the
northern Gulf of Mexico, USA. Marine Geology 210:63-78.

Swilling, W. R. and M. C. Wooten. 2002. Subadult dispersal in a monogamous species:
The Alabama beach mouse (Peronzyscus polionotus a~nmnobates). Journal of
Mammalogy 83:252-259.

Swilling, W. R., M. C. Wooten, N. R. Holler, and W. J. Lynn. 1998. Population
dynamics of Alabama beach mice (Peronzyscus polionotus a~nmnobates)
following Hurricane Opal. American Midland Naturalist 140:287-298.

Snyder, R. A. and C. L. Boss. 2002. Recovery and stability in barrier island plant
communities. Journal of Coastal Research 18:530-536.

Taylor, M. F. J., K. F. Suckling, and J. J. Rachlinski. 2005. The effectiveness of the
endangered species act: A quantitative analysis. Bioscience 55:360-367.

U. S. Fish and Wildlife Service. 1981. Standards for the development of habitat suitability
index models for use in the habitat evaluation procedure. U.S. Fish and Wildlife
Service, Division of Ecological Services. Manuscript 103. Washington, DC.
171pp.

U.S. Fish and Wildlife Service. 1987. Recovery plan for the Choctawhatchee, Perdido
Key and Alabama beach mouse. U.S. Fish and Wildlife Service, Atlanta,
Georgia. 45pp.

Van Horne, B. 1983. Density as a misleading indicator of habitat quality. Journal of
Wildlife Management 47:893-901.

Van Horne, B., G. S. Olson, R. L. Schooley, J. G. Corn, and K. P. Bumnham. 1997.
Effects of drought and prolonged winter on Townsend's ground squirrel
demography in shrubsteppe habitats. Ecological Monographs 67:295-315.

Vellinga, P. 1982. Beach and dune erosion during storm surges. Coastal Engineering
6:361-387.

Welch, N. E., and J. A. MacMahon. 2005. Identifying habitat variables important to
the rare Columbia spotted frog in Utah (USA): An information-theoretic
approach. Conservation Biology 19:473-481.


Zar, J.H. 1998. Biostatistical Analysis. Prentice Hall. New York, NY.
















BIOGRAPHICAL SKETCH

Alexander James Pries was born in Spooner, Wisconsin on May 26, 1980. Son of

James and Constance Pries, he grew up in St. Paul, MN, but often escaped to the northern

forests of Minnesota and Wisconsin during the summer months. It was there, during

hours of playing in the forests, lakes, and streams that he began to first observe and

appreciate natural ecosystems. In 1998, he enrolled at The College of Wooster in Ohio

and received a B.A. in biology in the spring of 2002. After graduation, he traveled to

Costa Rica, where he served as a teaching assistant in a course on tropical ecology for the

Organization for Tropical Studies. After this experience, he moved to Avon Park, FL,

where he worked as a research technician for the University of Florida on a proj ect

looking at features of habitat use by round-tailed muskrats (Neofiber alleni). In 2003, he

was hired by Archbold Biological Station to serve as a research technician for a

population assessment of Florida Scrub Jays. He began graduate school in August of

2003 at the University of Florida' s Department of Wildlife Ecology and Conservation,

from which he received his M. S. in 2006.




Full Text

PAGE 1

HURRICANE IMPACTS ON COASTAL DU NES AND SPATIAL DISTRIBUTION OF SANTA ROSA BEACH MICE ( Peromyscus polionot us leucocephalus ) IN DUNE HABITATS By ALEXANDER JAMES PRIES A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FUFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2006

PAGE 2

Copyright 2006 by Alexander James Pries

PAGE 3

iii ACKNOWLEDGMENTS I would like to acknowledge the support received from numerous organizations, groups, and people during the deve lopment of this project. Funding and logistic support was provided by the National Park Service and Eglin Air Force Base. The University of Florida, Milton campus and UF Department of Wildlife Ecology and Conservation provided equipment and additional funding. My committee members (Dr. Lyn C. Br anch, Dr. Debbie L. Miller, and Dr. George W. Tanner) went above and beyond the call of duty in helping with development and implementation of this research. Dr. M iller provided housing af ter my trailer broke and continued to graciously welcome me into her home during the endless process of data collection. I have enjoyed thoroughly our conversations on restoration ecology and ethics of being a scientist. Dr. Tanner was most insightful in steering me towards proper techniques for assessment of vegetation and ot her habitat features. Dr. Branch is a sound editor who masterfully checked and rechecked my thesis for clarity and scope. Thanks go to Riley Hoggard at Gulf Islands National Seashore and Bruce Hagedorn, Bob Miller, and Dennis Teague at Jackson Guard, Eglin AFB. These individuals served as important contacts and sources of support when materials were lacking or I had questions about the accessibility of certain sites. I am most thankful for their support. Many individuals assisted me during various phases of this project and I will attempt to name them all here. I apologize for those who I may miss but your deeds are not forgotten. Tanya Alva rez endured the hot conditions and was covered in black

PAGE 4

iv carbon powder for weeks. Mica Schneider and Li sa Yager created the initial cover layer of dunes (with over 750 polygons) on Santa Rosa Island that was massively important. Jonathan Shore and Cathy Hardin were fantas tic as temporary field technicians during the difficult conditions. I also offer apprecia tion to Bob Schooley and Arpat Ozgul for statistical assistance dur ing initial data analysis. Conve rsations with many graduate students including Jason Martin, Elizabet h Swiman, Ann George, Dan Thornton, and Traci Darnell helped to craft and refine appropriate research questions. Finally, I am thankful to my immediate fa mily and friends outside of the scientific or graduate school community. Your abi lity to listen and be a sounding board when things became difficult is a ma jor reason why I have been able to complete this research. Your support and patience are massively important to me and I dedicate this work to you.

PAGE 5

v TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iii LIST OF TABLES............................................................................................................vii LIST OF FIGURES...........................................................................................................ix ABSTRACT....................................................................................................................... ..x CHAPTER 1 INTRODUCTION........................................................................................................1 Beach Mice and Threats to Survival.............................................................................1 Habitat Use by Beach Mice..........................................................................................1 Coastal Dunes, Development, and Erosion..................................................................2 Dune Restoration and Protection for Beach Mice........................................................3 2 INFLUENCE OF DUNE STRUCTURE ON STORM-RELATED EROSION FOR FOREDUNES AND SECONDARY DUNES ON SANTA ROSA ISLAND, FLORIDA.....................................................................................................................5 Introduction...................................................................................................................5 Methods........................................................................................................................7 Study Area.............................................................................................................7 Characteristics of Hurricane Ivan..........................................................................8 Dune Mapping.......................................................................................................9 Statistical Analyses..............................................................................................10 Results........................................................................................................................ .12 Conditions before Hurricane Ivan.......................................................................12 Hurricane Ivans Impact on Fo redunes and Secondary Dunes............................12 Regression Trees.................................................................................................13 Discussion...................................................................................................................15 3 INFLUENCE OF HABITAT AND LANDSCAPE FEATURES ON SPATIAL DISTRIBUTION OF SANTA ROSA BEACH MICE IN TWO DUNE HABITATS BEFORE AND AFTER A HURRICANE.............................................24 Introduction.................................................................................................................24

PAGE 6

vi Methods......................................................................................................................26 Study Area and Habitat Mapping........................................................................26 Dune Occupancy.................................................................................................27 Predictor Variables: Vegetation Cover and Landscape Structure.......................29 Occupancy Models..............................................................................................30 Results........................................................................................................................ .32 Hurricane Impacts on Habitat Availability at EAFB..........................................32 Dune Occupancy.................................................................................................32 Habitat Models....................................................................................................33 Discussion...................................................................................................................34 4 CONCLUSIONS AND CONSERVATION IMPLICATIONS.................................42 Dune Erosion and Loss of Beach Mouse Habitat.......................................................42 Habitat Restoration for Beach Mice...........................................................................44 APPENDIX A DELINEATION OF DU NES IN THE FIELD...........................................................46 B CORRELATION MATRICES FOR VARIABLES BY HABITAT..........................47 C COMPARISON OF FRONTAL DUNES AT EGLIN AIR FORCE BASE AND GULF ISLANDS NATIONAL SEASHORE.............................................................49 D PREDICTORS OF CHANGE IN OCCUPANCY OF FRONTAL DUNES AFTER HURRICANE IVAN....................................................................................50 E CORRELATION MATRIX FOR STRUCTURAL FEATURES OF FRONTAL DUNES ON EGLIN AIR FORCE BASE..................................................................51 F CORRELATION MATRIX FOR STRUCTURAL FEATURES OF SECONDARY DUNES ON EGLIN AIR FORCE BASE.........................................52 LITERATURE CITED......................................................................................................53 BIOGRAPHICAL SKETCH.............................................................................................59

PAGE 7

vii LIST OF TABLES Table page 2-1 Means and standard errors for stru ctural variables measured to explain dune erosion in foredunes and secondary dunes on Santa Rosa Island............................17 2-2 Statistics for evaluation of dune char acteristics as predictors of dune loss as a result of Hurricane Ivan............................................................................................19 3-1 Means and standard errors for stru ctural and vegetation variables measured for modeling occupancy of frontal and scr ub habitat by Santa Rosa beach mice on Eglin Air Force Base (EAFB) and Gulf Islands National Seashore (GINS) on Santa Rosa Island, FL...............................................................................................38 3-2 AIC-based selection of site occupa ncy models of dune occupancy for Santa Rosa beach mice in frontal and scrub dune habitat...........................................................40 3-3 Relative importance (wsum), model-averaged parameter estimates, and unconditional standard errors for variab les used to model occupancy for beach mice in frontal and scrub habitat before and after Hurricane Ivan...........................41 B-1 Correlations for variables measured on 61 scrub dunes surveyed for beach mice before Hurricane Ivan (Jun. 2004 Sep. 2004).......................................................47 B-2 Correlations for variables measured on 61 scrub dunes surveyed for beach mice after Hurricane Ivan (Oct. 2004 Jan 2005)............................................................48 B-3 Correlations for variables measured on foredunes (Eglin Air Force Base, n = 11, and Gulf Islands National Seashore, n = 15) surveyed for beach mice after Hurricane Ivan. (Oct. 2004 Feb. 2005).................................................................48 C-1 Results of t-tests comparing vege tation, structure and landscape context for frontal dunes on Eglin Air Force Base and Gulf Islands National Seashore measured after Hurricane Ivan.................................................................................49 D-1 Mean values, standard errors, and ttest results for habita t variables on frontal dunes on EAFB that became unoccupied and for dunes that remained occupied after Hurricane Ivan.................................................................................................50 E-1 Correlations for structural and lands cape context variables measured on frontal dunes (N = 93) on Santa Rosa Isla nd prior to Hurricane Ivan.................................51

PAGE 8

viii F-1 Correlations for structural a nd landscape context variables measured on secondary dunes on Santa Rosa Island prior to Hurricane Ivan...............................52

PAGE 9

ix LIST OF FIGURES Figure page 2-1 Map of Santa Rosa Island, FL. The study area encompasses the section between Navarre and Fort Walton Beach.................................................................20 2-2 Cross validation rela tive error for regression tr ees for (a) foredunes and (b) secondary dunes to explain dune loss from Hurricane Ivan in relation to measured predictor variables....................................................................................21 2-3 Regression trees relating percentage of dune lost from Hurricane Ivan for (a) foredunes (N = 93) and (b) secondary dune s (N = 52) to physical features of dunes, spatial location of dunes with respect to where Hurricane Ivan made landfall, and width of island.....................................................................................22 2-4 Regression trees relating percentage of dune lost from Hurricane Ivan for (a) secondary dunes 0.25 ha (N = 61) and (b) seco ndary dunes <0.25 ha (N = 34) to dune features, spatial location, and island width..................................................23

PAGE 10

x Abstract of Thesis Presen ted to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Science HURRICANE IMPACTS ON COASTAL DU NES AND SPATIAL DISTRIBUTION OF SANTA ROSA BEACH MICE ( Peromyscus polionot us leucocephalus ) IN DUNE HABITATS By Alexander James Pries May 2006 Chair: Lyn C. Branch Cochair: Deborah L. Miller Major Department: Wildlife Ecology and Conservation I examined the impact of Hurricane Ivan on dune erosion and changes in spatial distribution of beach mice ( Peromyscus polionotus leucocephalus ) in two dune habitats on Santa Rosa Island, FL. Foredunes (i.e., fr ontal dunes) and sec ondary (i.e., scrub) dunes were mapped and surveyed for the pres ence of beach mice before and after the hurricane using polyvinyl chlo ride (PVC) tracking tubes. I also collected data on physical structure, vegetation, and landscape context for each dune during these two time periods. Regression trees were used to evaluate structur al features of dunes that explained patterns in dune erosi on as a result of Hurricane Iv an for the two dune types. I used site-occupancy models and an informationtheoretic approach to evaluate predictors of occupancy for beach mice in frontal and s econdary dune habitats before and after the hurricane. Hurricane Ivan removed 68.2% of frontal dune area surveyed for beach mice (Chapter 3) and 76.8% of a ll frontal dune area mapped (Cha pter 2). Secondary dunes

PAGE 11

xi surveyed for beach mice only lost 14.8% of thei r total area (Chapter 3) and all secondary dunes surveyed lost 19.3% of their area (Chapter 2). Dune erosion for frontal dunes was related inversely to distance from where th e eye of Hurricane Ivan passed over the island, dune height, and dune width. Dune erosion for large sec ondary dunes was reduced when dunes were located behind foredunes. Erosi on of small secondary dunes increased with distance of the dune from where the eye of the hurricane passed over the island. The reason for this pattern is unknown, but it may be related to spatial distribution of storm surge in the Santa Rosa Sound. Dune erosi on decreased with incr easing length and area of small secondary dunes. Beach mice o ccupied 100% of frontal dunes before the hurricane and 59% of these dunes after the hur ricane. Occupancy of scrub habitat by beach mice was not statistically different be fore and after the hurricane, but was higher (~70% of sites) than frontal dune occupanc y after the storm. Frontal dune occupancy was influenced largely by percent cover of woody vegetation and distance to nearest occupied dune. Probability of occupancy by beach mice in scrub habitat increased with an increase in dune area and amount of dune habitat surrounding the dune within 200 m. This study indicates that scrub habitat, which currently is not protected for beach mice, is important for mice because of the stochastic and severe impacts of storms on frontal habitat. With further removal of frontal dune habitat, scrub could become essential for long-term persistence of beach mice. The study suggests that restoration programs for frontal dunes set targets for the construction of dunes that are tall and wide. Dunes with these two features likely will mitigate storm surge associated with hurricanes and protect coastal dunes, mouse habitat, or in frastructure located further inland.

PAGE 12

1 CHAPTER 1 INTRODUCTION Beach Mice and Threats to Survival Beach mice ( Peromyscus polionotus spp.) are a complex of eight subspecies of the oldfield mouse, which occupy the coastal dune s of Alabama and Flor ida (Holler 1992). Two subspecies along the Atlant ic coast of Florida are fede rally protected and one is extinct. Gulf Coast populations of beach mice comprise five subspecies, four of which are listed as threatened or endangered (Potter 1985; Milio 19 98). The remaining subspecies along the Gulf coast, the Santa Rosa beach mouse ( Peromyscus polionotus leucocephalus ), is not yet listed beca use its geographic range includes several federally managed lands (Gore and Schaffer 1993). All populations of beach mice suffer from severe habitat loss from coastal developmen t and habitat disturbance from hurricanes (Gore and Schaffer 1993; Swilling et al. 1998). Additional threats to beach mice include predation by feral cats ( Felis silvestris ), competition with house mice ( Mus musculus ) in dunes around coastal development, and low population levels in late summer when hurricane activity is most prevalent (Hum phrey and Barbour 1981; Oli et al. 2001). Habitat Use by Beach Mice Beach mice are considered habitat speci alists on dune habitats with burrow locations strongly correlated to these habita t types (Blair 1951; Humphrey and Barbour 1981). Foredunes or frontal dunes, located imme diately adjacent to th e Gulf of Mexico, are believed to be optimal ha bitat for beach mice (USFWS 198 7). Mice also occur in secondary dunes (also known as scrub), which are located farther from the shoreline and

PAGE 13

2 are characterized by increased dominance of woody vegetation. Although densities of beach mice generally are highest in frontal habitat (Swilling et al. 1998), abundance of beach mice in scrub can increase after hurrica nes. Prior to Hurricane Opal, population abundance of Alabama beach mice ( P. p. ammobates ) on trapping grids in scrub habitat was less than frontal dune habitat (Swilling et al. 1998). Four months after the storm, abundance in scrub habitat was almost twice that of frontal dune habita t. Despite use of scrub by beach mice, this habitat is not de signated as critical habitat under USWFS recovery plans for Gulf coast beach m ouse populations (USFWS 1987). Additionally, little is known about what feat ures define suitable scrub habitat for beach mice or how much of this habitat is utilized by mice. Although hurricanes are a natural feature of coastal disturbance, their impacts work in concert with coastal development and anthropogenic habitat loss to impact habitat availability for beach mice (Holler 1992). Population models suggest that hurricane impacts pose a significant threat to all s ubspecies of beach mice (Oli et al. 2001). Hurricanes, in addition to di rectly destroying dunes, fragment remaining dunes and potentially change features of dunes that make them suitable for beach mice. Fragmentation of dunes from hurricanes ma y require beach mice to travel more frequently between remaining dune patches, e xposing them to predators. Habitat loss and fragmentation also may reduce landscape connectivity for beach mice, limiting their ability to recolonize frontal dunes after stor m impacts or force them to utilize more marginal habitats. Coastal Dunes, Development, and Erosion The coastal dunes that beach mice occupy are valued for their fauna and flora, natural beauty, and ability to protect human-made infrastructu re (Nordstrom et al. 2000;

PAGE 14

3 Martinez et al. 2005). Despite this, many co astal dune ecosystems have been changed irreversibly as a result of exploitation of natural resources and anthropogenic development (Martinez et al. 2005). Increases in storm severity and frequency are predicted and storm impacts wi ll continue to alter dunes, re duce infrastructure protection and disturb habitat for wildlife. As a result interest has increase d in the creation or restoration of dunes that will withstand storm impacts. Coastal dunes are formed from aeolian processes with dune development occurring where sediment is trapped by existing ve getation (Hesp 2004; Psuty 2005). Foredunes, located closest to the shoreline, are dynamic structures th at are influenced greatly by the flow of water, wind, and sediment with norma l environmental fluctuations and periodic storm events (Psuty 2005). Secondary dunes are created by sediment flows from existing foredunes or they may be old foredunes. These dunes are no longe r maintained by the processes that drive foredune morphology. Hurricanes alter dune ecosystems by bur ying native vegetatio n under centimeters of deposited sand and also change the conf iguration or presence of dune structures (Ehrenfeld 1990). Dune erosion occurs as a result of storm surge and waves repeatedly narrowing the dune face to cause an eventual breach or when storm surge overwashes a dune and pushes sediment landward (Hesp 2002; Judge et al. 2003). Dune erosion is influenced by duration and intensity of a stor m event; however, structural features of the dune also alter the risk of erosion. Dune Restoration and Protection for Beach Mice Increasing the amount of protected beach m ouse habitat generally is not an option as most coastal dunes in Florida already are in public lands or have been developed (Bird 2002). Therefore, other approaches such as habitat restoration may be important for

PAGE 15

4 long-term maintenance of beach mouse populatio ns. Although habitat restoration is cited as critically important for the recovery of beach mouse populations (USFWS 1987; Oli et al. 2001), little work has been conducted to id entify habitat and lands cape features that influence use of frontal or secondary dune ha bitat by beach mice. Restoration techniques have been developed to promote regeneration of physical structure to dunes after storms (Miller et al. 2001; 2003). These techniques can be used to create dunes with particular structural features (e.g., tall and wide), but identification of structural features of dunes that confer resistance against st orm-related erosion is limited. Restoration techniques that promote crea tion of dunes and dunes with key habitat requirements of beach mice may aid in management of existing protected habitats for these mice. My study contributes to this effort in the following ways: Examining the relationship between dune er osion as a result of Hurricane Ivan and the physical structure of frontal and secondary dunes (Chapter 2) Assessing the impact of Hurricane Ivan on the overall occupancy of frontal and secondary dune habitat by beach mice (Chapter 3) Identifying habitat variables at the patch and landscape scale that influence occupancy of frontal and secondary dunes by beach mice (Chapter 3). Chapter 2 and 3 are written as stand-alone papers for publication. Therefore, some background material is repeated in each chapter.

PAGE 16

5 CHAPTER 2 INFLUENCE OF DUNE STRUCTURE ON STORM-RELATED EROSION FOR FOREDUNES AND SECONDARY DUNES ON SANTA ROSA ISLAND, FLORIDA Introduction Coastal dunes are valued for their aesthetic beauty and their ability to protect human-made structures during storms (Nords trom et al. 2000; Nordstrom and Mitteager 2001). Dunes absorb wave energy, block storm surge, and reduce damage to infrastructure. Coastal dunes also are importa nt wildlife habitat (Mar tinez et al. 2005). Hurricanes and tropical storms have altere d coastal dunes on barri er islands along the northern portion of the Gulf of Mexico in the last decade (Stone et al. 2004). Increases in the severity and frequency of tropical cyclones are predicted and will further modify dune configuration, reduce infrastr ucture protection, and distur b wildlife habitat (Emmanuel 2005). As a consequence, creation and rest oration of dunes has become an important issue in coastal management strategies (Nor dstrom et al. 2000). Strategies for dune protection and restoration coul d benefit from information on physical and spatial factors that influence storm impacts on dunes. Impacts of storms on dune erosion are a function of storm characteristics and structural features of dunes. Dune er osion occurs when storm surge and waves repeatedly narrow a dune face, causing irregula r slumping of sediment and an eventual breach, or when overtopping by storm surge co mpletely overwashes a dune and pushes sediment landward (Hesp 2002; Judge et al. 2003). Although severi ty and length of a storm influence dune erosion (Kriebel et al. 1997; Sallenger 2000), key structural features

PAGE 17

6 of dunes (e.g., height, width) also provide pr otection against dune er osion (Judge et al. 2003). Laboratory research and numerical models of dune erosion are extensive (Vellinga 1982), but few studies have evalua ted importance of dune structure in stormrelated erosion in the field (but see Judge et al. 2003). Additionall y, past evaluations of dune erosion often have been limited to fore dune structures (i.e., dunes nearest to the high tide line). Coastal foredunes are formed from aeolia n processes with dune development occurring where sediment is trapped by ve getation (Hesp 2004; Psuty 2005). Secondary dunes generally are found landward of foredunes and develop from sediment originating on foredunes or they may be relict foredunes that are no longer c ontrolled by aeolian processes (Hesp 2004). Foredunes are differentia ted as either incipi ent or established. Incipient foredunes are low-lying developi ng dunes associated with pioneer plant communities. Established foredunes evolve from incipient dunes and are distinguished by presence of an intermediate plant comm unity, including woody species. These dunes have greater height and widt h than incipient dunes (Hesp 2002). Although the location and development of incipient dunes may change annually, development of large established foredunes takes decades, and these dunes remain in a relatively fixed position unless removed by storms or anthropogenic dist urbance. Evolution and maintenance of established foredunes is not determined solely by sediment flows but rather by a suite of additional factors like vegetation density a nd the frequency of wave and wind forces (Hesp 2004). Established foredunes and secondary dunes, by way of their size, should provide greater resistance to increased tide levels and storm events than incipient dunes.

PAGE 18

7 However, storm surge and waves associated with hurricanes of cat egory 3 or above on the Saffir-Sampson scale can cause even large (> 3 m tall) established foredunes to return to a more erosional form or to be destroye d (Hesp 2002). Effects of strong hurricanes on secondary dunes are less well documented. Impact of storm surge on secondary dunes may be less severe as these structures ar e no longer governed by sand exchange, storm tides or wave activity associated with fo redune development (Hesp 2004). Additionally, as a result of their spatial location behind wave-absorbing foredunes, dune erosion from storm events may be lower for secondary dunes. I assessed dune erosion along a barrier island ecosystem in the Gulf of Mexico after Hurricane Ivan. The objectives of this study were to examin e impacts of Hurricane Ivan on established foredune and secondary dunes and to evaluate structural features of dunes as predictors of dune vulnerability for th ese two dune types. I also examined dune erosion as a function of the landscape cont ext of the dune, includi ng island width at the location of the dune, distance to neighboring dun es and distance of the dune from the position where Hurricane Ivan pass ed over the island. Identifica tion of structural features that allow dunes to resist storm-related erosi on and evaluation of lands cape attributes that influence erosion are important for future ma nipulation of coastal dunes in a restoration context. Methods Study Area The study was conducted on Santa Rosa Island, which is a barrier island approximately 60 km long and 0.5 km wide, in the Gulf of Mexico. The study site is located on property owned and managed by Eglin Air Force Base (30' N, 81' W). This portion of the island is approximatel y 20 km long and includes the islands entire

PAGE 19

8 width (Fig. 2-1). This area contains severa l military structures and a paved road for military traffic but otherwise is undeveloped. A thorough description of Santa Rosa Islands geomorphology can be found in Stone et al. 2004. Foredunes are found near the high tide line and, in the absence of hurricane activity, can run con tinuously the length of the isla nd. Prior to Hurricane Opal (1995), mean dune height was 3.8 m (Stone et al. 2004). These dunes are dominated by sea oats ( Uniola paniculata ), cakile ( Cakile spp.), beach morning glory ( Ipomoea imperati) and seashore elder ( Iva imbricata ) but various woody spec ies can be present on foredunes in the absence of frequent distur bance. Secondary dune s are located behind foredunes on the bayside of th e island. Woody species domin ate these dunes, including false rosemary ( Ceratiola ericodes ), woody goldenrod ( Chrysoma pauciflosculosa ), scrubby oaks ( Quercus geminata ) and sand pine (Pinus clausa ). Between these two types of dunes is grassland dominated by maritime bluestem ( Schizachrium maritimum ) and bitter panic grass ( Panicum amarum ), interspersed with de nsely vegetated ephemeral wetlands. Characteristics of Hurricane Ivan Hurricane Ivan made landfall as a category 3 hurricane on 16 September 2004, west of Gulf Shores, Alabama, and approximately 100 km west of our study site. Storm surge from the hurricane was estimated at 3 4.5 m from Mobile, AL to Destin, FL (Stewart 2005), which encompassed all of Santa Rosa Island. Ivan was the most destructive hurricane to make landfall along the Gulf coast in 100 years with a majority of damage resulting from wave action a ssociated with unusually high storm surge (Stewart 2005).

PAGE 20

9 Dune Mapping Established dunes (foredunes, N = 93, and secondary dunes, N = 484) were delineated in the field after Hurricane Opal (1995). Because established dunes change very slowly over time, except when they ar e impacted by storms, these data could be used as a baseline for dune stru cture prior to Hurricane Iva n. Dunes were mapped again after Hurricane Ivan (2004). Geographic location of dune peri meters were recorded with a TRIMBLE GPS unit in UTMs (Universal Tranverse Mercator) and differentially corrected for < 1 m accuracy. Dunes were included if they were greater than 1.0-m high with woody vegetation or greater than 1.5-m high with grasses or other herbaceous vegetation. Dunes were consider ed distinct if they were separated by more than 3.0 m of sand. Dune height (m) was measured ever y 15 m along the long axis of each dune using a telescoping pole. Dune perimeters we re incorporated into ArcView 3.2 (ESRI 1996) and the following variables were calculated: dune ar ea (ha), dune width (perpendicular to the shoreline), length (parallel to the shor eline), and distance of each dune from the position where Hurricane Ivan made landfa ll. Coordinates for the position where Hurricane Ivan made landfall were obtained from the National Oceanic and Atmospheric Association (Stewart 2005). Aerial photographs taken in 1995 were overlaid on dune location in ArcView 3.2 to calculate island widt h at each dune location. I also recorded presence or absence of foredunes located se award of secondary dunes before Hurricane Ivan. Gap distance for each dune was calculate d as the average of the distance between the closest dunes located immediately to the west and east of the target dune. After Hurricane Ivan all remaining foredune s (N = 26) were remapped or recorded as completely destroyed (100% loss) if not found during remapping (N = 67). A random subset of small secondary dunes (< 0.25 ha, N = 34) were remapped after Hurricane Ivan.

PAGE 21

10 All large secondary dunes ( 0.25 ha, N = 61) were remapped. The percentage of each foredune or secondary dune lost from Hurri cane Ivan was calculat ed by subtracting the dunes area after Hurric ane Ivan from the post-Opal dune area and by dividing this value by the post-Opal dune area. Statistical Analyses For statistical analysis, I used data fro m all foredunes and all secondary dunes prior to Hurricane Ivan, and I used all fore dunes and a subset of secondary dunes sampled after the hurricane. Because all sma ll secondary dunes were not remapped after Hurricane Ivan, I determined the proportion of the landscape occ upied by large dunes ( 0.25 ha) and small dunes (< 0.25 ha) prior to Hurr icane Ivan. I used these proportions to determine the sample size for large and sma ll dunes in analyses. The total area of scrub dunes prior to Hurricane Ivan was 131.55 ha with large dunes making up 109.99 ha (83.6%) of this total. To ma intain the proportional area of the two dune types, I used the 34 small dunes randomly chosen for remapping after Hurricane Ivan and I randomly selected 18 large dunes from the larger pool we mapped. I used Pearson correlation coefficients to examine relationships among structural variables for dunes, spatial location, and isla nd width for all frontal dunes and secondary dunes. Variables were examined for normality prior to examining correlations between variables. For frontal and secondary dune s, data on percentage of dune loss were transformed using arcsine transformation, and dune area and dune height prior to Hurricane Ivan and dune area after Hurricane Ivan were transformed using logtransformation (Zar 1998). I used t-tests to examine diff erences in dune area and dune height between dune types before Ivan. Univar iate linear regression initially was used to identify structural or spatial variables that were important pr edictors of the percentage of

PAGE 22

11 dune erosion after Hurricane Ivan for foredune s and secondary dunes, and I used logistic regression to evaluate the importance of presence of foredunes on dune erosion in secondary dunes. Changes in dune area of frontal and secondary dunes with the impacts of Hurricane Ivan were examined with paired t-tests. All univariate tests were conducted in SPSS version 13.0 (SPSS Inc., 2004) and I rejected null hypotheses of no influence on dune loss when p < 0.05. Traditional multiple regression technique s may not work well when variables do not meet parametric assumptions or when re lationships between variables are complex or non-linear (Bourg et al. 2005). I wanted to simultaneously assess the influence of all predictor variables on dune erosion from Hurricane Ivan and examine relationships between physical features of dunes and sp atial location on dune erosion. I used classification trees to assess how multiple pr edictor variables explained the impacts of Hurricane Ivan on dune structure. This non-para metric approach spl its the dataset into smaller groupings with relatively homogeneous values of response va riables (Breiman et al. 1984). An advantage of classification trees is that they are simp le to create, provide intuitive descriptions of complex relationshi ps, and explain variance in a dataset in a manner similar to multiple regression or an alysis of variance procedures (Death and Fabricius 2000). I used the RPART package in R (R De velopment Core Team, 2003) to build and evaluate classification trees. Trees for foredunes were construc ted with the percentage of dune lost as the response vari able and the following predictor variables: dune area (ha), dune width (m), dune height (m), island wi dth (km), gap distance (m), distance from where the eye of Hurricane Ivan made landfa ll (km). All data except distance from the

PAGE 23

12 eye of the hurricane were from measurements made prior to the hur ricane. Classification trees for secondary dunes used the same res ponse and predictor variables as trees for foredunes, but also included presence or ab sence of a foredune before Ivan. I used a cross-validation procedure to evaluate the rate of misclassification as a function of tree size (e.g., number of groupings) to select trees that were not over-fit (Breiman et al. 1984). Results Conditions before Hurricane Ivan Before Hurricane Ivan, small dunes (< 0.25 ha) were numerous comprising 80 of the 93 (86.1%) foredunes and 403 of the 484 (83. 3%) secondary dunes. Secondary dunes were larger in area but not ta ller than foredunes prior to Hu rricane Ivan (dune area t = 2.265, df = 575, p = 0.03; dune height t = 1.020, df = 575, p > 0.10; Table 2-1). Foredune area was correlated highly with dune height (r = 0.63, p < 0.01), but was not correlated with west-east location as might be expected because Hurricane Opal made landfall west of the study s ite (r = 0.06, p > 0.5). Correlations only considering foredunes 0.25 ha or larger also indicated no relati onship between dune area and position on the islands landscape (r = -0.05, p > 0.5). Similarly, dune area for secondary dunes was correlated with dune height (r = 0.65, p < 0.01) and not correlated w ith west-east location on the landscape (r = 0.02, p = 0.69). Hurricane Ivans Impact on Foredunes and Secondary Dunes Dune area for foredunes and secondar y dunes was reduced significantly by Hurricane Ivan (foredunes, t = 5.160, df = 92, p < 0.01; s econdary dunes, t = 3.267, df = 51, p < 0.01, Table 2-1). Hurricane Ivans storm surge physically removed 76.8% of the foredune area. Of the orig inal 93 foredunes measured, 67 were destroyed completely.

PAGE 24

13 The 34 small secondary dunes sampled lost 42.1% of their total area. The 61 large dunes lost 14.8% of total area. Based on the proporti on of the area occupi ed by small and large dunes on the pre-Ivan landscape, the total es timated loss of secondary dune area with Hurricane Ivan is 19.3%. Reduction in dune ar ea to these dunes was significantly less than to foredunes (t = -9.953, df = 143, p < 0.01). Univariate analyses of the relationshi p between dune structure, dune location, island width, and dune loss for foredunes indicated that all variables except gap distance were related significantly to dune loss (Table 2-2). However, many of these variables were highly correlated making th ese tests difficult to interpre t (Appendix E). In contrast, for secondary dunes only dune height, dune width, dune length, and the presence of a foredune were related to dune loss. Distan ce from the eye of Hurricane Ivan was an important predictor of dune loss in foredunes, but not in secondary dunes (Table 2-2). Regression Trees Cross validation indicated the smallest classification trees to fit data from foredunes and secondary dunes without an increa se in misclassificati on error rate each had 5 branches (Fig. 2-2). Regression tr ees for foredunes and secondary dunes indicated that a different set of structural features were linked to dune erosion for dunes on the oceanfront and bayside of the island (Fig. 2-3). Percent of dune lost in foredunes after Ivan was related to dune structure (e.g., height and width) and the dist ance from where the eye of Ivan made landfall (Fig. 2-3a). The amount of variance (R2) in dune erosion explained by the classification tree was 78.9%. The regression tree for foredunes indicat ed that structural features influencing dune erosion in foredunes changed with dist ance of the dune from the location where Ivan passed over the island (Fig. 2-3a).

PAGE 25

14 The regression tree for secondary dunes sa mpled to represent proportional area of small and large dunes on the landscape indi cated that dune erosi on of secondary dunes was related to structural features of th e dune, their position on the landscape, and the presence of foredunes (Fig. 2-3b). This tree explained 76.3% of the variance in dune erosion for secondary dunes. The tree fi rst divided dunes by the width of a dune. For wider dunes, the presence or absence of a foredune was important in determining dune erosion. Dune erosion was lowest where fore dunes were present. Where foredunes were absent, dune erosion increased as distance from where Hurricane Ivan made landfall increased. For narrow secondary dunes, is land width was the only important factor influencing dune erosion. Dune erosion was greater where the island was wide. Island width and distance from where Hurricane Iv an made landfall ar e correlated (r = 0.46) and, thus, may provide some of the same information (Appendix F). When large (N = 61) and small (N = 34) secondary dunes were analyzed separately with regression trees, results were easier to interpret. The presence of foredunes reduced dune erosion for large secondary dunes and th is was the only important variable (Fig. 24a). However, this tree explained only 19.7% of the variance in dune erosion for large secondary dunes. Erosion of small seconda ry dunes was lowest for dunes nearer to where Hurricane Ivan made landfall, and no othe r variable appeared to be important in predicting dune erosion for thes e dunes (Fig. 2-4b). Dune length and area were related negatively to dune erosion for dunes at grea ter distances from where Hurricane Ivan made landfall. This tree explained 76.6% of the varian ce in dune erosion for small secondary dunes.

PAGE 26

15 Discussion My field study and mechanistic rese arch in the laboratory (Vellinga 1982) indicate that dune structure plays an important role in resistance of dunes to storm damage. In addition, this study clearly demons trates the influence of landscape context of dunes on their vulnerability to dune erosion, including spat ial location relative to a hurricanes eye and presence of other dune stru ctures. Identificati on of features that promote resistance to storm-related erosion can aid agencies in the classification of coastal areas that are especially vulnerable to future storm events. This information also can assist in defining targ ets for coastal restoration. Larger dunes on Santa Rosa Island expe rienced less erosion than smaller dunes from Hurricane Ivan. However, the importan ce of location of the dune on the landscape, and the specific structural features of dune s important in describing amount of erosion were different for foredunes and secondary dune s, suggesting the processes that act upon dunes during storms are different depending on distance from the shoreline. Much of the erosion for foredunes probably is a result of st orm surge. Foredunes that remained after Hurricane Ivan showed signs of sediment slumping, dead or uprooted vegetation, and blowouts; all of which are common effects of storm surge, wave action, and overwash. Under these conditions height of a dune is like ly to play a key role in resistance of dunes to erosion, as demonstrated by the importance of this variable in our regression trees for foredunes. Along exposed oceanfront beaches, the magnitude of storm surge and wave action decreases with distance from the edge of the hurricanes eyewall and damage to foredunes follows a similar spatial pattern. Secondary dunes on the isla nd that experienced erosi on lost sediment along dune edges from passing storm surge and not from continual wave action. Dune erosion for

PAGE 27

16 secondary dunes likely is influenced by stor m surge from the Gulf of Mexico merging with rising water levels in the Santa Rosa Sound, located behind the island. The presence of foredunes substantially reduces erosi on of secondary dunes. This observation reinforces the importance of foredunes as bu ffers of storm surge for coastal features located further landward and for prot ection of human-made structures. One non-intuitive result of this research is that small secondary dunes that were father from the eye of the hurricane and that were on the widest part of the island were subject to more erosion that small secondary dunes nearer to the eye of Hurricane Ivan and on the narrower part of the island. Storm damage on small secondary dunes increased from west to east. The island wide ned from west to east, and the Santa Rosa Sound narrowed as the island widened. When th e flow of storm surge is confined and water is shallow, high penetration distances have been observed for washover (Morton and Sallenger 2003). I hypothesize that as th e sound became narrower, the magnitude of storm surge on the bayside of the island in creased and smaller secondary dunes were impacted more strongly, resu lting in an inverse relations hip between storm damage and distance from the hurricane and an inverse relationship between storm damage and island width. Previous research on coastal dune systems ha s suggested that dune systems exist in two opposing states: one where dune structur e and vegetation comm unities are arranged by environmental gradients generated from normal wind and wave activity and one dominated by periodic but high levels of di sturbance (e.g., hurricane s and tropical storms; Synder and Boss 2002; Stallins and Parker 2003). I believe that coastal dunes on Santa Rosa Island are beginning to shift towards the latter state, though assessment of dune

PAGE 28

17 erosion after additional storms is needed to evaluate this statement. Prior to hurricane Opal in 1995, the islands s horeline contained continuous fo redunes (Stone et al. 2004). Repeated hurricane activity has eroded or destroyed many foredune structures. My analysis indicates that foredunes are importa nt in protecting sec ondary dunes and thus, further storm impacts may begin to aff ect secondary dunes more severely. The consequences of this changing coastal la ndscape are large for maintenance of humandeveloped infrastructure, success of restorat ion projects, and conservation of wildlife species that depend on coastal dune habitat.

PAGE 29

18 Table 2-1. Means and standard errors for st ructural variables meas ured to explain dune erosion in foredunes and secondary dunes on Santa Rosa Island from Hurricane Ivan. Differences between means before and after Hurricane Ivan were compared using paired-t tests adju sted with Bonferr onis correction for multiple tests ** p < 0.05. Dunes were sub-sampled to represent proportional ar ea of large (83.6%) and small (16.4%) dunes on the landscape (N = 52 with 34 small dunes and 18 large dunes). Variable Foredunes Secondary Dunes1 Before Ivan (N = 93) After Ivan (N = 93) Before Ivan (N = 52) After Ivan (N = 52) Dune area (ha) 0.14 (0.03)** 0.05 (0.01) 0.37 (0.09)** 0.26 (0.07) Dune height (m) 2.84 (0.15)** 0.89 (0.16) 3.28 (0.18)** 2.81 (0.15) Dune length (m) 42.2 (5.8)** 12.1 (2.9) 80.6 (14.5)** 64.3 (11.3) Dune width (m) 29.9 (2.6)** 10.7 (2.2) 56.6 (6.2)** 44.3 (6.4) Gap distance (m) 95.6 (25.9) NA 50.8 (6.8) NA

PAGE 30

19 Table 2-2. Statistics for eval uation of dune character istics as predictors of dune loss as a result of Hurricane Ivan. Variables we re assessed along a 20-km stretch of Santa Rosa Island, FL prior to the hurri cane. The significance of a variables role as a predictor of dune loss wa s evaluated using univariate linear regression or a two sample t-test when examining the presence of absence of a foredune before a secondary dune. Hypothe ses of no association were rejected at p < 0.05. Variable t p-value R2 Foredunes (N = 93) Dune area (ha) 5.21 <0.01 0.21 Dune height (m) 5.43 <0.01 0.25 Dune length (m) -2.08 0.04 0.05 Dune width (m) -5.03 <0.01 0.22 Gap distance (m) 0.40 0.68 0.01 Island width (km) 3.08 <0.01 0.09 Distance from Ivans eye (km) 6.09 <0.01 0.25 Secondary Dunes1 (N = 52) Dune area (ha) -1.65 0.10 0.05 Dune height (m) -2.39 0.02 0.10 Dune length (m) -2.35 0.02 0.10 Dune width (m) -3.50 0.01 0.20 Gap distance (m) 0.78 0.40 0.02 Island width (km) 1.05 0.30 0.02 Distance from Ivans eye (km) -0.24 0.81 0.01 Presence / absence of foredune 2.47 0.02 NA Dunes sub-sampled to represent proportiona l area of large (83.6%) and small (16.4%) dunes on the landscape (N = 52 with 34 small dunes and 18 large dunes).

PAGE 31

20 Fig. 2-1. Map of Santa Rosa Island, FL The study area encompasses the section between Navarre and Fort Walton Beach.

PAGE 32

21 a) b) Fig 2-2. Cross validation relative error for regression trees for (a) foredunes and (b) secondary dunes to explain dune loss from Hurricane Ivan in relation to measured predictor variables. I used the 1-SE rule (Breiman et al. 1984) to identify regression trees that had the sm allest number of branches but were closest to the overall minimum misclassi fication error (dotted line). Arrows point to the best sized regression tr ee for each dune type. The complexity parameter, cp, represents a balance be tween the complexity of a tree (i.e., more branches) and the costs of utilizing a simpler tree.

PAGE 33

22 a) b) Fig. 2-3. Regression trees relating percenta ge of dune lost from Hurricane Ivan for (a) foredunes (N = 93) and (b) secondary dune s (N = 52) to physical features of dunes, spatial location of dunes with respect to where Hurricane Ivan made landfall, and width of island. Data for secondary dunes are based on sampling of small (<0.25 ha) and large dunes ( 0.25 ha) according to their proportional area on the landscape. Numbers at the e nds of terminal nodes are the average percentage of dune lost for all observati ons in that group. N is the number of observations within that group.

PAGE 34

23 a) b) Fig. 2-4. Regression trees relating percenta ge of dune lost from Hurricane Ivan for (a) secondary dunes 0.25 ha (N = 61) and (b) secondary dunes <0.25 ha (N = 34) to dune features, spatia l location, and island widt h. Numbers at the ends of terminal nodes are the av erage percentage of dune lost for all observations in that group. N is the number of observations within that group.

PAGE 35

24 CHAPTER 3 INFLUENCE OF HABITAT AND LANDSCAPE FEATURES ON SPATIAL DISTRIBUTION OF SANTA ROSA BE ACH MICE IN TWO DUNE HABITATS BEFORE AND AFTER A HURRICANE Introduction Identification and protection of habitat ar e critical for species conservation and an integral part of many conservation programs. GAP analysis, for example, utilizes information on land cover and predicted species di stributions to identify habitats that are poorly represented in reserves (Flather et al. 1997). Habitat su itability index (HSI) models establish relationships between a spec ies distribution and habitat variables to create an index of suitable habitat that can be used to evaluate suitability of other areas for habitat management or protection (UWF WS 1981). To aid in species recovery, the Endangered Species Act provides for designa tion of critical habitat, which is the geographic area that contains physical and biological featur es necessary for conservation of the species (16 U.S.C. 1531 et seq.). Ho wever, the effectiveness of critical habitat designation is controversial as critical habitat often is defi ned using limited data or only anecdotes, or found to be not determinable (Hoe kstra et al. 2002; Tayl or et al. 2005). Habitat models are a common tool used to examine the role of habitat features in explaining the spatial distribution, density, or diversity of species occurring on a landscape and these models aid in the develo pment of conservation strategies (Segurado and Araujo 2004; Guisan and Thuiller 2005). In corporation of spatial structure of habitat (e.g., size, shape and spatial distribution of habitat patches) along with traditional assessments of habitat quality has improve d the functionality of models (Cox and

PAGE 36

25 Engstrom 2001). However, one problem with application of habitat models to conservation problems is that their conc lusions are based upon a limited range of conditions because they generally are deve loped over short time periods (Pearce and Ferrier 2000). Habitat availability and quality can shift rapidly with stochastic events (VanHorne et al. 1997; Carlsson and Kindvall 2001), and key featur es that determine spatial distribution or population density could change. Designation of protected habitats often does not consid er impacts of environmental stochasticity on distribution and persistence of species even though dist urbance may influence habitat turnover and ultimately impact population persistence thr ough impacts on habitat (Oli et al. 2001; Jonzen et al. 2004; Frank 2005; Schrott et al. 2005). We demo nstrate this issue with an analysis of beach mouse habitat along the Gulf Coast of Florida before and after Hurricane Ivan which made landfall in September 2004. Coastal dunes are among the most dynamic and threatened habitats world-wide (Martinez et al. 2005). Gulf Coast populations of beach mice comprise 5 subspecies, all of which are subject to extreme stochastic events in the form of tropical storms and hurricanes (e.g., Hurricanes Opal, 1995; Iva n, 2004; Dennis, 2005). Four of these subspecies are listed as threatened or endangered (Potter 1985 ; Milio 1998). The remaining subspecies, the Santa Rosa beach mouse ( Peromyscus polionotus leucocephalus ), is not yet listed beca use its geographic range includes several federally managed lands (Gore and Schaffer 1993). All su bspecies suffer from severe habitat loss from destruction of coastal sand dunes by de velopment, and this habitat loss is exacerbated greatly by hurricanes (Swilling et al. 1998). Optimal habitat for beach mice is believed to be frontal dune habitat with sparse vege tative cover of sea oats ( Uniola

PAGE 37

26 paniculata ) adjacent to the high tide line (USF WS 1987). Mice also occur in scrub dunes, which are located farther from the beach and are characterized by increased dominance of woody vegetation. These dune s provide refugia for mice during and immediately after storms but are viewed as marginal habitat because of lower population density in this habitat (Swilling et al. 1998). Three subspecies of beach mice on the Gulf Coast are covered under the same federal recovery plan. This plan calls for protection of dune communities within 152 m (500ft) of the high tide line and includes all frontal dunes but excludes scrub habitat in most ar eas (Potter 1985; Swilling et al. 1998). The St. Andrews beach mouse ( P. p. peninsularis ) is protected under it s own recovery plan, which states designation of cr itical habitat is not necessary for conservation of this species (Milio 1998). I examined impacts of Hurricane Ivan on the structure of fron tal and scrub dunes, compared occupancy patterns of beach mice in these two habitats, and determined how these occupancy patterns changed after the hurri cane. I also developed habitat models for predicting dune occupancy by Santa Rosa b each mice and evaluated whether the factors that influenced patterns of habitat occupanc y were similar for frontal (optimal) and scrub (marginal) habitats. I examined whether pr edictors of habitat o ccupancy changed after the hurricane. My study demonstrates pr oblems associated with narrowly defining critical habitat as optimal habitat, part icularly in systems characterized by high stochasticity. Methods Study Area and Habitat Mapping The study was conducted on Santa Rosa Isla nd, a barrier island approximately 46km long and 0.5-km wide, located in the Gulf of Mexico near Fort Walton Beach, FL

PAGE 38

27 (30' N, 81' W). My study area incorpor ated a 15-km section of the island on Eglin Air Force Base (EAFB) and a 10-km secti on of the island on Gulf Island National Sea Shore (GINS). Dune habitat was similar in these two areas and bot h sections contain a single paved road and only a few structures. Frontal dunes were orient ed parallel to high tide line and were dominated by sea oats ( Uniola paniculata ), cakile ( Cakile spp.), beach morning glory ( Ipomoea imperati) and beach elder ( Iva imbricata) and various woody species in the absence of frequent disturba nce. Scrub dunes were located on the bayside of the island and woody species dominate sc rub habitat, including false rosemary ( Ceratiola ericodes ), woody goldenrod ( Chrysoma pauciflosculosa ), scrubby oaks ( Quercus geminata ) and sand pine (Pinus clausa ). The area between frontal and scrub dunes consisted of gently rolling grassla nds interspersed with densely vegetated wetlands. EAFB dunes were mapped in the field befo re and after Hurricane Ivan by recording their perimeters using a TRIMBLE GPS unit a nd then differentially corrected for < 1 m accuracy. GINS dunes were mapped only after the hurricane. Data were incorporated into a cover layer in ArcView 3.2 (ESRI 1996). Dune Occupancy I surveyed for presence of beach mice in all frontal (N = 15) and scrub (N = 61) dunes equal to or larger than 0.25 ha on EAFB before Hurricane Ivan (June September 2004) and after the hurricane (October 2004 December 2004). Frontal dunes (N = 15) on GINS also were surveyed for beac h mice after Hurricane Ivan (December 2004 February 2005). Presence of beach mice in each dune was determined with tracking tubes that register footprints of mice that enter the tube. Tracking with tubes is less dependent upon weather and less labor intens ive than live trapping and, therefore,

PAGE 39

28 particularly useful for large scale survey s of distribution (Mab ee 1998; Glennon et al. 2002). Tracking tubes were construc ted with PVC pipe (33-cm long x 5-cm diameter) and elevated 5-7 cm off the ground to prevent access by ghost crabs ( Ocypode quadrata ). Dowels placed at either end of the tube allowe d mice, but not crabs, to climb to the tube. A paper liner was inserted into the bottom of each tube, and the tube was baited in the middle with rolled oats. Felt inkpads located at each end of the pa per liner were coated with a 2:1 mineral oil and carbon power solution (Mabee 1998). Hispid cotton rats ( Sigmodon hispidus ) leave footprints that are substantia lly larger than footprints of Santa Rosa beach mice. No other small rodents occur on undeveloped portions of the island (Gore and Schaffer 1993). Dunes less than 0.50 ha received eight tubes; dunes > 0.50 ha < 2.00 ha received 16 tubes; and dunes greater than > 2.00 ha received 32 tubes. Tracking tubes were placed at 15-m intervals along transects that bega n and ended at the dunes boundary and ran parallel to the long axis of the dune. The star ting point for the first transect was selected randomly and, when more than one transect was needed, parallel transects were established 15 m apart. During each tracking session, tracking tubes remained in a dune for five nights and were checked after each night. For many species, probability of detection during presence/absen ce surveys is less than one resulting in underestimates of occupa ncy, biased parameter estimates for habitat models, and incorrect estimates of populat ion persistence (Gu and Swihart 2004; Kry 2004). Therefore, I used recent statistical appr oaches for analysis of site occupancy that build on traditional capture-recapture methods and used repeated cen suses to calculate

PAGE 40

29 detection probability ( p ) and to estimate the proportion of sites that are occupied ( ) after accounting for detectability (MacKenzie et al 2002). To estimate detection probability within each habitat type, I re-sampled a random subset of scrub dunes (N = 30) with tracking tubes three times after initial prestorm surveys, and I resurveyed another random subset of scrub dunes (N = 30) and all frontal dunes on EAFB three times after initial post-storm surveys. Each repeat survey was conduc ted over 5 nights following the sampling protocol described above. Predictor Variables: Vegetation Cover and Landscape Structure I measured vegetation cover and dune he ight on scrub dunes before and after Hurricane Ivan. Surveys for th ese variables were not completed on frontal dunes before the hurricane hit and, therefore, data on these variables were analyzed for frontal dunes only post-hurricane. Vegetation cover was qua ntified using the line-intercept method (Bonham 1989) along three 50-m transects place d 20 m apart and perpendicular to the long axis of each dune. I recorded distances (cm) that sea oats, other herbaceous vegetation, woody vegetation, and open sand oc cupied along each transect and divided the distance for each cover class by total length of the transect to obtain percent cover for each cover class. I averaged data for the thr ee transects prior to analysis. Sea oats and many herbaceous species are important food sources for beach mice (Moyers 1996). Amount of open sand may be important fo r burrow construction and woody vegetation may stabilize dunes during storms and provides food and cover for foraging. I recorded dune height (m) by measuri ng height every 15 m al ong the long axis of each dune using a telescoping pole and then averag ed all values for each dune. Height of a dune may influence perception of dune habitat by beach mice moving through the landscape or influence the imp act of storm surge on dunes. I calculated dune area and

PAGE 41

30 amount of dune habitat surrounding each dune in ArcView 3.2 from the GIS database created from field mapping of dunes. I also calculated the dist ance to the nearest occupied dune as a measure of isolation. Habitat area may influence size of local populations (Hanski 1994). We used the BU FFER function in ArcView 3.2 to estimate the total area of dune habitat surrounding each dune at the foraging (200 m) and dispersal (1 km) scales of beach mice (Bird 2002; Swilling & Wooten 2002). The east-west coordinate (UTM) at the center of each dune was included in habitat models to examine how spatial location relative to the eye of Hurricane Ivan influenced dune occupancy by mice. The eye of the hurricane passed approxima tely 75 km west of the western end of GINS and 100 km west of the western end of EAFB. Occupancy Models I created and ranked a series of models with the program PRESENCE to identify variables that influenced distribution of beach mice in frontal and scrub habitats (MacKenzie et al. 2003). Correlations among variables were examined and correlated variables ( r > 0.60) were not included in the same model (Welch and MacMahon 2005), or if correlated variables were used in the same model, a regression was conducted for the two variables and residua ls were included in the model as an independent measure of one of the variables (C ooper & Walters 2002). Correlate d variables requiring this approach were pre-hurricane dune habitat within 1 km and pre-hurricane east-west coordinate for scrub dunes (r = 0.69), posthurricane dune habitat within1 km and posthurricane east-west coordinate for scrub dunes (r = 0.70), post-hurricane dune habitat within 200 m and post-hurricane distance to ne arest occupied dune for frontal dunes (r = 0.62), and post-hurricane dune habitat with in 1 km and post-hurricane east-west coordinate for frontal dunes (r = 0.69).

PAGE 42

31 Fifty-six candidate models were evaluate d for scrub dunes using a combination of variables measured before Hurricane Ivan and similar models were created with posthurricane data. The first eight base models included a combin ation of patch-level features (e.g., dune area, % cover of woody vegetati on, % cover of herbaceous vegetation, dune height). An additional 24 models were cr eated by adding distance to nearest occupied dune, the 200-m habitat buffer or 1-km habita t buffer to the original base models. Finally, I created another 24 models by includ ing the east-west coordinate in the base model + landscape context models. I developed 40 candidate models for frontal dunes on EAFB and GINS after Hurricane Ivan. All frontal dunes were occupi ed before Hurricane Ivan, so no model was created for this period. To reduce risk of an over-parameterized model, I restricted the total number of variables in a model to three. The first eight base models were the same as in scrub habitat (i.e., patchlevel features). An additiona l 32 models were created by including distance to nearest occupied dune, 200-m habitat buffe r, 1-km habitat buffer, or spatial coordinate to base models. I also modeled post-hurricane occupancy of frontal and scrub dunes on EAFB with pre-hurricane conditions to assess the role of prehurricane conditions on post-hurricane occupanc y. Models were created using the same procedure as described above. I used an Akaike Information Criterion (AICc) corrected for small sample bias to select the best model and rank the rema inder. I present AIC differences ( i = AICci minimum AICc), so that the best model has i = 0 (Burnham & Anderson 2002). Models with i 2 are considered competitive models. I also include Akaike weight (wi), which indicates relative lik elihood that model i is the best model. The relative importance of

PAGE 43

32 each habitat variable ( wsum) was obtained by summing wi for all models that contained this variable (Burnham & A nderson 2002). I performed model averaging to obtain parameter estimates and unconditional standard e rrors for each habitat variable of interest to reduce the bias of estimating paramete r effects from a single model (Burnham and Anderson 2002). When the confidence inte rval around a model-averaged parameter estimate is > 0, an increase in the variable significantly increases the probability of occupancy, and a value < 0 indicates that an increase in the variable decreases the probability of occupancy (Buskirk 2005; Mazero lle et al. 2005). Estimated probability of detection ( p) and overall occupancy rate ( ) also were obtained us ing this approach. Results Hurricane Impacts on Habi tat Availability at EAFB Hurricane Ivan significantly re duced mean area of both types of dunes (Table 3-1), but frontal dunes lost a much greater proportion of area. Stor m damage resulted in a loss of 68.2% of the total area of frontal dunes surveyed for beach mice, including complete destruction of four dunes. No scrub dunes we re destroyed entirely but the total area of scrub dunes surveyed for beach mice was reduced by 14.8%. Dune height also was reduced significantly for both dune types (Table 3-1). The amount of habitat within 200 m and 1km of a dune was reduced significantly for scrub dunes but not frontal dunes. However, frontal dunes alrea dy had little habitat within 200 m prior to the hurricane (Table 3-1). Dune Occupancy Beach mice were detected in 100% of frontal dunes and 72.1% of scrub dunes prior to the hurricane, and in 51.8% of front al dunes and 73.8% of scrub dunes after the hurricane. Probability of detection was high in all surveys. Site-occupancy models

PAGE 44

33 suggest that before the hurricane 75.1 5.5% (model-averaged estimate unconditional SE) of scrub dunes were occupied, with a de tection rate of 88.6 5.6%, and after the hurricane 78.6 4.9% of sites were occupi ed, with a detection rate of 90.1 3.1%. Differences in occupancy of scrub dunes be fore and after the hurricane were not significant (t = 0.5, df = 60, p > 0.10). Occ upancy in frontal dunes dropped to 59.7 5.1%, with a detection rate of 89.8 5.5% afte r the hurricane, and occupancy in frontal dunes was significantly lower than occupanc y of scrub dunes (t = 1.8, df = 42, p < 0.05). Habitat Models A combination of patch-level and landscapelevel features ranked high in models of occupancy of scrub dunes before and after the hurricane (Table 3-2 and 3-3). The strongest model for scrub dunes before the hurricane included dune area, percent woody vegetation cover, and amount of dune habita t within 200 m. No other models were competitive. After the hurricane, the same model was the strongest; however, the Akaike weight was much lower and several additional models were competitive (Table 3-3). All models with i 2 contained some combination of the variables in the best model except dune height and total herbaceous cover were included in several models. Ranking of variables based on the sum of their Akakie we ights revealed that amount of dune habitat within 200 m of scrub dunes was the most importa nt variable in expl aining probability of occupancy of scrub dunes by mice before and after the hurricane (T able 3-3), followed closely by dune area, and pe rcent woody vegetation cover. Before and after the hurricane, probability of beach mice occ upying a scrub dune increased as amount of habitat surrounding the dune w ithin 200 m increased and dune area increased (Table 3-3). Occupancy of scrub habitat by beach mice al so appeared to increase with increasing cover of woody vegetation before and after the hurricane, but this relationship was not

PAGE 45

34 statistically significant. T op models of post-hurricane occupancy in scrub dunes using pre-hurricane conditions retained the same suite of predic tor variables (Table 3-2). The strongest model for occupancy of fr ontal dunes after Hurricane Ivan included percent woody vegetation cover a nd distance to nearest occupi ed dune (Table 3-2). In contrast to scrub habitat, the amount of habitat surrounding a dune within 200 m was not a factor in any competitive models. The likelihood of occupancy increased with increasing cover of woody vegetation and this variable was the top ranked variable in models of occupancy (Table 3-3). Increasi ng distance to the nearest occupied dune also appeared to reduce the probability of occupanc y after the hurricane but this relationship was not statistically significant (Table 3-3). Dune height was the third ranked variable and an increase in dune height appears to increase occupancy by beach mice in frontal habitats but also was not st atistically significant (Table 3-3). When post-hurricane occupancy of frontal dunes on EAFB was modeled with variables related to the structural and landscape context of dunes prior to the hurricane, dune height and distance to the nearest occupied dune were the most important predictors of occupancy (Tables 3-2 and 3-3). The likelihood of occupancy of fr ontal dunes by beach mice after the hurricane increased with a greater dune height prior to the hurricane, and a gr eater distance to the nearest occupied dune prior to the hurricane decreased the likelihood of occupancy after the hurricane (Table 3-3). Models for occ upancy of frontal dunes, with pre and posthurricane habitat data, had a better fit than models of scrub dunes (Table 3-2). Discussion Frontal dunes near the hi gh tide line are subjected to major impacts during hurricanes. Prior to Hurricane Opal ( 1995), frontal dunes ran relatively continuously along the entire length of Sant a Rosa Island (Stone et al. 2004). This hurricane and

PAGE 46

35 subsequent tropical storms fr agmented frontal dunes. St orm surge from Hurricane Ivan removed close to 70% of the remaining frontal dunes. In contrast, no scrub dunes, which are located on the bay side of the island, were completely lost with Hurricane Ivan and reduction in area of scrub dunes occurred al ong dune edges from passing storm surge. Distance from the eye of the hurricane influe nced dune lost for fr ontal and scrub dunes along this portion of Santa Rosa Island (Chapt er 2). Tropical storms and hurricanes are predicted to be increasing in number and se verity (Emanuel 2005). Frontal habitat for beach mice will continue to be fragmented and removed if the interval between hurricanes and other tropical storms remain s shorter than the time required for dunes to develop. In contrast, my results suggest th at the amount and configuration of scrub dunes on this barrier island may remain relatively consistent. However, as buffering capacity provided by frontal dunes is lost, scrub dunes ma y suffer more impacts. Also, Hurricane Ivan was a category 3 hurricane; stronger hur ricanes could have greater impacts. Predictors of occupancy for beach mice in frontal habitat after the hurricane were closely tied to local habitat features (e .g., percent cover of woody vegetation and dune height) and proximity to other occupied dunes. Optimal beach mouse habitat generally is described as tall frontal dune s vegetated by sea oats and ot her herbaceous plants (Holler 1992). My habitat model indicates that woody ve getation cover also is important to mice, at least during hurricane cycles. Foraging experiments demonstrate that mice consume more seeds under vegetation cover than in the open (Bird 2002). Woody plants provide cover for foraging, serve as a food source for mice, and also may promote dune stability during storms (Moyers 1996; Musila et al. 2001 ). Similarly, dune height may be an important factor in dune stabil ity, particularly in preventing overwash by storm surge.

PAGE 47

36 Beach mice also are semi-fossorial and an increase in dune height may facilitate conditions appropriate for burrow construc tion. For frontal dunes on EAFB where the impact was severe, dune height prior to th e hurricane was a significant predictor of posthurricane occupancy, but after the hurricane th e importance of this variable was not as clear. Isolation explained post-hur ricane occupancy of frontal habitat, whether modeled with preor post-hurricane habi tat conditions. This observation likely reflects the history of disturbance and loss of fr ontal habitat on this island. Beach mice occupying frontal dunes prior to Hurricane Opal experienced a fairly continuous ha bitat where habitat quality might have been determined largel y by resource availabil ity or appropriate burrow conditions. The current fragmented frontal dunes are too small to support separate populations of beach mice, but ra ther may serve as resource patches for mice moving among dunes. The most important predictors of occ upancy before and after the hurricane for beach mice in scrub habitat were landscape f eatures related to habitat amount (i.e., dune area and amount of surrounding hab itat). Predictors of occupancy in scrub habitat were similar before and after the hurricane, presum ably because the impact of the hurricane on the structure of these dunes was minimal. Woody vegetation also may play a role in occupancy of scrub dunes by beach mice, but this relationship is not as clear as in frontal dunes. The amount of dune habitat surroundi ng scrub dunes is greater than for frontal dunes, which may indicate less isolation for these dunes. The importance of surrounding dune habitat for occupancy of scrub may reflect reduced habitat quality in scrub dunes. Alabama beach mice travel further distances to forage in scrub habitat than in frontal

PAGE 48

37 habitat during the winter and spring (Sn eckenberger 2002). If habitat quality and population density are lower in scrub dunes, la rger areas may be required to maintain mouse populations. Dune restoration after hurricanes primarily has focused on re-establishment of sea oats, which produces a lattice of rhizomes that accumulate sand and also is an important food plant for beach mice. My results suggest that restoration programs for frontal dunes also should include re-establishment of woody plants and promote increases in dune height. Beach mice also should benefit from re storation programs that reduce isolation of frontal dunes. The results of my study sugge st that optimal habitat fo r beach mice differs under different environmental condi tions. Lower occupancy of scrub habitat than frontal habitat by beach mice prior to the hurricane, and documentation of lower density in scrub habitat from other studies, sugge st that scrub habitat could be lower quality than frontal dunes under pre-hurricane conditions, though dens ity is not always a good indicator of habitat quality (Van Horne 1983). However, pers istence of scrub habi tat and maintenance of occupancy levels by beach mice through the hurricane in this habitat versus the severe loss of habitat and significant reduction in occ upancy of frontal habi tat suggest scrub is an essential habitat. Scr ub was an important refugia habitat for Alabama beach mouse populations during Hurricane Opal and a sour ce of dispersing i ndividuals after the hurricane (Swilling et al. 1998). Re-colonization of frontal habitats by beach mice after Hurricane Opal occurred within nine months (Swilling et al. 1998). I observed beach mouse tracks on previously unoccupied front al dunes in March 2005, approximately six months after Hurricane Ivan. These mice may have dispersed from scrub or neighboring

PAGE 49

38 frontal dunes. Given the inevitable loss of frontal dune s with hurricanes, incorporation of scrub habitat into conservation efforts for Gu lf Coast beach mice is warranted to ensure long-term population persistence. Scrub habita t, even as marginal habitat, will improve population persistence and lessen extinction risk as frontal habitat is further removed. The role of stochasticity and uncerta inty in management outcomes has been explored extensively with respect to impact s on population size and persistence of species of economic or conservation concern (Ellner and Fieberg 2003). Our study demonstrates the need to incorporate these factors in habitat planning and protection. Habitat availability for species in dynamic landscapes can change quickly and additional habitats may become critically important after st ochastic events (Car lsson and Kindvall 2001; Biedermann 2004). When the dynamics of la ndscape and population are not understood, protection of habitat should follow a conser vative approach. Failure to consider and protect habitats required under different environmental conditions may exacerbate the impacts of habitat loss and change on extinction risk.

PAGE 50

39 Table 3-1. Means and standard errors for st ructural and vegetation variables measured for modeling occupancy of frontal a nd scrub habitat by Santa Rosa beach mice on Eglin Air Force Base (EAFB) and Gulf Islands National Seashore (GINS) on Santa Rosa Island, FL. Variab les were assessed before and after Hurricane Ivan made landfall on 18 September 2005. Differences between means before and after Ivan were compared for dunes on EAFB using paired t-tests adjusted with Bonfferonis co rrection for multiple tests. ** p < 0.05. Scrub Dunes Mean ( SE) Frontal Dunes Mean ( SE) Variable1 EAFB before hurricane (N = 61) EAFB after hurricane (N=61) EAFB before hurricane (N=15) EAFB after hurricane (N=11) GINS after hurricane (N=15) Dune area (ha) 1.82 (0.38) 1.55 (0.37) ** 0.59 (0.09) 0.26 (0.06) ** 0.15 (0.03) Dune height (m) 4.64 (0.3) 3.32 (0.17 ) ** 4.01 (0.33) 3.13 (0.35) ** 3.24 (0.20) Dune habitat within 200 m (ha) 2.21 (0.28) 1.71 (0.19) ** 0.29 (.09) 0.22 (0.08) 0.24 (0.05) Dune habitat within 1 km (ha) 12.73 (1.19) 11.29 (1.03) ** 8.45 (1.48) 8.67 (1.44) 1.94 (0.29) Distance to nearest occupied dune (m) 176.1 (38.3) 174.2 (37.6) 219.2 (54.7) 161.9 (25.9)2 126.2 (39.4) Percent woody cover 19.6 (1.6) 19.9 (1.6) no data 6.5 (1.4) 3.3 (2.1) Percent total herbaceous cover 14.1 (1.4) 7.4 (0.8) ** no data 24.5 (2.9) 19.2 (2.3) East-West Coordinate (UTM) (m) 524519 (653) 524519 (653) 522959 (1337) 524208 (879) 506117 (1709) 1 Dune perimeter was correlated highly (p < 0.01) with dune area and was omitted from analyses. 2 Distance to nearest occupied dune dropped after the hurrica ne because the four dunes that were destroyed were very isolated (mean distance to nearest oc cupied dune for those dunes = 459.5 m). Pre-hurricane mean for distance to nearest occupied dune for the 11 dunes to survive Ivan = 131.7 m.

PAGE 51

40 Table 3-2. AIC-based selecti on of site occupancy models of dune occupancy for Santa Rosa beach mice in frontal and scrub dune habitat. K = the number of explanatory variables plus 1, i = AICci minimum AICci, wi = Akaike weights. Models with i 2 are presented. Habitat and conditions Period of occupancy Location Model K i wi R2a Scrub prehurricane Pre-hurricane EAFB / N =61 Dune area, habitat within 200 m, percent woody cover 4 0.00 0.42 0.242 Scrub posthurricane Posthurricane EAFB / N = 61 Dune area, habitat within 200 m, percent woody cover 4 0.00 0.18 0.264 Dune area, habitat within 200 m 3 0.25 0.16 Dune height, percent woody cover, habitat within 200 m 4 0.50 0.14 Percent woody cover, habitat within 200 m 3 0.94 0.12 Percent total herbaceous cover, habitat within 200 m 3 1.63 0.08 Dune height, habitat within 200 m 3 1.69 0.08 Dune area, percent total herbaceous cover, habitat within 200 m 4 1.98 0.07 Scrub prehurricane Posthurricane EAFB / N =61 Dune area, habitat within 200 m, percent woody cover 4 0.00 0.26 0.287 Dune habitat within 200 m 2 0.70 0.18 Percent woody cover, dune habitat within 200 m 3 1.53 0.12 Dune area, dune habitat within 200 m 3 1.60 0.12 Dune height, dune habitat within 200 m, percent woody cover 4 1.91 0.10 Dune area, percent total herbaceous cover, dune habitat within 200 m 4 1.91 0.10 Frontal posthurricane Posthurricane EAFB; GINS / N=27 Percent woody cover, distance to nearest occupied dune 3 0.00 0.58 0.467 Frontal prehurricaneb Posthurricane EAFB / N = 15 Dune height, distance to nearest occupied dune 3 0.00 0.73 0.489 a There is no R2 analogue for patch occupancy models, instead we used a max-rescaled R2 value as an approximate measure of strength of association for the top model in each candidate set (Nagelkerke 1991). All top models provided a significantly better fit than a base model w ith no environmental predictors (p < 0.05). b This model was developed with data on the structure and landscape context of dunes and does not include vegetation variables that are in other models.

PAGE 52

41 Table 3-3. Relative importance (wsum), model-averaged parameter estimates, and unconditional standard errors for variable s used to model occupancy for beach mice in frontal and scr ub habitat before and af ter Hurricane Ivan. Wsum was estimated by summing Akakie weights (wi) of all models with a variable of interest. **Confidence interv als do not contain 0 and indicate variable significantly influences occupancy. Habitat Wsum Parameter Estimate SE 90% C.I. Scrub pre-hurricane habitat Pre-hurricane occupancy Dune area (ha) ** 0.679 0.709 0.418 0.022 1.396 Dune height (m) 0.165 0.029 0.039 -0.035 0.093 Percent cover woody vegetation 0.642 1.717 1.221 -0.292 3.726 Percent cover herbaceous 0.128 0.003 0.182 -0.296 0.302 Distance to nearest occupi ed dune (m) 0.046 0.028 0.035 -0.03 0.086 Dune habitat within 200 m (ha) ** 0.888 0.719 0.351 0.142 1.296 Dune habitat within 1 km (ha) 0.017 0.003 0.007 -0.009 0.015 East coordinate (m) 0.000 Scrub post-hurricane habitat Post-hurricane occupancy Dune area (ha) ** 0.536 0.548 0.319 0.023 1.073 Dune height (m) 0.257 0.081 0.106 -0.093 0.255 Percent cover woody vegetation 0.498 1.024 0.943 -0.527 2.575 Percent cover herbaceous 0.197 -0.106 0.714 -1.281 1.069 Distance to nearest occupi ed dune (m) 0.023 0.009 0.012 -0.011 0.029 Dune habitat within 200 m (ha) ** 0.864 1.089 0.574 0.145 2.033 Dune habitat within 1 km (ha) 0.065 0.006 0.007 -0.006 0.018 East coordinate (m) 0.000 Scrub pre-hurricane habitat Post-hurricane occupancy Dune area (ha) ** 0.524 0.516 0.298 0.026 1.006 Dune height (m) 0.162 0.036 0.045 -0.038 0.11 Percent cover woody vegetation 0.498 1.218 1.048 -0.506 2.942 Percent cover herbaceous 0.163 -0.305 0.377 -0.925 0.315 Distance to nearest occupied dune (m) 0.000 Dune habitat within 200 m (ha) ** 0.936 1.022 0.461 0.264 1.780 Dune habitat within 1 km (ha) 0.010 0.129 0.093 -0.024 0.282 East coordinate (m) 0.000 Frontal post-hurricane habitat Post-hurricane occupancy Dune area (ha) 0.141 0.383 0.629 -0.652 1.418 Dune height (m) 0.652 1.027 0.893 -0.353 2.407 Percent cover woody vegetation** 0.969 16.355 9.854 0.146 32.564 Percent cover herbaceous 0.014 0.009 0.033 -0.045 0.063 Distance to nearest occupi ed dune (m) 0.769 -2.627 1.851 -5.672 0.418 Dune habitat within 200 m (ha) 0.032 0.013 0.068 -0.106 0.132 Dune habitat within 1 km (ha) 0.048 -0.055 0.089 -0.201 0.091 East coordinate (m) 0.071 0.006 0.009 -0.009 0.021 Frontal pre-hurricane habitat Post-hurricane occupancy Dune area (ha) 0.067 0.032 0.098 -0.129 0.193 Dune height (m) ** 0.730 0.485 0.151 0.237 0.733 Distance to nearest occupied dune (m) ** 0.828 -0.015 0.008 -0.028 -0.002 Dune habitat within 200 m (h a) 0.053 0.093 0.121 -0.106 0.292 Dune habitat within 1 km (ha) 0.025 0.050 0.060 -0.048 0.148 East coordinate (m) 0.000

PAGE 53

42 CHAPTER 4 CONCLUSIONS AND CONSERVATION IMPLICATIONS Habitat loss and fragmentation from coastal development and hurricanes are believed to be major threats to the long-term population pers istence of Gulf Coast beach mice (Holler 1992; Oli et al. 2001). Frontal dun es are protected as critical habitat for Gulf Coast beach mice, but they are disturbed greatly by hurricanes (Chapter 2). Scrub dunes (also known as secondary dunes), which are not currently protected under federal recovery plans, are impacted less by hurrica nes and have been proposed to serve as important refugia habitat for beach mice dur ing hurricanes (USFWS 1987; Swilling et al. 1998). Although restoration techniques exis t to promote regeneration of physical structure of coastal dunes after storms (M iller et al 2001; 2003) understanding of the features that confer resistance against storm er osion is limited. Also, prior to this study, little quantitative information was availabl e on: 1) how hurricanes impact habitat availability for beach mice, 2) utilization of scrub habitat by beach mice, 3) habitat features that predict occupa ncy of frontal and scrub dunes by beach mice, and 4) relative impacts of hurricanes on beach mouse occupancy of frontal versus scrub dunes. My study contributes to filling these gaps. Dune Erosion and Loss of Beach Mouse Habitat Frontal dunes received much greater impacts from Hurricane Ivan than scrub dunes, and larger dunes in both frontal and sc rub habitat experienced less erosion than small dunes Structural features that conferred re sistance against storm erosion differed for frontal and scrub dunes, suggesting that different processes act upon these two dune

PAGE 54

43 types. For frontal dunes, tall and wide dune s experienced the leas t amount of erosion from Hurricane Ivans high storm surge. Dune erosion for secondary dunes was influenced by storm surge from the Gulf a nd probably also by rising water levels in the Santa Rosa Sound, located behind the island. Secondary dunes experience less erosion when located behind a frontal dune. This observation highlights the importance of maintaining frontal dunes as buffers of storm surge Small secondary dunes located farther from the eye of Hurricane Ivan a nd located on the widest parts of the island experienced more erosion than small secondary dunes closer to the eye of Hurricane Ivan and on narrow parts of the island. The reas on for this pattern is unknown, but it may be related to storm surge in the narrow parts of the Santa Rosa Sound. When the flow of storm surge is confined and water is shallow, high penetration dist ances have been noted for washover (Morton and Sallenger 2003). Hurricanes will continue to fragment and reduce coastal dunes if the interval between hurricanes and other tr opical storms remains shorte r than the time required to redevelop dunes through natura l processes or restoration. Tall and wide frontal dunes are more resistant to storm erosion than smalle r frontal dunes and may continue to provide suitable habitat for beach mice if they maintain appropriate habitat conditions. However, Hurricane Ivan alone reduced the frontal dune habitat of beach mice by 76.8% in our study area As dunes become smaller with subsequent storms, erosion of frontal dunes may accelerate. In contrast, secondary dune habitat was reduced by only 19.3% by Hurricane Ivan, indicating that, in periods of high hurricane activity, scrub dunes provide more stable habitat for beach mice than frontal dunes However, removal of frontal dunes is likely to increase impacts of hurricanes on secondary dunes as the buffering

PAGE 55

44 capacity of frontal dunes is lost. Given the inevitable lo ss of frontal dunes from hurricanes, incorporation of scrub habitat in to conservation efforts for Gulf Coast beach mice is warranted Although scrub habitat has been cons idered marginal for beach mice, my data suggest that conservation of scr ub habitat will promote population persistence and lessen extinction risk as frontal dunes continue to be removed from the landscape. Landscape-scale research is needed to unders tand the interdependency of subpopulations of mice in frontal and scrub dunes, the conditi ons under which either of these habitats is optimal or marginal, and the relative contribut ions of each of these habitats to long-term persistence of beach mice populations. Habitat Restoration for Beach Mice Dune restoration for frontal dunes after storms typically has focused on the reestablishment of sea oats, which produces a la ttice of rhizomes that can quickly trap and accumulate sand. My results indicate that cover of woody vegetation is important for promoting occupancy of frontal dunes by beach mice and woody plants also may influence occupancy of scrub dunes. This observation has important implications for conservation and management of beach mouse habitat, as optimal habitat for beach mice generally is believed to cons ist of tall frontal dunes vege tated by sea oats and other herbaceous species (Holler 1992). My data suggest that restoration programs should incorporate the re-establishment of woody plants on frontal dunes Scrub dunes are dominated by woody vegetation, and habitat management strategies for beach mice should aim to maintain this vegetation. We do not know the exact mechanism by which woody vegetation influences occupancy of dunes by mice, but woody species provide cover and food for mice and may stabilize dunes during storms. More research will be required to understand these mechanisms and to identify key woody species for mice.

PAGE 56

45 Results of my study indicate that lands cape context is important for enhancing occupancy of dune habitats by beach mice rega rdless of dune type. Isolation restricts occupancy of frontal dunes and amount of dune habitat surrounding scrub dunes influences occupancy of these dunes. As dune systems are eroded by hurricanes, dune fragments become more widely separated by open sand that does not provide resources for beach mice (e.g., food and substrate for burrow construction) and mice are forced to move over large open areas to obtain resource s in different patches. Movement of mice also is critical for recolonizat ion of the landscape in areas where mice are extirpated during hurricanes and for recolonization of rest ored habitat. These movements are likely to entail considerable risk (e.g., increased risk of predation). Management efforts should aim to minimize isolation of dunes. Restor ation techniques that provide connectivity (i.e., facilitate movement) between fragmen ted frontal dunes or between frontal and secondary dunes also may benefit beach mice. Vegetation cover faci litates foraging of beach mice (Bird et al. 2004) and, presumably, would enhance movement by reducing risks associated with moving between fragment s of habitat that re main after hurricanes. Although my occupancy data provide genera l evidence that landscape connectivity is important for beach mice, factors limiting mouse movement (e.g., the degree to which large open sand gaps restrict movement) are unk nown and this would be a fruitful area of research for understanding the long-term persistence of beach mice in dynamic landscapes Finally, restoration techniques to pr omote increases in dune height for frontal dunes also would be beneficial for beach mice as taller dunes are more resistant to storm-related erosion and may facilitate conditions appropriate for burrow construction.

PAGE 57

46 APPENDIX A DELINEATION OF DUNES IN THE FIELD Critical definition of dunes fo r delineation in the field: Dunes were mapped if greater than 1 m high with woody vegetation or greater than 1.5 m with grasses and other herbaceous vegetation. Dune spurs were considered part of a dune if the cleft between dunes was less than 1.5 m in height. Dunes were considered to be separate if they were separated by more than 3 m.

PAGE 58

47 APPENDIX B CORRELATION MATRICES FOR VARIABLES BY HABITAT Table B-1. Correlations for va riables measured on 61 scr ub dunes surveyed for beach mice before Hurricane Ivan (Jun. 2004 Sep. 2004).

PAGE 59

48 Table B-2. Correlations for va riables measured on 61 scr ub dunes surveyed for beach mice after Hurricane Ivan (Oct. 2004 Jan 2005). Table B-3. Correlations for va riables measured on foredune s (Eglin Air Force Base, N = 11, and Gulf Islands National Seashore, N = 15) surveyed for beach mice after Hurricane Ivan. (Oct. 2004 Feb. 2005).

PAGE 60

49 APPENDIX C COMPARISON OF FRONTAL DUNES AT EGLIN AIR FORCE BASE AND GULF ISLANDS NATIONAL SEASHORE Table C-1. Results of t-tests comparing ve getation, structure and landscape context for frontal dunes on Eglin Air Force Base and Gulf Islands National Seashore measured after Hurricane Ivan. Variable t df p Dune area (ha) 1.915 24 0.06 Dune height (m) -0.287 24 0.77 Dune habitat within 200 m (ha) -0.209 24 0.84 Dune habitat within 1 km (ha) 5.305 24 <0.01 % woody vegetation cover -2.260 24 0.03 % total herbaceous cover -1.060 24 0.30 Distance to nearest neighbor (m) 0.697 24 0.49 Distance to nearest scrub dune (m) -1.975 24 0.06

PAGE 61

50 APPENDIX D PREDICTORS OF CHANGE IN OCCUPANCY OF FRONTAL DUNES AFTER HURRICANE IVAN Table D-1. Mean values, standard errors, and t-test results for habitat variables on frontal dunes on EAFB that became unoccupied and for dunes that remained occupied after Hurricane Ivan. Where variances were found to not be equal (p < 0.05), a student t-test with the assu mption of unequal variances was used. For all other variables, t-statistics and p-value are for student t-tests with the assumption of equal variances. Variable1 Unoccupied Mean (SE) Occupied Mean ( SE) t df p Before hurricane Dune Area (ha) 0.56 (0.12) 0.62 (0.16) -0.31 13 0.76 East-west coordinate (UTM) 522516 (1901) 523624 (1919) -0.39 13 0.70 Dune height (m) 3.44 (0.32) 4.87 (0.53) -2.53 13 0.03 Dune habitat within 200 m (ha) 0.25 (0.13) 0.37 (0.15) -0.61 13 0.55 Dune habitat within 1km (ha) 6.59 (1.34) 11.24 (2.89) -1.63 13 0.13 Distance to nearest occupied dune (m) 317.57 (87.41) 106.61 (28.51) 2.16 13 0.05 After hurricane2 Dune Area (ha) 0.11 (0.06) 0.28 (0.07) -1.96 13 0.07 East-west coordinate (UTM) 522410 (2160) 525033 (793) -1.14 13 0.28 Dune height (m) 1.52 (0.59) 3.18 (0.53) -2.06 13 0.06 Dune habitat within 200 m (ha) 0.08 (0.03) 0.30 (0.12) -0.91 9 0.39 Dune habitat within 1 km (ha) 7.56 (1.04) 9.29 (2.21) -0.44 9 0.67 Distance to nearest occupied dune (m) 210.12 (25.46) 134.34 (35.07) 1.49 9 0.17 Percent cover of woody vegetation 2.0 (1.4) 8.5 (4.8) -3.87 9 0.01 Percent cover of total herbaceous 23.0 (4.0) 25.0 (5.0) -0.33 9 0.75 1 Habitat data are presented for variables measured before and after the hu rricane. Occupancy data presented are da ta taken after the hurricane. 2 Analysis of dune area, dune height, and east-west coordinate for frontal dunes post-Ivan included the four dunes that were completely destroyed. These four dunes, however, were not included when assessing vegetation, distance to nearest occupied dune, or amount of surrounding habitat.

PAGE 62

51 APPENDIX E CORRELATION MATRIX FOR STRUCTURAL FEATURES OF FRONTAL DUNES ON EGLIN AIR FORCE BASE Table E-1. Correlations for structural a nd landscape context variables measured on frontal dunes (N = 93) on Santa Rosa Island prior to Hurricane Ivan.

PAGE 63

52 APPENDIX F CORRELATION MATRIX FOR STRUCTURAL FEATURES OF SECONDARY DUNES ON EGLIN AIR FORCE BASE Table F-1. Correlations for structural a nd landscape context variables measured on secondary dunes on Santa Rosa Island pr ior to Hurricane Ivan. Dunes were sub-sampled to represent proportiona l area of large (83.6%) and small (16.4%) dunes on the landscape (N = 52 with 34 small dunes and 18 large dunes).

PAGE 64

53 LITERATURE CITED Biedermann, R. 2004. Modeling the spatial dynamics and persistence of the leaf beetle ( Gonioctena olivacea ) in dynamic habitats. Oikos 107:645-653. Bird, B. L. 2002. Effects of predatory risk, ve getation structure, and artificial lighting on the foraging behavior of beach mice. MS Thesis. University of Florida, Gainesville, FL. 58 pp. Bird, B. L., L. C. Branch, and D. L. Mill er. 2004. Effects of coastal lighting on foraging behavior of beach mice. Conservation Biology 18(5):1435-1439. Blair, W. F. 1951. Population structure, soci al behavior and environmental relations in a natural population of the beach mouse ( Peromyscus polionotus leucocephalus ). Pages 1-47. Contributions from the Laboratory of Vertebrate Biology, University of Michigan. Ann Arbor, MI. Bonham, C. D. 1989. Measurements for terr estrial vegetation. Wiley. New York, NY. Bourg, N.A., W.J. McShea, and D.E. G ill. 2005. Putting a cart before the search: successful habitat prediction for a ra re forest herb. Ecology 86:2793-2804. Breiman, L., Friedman, J.H., Olshen, R. A., Stone, C.G. 1984. Classification and regression trees. Wadsworth In ternational Group. Belmont, CA. Burnham, K. P. and D. R. Anderson 2002. M odel selection and multimodel inference: a practical information-theoretic ap proach. Springer. New York, NY. Buskirk, J.V. 2005. Local and landscape influence on amphibian occurrence and abundance. Ecology 86:1936-1947. Carlsson, A. and O. Kindvall. 2001. Spatial dynamics in a metapopulation network: recovery of a rare grasshopper ( Stauroderus scalaris ) from population refuges. Ecography 24:452-460. Cooper, C. B. and J. R. Walters. 2002. Independent effects of woodland loss and fragmentation on Brown Treecreeper distribution. Biologi cal Conservation 105:1-10.

PAGE 65

54 Cox, J. and R. T. Engstrom. 2001. Influence of the spatial pattern of conserved lands on the persistence of a large populat ion of red-cockaded woodpeckers. Biological Conservation 100:137-150. Dahl, B.E. and Woodard, D.W. 1977. Constr uction of Texas coastal foredunes with sea oats ( Uniola paniculata ) and bitter panicum ( Panicum amarum ). International Journal of Biometeorology 21:267-275. Death, G. and K.E. Fabricius. 2000. Cla ssification and regre ssion trees: a powerful yet simple technique for ecologi cal data analysis. Ecology 81:3178-3192. Ehrenfeld, J.G.,1990. Dynamics and processes of barrier island vegetation. Critical Reviews in Aquatic Science 2:437-480. Ellner, S.P. and J. Fieberg. 2003. Using PVA for management despite uncertainty: effects of habitat, hatcheries, and harvest on salmon. Ecology 84:1359-1269. Emanuel, K. 2005. Increasing destructiveness of tropical cyclones over the past 30 years. Nature 436: 686-688. Environmental Systems Research Ins ititue. 1996. ArcView GIS, version 3.2. Environmental Systems Research Institute, Redlands, California, USA. Flather, CH, K.R. Wilson, D.J. Dean, a nd W.C. McComb. 1997. Id entifying gaps in conservation networks: Of indicators and uncertainty in geographic-based analyses. Ecological Applications 7:531-542. Frank, K. 2005. Metapopulation persistence in heterogeneous landscapes: Lessons about the effect of stochastic ity. American Naturalist 165:374-388. Glennon, M. J., W. F. Porter, and C. L. De mers. 2002. An alternative field technique for estimating diversity of small-mammal populations. Journal of Mammalogy 83:734-742. Gore, J. A. and T. L. Schafer. 1993. Distri bution and conservation of the Santa Rosa beach mouse. Proceedings Annual Conf erence Southeast Association of Fish and Wildlife Agenices 47:378-385. Gu, W. D. and R. K. Swihart. 2004. Absent or undetected? Effect s of non-detection of species occurrence on wildlife-ha bitat models. Biological Conservation 116:195-203. Guisan, A. and W. Thuiller. 2005. Predicting sp ecies distribution: offering more than simple habitat models. Ecology Letters 8:993-1009.

PAGE 66

55 Hallermier, R.J. and P.E. Rhodes. 1988. Gene ric treatment of dune erosion for 100-year event. Proceedings of the 20th International Conferen ce on Coastal Engineering, ASCE, New York, 1107-1115. Hanski, I. 1994. A practical model of me tapopulation dynamics. Journal of Animal Ecology 63:151-162. Hesp, P. 2002. Foredunes and blowouts: in itiation, geomorphology, and dynamics. Geomorphology 48:245-268. Hesp, P. 2004. Coastal dunes in the tropi cs and temperate regions; Location, formation, morphology and vegetation processes. Pages 29-49 in M. L. Martinez, and N.P. Psuty, editor. Co astal Dunes, Ecology and Conservation. Springer-Verlag. Berlin. Hoekstra, J. M., W. F. Fagan, and J. E. Brad ley. 2002. A critical role for critical habitat in the recovery planning process? Not yet. Ecological Applications 12:701-707. Holler, N.R. 1992. Perdido Key Beach Mous e. In Rare and Endangered Biota of Florida, Volume 1. Mammals (S. R. Humphrey ed.). University Press of Florida, Gainesville, FL 52pp. Humphrey, S. R. and D. B. Barbour. 1981. St atus and habitat of 3 subspecies of Peromyscus polionotus in Florida. Journal of Mammalogy 62:840-844. Jonzen, N., C. Wilcox, and H. P. Possi ngham. 2004. Habitat selection and population regulation in temporally fluctuati ng environments. American Naturalist 164:E103-E114. Judge, E.K., M.F. Overton, and J.S. Fisher 2003. Vulnerability indicators for coastal dunes. Journal of Waterway, Port, Co astal, and Ocean Engineering 129:270-278. Kery, M. 2004. Extinction rate, estimates for plant populations in revisitation studies: importance of detectability. Conservation Biology 18:570-574. Kriebel, D.L., R. Dalrymple, A. Pratt, and V. Sakovich. 1997. A shoreline risk index for northeasters. Proceedings of the AS CE International Conference on Natural Disaster Reduction, AS CE, New York, 251-252. Mabee, T. J. 1998. A weather-resistant tr acking tube for small mammals. Wildlife Society Bulletin 26:571-574. MacKenzie, D. I., J. D. Nichols, G. B. L achman, S. Droege, J. A. Royle, and C. A. Langtimm. 2002. Estimating site occupancy rates when detecti on probabilities are less than one. Ecology 83:2248-2255.

PAGE 67

56 MacKenzie, D. I. J. D. Nichols, J. E. Hine s, M. G. Knutson, and A. B. Franklin. 2003. Estimating site occupancy, colonization, a nd local extinction when a species is detected imperfectly. Ecology 84:2200-2207. Martinez, M. L., N.P. Psuty., and R.A. Lubke. 2005. A perspective on coastal dunes. Pages 1-10 in M. L. Martinez, and N.P. Psuty, editor. Coastal Dunes, Ecology and Conservation. Springer-Verlag. Berlin. Mazerolle, M.J., A. Desrochers, L. Rochef ort. 2005. Landscape characteristics influence pond occupancy by frogs after accounting for detectability. Ecological Applications 15:824-834. Mendelssohn, I.A., M.W. Hester, F.J. Montef emante, and F. Talbot. 1991. Experimental dune building and vegetative st abilization in a sand-defic ient barrier island setting on the Louisiana Coast. Journal of Coastal Research 7 :137-149. Milio, J.F. 1998. Endangered and Threatened Wildlife and Plants; Determination of Endangered Status for the St. Andrew Beach Mouse. U.S. Fish and Wildlife Service. Federal Register 63:70053-70062. Miller, D.L. and M. Thetford. 2001. Evalua tion of sand fence and vegetation for dune building following overwash by Hurricane Opal on Santa Rosa Island, Florida. Journal of Coastal Research 17:936-948. Miller, D. L., L. Yager, M. Thetford, and M. Schneider. 2003. Potential use of Uniola paniculata rhizome fragments for dune rest oration. Restoration Ecology 11:359369. Morton, R.A. and A.H. Sallenger. 2003. Morpho logical impacts of extreme storms on sandy beaches and barriers. Journal of Coastal Research 19:560-573. Moyers, J. E. 1996. Food habits of the Gulf Coast subspecies of beach mice ( Peromyscus polionotus spp.). MS Thesis. Auburn University, Auburn, AL. 54pp. Musila, W.M., J.I. Kinyamario, and P.D. Jungerius. 2001. Vegetation dynamics of coastal sand dunes near Malindi, Ke nya. African Journal of Ecology 39:170177. Nagelkerke, N. J. D. 1991. A note on a ge neral definition of the coefficient of determination. Biometrika 78:691-692. Nordstrom, K.F., R. Lampe., and L.M. Vandemark. 2000. Reestablishing naturally functioning dunes on developed coasts. Environmental Management 25:37-51.

PAGE 68

57 Nordstrom, K.F. and W.A. Mitteager. 2001. Pe rceptions of natural and restored beach and dune characteristics by high school students in New Jersey, USA. Ocean and Coastal Management 44: 545-559. Oli, M. K., N. R. Holler, and M. C. W ooten. 2001. Viability analysis of endangered Gulf Coast beach mice ( Peromyscus polionotus ) populations. Biological Conservation 97:107-118. Pearce, J. and S. Ferrier. 2000. Evaluating the predictive performance of habitat models developed using logistic regres sion. Ecological Mode lling 133:225-245. Potter, J. C. 1985. Endangered and Threatened Wildlife and Plants; Determination of Endangered Status and Critical Habita t for Three Beach Mice. U.S. Fish and Wildlife Service. Fede ral Register. 50:23872-23889. Psuty, N.P. 2005. The coastal foredune: A morp hological basis for regional coastal dune development. Pages 11-27 in M. L. Ma rtinez, and N.P. Psuty, editor. Coastal Dunes, Ecology and Conservati on. Springer-Verlag. Berlin. R Development Core Team. 2003. R: A langua ge and environment for statistical computing. R Foundation for Statistical Computing. Vienna, Austria. Sallenger, A.H., Jr. 2000. Storm impact scale for barrier islands. Journal of Coastal Research 16:890-895. Schrott, G. R., K. A. With, and A. T. W. King. 2005. On the importance of landscape history for assessing extinction ris k. Ecological Appli cations 15:493-506. Segurado, P. and M.B. Araujo. 2004. An eval uation of methods for modeling species distributions. Journa l of Biogeography 31:1555-1568. Sneckenberger, S. I. 2001. Factors influenci ng habitat use by the Alabama beach mouse ( Peromyscus polionotus ammobates ). MS Thesis. Auburn University, Auburn, AL. 101pp. SPSS Inc. 2004. SPSS Base 13.0 for Windows Users Guide. SPSS Inc., Chicago IL. Stallins, J. A. and A. J. Parker. 2003. The influence of complex systems interactions on barrier island dune vegetation patte rn and process. Annals of the Association of American Geographers 93:13-29. Stewart, S.R. 2005. Tropical Cyclone Re port: Hurricane Ivan 2-26 September 2004. NOAA National Weather Service, NHC/TPC.

PAGE 69

58 Stone, G.W., B. Liu, D.A. Pepper, and P. Wang. 2004. The importance of extratropical and tropical cyclones on th e short-term evolution of barrier islands along the northern Gulf of Mexico, USA. Marine Geology 210:63-78. Swilling, W. R. and M. C. Wooten. 2002. Subadult dispersal in a monogamous species: The Alabama beach mouse ( Peromyscus polionotus ammobates ). Journal of Mammalogy 83:252-259. Swilling, W. R., M. C. Wooten, N. R. Holler, and W. J. Lynn. 1998. Population dynamics of Alabama beach mice ( Peromyscus polionotus ammobates ) following Hurricane Opal. Ameri can Midland Naturalist 140:287-298. Snyder, R. A. and C. L. Boss. 2002. Recove ry and stability in barrier island plant communities. Journal of Coastal Research 18:530-536. Taylor, M. F. J., K. F. Suckling, and J. J. Rachlinski. 2005. The effectiveness of the endangered species act: A quantitati ve analysis. Bioscience 55:360-367. U.S. Fish and Wildlife Service. 1981. Standard s for the development of habitat suitability index models for use in the habitat eval uation procedure. U.S. Fish and Wildlife Service, Division of Ecological Se rvices. Manuscript 103. Washington, DC. 171pp. U.S. Fish and Wildlife Service. 1987. Recove ry plan for the Choctawhatchee, Perdido Key and Alabama beach mouse. U.S. Fish and Wildlife Service, Atlanta, Georgia. 45pp. Van Horne, B. 1983. Density as a misleading indicator of habitat quality. Journal of Wildlife Management 47:893-901. Van Horne, B., G. S. Olson, R. L. Schooley, J. G. Corn, and K. P. Burnham. 1997. Effects of drought and prolonged wi nter on Townsend's ground squirrel demography in shrubsteppe habi tats. Ecological Monographs 67:295-315. Vellinga, P. 1982. Beach and dune erosion during storm surges. Coastal Engineering 6:361-387. Welch, N. E., and J. A. MacMahon. 2005. Iden tifying habitat variables important to the rare Columbia spotted frog in Ut ah (USA): An information-theoretic approach. Conservation Biology 19:473-481. Zar, J.H. 1998. Biostatistical Anal ysis. Prentice Hall. New York, NY.

PAGE 70

59 BIOGRAPHICAL SKETCH Alexander James Pries was born in S pooner, Wisconsin on May 26, 1980. Son of James and Constance Pries, he grew up in St. Paul, MN, but often escaped to the northern forests of Minnesota and Wisconsin during th e summer months. It was there, during hours of playing in the forests, lakes, and streams that he began to first observe and appreciate natural ecosystems. In 1998, he en rolled at The College of Wooster in Ohio and received a B.A. in biol ogy in the spring of 2002. After graduation, he traveled to Costa Rica, where he served as a teaching assi stant in a course on tropical ecology for the Organization for Tropical Studies. After this experience, he moved to Avon Park, FL, where he worked as a research technician for the University of Florida on a project looking at features of habitat use by round-tailed muskrats ( Neofiber alleni ). In 2003, he was hired by Archbold Biological Station to serve as a research technician for a population assessment of Florida Scrub Jays. He began graduate school in August of 2003 at the University of Floridas Depart ment of Wildlife Ecology and Conservation, from which he received his M.S. in 2006.