Citation
Predicting Soil Phosphorus Storage in Historically Isolated Wetlands within the Lake Okeechobee Priority Basins

Material Information

Title:
Predicting Soil Phosphorus Storage in Historically Isolated Wetlands within the Lake Okeechobee Priority Basins
Creator:
MCKEE, KATHLEEN
Copyright Date:
2008

Subjects

Subjects / Keywords:
Constructed wetlands ( jstor )
Ditches ( jstor )
Highlands ( jstor )
Land use ( jstor )
Pastures ( jstor )
Phosphorus ( jstor )
Polygons ( jstor )
Soils ( jstor )
Vegetation ( jstor )
Wetlands ( jstor )
Lake Okeechobee ( local )

Record Information

Source Institution:
University of Florida
Holding Location:
University of Florida
Rights Management:
Copyright Kathleen Mckee. 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:
5/1/2005
Resource Identifier:
71230808 ( OCLC )

Downloads

This item has the following downloads:

mckee_k ( .pdf )

mckee_k_Page_164.txt

mckee_k_Page_195.txt

mckee_k_Page_130.txt

mckee_k_Page_205.txt

mckee_k_Page_055.txt

mckee_k_Page_223.txt

mckee_k_Page_203.txt

mckee_k_Page_063.txt

mckee_k_Page_112.txt

mckee_k_Page_151.txt

mckee_k_Page_032.txt

mckee_k_Page_018.txt

mckee_k_Page_225.txt

mckee_k_Page_001.txt

mckee_k_Page_147.txt

mckee_k_Page_022.txt

mckee_k_Page_004.txt

mckee_k_Page_125.txt

mckee_k_Page_019.txt

mckee_k_Page_069.txt

mckee_k_Page_074.txt

mckee_k_Page_201.txt

mckee_k_Page_106.txt

mckee_k_Page_235.txt

mckee_k_Page_057.txt

mckee_k_Page_232.txt

mckee_k_Page_190.txt

mckee_k_Page_231.txt

mckee_k_Page_241.txt

mckee_k_Page_052.txt

mckee_k_Page_059.txt

mckee_k_Page_122.txt

mckee_k_Page_003.txt

mckee_k_Page_185.txt

mckee_k_Page_035.txt

mckee_k_Page_138.txt

mckee_k_Page_224.txt

mckee_k_Page_064.txt

mckee_k_Page_065.txt

mckee_k_Page_118.txt

mckee_k_Page_152.txt

mckee_k_Page_007.txt

mckee_k_Page_058.txt

mckee_k_Page_176.txt

mckee_k_Page_186.txt

mckee_k_Page_098.txt

mckee_k_Page_144.txt

mckee_k_Page_050.txt

mckee_k_Page_051.txt

mckee_k_Page_014.txt

mckee_k_Page_206.txt

mckee_k_Page_212.txt

mckee_k_Page_116.txt

mckee_k_Page_026.txt

mckee_k_Page_041.txt

mckee_k_Page_067.txt

mckee_k_Page_218.txt

mckee_k_Page_101.txt

mckee_k_Page_113.txt

mckee_k_Page_092.txt

mckee_k_Page_103.txt

mckee_k_Page_023.txt

mckee_k_Page_042.txt

mckee_k_Page_119.txt

mckee_k_Page_062.txt

mckee_k_Page_189.txt

mckee_k_Page_208.txt

mckee_k_Page_216.txt

mckee_k_Page_179.txt

mckee_k_Page_184.txt

mckee_k_Page_234.txt

mckee_k_Page_131.txt

mckee_k_Page_175.txt

mckee_k_Page_048.txt

mckee_k_Page_150.txt

mckee_k_Page_215.txt

mckee_k_Page_097.txt

mckee_k_Page_045.txt

mckee_k_Page_167.txt

mckee_k_Page_070.txt

mckee_k_Page_028.txt

mckee_k_Page_136.txt

mckee_k_Page_111.txt

mckee_k_Page_153.txt

mckee_k_Page_107.txt

mckee_k_Page_230.txt

mckee_k_Page_207.txt

mckee_k_Page_159.txt

mckee_k_Page_017.txt

mckee_k_Page_083.txt

mckee_k_Page_169.txt

mckee_k_Page_016.txt

mckee_k_Page_171.txt

mckee_k_Page_078.txt

mckee_k_Page_068.txt

mckee_k_Page_005.txt

mckee_k_Page_013.txt

mckee_k_Page_161.txt

mckee_k_Page_143.txt

mckee_k_Page_002.txt

mckee_k_Page_181.txt

mckee_k_Page_006.txt

mckee_k_Page_220.txt

mckee_k_Page_075.txt

mckee_k_Page_134.txt

mckee_k_Page_132.txt

mckee_k_Page_211.txt

mckee_k_Page_139.txt

mckee_k_Page_037.txt

mckee_k_Page_120.txt

mckee_k_Page_202.txt

mckee_k_Page_193.txt

mckee_k_Page_226.txt

mckee_k_Page_155.txt

mckee_k_Page_158.txt

mckee_k_Page_044.txt

mckee_k_Page_196.txt

mckee_k_Page_187.txt

mckee_k_Page_108.txt

mckee_k_Page_178.txt

mckee_k_Page_121.txt

mckee_k_Page_077.txt

mckee_k_Page_087.txt

mckee_k_Page_031.txt

mckee_k_Page_072.txt

mckee_k_Page_047.txt

mckee_k_Page_148.txt

mckee_k_Page_238.txt

mckee_k_Page_100.txt

mckee_k_Page_166.txt

mckee_k_Page_140.txt

mckee_k_Page_115.txt

mckee_k_pdf.txt

mckee_k_Page_145.txt

mckee_k_Page_129.txt

mckee_k_Page_157.txt

mckee_k_Page_020.txt

mckee_k_Page_156.txt

mckee_k_Page_094.txt

mckee_k_Page_165.txt

mckee_k_Page_222.txt

mckee_k_Page_123.txt

mckee_k_Page_104.txt

mckee_k_Page_053.txt

mckee_k_Page_033.txt

mckee_k_Page_239.txt

mckee_k_Page_170.txt

mckee_k_Page_021.txt

mckee_k_Page_168.txt

mckee_k_Page_214.txt

mckee_k_Page_079.txt

mckee_k_Page_199.txt

mckee_k_Page_025.txt

mckee_k_Page_034.txt

mckee_k_Page_073.txt

mckee_k_Page_089.txt

mckee_k_Page_084.txt

mckee_k_Page_049.txt

mckee_k_Page_011.txt

mckee_k_Page_008.txt

mckee_k_Page_237.txt

mckee_k_Page_054.txt

mckee_k_Page_146.txt

mckee_k_Page_030.txt

mckee_k_Page_009.txt

mckee_k_Page_194.txt

mckee_k_Page_012.txt

mckee_k_Page_227.txt

mckee_k_Page_081.txt

mckee_k_Page_086.txt

mckee_k_Page_183.txt

mckee_k_Page_027.txt

mckee_k_Page_210.txt

mckee_k_Page_036.txt

mckee_k_Page_071.txt

mckee_k_Page_177.txt

mckee_k_Page_091.txt

mckee_k_Page_137.txt

mckee_k_Page_080.txt

mckee_k_Page_182.txt

mckee_k_Page_046.txt

mckee_k_Page_015.txt

mckee_k_Page_173.txt

mckee_k_Page_082.txt

mckee_k_Page_088.txt

mckee_k_Page_219.txt

mckee_k_Page_217.txt

mckee_k_Page_204.txt

mckee_k_Page_038.txt

mckee_k_Page_174.txt

mckee_k_Page_188.txt

mckee_k_Page_213.txt

mckee_k_Page_197.txt

mckee_k_Page_066.txt

mckee_k_Page_117.txt

mckee_k_Page_090.txt

mckee_k_Page_110.txt

mckee_k_Page_043.txt

mckee_k_Page_093.txt

mckee_k_Page_127.txt

mckee_k_Page_154.txt

mckee_k_Page_172.txt

mckee_k_Page_114.txt

mckee_k_Page_109.txt

mckee_k_Page_236.txt

mckee_k_Page_191.txt

mckee_k_Page_200.txt

mckee_k_Page_056.txt

mckee_k_Page_233.txt

mckee_k_Page_240.txt

mckee_k_Page_024.txt

mckee_k_Page_209.txt

mckee_k_Page_102.txt

mckee_k_Page_160.txt

mckee_k_Page_126.txt

mckee_k_Page_162.txt

mckee_k_Page_124.txt

mckee_k_Page_040.txt

mckee_k_Page_128.txt

mckee_k_Page_010.txt

mckee_k_Page_105.txt

mckee_k_Page_060.txt

mckee_k_Page_099.txt

mckee_k_Page_076.txt

mckee_k_Page_141.txt

mckee_k_Page_198.txt

mckee_k_Page_221.txt

mckee_k_Page_135.txt

mckee_k_Page_142.txt

mckee_k_Page_228.txt

mckee_k_Page_029.txt

mckee_k_Page_180.txt

mckee_k_Page_229.txt

mckee_k_Page_085.txt

mckee_k_Page_149.txt

mckee_k_Page_192.txt

mckee_k_Page_061.txt

mckee_k_Page_133.txt

mckee_k_Page_039.txt

mckee_k_Page_096.txt

mckee_k_Page_163.txt

mckee_k_Page_095.txt


Full Text












PREDICTING SOIL PHOSPHORUS STORAGE IN HISTORICALLY ISOLATED
WETLANDS WITHIN THE LAKE OKEECHOBEE PRIORITY BASINS
















By

KATHLEEN A. MCKEE


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

UNIVERSITY OF FLORIDA


2005

































Copyright 2005

by

Kathleen A. McKee


































To my nieces and nephews Andrea, Matt, Robert, Jonathan and Andrew, for whom I
hope to incite learning.





"The learning process is something you can incite, literally incite, like a riot."
--Audre Lorde















ACKNOWLEDGMENTS

This thesis would not be possible without the gracious support of at least 54

landowners from Okeechobee, Martin, Highlands, and St. Lucie Counties, many of whom

fostered a stimulating exchange between science student and land steward. Especially

helpful were Pete Beaty, Louis Larson, J.C. Bass, and Norman Ephraim: expert cattle

ranchers, and lovers of the land. Other land owners included Keith Rucks, Williamson

Cattle Co., Triple Cross Horse and Cattle, Triple Diamond Ranch, Allen Smith, Allen

Lewis, Edwin Walpole, the Duncans, Jerald Newcomer, Taylor Creek Ranch,

B&E Ranch and Grove, Grassy Island Ranch, Pete Clemons, Patrick Luna, Ralph Palaez,

James Fowler, Seller Prescott, Robert Arnold, Lykes Brothers, Danny Fairclaw, Robert

Edwards, Sacramento Farms, Haynes Williams, Nano Corona, Willoway Ranch, Richard

Smith, David Durango, Richard Hales, Harvey Cattle Co., Lois Johnson, the South

Florida Water Management District, M Cross Ranch, Marion Wagner, Linda Harvey,

Hamrick and Sons, Roy Hancock, Kirton Dudley, Bill Ritchie, and Larry Overton. The

MacArthur dairy staff were especially open and helpful hosts. Other supportive dairies

include Davie Dairy, Flying G Dairy, Larson Dairies and HW Rucks.

The support of landowners was expertly negotiated by Mitch Flinchum (University

of Florida Institute of Food and Agricultural Services, Extension) and Linda Crane

(Florida Department of Agriculture and Consumer Services). Ed Dunne and Charles

Campbell (along with many other volunteers) worked from dark morning hours, filling

ice chests, until dark evening hours, pulling soil cores. Erin Colburn at the South Florida









Water Management District provided invaluable, up-to-date GIS information and

support. Biogeochemist Erin Bostic (of the UF Wetland Biogeochemistry Lab)

coordinated the processing of over 3000 samples, trained staff, and executed laboratory

methods with quality control as her top priority.

I thank Sabine Grunwald (my major professor) who gave me encouraging support

and independence while keeping me moving on track. Committee member Sue Newman

provided a strong scientific eye to this thesis. Special thanks go to Dr. Mark Clark for his

multi-lateral support in many areas of my graduate experience, including field equipment,

vegetation identification, wetland ecology, lab logistics, volunteers, landowners and data

interpretation. He was always available when I needed a compassionate ear. Post-

doctoral associates Ed Dunne, Greg Bruland, and Matt Cohen were invaluable guides of

statistics, phosphorus, and the scientific process. I thank the faculty and staff of the Soil

and Water Science Department, who were very helpful, and interested in answering my

many biogeochemistry and soil questions. Lab-mates Rosanna Rivero, Sanjay Lamsal,

Aarthy Sabesan, Adrien Mangeot and Vinay Ramasundaram were a huge source of

support and friendship on a daily basis.

Funding was provided by South Florida Water Management District, Florida

Department of Agriculture and Consumer Services, and the Florida Department of

Environmental Protection.

Finally, my parents could not have been more supportive of my decision to leave a

secure career, to try my hand at something new and ever more valuable to me.
















TABLE OF CONTENTS

page

A C K N O W L E D G M E N T S ................................................................................................. iv

LIST OF TABLES ............................................................................. x

LIST OF FIGURES .............. ................................. ............. ........... xiii

ABSTRACT ........ .............. ............. ...... ...................... xvi

CHAPTER

1 IN TR OD U CTION ............................................... .. ......................... ..

N eed for Research ............... ................. .................................... .2
O bjectiv e s ................................................................... ................................. . .4

2 SO IL PH O SPH O R U S ................................................................. ........................ 6

P hosphorus in W wetlands ............................................................. ......... ...........7
R ole of W wetlands for R etaining P................................... .................................... 9

3 PHOSPHORUS OBSERVATIONS IN NONRIPARIAN WETLANDS ..................11

Intro du action ...................................... ................................................ 1 1
L an d U se ...................................... ............................... ................ 12
Hydrology .............................. ...... .. .........................13
Phosphorus Gradient and Hydrologic Zones...................................................13
H y p oth eses ........................................................................15
O objectives ................................................................ .. .... ......... 15
M materials and M methods ....................................................................... .................. 15
Study A rea .................................................................................................. 15
Spatial D ata ........................................18
Generating Layers for Site Selection.................... ........................................ 20
Step 1: Creating riparian and nonriparian wetland layers............................22
Step 2: Creating nonriparian emergent marsh, forested and scrub shrub
w etlan d s .................................................................... 2 4
Step 3: Creating land-use layers .................... ................................25
Step 4: Create six treatment layers...................... ........ .............. 27
Step 5: Stratified random sampling ................ ...... ........ ... ........... ........ 28









Field Sampling Methods and Site Descriptions ...............................................29
Laboratory A nalyses .................. ............................. .. .. .. .. ........ .... 40
D ata M anagem ent......... .............................................................. ..... ... ....4 1
S statistical A n aly ses........... .......................................................... .. .... .. .. ... 4 2
R results ......................... ..... ..... .. .... ... .. .................... ............... 45
Wetland Vegetative Communities and Surrounding Land Uses.........................45
Biogeochemical M measures in Hydrologic Zones.............................................45
Combined Effects of Land Use, Vegetative Community, and Wetland Size......46
Correlations between TP and M etals........................................ ............... 47
E effects of D itches ......... .............................................................. ........ 47
Wetland Phosphorus Storage...................... ....... ..............48
D discussion ............. .............................................................................................63
Z onal T P gradients ..................... .. .......................... ........ .... ...........64
Hydrologic Connectivity .................................. .....................................69
Land Use Differences .......... ........ ......... .. .... ........................ 70
Other Findings .............. .... ..... ................... ............... 71
Storage ....................... .............................72
Future Research ......... .... ........ .. .. .. ......... ......... .. ....... 72
C o n c lu sio n s........................................................................................................... 7 3

4 UPSCALING TOTAL PHOSPHORUS TO UNSAMPLED WETLANDS .............76

In tro d u ctio n ............... ........ ............... ......... ........................................................ 7 6
Environmental Variables .......... ..... ........................................ 76
Spectral D ata ....................................... .... ........ ........ 77
Ecological Response Variables and Classification Trees..............................79
H y p o th e se s ................................................................ 8 1
Objectives ............ 8............ ........ ..........81
M a te ria ls .............................................................................. 8 1
L an d U se ......................................................... 8 1
National Wetland Inventory ................... ........ ...............82
Soil Survey Geographic (SSURGO) Data Set ..............................................83
Landsat7 ETM+ Spectral Data .......................................... ................86
M methods ............................................................. 89
Upscaling M ethodology .......................... ......................................89
Collection of Independent Variables ......................................................90
Field data ................................. ............................ .... ....... 92
W ater regim e ................. ........ .........................................92
D instance to features .............................................. ............... 92
Creating Buffers around W wetlands .......................................... 95
Pre-Processing the Landsat7 ETM+ Image .............................................. 99
Calculating Spectral Indices from the Landsat Image .................................... 103
C classification T rees ............................................................................ 106
Classifying unsampled wetlands ............................... ................109
Edge percent transfer function ...................... ....................................... 110
Calculating Total Storage ................................................... ........................ 111
Results .................. .................. ....... ....113









Im portant Predictors for TP ......... ................................................ ..................113
Classification Trees for Upscaling TP Outside the Landsat Extent .................14
U scaling R results .................. ........................................... ...... ... 114
D discussion ..................................................................... ............ ......... 123
Prediction within the Landsat7 ETM+ Extent ................. ........................124
Prediction outside the Landsat7 ETM+ Extent .............................................126
Storage ............... ........................ .........................130
Future R research ......... ..... ...... .. ... ........ .. ..... ................ 131
C o n c lu sio n s......................................................................................................... 1 3 2

5 SYNTHESIS AND IMPLICATIONS............................................. ..... ..................134

APPENDIX

A SPATIAL DATA M ETADATA ........................................ ......................... 140

B G L O SSA R Y O F G IS TER M S ............................................................ ...............143

C WETLAND VEGETATIVE COMMUNITY DISTRIBUTIONS ..........................145

D SA M P L E D SIT E S ......................................................................... ..................... 14 6

E SAM PLING PERIOD RAINFALL................................... ..................................... 149

F ADDITIONAL SOIL BIOGEOCHEMICAL DATA ................... .................. 150

G LAND SAT7 ETM + HEADER FILE ............................................ .................. 163

H SOILS IN UPLANDS ADJACENT TO HISTORICALLY ISOLATED
W E T L A N D S ...................................................... ................. 16 5

I OUTPUT FILES OF FINAL CLASSIFICATION TREES ...................................168

C enter Soils in L andsat A rea ......... ................. .................................................... 168
Edge Soils in Landsat Area .......... ........... ................. .......................... 178
Center Soils in NonLandsat Area ....... ...................... .........................188
Edge Soils in NonLandsat Area ....... ........ ......... ....................... 196

J SCRIPTS FOR CLASSIFYING UNSAMPLED WETLANDS AND
CALCULATING STORAGE ............................................................................ 205

Classifying Unsampled Wetlands... .. ................. ........................205
C enters in L andsat A rea ............................................. ............ ............... 205
Edges in Landsat A rea........................ ................... ............... ............... 206
Centers in nonLandsat A rea ........................................ ......................... 206
Edges in nonLandsat A rea ....................................................... .... ........... 206









Calculating Storage in Unsampled W wetlands ................................. ............... 207
E d g e s ...........................................................................2 0 7
C enters ........... ..... ....... ..................................................... ... 207

LIST O F R EFEREN CE S ......... ......... ........ .......... ......................... ............... 208

B IO G R A PH IC A L SK E T C H ........................................ ............................................224
















LIST OF TABLES


Table page

3-1 Frequency, proportion and area of wetlands sampled by land use ...................32

3-2 Distribution of vegetation classes in hydrologic zones of sampled wetlands........32

3-3 Sampling scheme of historically isolated wetlands by land-use type and
vegetative com m unity ............................................................................ 36

3-4 Distributions of National Wetland Inventory polygons by vegetative
com m unity and land use ............................................... ............................. 49

3-5 Physico-chemical properties among hydrologic zones........................................50

3-6 Total P comparisons among hydrologic zones ............................................... 51

3-7 One MHC1 extractable P comparisons........ .................................52

3-8 M etal comparisons among hydrologic zones............................................. 52

3-9 Effects of land use and vegetative community and wetland area on soil
ch aracteristic s ............... ........... ............ ........ .............................. 5 3

3-10 Estimated unweighted means and standard errors of pH ........... ...............53

3-11 Effects of land use and vegetative community and wetland area on center
so il T P .................................................................................. 5 6

3-12 Effects of land use and vegetative community and wetland area on edge
so il T P .................................................................................. 5 9

3-13 Spearman correlations for wetland center soil TP and biogeochemical
param eters ............................................................... ... .... ........ 61

3-14 Spearman correlations for wetland edge soil TP and biogeochemical
param eters ............................................................... ... .... ........ 61

3-15 Total P stored in surface soils of sampled wetlands among land uses...................62

4-1 Landsat7 ETM + spectral bands. ........................................ ........................ 86









4-2 Non-tidal water regimes of wetlands in the National Wetland Inventory. ............92

4-3 Linear transformation coefficients to produce three Tasseled Cap indices.........105

4-4 Predictors introduced into the classification tree building process..................108

4-5 Mean TP and bulk density of sampled wetland soils by hydrologic zone
a n d T P c la ss ................................................... ............ ................ 1 12

4-6 Important variables for the center soils classification tree using spectral data....116

4-7 Prediction success table for the center soils classification tree ..........................116

4-8 Important variables for the edge soils classification tree using spectral data......117

4-9 Prediction success table for the edge soils classification tree using
spectral data ............................... ............................... ........ 117

4-10 Important variables for the center soils classification tree excluding
sp ectral d ata ................................................. .......... ............................. 1 18

4-11 Prediction success table for the center soils classification tree excluding
sp ectral d ata ................................................. .......... ............................. 1 18

4-12 Important variables for the edge soils classification tree excluding
sp ectral d ata ................................................. .......... ............................. 1 19

4-13 Prediction success table for the edge soils classification tree excluding
sp ectral d ata ................................................. .......... ............................. 1 19

4-14 Predicted TP storage in surface soils and descriptive statistics of unsampled
historically isolated wetlands by size and land use ................ ........ ..........119

A-1 Projection information for GIS data layers................ ......................140

C-l All National Wetland Inventory polygons within the four priority basins by
vegetative community ........... .. ......... ......................... 145

D-l Sampled wetland locations and selected characteristics............... ............... 146

E-l Rainfall record by month of sampling period................................................149

F-l Comparisons of P extractions among hydrologic zones.............. .............. 151

F-2 Observed values and comparisons of P extractions among land uses ...............152

F-3 Observed values and comparisons of metals among land uses.......................... 154

F-4 Pearson correlations of selected soil biogeochemical parameters .......................156









F-5 Pearson correlations of selected biogeochemical parameters in wetland center
soils in im proved pastures ........................................................ ............. 157

F-6 Pearson correlations of selected biogeochemical parameters in wetland center
so ils in d a irie s ................................................................... 1 5 8

F-7 Pearson correlations of selected biogeochemical parameters in wetland center
soils in unimproved pastures / rangelands. .................................. ............... 159

F-8 Pearson correlations of selected biogeochemical parameters in wetland edge
soils in im proved pastures ........................................................ ............. 160

F-9 Pearson correlations of selected biogeochemical parameters in wetland edge
soils in dairies. .......................................................................... 16 1

F-10 Pearson correlations of selected biogeochemical parameters in wetland edge
soils in unim proved pastures / rangelands ..........................................................162

H-1 Soil series in the components of prevalent map units of upland areas ..............166
















LIST OF FIGURES


Figure pge

2-1 Selected P cycle components in wetlands.... ....................................10

3-1 Lake Okeechobee with 41 watershed basins and county lines. ..........................16

3-2 Four priority basins, counties and waterways................................... ............... 19

3-3 Land use in the priority basins ....................... ......... ................. 20

3-4 Procedure for creating spatial layers and selecting sample sites .........................21

3-5 Selecting riparian w wetlands ...................... .... ......... .................... ............... 23

3-6 All NWI wetlands classified as riparian and nonriparian.....................................24

3-7 Nonriparian wetlands by wetland vegetative community.................................25

3-8 Three steps to generalize polygons by vegetative community ...........................26

3-9 Reclassification of land-use wetland polygons.................................................27

3-10 Land-use layer showing land-use areas of interest in the priority basins after
land-use reclassification .............................. ........................... ............... 28

3-11 Sampled wetlands in the four priority basins of the Lake Okeechobee
w watershed. .......................................... ............................ 30

3-12 Percentage of sites by land use per vegetative community..............................31

3-13 An example of zonation within an isolated wetland............... ........................33

3-14 An example of the "mixed transitional" vegetation class.............................. 34

3-16 Detail areas of DOQQ photos showing hydrological connection types ...............37

3-17 Three intensities of man-made drainage ditches............................ .....................38

3-18 D itch class by land use................................................. .............................. 39

3-19 D itch num ber by land use ............................................ .............................. 39









3-20 D itch intensity by land use.......................................................... ............... 40

3-21 Soil bulk density and percent organic matter comparisons among vegetative
com m unities ..........................................................................54

3-22 Soil bulk density and percent organic matter by detailed land use showing
m ean values .............. ...... .... ...... .................. ............ ......... 55

3-23 Scatterplot of center soil total phosphorus versus wetland size...........................56

3-24 Plots of total phosphorus in center soils showing interaction of vegetation
com m unity and land use. .............................................. ............................. 57

3-25 Total phosphorus comparisons of center soils showing means ...........................58

3-26 Scatterplot of edge soil TP versus wetland size...............................................59

3-27 Total phosphorus comparisons of edge soils showing means. ...........................60

3-28 Total P in center soils by ditch class ..................................................... ..........62

3-29 Land uses and TP stored in surface soils .................................... ............... 63

3-30 Cartoon comparing relative magnitudes of wetland edge and center
soil TP am ong land uses. .............................................. ............................. 65

4-1 Land uses of the priority basins. ......................................................... .... .......... 82

4-2 National Wetland Inventory polygons ........................................................83

4-3 Soil orders in the four priority basins. ..................................................... 84

4-4 National wetland inventory wetland polygons shifted in multiple
directions from the SSURGO map unit polygons .............................................85

4-5 Landsat7 ETM+ image of path 16/41 from March 24, 2003 ..............................88

4-6 April 5, 2004 photo of a historically isolated wetland ..........................................88

4-7 Steps for collecting sampled wetland independent variables. ............................91

4-8 Distance-to-high-intensity-area raster ........................................... ...............93

4-9 D istance-to-m ajor-roads raster........................................................... .... .......... 94

4-10 D istance-to-w aterw ays raster .................................................................... ....... 95

4-11 Landsat7 ETM+ false-color composite....................................... ............... 97









4-12 Steps for creating buffers and associating them with the wetland ID....................98

4-13 The Landsat7 ETM+ subset image of study area..................................... 101

4-14 Landsat7 ETM + image and NW I polygons....................................................... 102

4-15 Nonriparian polygons before and after manual alignment ...............................103

4-16 Wetland area confidence intervals by edge percent ...........................................110

4-17 Scatterplot of edge percent vs. wetland area.................................................111

4-18 Classification tree for sampled wetland center soils using Landsat ..................115

4-19 Classification tree for sampled wetland edge soils using Landsat .....................116

4-20 Classification tree for sampled wetland center soils ....................... ..............117

4-21 Classification tree for sampled wetland edge soils.....................................118

4-22 Classification distributions of TP in wetland hydrologic zones ........................120

4-23 Unsampled historically isolated wetlands indicating TP class in center and
edge soils as predicted by two sets of classification trees ................................. 121

4-24 Predicted TP in the surface soils of historically isolated wetlands ...................122

4-25 Total P class distribution of sampled center soils among soil subgroups of
m ajor upland com ponents. .................................................................................. 127

4-26 Total P class distribution of sampled center soils among land uses ..................128

4-27 Scatterplot of TP in sampled center soils by wetland perimeter..........................129

5-1 Total P storage by land use: contrasting models.........................................139

E -1 M ay, 2003 precipitation m ap .................................................................... ... 149















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

PREDICTING SOIL PHOSPHORUS STORAGE IN HISTORICALLY ISOLATED
WETLANDS WITHIN THE LAKE OKEECHOBEE PRIORITY BASINS

By

Kathleen A. McKee

May 2005

Chair: Sabine Grunwald
Cochair: Mark W. Clark
Major Department: Soil and Water Science

In South Florida's Lake Okeechobee, the problem of eutrophication is largely

caused by phosphorus (P) pollution in runoff from dairies and beef-cattle pastures in four

basins of the watershed. A recommendation for mitigating high P loads to surface waters

is to restore the hydrology of isolated wetlands which have been extensively ditched. A

synoptic sampling of surface soils (0-10 cm) in 118 wetlands and surrounding uplands

aimed to characterize historically isolated wetlands and relationships between total P

(TP) and landscape-scale variables. A random sampling scheme, stratified by land use

and wetland vegetative community, was used. It was hypothesized that a decreasing TP

gradient exists from wetland centers, to wetland edges, to surrounding uplands, and that

this storage varies with land use and ditching intensity. Overall, wetland center soils

stored significantly more TP (median: 24.9 g m-2) than edges (median: 16.5 g m-2) and

uplands (median: 12.5 g m-2). Wetland dairy (median: 35.3g m-2) and improved pasture









(median: 22.7 g m-2) soils stored more TP than wetland soils in unimproved pastures and

rangelands (median: 20.4 g m-2). Ditch intensity had a significant effect on center soils,

with more TP (g m-2) stored in wetlands with larger ditches. While there were no

significant differences in the amounts of calcium, aluminum, iron or magnesium, organic

matter content was significantly higher in center soils (median: 20.4%) than in edges

(median: 11.8%), and is a controlling factor for TP storage in these wetlands.

To upscale these findings to the basin-scale, ancillary spatial data were used to

predict TP stored in the all historically isolated wetlands in the priority basins. It was

hypothesized that TP can be predicted using spectral data from a Landsat7 ETM+ image,

upland soil taxonomic data, land use, landscape metrics and wetland characteristics.

Classification trees were generated using the TP (mg kg-1) measures of the wetland center

and edge soils to predict TP condition (high or low) of unsampled wetland center and

edge soils. Spectral data reflecting vegetation and soil moisture were the most important

predictors in the classification trees which had overall cross-validation accuracy rates of

76%. Sampled bulk density and TP (mg kg-1) means were used to predict TP storage in

unsampled wetlands. It was predicted that 2,736,563 400,568 kg (292.3 kg ha-1) are

being stored in the surface soils of all historically isolated wetlands in the four priority

basins. By restoring the hydrology of these wetlands, more P would be stored in upland

and wetland soils. Increased wetland area and hydraulic retention times would enhance

organic matter accretion and other P-binding mechanisms. These findings support the

idea of hydrologic restoration as a useful management practice to reduce P loads to

surface waters. The landscape-scale predictive model will help land managers target high

P areas and estimate the effectiveness of future hydrological restoration efforts.














CHAPTER 1
INTRODUCTION

Natural eutrophication in freshwater lakes is the accumulation of sediments,

nutrients, and organic matter from the surrounding watershed, with time (Wetzel, 1983).

Anthropogenic eutrophication is the acceleration of this process due to human activities,

and is a major problem in agricultural watersheds (Sharpley et al., 2000). Phosphorus (P)

is often recognized as the limiting nutrient in freshwater aquatic systems, because of its

key role in primary production (Syers et al., 1973; Allen et al., 1982; Sonzogni et al.,

1982; Wetzel, 1983). Phosphorus has been identified as the key element contributing to

eutrophication in Lake Okeechobee in South Central Florida (Davis and Marshall, 1975;

Federico et al., 1981). This has resulted in frequent algal blooms, detrimental changes in

biological communities, and impaired use of water resources (South Florida Water

Management District (SFWMD), 1993). Most of the P entering the lake from the

watershed is from nonpoint sources (Federico et al., 1981; Fluck et al., 1992; Florida

Department of Environmental Protection (FDEP), 2001; Hiscock et al., 2003).

Lake Okeechobee is a shallow, subtropical lake, with a surface area of 1,700 km2,

and a drainage basin of about 12,000 km2 (SFWMD, 1993). In 1987, Florida adopted the

Lake Okeechobee Surface Water Improvement and Management (SWIM) Plan

(SFWMD, 1993). The major goal of the plan is to improve lake water quality through a

watershed approach. The SWIM Plan identified four of the 41 basins as "priority

basins," because they were contributing most of the P loads to the lake (35%), while

occupying 12% of the land area (FDEP, 2001). These basins are known as S-191 (Taylor









Creek/Nubbin Slough), S-65D, S-154, and S-65E (three Kissimmee River basins). Land

use in the priority basins is dominated by beef-cattle ranches and dairies. Both of these

land uses are thought to be responsible for large P exports to the lake via stormwater

runoff from grazing pastures and cattle-feeding areas (MacGill et al., 1976; Allen et al.,

1982; Anderson and Flaig, 1995). Dairies are considered the most intensive land use for

nutrient input in the basins (Reddy et al., 1995).

The Clean Water Act (CWA) section 303 (d) lists Lake Okeechobee as an

"impaired surface water" and requires the development of pollutant total maximum daily

loads (TMDLs). The TMDL goal for P to Lake Okeechobee is 140 t per year by 2015

(FDEP, 2001). Between 1995 and 2000, the lake received an average of 640 t of P per

year (FDEP, 2001). In 2002, the P load measured into Lake Okeechobee was 543 t

(SFWMD et al., 2004a). The SFWMD's Works of the District permitting plan set P-

loading targets for each basin in order to reach the TMDL goal, but these goals are still

being exceeded in basins S-191 and S-154 (US Army Corps of Engineers (USACE) and

South Florida Water Management District (SFWMD), 2003). Phosphorus measured in

basins S-65D and S-65E are thought to be below targets because of effective agricultural

best management practices, but uncertainty remains about the future fate and transport of

P in all four of the basins (USACE and SFWMD, 2003).

Need for Research

Continued improvements are needed to reduce nonpoint source P pollution from

agriculture in these basins. It is known that P retention by wetlands helps reduce P loads

to receiving waters (Mitsch, 1992; Reddy et al., 1996a; Nairn and Mitsch, 1999;

Braskerud, 2002; Coveney, 2002; Kuusemets and Mander, 2002; Shan et al., 2002).

Wetland systems between agricultural land and water bodies can help improve water









quality (Sharpley, 1995; Reddy et al., 1996b; Sharpley et al., 2000; Tate et al., 2000).

Boggess et al. (1995) concluded that 17% of total P inputs to the northern watershed of

Lake Okeechobee (containing the priority basins) was stored in wetlands, and that 60% of

the mass transported from the uplands via surface and subsurface water flow was stored

in wetlands. They also concluded that basins with greater wetland areas could help

reduce P loading from uplands to surface water systems. Hiscock et al. (2003) used a

modeling approach to estimate that 32% of P runoff from uplands in the northern

Okeechobee watershed was stored in wetlands, and that 8% of all P imported to the

watershed was stored in wetlands. Reddy et al. (1996a) estimated that 70% of the P

imported into the Lake Okeechobee watershed is stored in uplands; and an additional

18% is stored in wetlands and streams, with the remaining balance being in livestock and

feed. While wetlands store P, much of what is not retained is exported downstream via

drainage systems and streams (Graetz and Nair, 1995).

During the 1960s, 16,000 to 20,000 ha of floodplain wetlands were drained for the

development of agriculture (Loftin et al., 1990). In the last 50 years, drainage ditches

have been installed to drain lands for grazing pasture (Flaig and Reddy, 1995). The Lake

Okeechobee Protection Act of 2000 states that about 45% of the wetlands north of the

lake have been ditched (SFWMD et al., 2004a). Part of the Lake Okeechobee Protection

Plan (LOPP), a response to the act, suggests that it may be necessary to "reclaim isolated

wetlands on pasture lands" in order to reduce P loads and attenuate peak water flows

during storm events to the lake (SFWMD et al., 2004a).

In our study, isolated wetlands were defined as depressional wetlands completely

surrounded by upland, which may be hydrologically connected to other wetlands and









waterbodies through ground-water flows or intermittent overflows (spillovers) (Tiner,

2003). Natural hydrologic influences on these wetlands are principally precipitation,

groundwater, and local runoff Man-made connections to other surface water bodies are

often associated with these wetlands, so they are referred to as historically isolated or

nonriparian wetlands.

Flaig and Havens (1995) estimated that 15% of wetlands in this area are isolated.

According to the National Wetland Inventory (NWI) (U.S. Fish and Wildlife, 2002),

wetlands cover 18% of the four priority basins, and 59% of those are historically isolated.

While some studies have investigated the role of riparian wetlands in the study area

(Reddy et al., 1995; Reddy et al., 1996a), few studies have addressed P retention in the

historically isolated wetlands. Moreover, little is known about the biogeochemical,

vegetative, and hydrologic characteristics of these wetlands. To predict the success of

reclaiming historically isolated wetlands, it is important to quantify the land area extent

of these wetlands, determine how much total P (TP) they are currently storing, and

investigate watershed and field-scale factors that affect P storage. This study focused on

soil TP storage, as most of the long-term P storage in wetlands occurs in soils (Johnston,

1991; Reddy 1996a).

Objectives

Objectives of my study were as follows:

* Sample and characterize the historically isolated wetlands in dairies and beef-cattle
pastures of the priority basins: land area, vegetation community, surrounding land
use, hydrologic connectivity, soil physicochemical characteristics, and P content of
surface soils.

* Use sampling data to predict the TP stored in surface soils of all historically
isolated wetlands in dairies and beef-cattle pastures of the priority basins.






5


These objectives and related hypotheses are discussed in Chapters 3 and 4,

respectively. Chapter 5 provides a synthesis of both chapters and discusses the

implications the results may have for land managers and researchers investigating the

possibility of hydrologic restoration of historically isolated wetlands in the priority basins

of the Lake Okeechobee watershed.














CHAPTER 2
SOIL PHOSPHORUS

Soil phosphorus (P) exists in inorganic and organic forms that can either be in

particulate or dissolved forms. Native sources of P include phosphate rock like apatite

(inorganic) and dying cells (organic). Common allochthonous sources of P in

agriculturally dominated watersheds are commercial fertilizer (inorganic P) and cattle

feed (organic and inorganic P) (Reddy et al., 1999a). Graetz and Nair (1995) found that

18-20% of the P in upland surface soils of the northern Lake Okeechobee watershed was

organic, compared to 25-50% in the subsurface spodic horizons, and suggested that this

was due to leaching of organic P from surface horizons to subsurface layers. This form

of P, which is not readily adsorbed to negatively charged soil particles, is susceptible to

leaching from sandy upland soils (Reddy et al., 1996b). Organic P can be transported

from upland soils of this region to a wetland system through lateral groundwater and

surface flow (Graetz and Nair, 1995; Campbell et al., 1995). In most cases, organic

forms of P are not directly available for biological uptake. Enzymes called phosphatases

are produced by many plants, algae, and microbial species to mineralize organic P into

inorganic phosphate anions (e.g., H2P04-), making it bioavailable (Troeh and Thompson,

1993; Brady and Weil, 1999).

Dissolved inorganic P (DIP) is directly bioavailable, so P concentrations in this

form are of greatest concern for water quality (Reddy et al., 1995). The adsorption of

DIP to soil particles makes it unavailable for biological uptake and is of interest for

mitigating excess P in aquatic systems. Organic matter is predominantly negatively









charged so has little capacity to bind DIP. However, DIP can react with organometallic

molecules: Fe+3 and Al+3 cations that are associated with organic matter molecules (Zhou

et al., 1997). Phosphate ions can also be adsorbed to positively charged sites on clays

and metal hydroxides (Brady and Wiel, 2002). Another adsorption process is ligand

exchange, whereby a phosphate ion replaces a surface hydroxyl (OH-) on metal

hydroxides (Rhue and Harris, 1999). After longer periods of time, and with sufficient P

concentrations, P can diffuse into amorphous and poorly crystalline metal oxides (Rhue

and Harris, 1999). Once critical P concentrations are exceeded, precipitation with Fe,

Al, or Mn, (to form insoluble hydroxyl phosphates) can occur (Brady and Weil, 2002).

Phosphorus in Wetlands

Wetland soils can provide long-term storage of P with a turnover time of 96 years

in contrast to herbaceous plants turnover time of two years (Johnston, 1991). IfP inputs

to wetland components are greater than outputs, a wetland is a sink for P (Johnston,

1991). Phosphorus retention in wetlands is not considered long term storage unless soil

or litter is accumulating with time because the number of P adsorption sites on soil and

organic matter is finite (Johnston, 1991; Reddy et al., 1999b). The ability of wetlands to

retain P depends on soil physico-chemical properties, the amount of P that has been

loaded to the system, and hydraulic retention time (Reddy et al., 1998). These factors

control P fluxes among components such as vegetation, soils, plant litter, and overlying

water (Johnston, 1991). Some major aspects of the P cycle are shown in Figure 2-1.

Dissolved inorganic P in the water column and soil can be taken up by vegetation

and other living organisms. When plant leaves senesce and organisms die, organic P can

accumulate in layers of detritus when mineralization rates slow down due to anaerobic

conditions in the flooded soils. This can be a major long term storage component of P in









wetland ecosystems (Reddy et al., 1993; Reddy et al., 1996b). If the P concentration of

water loaded to a wetland is higher than the P concentration in interstitial soil water, P

can diffuse from the water column to the soil porewater. If the P concentration of loading

waters is lower, P will be released from the soil porewater and the wetland may act as a

source for P (Reddy et al., 1995). If similar gradients exist between the porewater and

the soil particles, P can either be sorbed or desorbed by soil particles. IfP is adsorbed to

soil particles for long periods of time, they may diffuse into the particle making it less

bioavailable. This can be a long-term form of P storage until the gradient is reversed for a

long period of time.

At the soil-water interface, there can be an oxidized layer where P can precipitate

with ferric iron (Fe+3) (Patrick and Khalid, 1974). However, deeper in the soil profile,

low redox potential (below 120 my) can cause reducing conditions that reduce Fe+3 to

ferrous form (Fe+2), releasing previously bound P (Faulkner and Richardson, 1989).

Phosphorus can also precipitate with aluminum (Al) to form aluminum phosphate which

is stable in acidic soils (Richardson, 1999). This reaction is not affected directly by redox

conditions, but rather by pH (Reddy et al., 1998; Kadlec and Knight, 1996; Mitsch and

Gosselink, 2000). In alkaline wetland environments, P can bind with Ca and Mg.

Changes in soil and surface water pH can cause these compounds to become unstable,

and release P (White and Taylor, 1977). Two examples of processes affecting pH are

algal growth, which may increase water column pH, and decomposition of organic

matter, which releases organic acids to lower the pH (Mitch and Gosselink, 2000; Troeh

and Thompson, 1993). Liming and manure application to pastures may also raise pH.









Low redox conditions can slow anaerobic microbial decomposition processes

(Levesque and Mathur, 1979) leading to the accumulation of organic matter. In some

wetlands, organic matter accumulation can be a major mechanism for long term P storage

(Richardson, 1999; Reddy et al., 1996b).

Aluminum and Fe hydroxides in amorphous and poorly crystalline forms (oxalate-

extractable) govern P sorption in many wetland soils and stream sediments (Reddy et al.,

1995; Richardson, 1999; Patrick and Khalid, 1974) because of the high surface areas that

provide many P binding sites (Rhue and Harris, 1999). Flaig and Reddy (1995) described

the different fractions of total P stored in wetland soils of different land uses in the

Okeechobee basins. Phosphorus associated with Fe- and Al-oxyhydroxides (amorphous

forms) accounted for 17-43% of the TP in wetland soils. Calcium- and magnesium-

bound soil P accounted for 50% of TP in a dairy wetland, which received dairy waste for

over 20 years. At nondairy sites within the basin they found that Ca- and Mg-bound P

was negligible.

Role of Wetlands for Retaining P

Most wetlands that are restored or created have goals of improving water quality

and habitat (Mitsh and Gosselink, 2000). Treatment wetlands have been constructed all

over the world for the purpose of mitigating nonpoint source pollution in agricultural

watersheds (Mitsch and Gosselink, 2000). In a review of treatment wetlands, Kadlec and

Knight (1996) found that treatment wetlands (on average) removed 31% of the total

phosphorus, and 41% of the inorganic phosphorus loaded to them. Other studies have

shown that natural wetlands can remove P from downstream waters (Yin and Lan, 1994;

Richardson, 2003; Nessel and Bayley, 1984), but can also be a source for P (Raisin and

Mitchell, 1995; Reddy et al., 1995) (depending on the P concentration of loading waters).










The two major compartments of P storage in wetlands, are the accumulation of

organic matter, and the adsorption sites on soil particles (Richardson, 1999). As wetlands

in the Okeechobee basins continue to be drained, detrital layers will oxidize and release P

to surface waters. As more P is loaded to wetlands through continuing agricultural

practices, fewer binding sites for P will be available in wetland systems, and the wetlands

can become a source for P when loaded with low P-concentration waters. In the sub-

basins of the Lake Okeechobee watershed, Reddy et al. (1996a) concluded that 45% of

the P storage capacity of wetlands was still available. Wetlands are in need of further

study to determine the mass of P that could potentially be stored in them, and what kinds

of land management practices may enhance P storage in wetland soils.


vegetative assimilation of DIP

S. leaf senescence, dying organisms


particulate P, ----- surface outflow


if acidic, P binds
with Al andFe;
Ca-, Mg-boundP released


if alkaline, P
binds with
Ca and Mg;
Fe-, Al-boundP
released
Adsorption to
organometallic complexes


Organic matter t
(anaerobic) accumulation to ..subsurface
store organic P outflow
Fe reduction, P release

Figure 2-1. Selected P cycle components in wetlands (not to scale). The dotted line
varies vertically with hydroperiod. DIP = dissolved inorganic P.














CHAPTER 3
PHOSPHORUS OBSERVATIONS IN NONRIPARIAN WETLANDS

Introduction

A recommendation for mitigating high P loads to surface waters of the four priority

basins of the Lake Okeechobee watershed is to restore the hydrology of historically

isolated wetlands (SFWMD et al., 2004a; Bottcher et al., 1995). It is estimated that 45%

of these wetlands have been ditched to reduce the wetland area, and to increase grazing

pasture area (SFWMD et al., 2004a). Many studies of upland soils in the Okeechobee

basins have been conducted (Mansell et al., 1991; Graetz and Nair, 1995; Nair et al.,

1998; Graetz et al., 1999; Nair et al., 1999; Villapando and Graetz, 2001; Nair and

Graetz, 2002; Pant and Reddy, 2003). Several studies of riparian wetlands have been

conducted (Scinto, 1990; Reddy et al., 1995; Reddy et al., 1996a; Reddy et al., 1996b;

Reddy et al., 1998), but there is little information about historically isolated wetlands in

the Lake Okeechobee basins (Steinman et al., 2003; Sperry, 2004). These wetlands

represent over half of the wetland area in the four priority basins (based on an analysis

described in this chapter). As a result, they potentially represent a large storage

compartment for P, which is currently unknown.

Understanding the role of these wetlands in P storage at different scales, is

important for assessing long-term efforts to reduce P loads to Lake Okeechobee. It is

important to characterize the extent of these wetlands, and determine how much total

phosphorus (TP) the soils are currently storing. Factors that influence P storage at

multiple scales in historically isolated wetlands, have not been addressed in this region.









This chapter will investigate factors that affect TP storage in wetlands at field- and

watershed-scales.

Land Use

Landscape conditions at watershed-scale can influence nutrient loadings to surface

waters (Hunsaker and Levine, 1995; Borsuk, et al., 2002). It is widely accepted that the

land use and land cover (LULC) of a landscape have a major effect on water quality

(Herlihy et al., 1998; Cuffney et al., 2000; Berka et al., 2001). Strong relationships exist

between LULC and water quality as related to P (Hall and Schreier, 1996; Bolstad and

Swank, 1997; Wear et al., 1998). Reddy et al. (1996a) found that TP stored in upland

surface soils (ranging from 18-24 cm in depth) of the Lake Okeechobee watershed

increased with intensity of land use: mean values of TP were 30 mg kg-1 in native

unimpacted (rangeland) areas, followed by 46 mg kg-1 in forage areas (unimproved

pastures), 84 mg kg-1 in improved pastures (fertilized, artificially drained and possibly

planted with forage grasses), and 1,792 mg kg-1 in intensive areas (dairy pastures). Total

P concentrations in the A horizon of manure-impacted upland soils of the priority basins

were 34 kg P ha-1 in nonimpacted (rangeland) soils, 165 kg P ha-1 in improved pasture

soils and 1,680 kg P ha-1 in high manure-impacted dairy soils (Graetz and Nair, 1995).

Up to 80% of TP had the potential to leave heavily-manured upland regions, while less

than 10% of the TP was likely to leave a low manure-impacted pasture soil (Graetz and

Nair, 1999). This suggests that more TP is transported to wetlands and ditches of more

intensely manure-impacted soils. Studies of historically isolated wetland soils in the

Lake Okeechobee watershed found that TP in wetland soils was higher in areas with

more intensively grazed and fertilized pastures (Steinman et al., 2003; Sperry, 2004).









Sperry (2004) found that ditched wetlands in improved pastures released 5 to 7 times

more TP in surface water runoff than wetlands in semi-native pastures.

Hydrology

Studies have demonstrated that wetlands with surface connections to downstream

waterways export nutrients (Nessel and Bayley, 1984; Yin and Lan, 1994; Raisin and

Mitchell, 1995; Richardson, 2003; Mwanuzi et al., 2003; Sperry, 2004). Surface

connections may increase water velocities in wetlands to suspend and transport P-laden

sediments (Phillips, 1989; Mitsch and Gosselink, 2000). Draining of wetlands can reduce

their size and expose previously submerged organic matter which may oxidize and

release labile P to surface waters (Levesque and Mathur, 1979; Whigham and Jordan,

2003;). Campbell et al. (1995) concluded that surface drainage (ditches and canals) could

be a significant mechanism for P transport to Lake Okeechobee.

Phosphorus Gradient and Hydrologic Zones

Properties of upland soils in the Lake Okeechobee watershed, which are mainly

Spodosols, have important influences on the transport of P to wetland areas. The A

horizon is sandy with some organic matter (2-5%) and is generally 15 to 20 cm thick

(Graetz and Nair, 1995; Lewis et al., 2001). The E horizon consists primarily of white

sand as a result of the eluviation of clay and organic matter to the Bh (spodic) horizon

(Graetz and Nair, 1995). The A horizon has very low P retention capacity and the E, or

albic, horizon has almost none (Mansell et al., 1991). Phosphorus binds with

organometallic complexes in the spodic horizon which creates a stable pool of relatively

insoluble P (Yuan, 1966; Reddy et al., 1996a). While the spodic horizon is thought to

have a high retention capacity for P, drainage water may not come into contact with this

horizon, as water table levels are above it during much of the rainy season (Graetz et al.,









1999; Blatie, 1980). The poor P retention and high water permeability of these sandy

acidic surface soils (as well as the route of P transport via surface and subsurface flow

along eluviated horizons) make these soils ineffective for P removal, which leads to a

high P concentration in the drainage water (Campbell et al., 1995). Thus upland soils

adjacent to wetlands should have less P than the more organic wetland soils that receive

drainage waters.

Phosphorus reactivity in wetland soils is known to depend on hydrologic conditions

(Mitsch and Gosselink, 2000). Wetland centers typically have more undecomposed

organic matter because of longer hydraulic retention times compared to wetland edges

and this can be the dominant long-term P storage pool in wetlands (Richardson, 1999;

Reddy et al., 1996b). Higher amounts of amorphous, and less crystalline forms of iron

(Fe) and aluminum (Al) (which are thought to be the most important agents of P sorption

in acidic soils) are also found in soils that are flooded for longer periods (Reddy et al.,

1995; Richardson, 1985; Patrick and Khalid, 1974). In a study of 12 improved pasture

wetlands in a beef ranch in the Lake Okeechobee watershed, surface soil (0-15 cm) TP in

wetland centers was significantly higher (242 mg kg-1) than wetland edges (171 mg kg-1)

(Sperry, 2004). When these numbers were adjusted for bulk density, there was no

significant difference (Sperry, 2004). Characterizing differences of P storage in wetland

hydrologic zones may be important for assessing the potential increase in P storage due to

hydrologic restoration.

The surface soil layers of the study area have been shown to contain the most TP

and higher P sorption capacities (Scinto, 1990; Nair et al., 1995; Rechcigl and Bottcher et

al., 1995). My study surveyed the surface soils of historically isolated wetlands in order









to characterize TP storage and investigate relationships between TP and variables at the

field- and watershed-scales. Information collected in this chapter was used in Chapter 4

to predict TP storage in all historically isolated wetlands of the four priority basins.

Hypotheses

* A zonal gradient exists within historically isolated wetlands, with TP storage
decreasing from wetland centers, to wetland edges, to uplands.

* Hydrologic connectivity affects TP storage, which decreases, as ditching intensity
increases.

* Adjacent land use affects TP stored in wetlands, with highest TP stored by
wetlands surrounded by dairies, followed by improved pastures, followed by
unimproved pastures and rangelands.

Objectives

* Characterize extent, vegetative communities and surrounding land uses of
historically isolated wetlands in dairies and beef-cattle pastures.

* Survey biogeochemical characteristics of historically isolated wetland surface soils.

* Investigate relative TP storage in the surface soils of two wetland hydrologic zones
(center and edge) and surrounding upland.

* Investigate relationships between wetland surface soil TP and the associated
hydrologic connectivity, vegetative community, and surrounding land use.

Materials and Methods

Study Area

The four priority basins are north of Lake Okeechobee and cover an area of

121,000 ha, with 101,200 ha (84%) in Okeechobee County (Figure 3-1). The climate is

subtropical, with a summer wet season and a winter dry season. The rainfall average,

since 1948, is 127 cm per year in the southeastern part of the study area (Southeastern

Regional Climate Center, 2004). Most of the rainfall (75%) occurs during the wet

season, between May and September (Shih, 1983). There is little topographic relief (0-










2% slopes), and the water table is near or at the soil surface during much of the wet

season (Blatie, 1980). It can drop to 2 m below the surface during dry months (Knisel et

al., 1985). The underlying geology is unconsolidated marine sediments, primarily sand

and gravel, with clay lenses (Parker, 1955). A series of three confined aquifers occurs at

depths greater than 40m: the Surfical Aquifer System, the Intermediate Aquifer System

and the Floridan Aquifer (Miller, 1997). Major streams that drain to Lake Okeechobee

include the Kissimmee River, Fisheating Creek, and Taylor Creek (FDEP, 2001).


Figure 3-1. Lake Okeechobee with 41 watershed basins and county lines in Florida,
USA. The study area consists of four priority basins colored orange.

According to an analysis of the Florida National Wetland Inventory (NWI) (US

Fish and Wildlife Service (USFWS), 2002) (described later in this chapter), 18% of the


other basins of the Lake Okeechobee watershed
Kilometers
0 100 200









area consists of wetlands, but this was historically estimated to be 25% (McCaffery et al.,

1976). Forty-one percent of the areal coverage of wetlands is associated with natural

streams, and the rest are mostly small (typically less than 1 ha), depressional, nonriparian

wetlands. They are usually dominated by emergent macrophytes, forest, or scrub-shrub

vegetation. It has been estimated that 45% of the wetlands in the study area are

artificially drained (SFWMD et al., 2004a).

Upland soils of the Lake Okeechobee watershed are mainly Spodosols formed in

sandy marine sediments (Graetz and Nair, 1995) which have a naturally low native P

content (Hodges et al., 1967). Entisols are also common, having weakly formed Bh

horizons (Lewis et al., 2001). According to soil data of the area (U.S. Department of

Agriculture /Natural Resource Conservation Service (USDA/NRCS), 1995), historically

isolated wetland soils are dominated by Aquepts and Aquents and to a lesser extent,

Aquolls and Aqualfs. The Spodosols are mostly Alaquods. Immokalee (sandy, siliceous,

hyperthermic Arenic Alaquod) and Myakka (sandy, siliceous, hyperthermic Aeric

Alaquod) are the most common soil series occurring in the upland areas. Alaquods are

wet, sandy soils that have a fluctuating water table that moves iron out of the soil profile

(Soil Survey Staff, 1999). They have high humus content, and the spodic horizon is

mostly an accumulation of Al and organic matter. Despite the high hydraulic

conductivities of these soils (>16 cm hr1), drainage is poor (Lewis et al., 2001).

Drainage has been improved with extensive ditching of fields and wetlands to convey

stormwater runoff towards Lake Okeechobee (Haan, 1995). Ditching densities in the

landscape increase with land-use intensity from unimproved pastures to improved

pastures to intensively managed pastures (dairies) to row crops (Heatwole, 1986).









Based on 2003 land-use data (SFWMD, 2003), the four priority basins are 64.2%

agriculture, 7.9% urban, 15.1% wetlands, 6.2% upland forests, 3.6% is transportation /

utilities, and 1.5% water. Agriculture consists of 47.5% improved pastures, 7.2% dairies,

and 6.2% unimproved pastures and rangeland. Citrus, row crops, and sod farms make up

3.0%, 2.0%, and 0.5% respectively. Dairies and beef-cattle operations of the study area

contribute the majority of P to Lake Okeechobee in equal amounts, but dairy inputs are

much more concentrated (more P per unit area) (MacGill et al., 1976; Allen et al., 1982;

Anderson and Flaig, 1995; Reddy et al., 1995).

Spatial Data

Spatial data layers were assembled and analyzed in a geographic information

system (GIS) to characterize wetlands and select sampling sites. The GIS software

ArcGIS version 8.3 that consists of ArcCatalog, ArcMap and ArcToolbox modules

developed by Environmental Systems Research Institute (ESRI Inc.; Redlands, CA) was

used. The layers included watershed boundaries, counties, waterways, land use,

wetlands, and digital orthorectified quarter quadrangles (DOQQ) aerial photos (metadata

is in Appendix A). Watershed boundaries delineated the four priority basins and

waterways consisted of canals, streams and major tributaries (Figure 3-2).

Land use was based on 2001 conditions (Figure 3-3) and employed three levels of

land-use codes, according to the Florida Land Use and Cover Classification System

(FLUCCS) (Florida Department of Transportation (FDOT), 1999). The NWI provided

the basis of spatial wetland information in this study. Associated with each NWI

polygon, was a hierarchical wetland classification system with four levels: system,

subsystem, class and subclass. The class represented the appearance of the wetland in









terms of vegetation substrate (USFWS, 2002), and in this thesis it is referred to as the

vegetative community.


Figure 3-2. Four priority basins, counties and waterways.

The 1-meter resolution 1999 DOQQ photos were used throughout this study for the

verification of polygons of other data layers including wetlands, land use and hydrologic

connectivity. All layers were projected in High Precision Geodetic Network (HPGN),

1983 North American Datum (NAD 83), Albers Equal Area projection. Projection

parameters are listed in Appendix A. A glossary of GIS terms used in this text can be

found in Appendix B. More information about GIS concepts and map projections can be

found in Bolstad (2002).











































Figure 3-3. Land use (2001) in the priority basins. White areas represent other land uses
(e.g., cropland, residential, upland forests, commercial) (SFWMD, 2003).

Generating Layers for Site Selection

The goal of the GIS analysis was to create 6 layers from which to select samples

based on population proportions of representative wetland vegetative communities within

the most common land uses. The land use treatment had 3 levels: dairy (DAI), improved

pasture (IMP) and unimproved pasture / rangeland combined (UNIMP). The vegetative

community treatment had two levels: emergent marsh (EM) and forested / scrub-shrub









combined (FOSS). These two treatments resulted in 6 categories. Figure 3-4

summarizes the steps to create these GIS layers which are described in detail in the

following sections.


Figure 3-4.


Procedure for creating spatial layers and selecting sample sites. EM =
emergent marsh, FO = forested, SS = scrub-shrub, DAIR = dairy, IMP
improved pasture, UNIMP = unimproved pasture/rangeland.









Step 1: Creating riparian and nonriparian wetland layers

The NWI wetland layer was used to create two layers representing riparian and

nonriparian wetlands. Within ArcMAP, NWI and the three waterway layers (canals,

streams and major tributaries) were clipped to the four basin boundaries. Wetlands that

were cut off at basin edges during the clipping were deleted to omit them from possible

selection as sampling sites, but were included in calculations to quantify wetlands in the

study area.

To build the riparian wetland layer, the three hydrography layers were assembled

with the NWI wetland layer. Polygons that fell within 50 m of canals, 25 m of major

tributaries or 25 m of streams, were selected and saved as a new layer. These distances

were chosen based on trial and error attempts to capture the most riparian wetlands, and

the least number of nonriparian wetlands (based on visual verification using the DOQQs).

Wetlands that followed stream direction, those connected to waterways, and those

forming chain patterns with interconnecting natural streams, were considered riparian.

Wetlands that were contiguously connected to the riparian wetlands were also selected

(Figure 3-5A). These were then merged to the first set of riparian wetlands to create a

new set of riparian wetlands. The remaining wetlands were saved as the nonriparian

wetland layer.

Because the stream and major tributary layers included both natural and man-made

hydrological features (i.e., small canals and ditches), some wetlands were misclassified as

riparian during the process described above. Polygons adjacent to or cut through by

man-made ditches were reassigned to the nonriparian wetland layer, based on visual

interpretation of DOQQs (Figure 3-5B). On the other hand, some wetlands were










misclassified as nonriparian, since some streams (evident on the DOQQs) were not

represented by any of the hydrography layers.















SNWI polygons 0 250 500 Meters rparan wetlands 0 0 25 0 5
streams and major tributaries --prevous se s npa
25m buffer
nparian wetlands ma but
A B

Figure 3-5. Selecting riparian wetlands. A) NWI polygons intersecting stream and
major tributary buffers (25 m wide) plus wetlands contiguous to those. B)
Yellow polygons and green polygons represent wetlands classified based on
distance to waterways. Orange polygons were originally selected as riparian
based on distance to waterways but then reclassified as nonriparian because
the waterway was a manmade structure.

Both layers were visually verified: wetlands following stream direction, those

connected to waterways, or those forming chain patterns with interconnecting natural

streams, were considered riparian wetlands. If a polygon was cut from one layer, it was

saved into the other one so that both layers were mutually exclusive and comprehensive

(Figure 3-6).

Because EM, FO and SS wetlands represented 88% of the historically isolated

wetlands, polygons other than these were selected and deleted leaving an "EM-FO-SS"

nonriparian wetland layer (Figure 3-7). The deleted vegetative communities included

aquatic bed, unconsolidated shore, open water, and unconsolidated. Distributions of all

NWI wetlands in the study area are listed by vegetative community in Appendix C.












nonriparian

riparian
basins


0 5 10 21


Lake Okeechobee


Figure 3-6. All NWI wetlands classified as riparian and nonriparian.

Step 2: Creating nonriparian emergent marsh, forested and scrub shrub wetlands

Within the EM-FO-SS nonriparian wetland layer, there were neighboring polygons

of the same type, but they had different hydrological regimes according to data in the

NWI (Figure 3-8A). For purposes of selecting a random sample based on vegetative

community, these polygons were dissolved based on the vegetative community. To do

this, the polygon layers were first converted to a vegetative community raster with spatial

resolution of 5 m (values were EM, SS or FO) (Figure 3-8B). The raster was then

converted to polygons (Figure 3-8C), essentially dissolving neighboring wetland









polygons of the same vegetative community type together. Wetlands of type EM were

exported into a new separate layer. There were few FO and SS sites, so wetlands of these

similar vegetative community types were exported into one FOSS layer.


Figure 3-7. Nonriparian wetlands by wetland vegetative community.

Step 3: Creating land-use layers

To select nonriparian EM and FOSS wetlands based on the land use they overlaid,

three new land-use layers were needed: IMP, UNIMP and DAIR. The DAIR layer

already existed as a separate layer. New IMP and UNIMP layers needed to be created

but IMP and UNIMP polygons were interrupted by wetland land-use polygons (WETL)









(Figure 3-9A). These land-use polygons often corresponded with NWI polygons but the

NWI layer was more complete and accurate. In order to identify NWI wetlands that

coincided to IMP or UNIMP land uses, the WETL polygons had to be incorporated into

the IMP or UNIMP layers.


n flssi
-SS


B


Figure 3-8. Three steps to generalize polygons by vegetative community (EM, FO or
SS). A) Original nonriparian polygons. B) Conversion to 5 m raster cells. C)
Conversion back to polygons.

The 2001 land-use polygons were reclassified as IMP, UNIMP or WETL according

to FLUCCS codes. Polygons with FLUCCS level-2 code equal to 211 were classified as

IMP. Those with FLUCCS level-2 code equal to 212 (unimproved pastures) or 213

(woodland pastures), or FLUCCS level-1 code equal to 300 (rangeland), were classified

as UNIMP. Polygons with FLUCCS level-1 code equal to 600 were classified as WETL.


lob









The WETL polygons that were completely contained by the IMP land-use layer were

selected, exported, and then merged with the IMP land-use layer (Figure 3-9B). The

WETL polygons occurring at the edges of the land-use polygons (that were at least 50%

within the land-use polygon based on visual judgement) were selected, exported, and

merged with the IMP land-use layer. These steps were repeated for the UNIMP layer.

These steps were not necessary for the DAIR layer, because all polygons within it were

already classified as DAIR. Figure 3-10 shows the three final land-use layers.















Improved land use
Island use wetlands

Figure 3-9. Reclassification ofland-use wetland polygons. A) Before reclassification, a
NWI wetland (circled in yellow) did not intersect IMP land use. B) After
reclassification, the NWI wetland (circled in yellow) intersected with IMP
land use.

Step 4: Create six treatment layers

Six new NWI wetland layers were created by selecting EM or FOSS wetlands that

were at least 50% contained within each of the three land-use layers. It was ensured that

the six new wetland layers (EM-DAIR, FOSS-DAIR, EM-UNIMP, FOSS-UNIMP, EM-

IMP, FOSS-IMP) were mutually exclusive.











































Figure 3-10. Land-use layer (2001) showing land-use areas of interest in the priority
basins after land-use reclassification. White areas represent other land uses
(e.g., cropland, residential, upland forests, commercial) (SFWMD, 2003).

Step 5: Stratified random sampling

Proportions represented by each of the six wetland strata (wetland type combined

with land-use type) were determined based on population polygon counts. Based on a

minimum sample size of 108, the number of wetlands to be sampled within each

treatment was determined. This minimum number was a result of an initial sampling

design based on three factors, each with three levels: land use, vegetative community and









hydrologic-connection types. These 27 treatments would have a minimum of 4 samples

each, thus the number 108. Since no a priori information was available for the wetland

hydrological connections, this became a random factor in the current sampling scheme,

and the number 108 remained as a commitment to the project's funding agency.

Random numbers for each of the 6 wetland layers were generated using SPSS

software version 11.5 for Windows (SPSS, Inc, Chicago, IL). The quantity of sites

selected was proportional to the populations of each strata in the sampling scheme.

These random numbers were used to select polygons based on polygon identification

numbers (unique, GIS-assigned integer attributes). A minimum of four sites per category

was adopted, with an added 25% in each category. This would provide extra sites if

wetlands did not exist at the designated geographic coordinates or if wetlands were

inaccessible. This produced a total number of 145 possible sampling sites.

Field Sampling Methods and Site Descriptions

I sampled 118 nonriparian wetlands within the research land uses between May 19

and November 19, 2003 (Figure 3-11). Details about site locations, sampling dates and

selected attributes are listed in Appendix D. Rainfall during the sampling period is

described in Appendix E.

Land use and vegetative community. Land uses were assigned to sites based on a

combination of the GIS land-use map and visual observation at the sites. Improved

pasture was defined as land which has been cleared, tilled, reseeded with forage grasses,

and periodically treated with brush control and fertilizer application (FDOT, 1999).

Unimproved pastures were defined as cleared land with major stands of trees and brush,

where native grasses have been allowed to develop, and is typically not managed with

brush control or fertilizer (FDOT, 1999). Rangeland was defined as land where the









dominant vegetation is predominantly native grasses, forbs, and shrubs, and is capable of

being grazed (FDOT, 1999). Management may include brush control, but is not irrigated,

cultivated, or fertilized. Dairy land was defined as any area pertaining to a commercial

dairy and has a wide range of usage intensity (based on fertilization or stocking rates)

including ungrazed fields, fertilized hayfields, sprayfields, feeding pastures, and cow-

barn areas. Any of these three land-use types may have man-made drainage systems.


Ni U EM-DAIR

0 rEM-IMP
S EM-UNIMP
D* FOSS-DAIR

E- E E- FOSS-IMP
D* FOSS-UNIMP
S, ^ o -- basins












Lake Okeechobee / E
0 5 10 2E ?im Eler




Figure 3-11. Sampled wetlands in the four priority basins of the Lake Okeechobee
watershed. EM= emergent marsh, FOSS forested / scrub-shrub, NIMP =
unimproved pasture / rangeland, IMP = improved pasture, DAIR = dairy.

A distinction between very improved (IMP1) and less improved (IMP2) pastures

was made and assigned to sites, in order to compare TP content among these additional









land-use classes. Less improved pastures had less Bahia grass and more native grasses,

or, had scrub present. A similar distinction was made for the unimproved pastures /

rangeland class. Unimproved pastures were classes as UNI1, and rangelands (which had

more scrub vegetation) were classed as UNI2. Within dairies, there was a variety of land

uses. Some areas were grazed by beef cattle (DABE), and some were pastures where

dairy cows sometimes received feed (DAPA). Some were in, or adjacent to, sprayfields

(DASP) or hayfields (DAHA). Several were in areas that were not currently being

managed with grazing or hay activities, and looked similar to unimproved pastures

(DAUN) (Table 3-1).

Many EM sites had large areas (more than 75% of the system) of open water. Due

to this variability, a third class was defined as emergent marsh/open water (EW). No EW

wetlands were sampled within the UNIMP areas and more than half the FOSS wetlands

were sampled within dairies (Figure 3-12).

75%
Land Use
i Dairy
E Improved
Unimproved
50%
Bars show percent

I.

25%




0%
EM EW FS
Vegetation Community

Figure 3-12. Percentage of sites by land use per vegetative community. EM= emergent
marsh, EW=emergent marsh/open water, F S=forested/scrub-shrub.











Table 3-1. Frequency, proportion and area of wetlands sampled by land use.
Land Use Freq. Percent Total ha
Dairy 21 17.8 32.86
Dairy Beef (DABE) 4 4.2
Dairy Hayfield (DAHA) 6 5.0
Dairy Pasture (DAPA) 4 3.4
Dairy Sprayfield (DASP) 1 .8
Dairy Unimproved (DAUN) 6 5.0
Improved pasture 85 72.0 97.14
Very improved (IMP1) 51 42.9
Less improved (IMP2) 34 28.6
Unimproved pasture / rangeland 12 10.2 24.4
Unimproved Pasture (UNI1) 7 5.9
Rangeland (UNI2) 5 4.2
Total 118 100.0 154.4

Hydrologic zone delineation and wetland size. Vegetation and topography

change were used to distinguish hydrologic zones. Functional vegetation classes (often

found exclusively within a particular hydrologic zone) were defined and used to

demarcate boundaries between upland, edge, and center zones (Table 3-2).

Table 3-2. Distribution of vegetation classes in hydrologic zones of sampled wetlands.
Center Edge Upland
---------------------------------%-----------------------------
Aquatic vegetation 4.2
Open water 12.7 --
Panicum hemitomum Shult. 11.9 --
Trees 5.1 0.8 3.4
Shrubs 2.5 7.6 9.3
Juncus effuses L. 6.8 43.2 --
Pontedaria cordata L. 25.4 --
Polygonum spp. 17.8 -- --
Other native herbaceous 5.9 9.3 4.2
Forage grassesb -- 8.5 63.6
Other grasses' 7.6 14.4 19.5
Mixed transitional 0 16.1 --
a For example, Nuphar luteum. and Nymphea spp.
b Often Paspalum notatum Fluegge (Bogdan) (Bahia grass); also Cynodon spp. (stargrass)
and Hermarthria spp. (limpograss).
SFor example, Andropogon spp. and Panicum spp.
d Refers to a mix of small herbaceous plants which often contained Hydrocotyle spp.









The transition between zones was often diffuse, due to overlap of species.

Demarcation between upland and edge zones was often based on the waterward extent of

Bahia grass (Paspalum notatum Fluegge (Bogdan). In most IMP emergent marshes, Soft

Rush (Juncus effusus L.) was the dominant species in the edge zone, and could often be

used to demarcate the outer or inner extent of that zone (Figure 3-13). A complex of

emergent macrophytes, which often included Hydrocotyle spp., Centella spp. and

Eleocharis baldwinii (Torr.) Chapm., was often the outermost wetland edge zone

vegetative community, and was referred to as "mixed transitional" (Figure 3-14).























Figure 3-13. An example of zonation within an isolated wetland. In this example, the
center is the area containing surface water, Polygonum hydropiperoides
Michx., and Pontedaria cordata L. The edge consists of Juncus effuses L.
and "mixed transitional." The edge ends where Paspalum notatum Fluegge
(Bogdan) (Bahia grass) dominates.

Although demarcation lines between zones were not always easily defined,

vegetation differences were easily identified, providing a high degree of confidence that

samples were representative of the targeted hydrologic zone. The outer wetland edge was









usually walked and recorded using a Pathfinder Pro XR (Trimble Navigation LTD,

Sunnyvale, CA) GPS receiver. The distribution of wetland sizes was right-skewed. The

median area of sampled wetlands was 0.9 ha for EM, 1.5 ha for FS and 1.0 ha for EW.



















Figure 3-14. An example of the "mixed transitional" vegetation class showing
Hydrocotyle umbellata L., Eleocharis baldwinii (Torr.)
Chapm. and Diodia virginiana L.

Soils. Four 10 cm soil cores were extracted and composite for each hydrologic

zone (Figure 3-15). If an outlet ditch was present, a three-core composite was collected

from the base of the ditch beyond the wetland edge. Outlet ditches were identified by

evidence of water flowing from the wetland or topography change that indicated outward

flow. Upland cores were taken between three and twenty meters from the wetland edge.

To take a soil core, thick-walled (0.3 cm) polycarbonate tubes (7.6 cm internal

diameter) sharpened at one end were hammered in with rubber mallets and extracted by

hand. The top 10 cm of the soil core was extruded upward into a separate 10 cm section

of tubing and then placed into a zip closure plastic bag. The extruder was a PVC pipe

fitted with a rubber stopper designed to prevent water from draining from the sample.










Living vegetation was cut off at the soil surface with a knife. Thicker portions of

flocculent layers were kept as part of the 10 cm depth. Litter on the cores was discarded

and roots were left intact.

If a large wetland had more than one center area, one of the areas and the adjacent

edge and upland was sampled. On two occasions when the wetland was surrounded by

wet prairie, the wet prairie was sampled as upland. Samples were immediately stored on

ice and transported to the laboratory within 80 hours, and usually within 36 hours.

u











u = upland
\ u



de = edge

d c = center
d = ditch



Figure 3-15. Zones of soil sample collection.

Comparison to initial sampling scheme. Table 3-3 shows sampled wetlands

compared to the initial sampling scheme. Forested / scrub-shrub sites were under-

represented in the IMP and UNIMP land-use types, in part, because upon visiting these

sites, they were actually riparian or non-wetlands or the land use had changed.

Unimproved pastures and rangelands were also under-represented because many had

been converted to improved pastures or row crops.









Table 3-3. Sampling scheme of historically isolated wetlands by land-use type (2001)
and vegetative community. Proportions based on wetland polygon count.
EM = emergent marsh, FOSS = forested / scrub shrub, EW = emergent
marsh / open water.
Land use Vegetative Polygon Minimum Randomly Sampled
Community proportion samples selected**
Dairy (FOSS) 1.0 1- 4 6 6
Dairy (EM) 10.1 11 15 15 (6 EW)
Improved pasture (FOSS) 5.0 5 7 4
Improved pasture (EM) 68.0 78 96 81 (15 EW)
Unimproved pasture / 1.5 2 4 6
rangeland (FOSS)
Unimproved pasture /
anea 10.2 11 15 11 (0 EW)
rangeland (EM)
Total 108 145 118
Determined as percent of 108 minimum samples; minimum per category is 4.
** After adding 25%.

Hydrologic connectivity. Hydrologic connectivity was characterized as isolated,

flow-through, head-of-ditch, tangent, subsurface or end-of-ditch (Figure 3-16). Classes

were defined based on effects of connectivity on nutrient export. Isolated wetlands had

no ditch associated with them. Flow-through wetlands had at least 2 ditches which

influenced the center of the wetland. Head-of-ditch wetlands had at least one ditch

carrying water away from the wetland. Tangent wetlands were connected to a ditch

which did not directly influence the center of the wetland. Subsurface site edges were

within 30 m of a stream or major ditch (USDA/NRCS, 1982). End-of-ditch wetlands had

one or more ditches bringing water to the wetland (some wetlands were being drained

into others). Ditches were also classified according to intensity: 3 (major), 2

(intermediate) or 1 (minor) (Figure 3-17). Minor ditches were unmaintained, shallow

(less than 15cm deep), and narrow (less than Im). Intense ditches were wide (more than

Im) and deep (more than 40 cm). Intermediate ditches were in between. Less intense

(smaller) ditches were usually more vegetated.

























































E F

Figure 3-16. Detail areas of DOQQ photos showing hydrological connection types. A)
isolated (a fence is on the left side), B) head-of-ditch, C) flow-through, D)
subsurface connection, E) tangent, and, F) end-of-ditch.


Meters
0 25 50 100


Meters
0 25 50 100

























VC A









t 'NI




I FALB















C

Figure 3-17. Three intensities of man-made drainage ditches: A) minor B) intermediate
C) major.












About half of the sampled wetlands had no surface connection to waterways (they


were classified as subsurface or isolated), and 32% of them were isolated (Figure 3-18).


Improved pasture wetlands had more ditches than wetlands in other land uses. The


majority of wetlands in DAIR and UNIMP areas had one or no ditches (Figure 3-19). In


the IMP areas, about half of the wetlands had intensity equal to two or greater (Figure 3-


20).


100%
90% 2 1 2 20
80%
70%- -- 13- 23 Tangent
60% -10S ubsurface
O Isolated
50%-00 - --3
50% 3 13 9 eo Head of Ditch
40% 3 Flowthrough


10% 2 4
30 End of Ditch











Figure 3-18. Ditch class by land use. Numbers in the bars represent number of wetlands.


100% 5 5
90% 2 9 6 1 6 19
80% -
70%
60%
E2 2 ditches
50%1
40%
4O ditches
30%
20%


0%










Figure 3-19. Ditch number by land use. Numbers in the bars represent number of

wetlands.











100%
90% -
80% -
70% 3 1 Int = 3
60% Int 2
50% 1i
40%0 Int = 1
30% Int = 0
20%
10%
0%














Composite soil samples were manually homogenized. Roots larger than 2 mm in

diameter, and live vegetation were removed. A subsample of the remaining composite

soil sample was weighed, and pH was measured (usually within 48 hours) using a 1:1,

soil to water, ratio (20g of soil to 20 mL of distilled, deionized water (DDI)). Moisture

content was determined as the difference between the wet and dry weights of an oven-

dried (70C for three days) sample. Bulk density was calculated on a dry-weight basis

using the known volume of soil cores. A subsample of each homogenized soil was

placed in open tubs in a greenhouse to air-dry for approximately 3 weeks. The samples

were then hand-mixed once a week to facilitate drying. After air-drying, the samples

were brought back to the laboratory and placed in a 40C oven for 6 hours to stabilize the

soil moisture contente. Finally, samples were machine-ground and passed through a #100

mesh (0.15 mm openings) sieve.

Total P was determined on a 70C dried sample using the ignition method


(Anderson, 1976). Soils were ashed at 5500C for 4 hours in a muffle furnace. Loss on









ignition (LOI) was determined as the difference between the initial soil weight and ash

weight and is an estimate of organic matter content. Ashed samples were then digested

with 6MHC1 and filtered using Whatman #41 filter paper. Digestate was kept at room

temperature (20C) until being analyzed calorimetrically for soluble reactive P (SRP)

using the automated ascorbic acid method (Method 365.1; US EPA, 1993).

An estimate of total inorganic P (HCL-Pi) was determined after 0.5 g of soil was

extracted with 25 mL of 1 MHC1 (Reddy et al., 1998). This extraction was developed

based on an empirical relationship between total inorganic P (determined from a P

fractionation scheme) and HCl-Pi of organic soils collected from the Florida Everglades

(Reddy et al., 1998). Soil solutions were filtered through 0.45 tm membrane filters after

centrifuging (6000 rpm 10 minutes). The extracts were stored at 40C, and analyzed

calorimetrically for soluble reactive P (SRP) using the automated ascorbic acid method

(Method 365.1; US EPA, 1993). These soil extracts were also analyzed on an

Inductively Coupled Plasma Spectrophotometer (ICP) (US EPA, 1984) to determine total

calcium (Ca) and magnesium (Mg) concentrations.

To determine oxalate-extractable Al (Alox) and Fe (Feox), which represent

amorphous forms, soils were extracted in the dark with 0.175 M ammonium oxalate and

0.1 M oxalic acid at a soil to solution ratio of 1:40 for 4 hours (McKeague and Day,

1966). Extracts were filtered through 0.45 tm membrane filters, and analyzed using an

ICP (US EPA, 1984).

Data Management

Polygons and points from the GPS data collected in the field that delineated

sampled wetlands were converted to shapefiles in ArcGIS. If GPS data was incomplete

for a site, wetland shapefiles were created based on DOQQ interpretation and NWI









polygons. Using Xtools (DataEast LLC, Novosibirsk, Russia), area, perimeter and XY

coordinates were determined. This information was imported into a MS Access database

(Microsoft, Redmond, WA), and field notes about vegetation, land use, ditches, and edge

percent were added for each wetland record. Data from the laboratory analyses were also

stored in the relational database. Data queries of data were generated for statistical

analyses.

Statistical Analyses

Statistical analyses were performed using SPSS. Descriptive statistics of observed

values were listed for physical properties, pH, TP, and metals. Means of pH values were

calculated from H ion concentrations. Total P, HC1-Pi and metals were bulk density-

-2
adjusted to calculate g m-2 in the top 10 cm of soil by multiplying the bulk density

(g cm-3) by the respective sample's mass-based measure (mg kg -1) and multiplying by

10. Parametric comparisons among means were made on normalized data. Most data

were normalized by a natural log transformation. Bulk density and mass-based TP (mg

kg-1) were transformed by the square-root function. Data for each level to be compared

was tested for normality using the Kolmogorov-Smirnov test at a 95% confidence level

before a parametric test was run.

Differences among hydrologic zones. Descriptive statistics were listed by

hydrologic zone and comparisons among center, edge and upland zones were made using

one-way analysis of variance (ANOVA). If variances were equal according to Levene's

test (p < 0.05), least significant difference (LSD) post-hoc pairwise comparisons were

used. Otherwise, Games-Howell comparisons were used (SPSS, 2002).

Combined effects. For determining the effects of categorical factors on bulk

density, percent organic matter and TP, a generalized linear model using univariate









factorial analyses of variance (ANOVA) and Type III sums of squares were computed

with normalized dependent variables. Univariate factorial ANOVAs test the effects of

two or more categorical factors on one continuous dependent variable, as well as the

interaction between the two factors (Underwood, 1997; SPSS, 2002). The Type III sum

of squares method uses unweighted, or estimated means, (not biased towards a group

with the largest sample size) so can be used with different, and unproportional sample

sizes (SPSS, 2002).

Factorial ANOVAs were used to determine the effects of two fixed factors

(vegetative community and land use) on pH, bulk density, percent organic matter, and TP

(mg kg-1 and g m-2). Wetland area was a covariate (this was natural-log transformed).

Mixed model factorial ANOVAs were used to determine the effects of hydrologic

connectivity (a random factor), land use (a fixed factor), and wetland area (a covariate)

on TP (g m2) in wetland centers and edges and wetlands as a whole. The ditching

variables tested were: a number-of-ditches dummy, (value is 0 if number of ditches = 0, 1

otherwise) and a ditch-intensity dummy (value is 0 if intensity < 2, 1 otherwise). Dummy

variables were used in order to have enough degrees of freedom for the analyses.

If the interaction of two factors was found to be significant, conclusions about one

factor were made separately at each level of the other factor (Ott and Longnecker, 2001).

Error bar graphs were used to illustrate interacting variables (Ott and Longnecker, 2001;

Kinnear and Gray, 1999). Center and edge soils were analyzed separately to maintain

independence of samples. Homogeneity of variances was checked with Levene's test (p

< 0.05), but if there were at least 6 cases within each group, heteroscedasticity was

accepted (Underwood, 1997). The F-statistic and p-value were reported, and Bonferroni









pairwise comparisons of estimated marginal unweightedd) means of factor levels were

performed. Both the factorial ANOVAs and the Bonferroni pairwise comparisons were

tested at the 90% confidence level due to the coarse resolution, and high variability of the

landscape-scale treatments being tested.

Boxplots and error bars (showing 95% confidence interval of the mean) were used

to display observed distributions of data among treatments. The top of the box is at the

75th percentile, and the bottom of the box is at the 25th percentile. The box represents the

inter-quartile range and the horizontal line through the box represents the median. The

ends of the whiskers represent the largest and smallest values that are not outliers. An

outlier is represented by the symbol 0, and is defined as a value that is smaller (or larger)

than 1.5 times the inter-quartile range from the 25th (or 75th) percentile. An extreme

value (represented by the symbol *), was defined as a value that is smaller (or larger) than

3 times the inter-quartile range from the 25th (or 75th) percentile.

Correlations. Spearman correlations (rs) between TP and metals (mg kg-1) and

between TP (mg kg-1) and percent organic matter in center and edge soils were computed

(for all wetlands and by land use). Spearman correlations are computed from ranks, so

they express the proportion of variability accounted for between non-normally distributed

parameters (SPSS, 2002).

Wetland TP storage. Storage of TP (g m-2) in the top 10 cm of each wetland as a

whole was determined by multiplying the TP (g m-2) of each hydrologic zone by the

respective percentage of each zone, and summing the products. A univariate factorial

ANOVA and Bonferroni pairwise comparisons of treatments were performed at the 90%

confidence level with land use and vegetation community as fixed factors and wetland









area as a covariate. Another factorial ANOVA was performed to test if land use had an

effect on wetland size. Total storage of TP (kg) was calculated by multiplying that

number (g m-2) by the total area (m2) of the wetland and dividing by 1000 to convert g to

kg. Maps of land use and TP (kg) within each sampled wetland were generated using

ArcMap.

Results

Wetland Vegetative Communities and Surrounding Land Uses

Fifty-nine percent of the wetlands in the study area were historically isolated, and

56% of those were emergent marsh, forested, or scrub-shrub wetlands (Table 3-4). Most

historically isolated wetlands occurred within IMP areas.

Biogeochemical Measures in Hydrologic Zones

Physical properties and pH were significantly different among hydrologic zones

(Table 3-5). From centers to edges to uplands, bulk density increased while organic

matter, moisture content and pH decreased.

Overall, median TP (mg kg-1) was three times as high, and TP (g m-2) was 1.5 times

as high in wetland centers than in edges (Table 3-6). Total P by volume was significantly

higher in wetland centers compared to edges in IMP land-use areas, but not in DAIR or

UNIMP areas. The amount of volume-based HCl-Pi was similar in all hydrologic zones

(Table 3-7). In wetland center soils, median percent organic matter was 26.6% in DAI,

18.2% in IMP and 28.4% in UNIMP. For edges it was 15.6% in DAI, 11.4% in IMP, and

12.6% in UNIMP. There were no significant differences in metals between center and

edge hydrologic zones (Table 3-8).

Comparisons of other P extractions and metals among hydrologic zones and land

uses are in Appendix F.









Combined Effects of Land Use, Vegetative Community, and Wetland Size

Bulk density, organic matter and pH. There was a significant effect of

vegetative community on bulk density and percent organic matter in wetland centers and

edges (Table 3-9 and Figure 3-21). Land use had no significant effect on bulk density or

organic matter in either hydrologic zone. Boxplots by detailed land use are shown in

Figure 3-22. Land use and vegetative community each had a significant effect on H

concentration (Table 3-9). Estimated means and standard error per category were

converted to pH (Table 3-10). Wetland size had no significant effect on bulk density,

organic matter or pH.

Center TP. The interaction of vegetation community and land use had a

significant effect on center TP (mg kg-1 and g m-2) in two factorial ANOVAs (Table 3-

11). Wetland size had no significant effect on mass-based TP (mg kg-1) but

volume-based TP (g m-2) was lower in larger wetlands (Table 3-11 and Figure 3-23). The

fixed factor interactions are illustrated by the dotted lines having opposite-signed slopes

in Figure 3-24. When analyzed for the 3 levels of vegetation, land use only had a

significant effect for EM wetlands, and when analyzed for each level of land use,

vegetation community only had a significant effect within IMP wetlands (Table 3-11 and

Figure 3-25).

Edge TP. The interaction of vegetation community and land use on wetland edge

did not have a significant effect on soil TP. Wetland size had a significant effect: larger

wetlands had less TP per unit of soil. (Table 3-12 and Figure 3-26). Vegetation

community had a significant effect on mass-based TP (mg kg-1) but not on volume-based

TP (g m-2) (Table 3-12 and Figure 3-27). Land use had a significant effect on both

(Figure 3-27).









Correlations between TP and Metals

Overall, total P (mg kg-1) was most highly correlated with organic matter

percentage in wetland centers and edges (Tables 3-13 and 3-14). In center soils, Alox was

more highly correlated with TP, and in wetland edges, Mg and Feox were more highly

correlated with TP.

Effects of Ditches

Total P storage (g m-2). The ditch number dummy did not have significant effect

on TP g m-2) in centers, and neither dummy variable had a significant effect on edge soils

or on the wetlands as a whole. The ditch intensity dummy variable had a significant

effect on wetland center TP (g m-2) in a mixed model factorial analysis (F = 8.482; p =

0.004) but there was no significant difference between the two estimated means in the

Bonferroni comparisons at the 90% confidence level. The analysis was repeated for only

IMP land-use areas to reduce variability. The ditch intensity dummy variable had a

significant effect on center TP (g m-2) (F = 3.737; p = 0.057) and the two class estimated

means of the two classes were significantly different (p = 0.057). The estimated mean TP

-2
for ditch intensity class 0 (ditch intensity of 1 or less) was 22.02 4.83 g m-2 and for

-2
class 1 (ditch intensity of 2 or more) was 29.50 7.27 g m-2

Ditch class and TP (mg kg-1). There were not enough samples per class to

conclude that TP means were significantly different among ditch classes but interestingly,

TP (mg kg-1) in center soils varied similarly according to ditch class, regardless of

vegetative community or land use. The differences in TP among ditch classes indicated

that the TP in head-of-ditch, flowthrough, and end-of-ditch wetlands may be higher than

in isolated, tangent, and subsurface wetlands (Figure 3-28). These relationships generally









persisted among different land uses and vegetation types. This pattern did not exist for

edge soils.

Wetland Phosphorus Storage

The TP (g m-2) stored in the top 10 cm of wetlands as a whole was significantly

affected by land use (F = 4.381; p= 0.015) and wetland area (F = 4.598; p = 0.007) but

not by vegetative community. Wetland sizes among land uses were not significantly

different (p < 0.10). Total P storage and size statistics for sampled wetlands are listed in

Table 3-15. Dairy wetlands represented 18% of the sampled sites (by count) and were

storing 29% of the P. Improved pasture wetlands represented 72% of the sites and were

storing 62% of the P. The unimproved pasture and rangeland wetlands represented 10%

of the sites and stored 9% of the TP. A map of TP (g m-2) stored in the top 10 cm of

sampled wetlands is shown in Figure 3-29.





a Historically isolated wetlands.
b Percentage of category above it.
c Percentage of all EM+FOSS in research LU.


Table 3-4. Distributions of National Wetland Inventory polygons within the four
priority basins by vegetative community and land use (LU). Land-use data
is from 2003. Research LU refers to UNIMP (unimproved
pasture/rangeland), IMP (improved pastures) and DAIR (dairy). Vegetative
communities are EM (emergent marsh), and FOSS (forested / scrub-shrub).
Polygon Polygon Total ha Ha %
count %
All NWI 9,257 -- 21,649 --
Riparian 1,894 20.5 8,926 41.2
Nonripariana 7,363 79.5 12,723 58.8
EM+FOSS 6,471 87.9 b 12,121 56.0 b
In research LU 5,095 78.7 b 9,887 81.6 b
UNIMP 643 12.6 c 2,101 21.2 c
EM 588 11.5 c 1,911 19.3 c
FOSS 55 1.1 c 190 1.9 c
IMP 3,895 76.4 c 6,792 68.7 c
EM 3,571 70.0 c 6,228 63.0 c
FOSS 324 6.4 c 564 5.7 c
DAIR 557 10.9 c 994 10.1 C
EM 495 9.7 c 878 8.9 C
FOSS 62 1.2 116 1.2 c





Loss on ignition
Center
Edge
Upland
Ditch
Moisture content
Center
Edge
Upland
Ditch
pHb


---------------------------- (%) ----------------------
26.3 20.8 a 20.4 1.9 3.0 90.8
16.2 12.7 b 11.8 1.2 1.4 63.8
11.1+ 7.3 c 9.5 0.7 2.1 43.3
14.9 17.0 9.3 1.5 1.3 93.1
------------- ----- (%) ----------------------------
51.8 17.1 a 51.1 0.02 19.6 82.6
39.5 11.5 b 36.0 0.01 19.6 79.4
28.5 7.7 c 28.8 0.01 7.3 52.4
36.2 14.2 34.0 0.36 0.1 79.4


Center 117 5.0 + 4.8 a 5.4 0.07 4.0 7.5
Edge 117 4.9 4.9 b 5.2 0.05 4.3 6.7
Upland 116 4.7 4.7 b 4.8 0.06 3.7 7.9
Ditch 60 5.1 + 4.9 5.4 0.08 4.1 7.4
a Pairwise comparisons based on square-root transformed data.
b Pairwise comparisons based on H concentrations.


Table 3-5. Physico-chemical properties among hydrologic zones based on one-way
ANOVA of natural log transformed data and Games-Howell post-hoc
comparisons. Mean values are followed by std. dev. and letters, which when
the same, indicates the means are not different (p < 0.005). Ditch samples
are not included in the comparisons. SE = one standard error of the mean.
n Mean Median SE Min. Max.
a -3
Bulk densitya ------------------g cm-3-----
Center 117 0.67 0.33 a 0.65 0.03 0.16 1.42
Edge 117 0.90 + 0.25 b 0.95 0.02 0.22 1.40
Upland 116 1.04 0.18 c 1.06 0.02 0.56 1.44
Ditch 60 0.79 0.27 0.80 0.03 0.10 1.35










Table 3-6.


All wetlands
Center
Edge
Upland
Ditch
All wetlands
Center
Edge
Upland
Ditch
DAIR
Center
Edge
Upland
Ditch
DAIRa
Center
Edge
Upland
Ditch
IMP
Center
Edge
Upland
Ditch
IMP
Center
Edge
Upland
Ditch
UNIMP
Center
Edge
Upland
Ditch
UNIMPa
Center
Edge
Upland
Ditch


Total P comparisons among hydrologic zones based on one-way ANOVA
and Games-Howell post-hoc comparisons (TP by volume was natural log
transformed and TP by mass was square root transformed). Mean values are
followed by std. dev. and letters which, when the same, indicates the means
are not different (p < 0.05). Ditch samples are not included in the
comparisons. SE = one standard error of the mean.


Mean Median
------------------------------- mg kg
671.5 622.9 a 560.5
320.3 380.9 b 187.4
199.4 212.6 c 125.7
304.0 316.8 183.9
---------------------------- gm2 (0 -
34.4 28.0 a 24.9
23.0 21.3 b 16.5
19.4 19.1 b 12.5
18.2 19.6 11.6
------------------------------- mg kg-
1080.5 + 1083.6 a 1004.6
638.3 659.9 ab 315.3
322.9 346.1 b 189.5
478.4 508.3 257.7
---------------------------- gm-2 (0-
45.6 40.6 a 36.5
36.7 36.0 a 23.5
27.8 24.5 a 18.1
32.9 35.8 23.2
------------------------------- mg kg-
576.0 415.6 a 540.0
254.4 245.2 b 172.6
181.3 245.2 c 172.6
253.3 246.8 169.1
---------------------------- gm-2 (0 -
33.3 24.6 a 22.3
20.1 15.1 b 15.4
18.4 18.2 b 12.3
16.9 16.4 10.9
------------------------------- mg kg


502.7
219.4
110.4
276.4


282.9 a
200.0 b
54.7 b
329.0


413.0
169.2
99.5
171.2
1


---------------------------- gm (0 -
22.9 17.4 a 17.9
19.1 + 15.6 a 17.6
12.0 5.3 a 11.6
9.4+ 9.1 6.6


SE Min.


Max.


56.9 35.9 4324.8
35.4 30.8 2606.2
19.7 39.3 1291.0
38.2 6.2 1337.4
10 cm) -------------------------


2.6
23.0
19.4
18.2


188.3
152.4
96.3
110.2


-------------------------------------
236.5 66.2 4324.8
144.0 49.0 2606.2
75.5 74.6 1291.0
192.1 51.8 1337.4
10 cm) -------------------------
8.9 8.1 188.3
7.9 6.6 153.4
27.8 8.0 96.3
13.5 6.8 110.1
-------------------------------------
45.6 35.9 1655.2
26.9 37.8 1190.8
18.4 48.2 837.5
35.6 18.5 1083.5
10 cm) -------------------------
2.7 4.5 108.8
1.7 3.9 85.2
2.0 4.6 91.0
2.4 2.0 70.1


81.7
57.7
15.8
147.1


159.2
30.8
39.3
6.2


1000.8
777.2
212.5
793.4


10 cm) -------------------------
5.0 5.7 66.0
4.5 4.0 65.2
1.5 4.2 20.2
4.1 0.8 21.6


a Pairwise comparisons based on LSD pos-hoc procedure (p < 0.05).










Table 3-7. 1 MHC1 extractable P comparisons based on one-way ANOVA and Games-
Howell post-hoc comparisons of natural log transformed data. Mean values
are followed by std. dev. and letters which, when the same, indicates the
means are not different (p < 0.05). Ditch samples are not included in the
-2
comparisons. Median percent of TP is based on g m2.
n Mean Median Mean Median Median


---------- mg kg-1
119.1 317.9a
87.3 253.9 b
41.0 46.7 b
43.2 49.8


53.0
31.7
25.5
25.6


--- g m2 (0-10 cm) ---
6.83 17.5 a 2.9
6.14 15.6a 2.9
3.95 4.17 a 2.7
2.74 2.87 1.8


% of TP
---- % ---
13
18
21
20


Table 3-8. Metal comparisons among hydrologic zones based on a one-way ANOVA
and LSD post-hoc comparisons of natural log transformed data. Mean
values are followed by std. dev. and letters which, when the same, indicates
the means are not different (p < 0.05). Ditch samples are not included in the
comparisons.


Mean
------------ mg kg-1

674.1 657.9 a
513.5 874.8 ab
537.8 926.4b
297.6 375.7

911.8 + 1194.2 a
971.3 1851.0 ab
510.4 674.0 b
475.4 691.0

1383.3 + 1951.9 a
949.8 1403.9 a
858.8 959.2 a
476.8 551.3

674.1 657.9 a
513.5 874.8 a
391.6 926.4b
297.6 375.7


Median
------------

465.9
274.1
236.4
210.3

488.0
287.3
248.0
248.1

715.3
524.5
505.2
278.6

465.9
274.1
155.8
210.3


Mean Median
g m-2 (0-10 cm) ------


33.3 27.0 a
38.5 + 44.4 a
49.7 + 74.6 a
18.2 + 36.0

45.2 48.5 a
61.6 + 88.5 a
48.8 + 58.2 a
31.5 36.0

69.9 + 87.1 a
73.2 82.3 a
80.9 76.7 b
30.7 + 29.2

48.2 + 83.7 a
54.6 + 88.7 a
36.3 53.4 b
23.8 + 33.8


25.5
24.9
23.9
14.9

29.4
26.3
26.5
17.6

38.6
38.4
54.7
19.0

22.9
21.8
16.0
10.9


Center
Edge
Upland
Ditch


Oxalate Al
Center
Edge
Upland
Ditch
Oxalate Fe
Center
Edge
Upland
Ditch
HC1 Ca
Center
Edge
Upland
Ditch
HC1 Mg
Center
Edge
Upland
Ditch


117
116
116
60

117
116
116
60

117
116
116
60

117
116
116
60










Effects of land use (LU) and vegetative community (VC) (fixed factors) and
wetland area (covariate) on soil characteristics of wetland centers and edges
based on a factorial ANOVA (p < 0.10) and Bonferroni pairwise
comparisons (p < 0.10). ns = not significant, DAI= dairy, IMP = improved
pasture, UNI = unimproved pasture/rangeland, EM = emergent marsh, EW
= emergent marsh/open water, FS = forested / scrub-shrub.
----------------Centers-------------- ----------------Edges----------
Level Level
F p comparisons F p comparisons
comparisons comparisons


Bulk Density
VC
LU
VC*LU
area
Organic Matter
VC
LU
VC*LU
area


6.626 0.008 EW > EM > FS


9.535 0.000 FS, EM > EW


5.303 0.006 EM, EW > FS


6.644 0.002 FS, EM > EW


EW > EM > FS
DAI > IMP > UNI


5.349
5.269


0.006
0.007


EW > EM > FS
DAI > IMP, UNI


Estimated unweighted means and standard errors of pH among treatments
based on a factorial ANOVA and Bonferroni pairwise comparisons (p <
0.01). Letters compare levels of land use or vegetative community within a
hydrologic zone. The same letter indicate the means are not different.
Fixed factors (vegetative community and land use) each had significant
effects (p < 0.01) but their interaction and the covariate (wetland area) did
not (p < 0.10).


Treatment and levels
Land use
Dairy
Improved
Unimproved pasture / rangeland
Vegetative community
Emergent marsh
Emergent marsh / open water
Forested / scrub-shrub


Center soil Edge soil
n --------------pH--------------


5.81 + 0.28 a
5.35 0.24 b
4.49 0.62 c

5.36 + 0.20 a
6.18 0.30b
4.67 0.50 c


5.48 0.22 a
5.07 0.20b
4.65 0.51 b

5.22 0.15 a
5.79+ 0.24b
4.73 + 0.39 c


Table 3-9.








Factor


12.946
5.309


VC
LU
VC*LU
area


0.000
0.006


Table 3-10.































n=86 n=21 n=11

EM EW FS
Vegetation Community


n=85


n=21


n=11


EM EW FS
Vegetative Community


S5(0-


0
S25-



0-


n=86


n=12


EM EW FS
Vegetation Community


75-




-
50-


r-

0 25-



0-
0-


a







0







n=86


5


0

-


n=11


E EW FS
Vegetative Community


Figure 3-21. Soil bulk density and percent organic matter comparisons among vegetative
communities showing means. A) Wetland centers. B) Wetland edges. Error
bars show 95% confidence interval of the mean. Data with the same letter
above the boxplots are not different, based on factorial ANOVAs (fixed
factors were land use and vegetative community, covariate was wetland
area) and Bonferroni pairwise comparisons (p < 0.10). Only vegetation
community had a significant effect (p < 0.10).


E
u 1.00-


" 0.75-


S0.50-


0.25-


1.25
E
-
o 1.00


"5 0.75


- 0.50


0.25-











1.5(


cn
E 1.0(



" 0.5(
D


n=4 n=6 n=4 n=1 n=6n=51n=34n=7 n=5

DABE DAPA DAUN IMP2 UNI2
DAHA DASP IMP1 UNI1
Land Use (detail)


n=4 n=6 n=4 n=1 n=6 n=61n=33n=7 n=6

DABE DAPA DAUN IMP2 UNI2
DAHA DASP IMP1 UNI1
Land Use (detail)


n=5 n=6 n=4 n=1 n=6 n=51n=33n=7 n=5

DABE DAPA DAUN IMP2 UNI2
DAHA DASP IMP1 UNI1
Land Use (detail)


-n n=6 n=4 n=1 n=6 n=51n=33n=7 n=5

DABE DAPA DAUN IMP2 UNI2
DAHA DASP IMP1 UNI1
Land Use (detail)


Figure 3-22. Soil bulk density and percent organic matter by detailed land use showing
mean values. A) Wetland centers. B) Wetland edges. Error bars show 95%
confidence interval of the mean. DABE = dairy beef, DAHA = dairy hay,
DAPA = dairy pasture, DASP = dairy sprayfield, DAUN = dairy
unimproved field, IMP1 = more improved pasture, IMP2 = less improved
pasture, UNI1 = unimproved pasture, UNI2 = rangeland.










Effects of land use (LU) and vegetative community (VC) (fixed factors) and
wetland area (covariate) on center soil TP (mg kg- and g m-2) based on
factorial ANOVAs (p < 0.10) and Bonferroni pairwise comparisons (p <
0.10). ns = not significant, DAI = dairy, IMP = improved pasture, UNI=
unimproved pasture/rangeland, EM = emergent marsh, EW = emergent
marsh/open water, FS = forested / scrub-shrub.
-------------- mg kg--------------- -----------g m-2 (0-10 cm)-----------


Table 3-11.








Factor
VC*LU


F p Comparisons
2.231 0.089


3.813a 0.024
3.333b 0.041
6.677 0.011


a Effect is only significant within IMP land-use areas.
b Effect is only significant within EM wetland vegetative communities.


FS >EM
DAI > UNI
smaller > larger


o










o
OO .



2o

S 2 4 6 8 10 12 14


Wetland size (ha)


Figure 3-23. Scatterplot of center soil total phosphorus (g m-2) versus wetland size (ha).


F p Comparisons
2.873 0.040
4.002a 0.022 FS>EM,EW
4.173b 0.019 DAI > IMP, UNI
ns


Area














EM





n=9


n=6 n=11

dairy unimprov
improved
land use


EW


n= n= 15
n=6

dairy unimprov
improved
land use


FS


n=4


n=6


5 2000

1500

1000
in
0
S500'

00
So

p-


Error Bars show Mean +/- 1.0 SE
Dot/Lines show Medians


E
U
o EM EW FS
75.00


E 50.00 ,
0) n=9" ,,, n n 4 "
A n = 9
S25.00 = n=15 n=6
^ 25.00 n- --
n=6
Sn=11 n=1

S0.0oo
dairy unimprov dairy unimprov dairy unimprov
improved improved improved
land use land use land use

Error Bars show Mean +/- 1.0 SE
Dot/Lines show Medians B


Figure 3-24. Plots of total phosphorus (TP) in center soils showing interaction of
vegetation community and land use. A) mg kg-1 B) g m-2. EW =emergent
marsh / open water, FS = forested / scrub-shrub, EM = emergent marsh.
While EW and FS wetlands have lower TP within dairies compared to
improved pastures, EM wetlands have higher TP within dairies.


dairy unimprov
improved
land use

















E 150-
o.
o
C-
S100-

_c
-,-
E
3 50-
a-
I-


4000'
E
o

o
S3000-
a-

2000-

,7,
2 1000
E
a-
I-










o 1500'
o

a-

E 000-



E 500-
0-
I.-


Unimprov



b


42


13U-


I I
Dairy Improved
Land Use


a a


n=64


n=16 n=4


EM EW FS
Vegetative Community


0







n=9 n=64
Dairy Improved
Land Use


b








0






n=11
Unprov
Unimprov


a ab b





0
C-








n=64 n=15 n=4

EM EW FS
Vegetative Community


Figure 3-25. Total phosphorus comparisons of center soils showing means. A) Emergent
marsh soils by land use. B) Improved pasture soils by vegetative
community. Error bars show 95% confidence interval of the mean. Data
with the same letter above the boxplots are not different, based on factorial
ANOVAs (fixed factors were vegetative community and land use, covariate
was wetland area) and Bonferroni pairwise comparisons (p < 0.10). The
interaction of vegetation community and land use had a significant effect (p
< 0.05). EM = emergent marsh, EW = emergent marsh / open water, FS =
forested / scrub-shrub.


n=64 n=11










Effects of land use (LU) and vegetative community (VC) (fixed factors) and
wetland area (covariate) on edge soil TP (mg kg- and g m-2) based on
factorial ANOVAs (p < 0.10) and Bonferroni pairwise comparisons (p <
0.10). ns=not significant, DAI = dairy, IMP = improved pasture, UNI =
unimproved pasture/rangeland, EM = emergent marsh, EW = emergent
marsh/open water, FS = forested/scrub-shrub.
----- ------ mg kg --------------------- g m-2 (0-10 cm) -------
F p Comparisons F P Comparisons


Table 3-12.








Factor
VC*LU
VC
LU
area


FS > EM
DAI > IMP
smaller > larger


2.845
8.430


ns
ns
0.063
0.004


DAI > IMP
smaller > larger


140

120

100

80

60

40

20

0


o 2 4 6 8 10 12 14

Wetland size (ha)


Figure 3-26. Scatterplot of edge soil TP (g m-2) versus wetland size (ha).


ns
0.066
0.056
0.009


2.788
2.961
6.995


0

1


o
C0



0 0D
D 00
^o o
^B~P DD
^ D o_____________D


I












a b
0








-- *





n=21 n=83
I I
Dairy Improved
Land Use


ab
ab 150.0-


E
0

100.0-



E
-50.-
I,.


E
o

" 2000-
o.
0



S1000-
E
a.
I-


a b ab







oo
*B 0


0
S-- 6



n=84 n=21 n=11

EM EW FS
Vegetative Community


8


.1


S n=21 n=83 n=12
I I I
Dairy Improved Unimprov
Land Use

a a a




0



0
0






n=84 n=21 n=11

EM EW FS
Vegetative Community


Figure 3-27. Total phosphorus comparisons of edge soils showing means. A) Land use.
B) Vegetative community. Error bars show 95% confidence interval of the
mean. Data with the same letter above the boxplots are not different, based
on a factorial ANOVA (fixed factors were vegetative community and land
use, covariate was wetland area) and Bonferroni pairwise comparisons (p <
0.10). The interaction of vegetation community and land use had no
significant effect. EM = emergent marsh, EW = emergent marsh / open
water, FS = forested / scrub-shrub.


OUUU


n=12
Unimprov


LZUUU


E
0 1500'
o




a-
0
. 1000'



E 500'
a-










Table 3-13. Spearman correlations and significance (2-tailed) for wetland center soil TP
and biogeochemical parameters (mg kg1 except organic matter) by land use.
IMP = improved pastures, DAIR = dairies, UNIMP = unimproved pastures /
rangeland.


% organic matter


HCl-extr Ca


HCl-extr Mg


Oxalate-extr Al


Oxalate-extr Fe


All
117
0.632
0.000

0.374
0.000

0.476
0.000


0.570
0.000

0.472
0.000


IMP
83
0.574
0.000

0.334
0.002

0.479
0.000

0.576
0.000

0.494
0.000


DAIR
22
0.868
0.000

0.687
0.000

0.415
0.055

0.751
0.000

0.513
0.017


UNIMP
12
0.455
0.138

-0.217
0.499

0.273
0.391

0.007
0.983

0.322
0.308


Table 3-14. Spearman correlations and significance (2-tailed) for wetland edge soil TP
and biogeochemical parameters (mg kg-1 except organic matter) by land use.
IMP = improved pastures, DAIR = dairies, UNIMP = unimproved pastures /
rangeland.


% organic matter


HCl-extr Ca


HCl-extr Mg


Oxalate-extr Al


Oxalate-extr Fe


All
117
0.748
0.000


0.179
0.055

0.657
0.000

0.385
0.000

0.598
0.000


IMP
83
0.717
0.000


0.118
0.291

0.577
0.000

0.363
0.001

0.528
0.000


DAIR
22
0.786
0.000


0.223
0.330

0.617
0.003

0.379
0.090

0.640
0.002


UNIMP
12
0.727
0.007


0.042
0.897

0.783
0.003

0.224
0.484

0.783
0.003











em marsh


ITI


. 2000.00
0
U)
0 1500.00
0
c 1000.00
03
500.00
E
0.00'
I-


S2000.00
0
U,

S1500.00

.c
- 1000.00
0)

S500.00
I-


lI


end of ditch isolated e
flowthrough subsurface
headofditc tangent
ditch class


em marsh / water







I1 T


improved


III


m i1i


nd ofditch isolated e
flowthrough subsurface
head of ditc tangent
ditch class


forest / scrub











A
S imp ro A
unimprov


LJ
nd of ditch isolated
flowthrough subsurface
head of ditc tangent
ditch class

Error Bars show Mean +/- 1.0 SE
Circles show median


Figure 3-28.



Table 3-15.


Total P (mg kg'-) in center soils by ditch class. A) Varying similarly among
vegetative communities. B) Varying similarly among and land uses.

Total P (TP) stored in surface soils of sampled wetlands (154 ha) among
land uses. Mean values are followed by std. dev. and letters, which when
the same, indicates the means are not different (p < 0.10). Comparisons are
based on a factorial ANOVA (fixed factors were vegetative community and
land use) and Bonferroni pairwise comparisons of unweighted means. DAIR
= dairies, IMP = improved pastures, UNIMP = unimproved pastures /
rangeland.


Land Mean Med. Mean size Med. Mean Med. Total
Use TP TP size TP TP TP
(n)
---- g m2 (0-10 cm)-- ---------- ha ------------------ kg ---------
DAIR 45.6 40.0 a 35.3 1.56 1.11 a 1.22 628.5 522.1 437.3 11,701
(21)
IMP 30.1 + 19.7 a 22.7 1.15 + 1.13 a 0.90 325.6 + 424.0 212.9 24,946
(84)
UNIMP 19.7 + 9.4 b 20.4 2.03 + 3.67 a 0.54 188.1 + 148.1 125.6 3,566
(12)
40,214


dairy











Priority basins
unimproved pastures / rangeland
improved pasture
dairies
TP (g/m2) in top 10 cm
5-14
14-22
0 22-36
0 36-58
S58- 185


0 5 10


20


Lake 1h'eechob'ee


Figure 3-29. Land uses (2003) and TP (g m-2) stored in surface soils (0-10 cm) of
sampled wetlands within the four priority basins of the Lake Okeechobee
watershed.

Discussion

This study aimed to assess general trends of TP storage within historically isolated

wetlands of four sub-basins of the Lake Okeechobee watershed across a range of land

uses (watershed-scale), for different ditching magnitudes (field-scale) and among

hydrologic zones within wetlands (field-scale). The TP (g m-2) in surface soils (0-10 cm)

reported in this study are comparable to those reported by studies of landscape wetlands


0O
06600


0 0



0
o









and uplands in the study area (Sperry, 2004; Reddy et al., 1996a; Nair and Graetz, 2002)

and wetlands in agricultural areas of North Carolina (Bruland et al., 2003).

Redox potential and pH are known to control P movement in wetlands (Richardson,

1999). Low redox potential slows microbial decomposition leading to organic matter

accretion (Levesque and Mathur, 1979) which may be the most important mechanism for

P storage in more organic wetland soils (Axt and Walbridge, 1999; Richardson, 1999).

In flooded soils containing Fe, low redox potential can cause ferric iron to be reduced to

the soluble ferrous form releasing P that was bound to it (Faulkner and Richardson,

1989). Lower redox potentials that may exist in the wetland centers can cause the

transformation of crystalline forms of Fe and Al to more amorphous forms (oxalate

extractable) which have more sites for P sorption and are more significant sorption agents

for P than ferric iron (Reddy and Smith, 1987; Rhue and Harris, 1999).

Zonal TP gradients

Data showed that significant zonal TP gradients exist from wetland centers

(median: 560.5 mg kg-1), to wetland edges (median: 187.4 mg kg-1) and surrounding

uplands (median: 125.7 mg kg-1). The gradient of volume-based TP was only significant

between the wetland center (median: 24.9 g m-2) and edge (16.5 g m-2) so a portion of the

first hypothesis was accepted. These differences were significant in IMP areas but not in

DAIR or UNIMP (Figure 3-30). Other studies have reported increased TP storage in

more saturated soils within wetlands (Scinto, 1990; Sperry, 2004; Reddy et al., 1996a).

The difference in IMP wetlands may be explained by ditches which are larger in size and

number in this land use compared to the other land uses. Wetland centers act as

hydrologic sinks accumulating water and other material from surrounding areas. Ditches

may transport dissolved and particulate P to the centers of these wetlands bypassing the









buffering effects of the edges. The lack of significant differences in edges and centers

within the other two land uses may be also a result of the higher TP variability combined

with a smaller number of samples in those land uses.


Unimproved Pastures
Unimproved Pas s Dairy Land Improved Pastures
/ Rangeland

45.6 40.6 g m2
33.3 24.6 g m2
22.9 17.4 g2 m
n 21
n-=12 n= 84
.1 15.6 20.115.1 g
36.7 36.0 g m-2


Figure 3-30. Cartoon comparing relative magnitudes (means and std. dev.) of wetland
edge and center soil TP (g m-2) among land uses (only TP in IMP hydrologic
zones are significantly different). Lines connected to centers represent
relative ditching.

The relationships between P and other biogeochemical properties help explain the

zonal TP gradient in the wetlands. Correlations between TP and organic matter were

strong in wetland centers (rs: 0.632) and in edges (rs: 0.748). Richardson (1999) pointed

out that mineral sorption mechanisms are more important in mineral soils. In this study,

the median organic matter in edges (12%) was significantly lower than that in centers

(20%). These values are somewhat higher than the median organic matter found in 14

EM wetland soils (0-5 cm) of similar size in Pennsylvania which was 12% (Campbell et

al., 2002) but similar to that reported in a study of similar wetlands in the same study area

(Sperry, 2004). Bruland and Richardson (2004) reported 9% organic matter in wet area

soils (0-15 cm) of 11-year-old constructed wetlands and about 6% in the edge areas.









More than half of the P in this study is estimated to be organic P. A 1 MHCI

extraction was used to estimate inorganic P to be less than 20%, which would mean that

more than 80% is organic P. This extraction is a conservative estimate of inorganic P

(K.R. Reddy, personal communication, November 22, 2004) and was developed for

organic wetland soils in calcareous systems that do not receive manure (Reddy et al.,

1998). It is helpful to look at a study of IMP and UNIMP isolated wetland soils in the

study area in which organic P of surface soils (0-15 cm) was estimated to be 62% based

on the difference of inorganic P (determined by a fractionation scheme) and total P

(Sperry, 2004).

Since the majority of the P stored in the wetlands is organic P, accretion

mechanisms may be more important than sorption mechanisms in these historically

isolated wetlands. The significantly higher TP in wetland centers corresponds with the

significantly higher organic matter content in wetland centers but the higher P in centers

cannot be attributed to more metals since their quantities are not significantly different

between hydrologic zones. While those metals may be playing a role in retaining P,

sorption mechanisms with the organic matter-metal complexes may also be an important

mechanism for P retention. Petrovic and Kastelan-Macan (1996) suggested that

phosphate (H2P04-) can be completed with Ca, Mg, Fe and Al cations that are bound to

negatively charged humic substances. This is sometimes referred to as bridging

(Richardson, 1999). Other studies reported that Al complexation with organic matter

played an important role in P sorption in surface wetlands soils (Axt and Walbridge,

1999; Haynes and Swift, 1989). More evidence for the important role of organic matter is

shown by the higher correlation between TP and organic matter compared to metals.









Scinto (1990) also found the highest correlation with TP to be organic matter in forested

dairy wetlands within the study area. Though the correlations between TP and metals

were not as high as between TP and organic matter, the trends are discussed below.

Some studies have shown that amorphous Al and Fe hydroxides govern P sorption

in acidic wetland soils and stream sediments (Patrick and Khalid, 1974; Reddy et al.,

1995; Richardson, 1999; Axt and Walbridge, 1999) whereas reactions with Ca and Mg

are important binding mechanisms in more alkaline systems (Richardson, 1999). Total P

in wetland edges was more correlated with Feox (rs: 0.598) than with Alox (rs: 0.385) and

conversely the TP in wetland centers was more correlated with Alox (rs: 0.570) than with

Feox (rs: 0.472) with the exception of UNIMP. In wetland edges, total Al may represent a

significant potential pool of available P adsorbing sites if the wetland were to be

expanded in size by hydrological restoration. This would expose crystalline Al to

flooding cycles and convert it to amorphous Al which has more sites for P adsorption

than the crystalline form (Rhue and Harris, 1999).

In the edges where soils have longer periods of oxidizing conditions than in

centers, there will be more Fe in the oxidized form which binds P (Patrick and Khalid,

1974). In wetland centers, more amorphous Al with high P-binding capacity is expected

than in edges. The results also showed a correlation between TP and Mg (rs: 0.657) in

wetland edges and a weaker correlation (rs: 0.476) in wetland centers. Because of the

acidic nature of the soils in which Mg phosphate would not be stable, it is possible that

the Mg is providing a bridge between phosphate and humic substances. Correlations in

edge soils might be due to the influences of adjacent land uses whereas correlations in

center soils might be more influenced by the import and export of materials by ditches









(confounded by the transport of material from the upland through the edge into the

center). Higher pH along with higher manure input to DAIR wetlands may explain why

there was a relatively strong correlation between TP and Ca (rs: 0.687) in the centers of

those wetlands.

Interestingly, results of this study indicated that organic P was lower in IMP

wetlands (18.2%) compared to DAIR (26.6%) or UNIMP (28.4%). Cattle may graze

more wetland vegetation in IMP areas compared to other land uses since fewer cows are

stocked in UNIMP areas (Steinman et al., 2003) and cows in DAIR areas receive

commercial feed (SFWMD et al., 2004a) and are excluded from many of the wetlands.

Herbivory of aquatic vegetation may reduce the amount of organic matter being accreted

and the reduced vegetation combined with physical impact to soils by cattle may cause

some sedimentation in the wetland. Studies have shown that vegetation around wetlands

and vegetated ditches reduces sedimentation in wetlands suggesting that less particulate

material (e.g., particulate P) is imported into the wetland (Fiener and Auerswald, 2003;

Hook, 2003). A study by Clary (1999) indicated that the degree of sedimentation is

related to the amount of grazing pressure on a wetland (Clary, 1999).

Total P (g m-2) stored in upland and wetland surface soils of unimproved pastures /

rangelands was similar to that reported by other studies in the same land uses in the Lake

Okeechobee watershed (Sperry, 2004; Reddy et al., 1996a; Graetz and Nair, 1995;

Scinto, 1990). Those studies reported higher values of TP in dairy soils than shown in

this study (mean: 45.6 g m-2). While those studies focused sampling on high intensity

dairy areas, this study showed a greater variability in sampled dairy areas that ranged









from ungrazed fields to pastures where a large number of cows received supplemental

feed.

Hydrologic Connectivity

Studies have demonstrated that wetlands with surface connections to downstream

waterways can export nutrients (Nessel and Bayley, 1984; Yin and Lan, 1994; Raisin and

Mitchell, 1995; Richardson, 2003; Mwanuzi et al., 2003; Sperry, 2004). This study

hypothesized that wetlands with more ditching would have less TP, but the inverse was

true. Center soils of improved pasture wetlands with deeper and wider associated ditches

had significantly more TP than wetlands with less intensive ditching while controlling for

land use and wetland area. The number of ditches did not have a significant effect on TP.

In this study, fertilizer and cattle density information could not be collected. More

intense ditches typically have less vegetation and faster water flow than less intense

ditches. Vegetation in ditches filters particulate P and serves to slow down flow

velocities (Fiener and Auerswald, 2003). This enhances hydraulic retention time in the

ditches, and allows sediments to fall out of flowing waters to reduce the amount of

particulate P being transported. More intense ditches may bypass the buffering effects of

edges bringing more P to wetland centers. It is also possible that ditch intensity is a

proxy for fertilization and cattle stocking intensity. A pasture with more intense ditching

represents a significant economic investment which may be accompanied by additional

fertilizer and more intense stocking of cattle.

Interestingly, TP (mg kg-1) varied similarly by ditch class regardless of land use or

vegetative community. Flowthrough, end-of-ditch and head-of-ditch wetlands generally

had more TP than subsurface, tangent and isolated wetlands. Water flow may be

transporting more particulate and dissolved P into wetland centers of flow-through and









end-of-ditch wetlands. Head-of-ditch and isolated wetlands may have more vegetation to

generate more organic matter that accretes and stores more TP. Isolated wetlands may be

in areas with less intensive grazing. Tangent wetlands have ditches that do not intersect

the centers hence P being transported by ditches may bypass the wetland center. Based

on these findings it can be speculated that wetlands with less ditching are in pastures with

less cattle activity. In improved pastures, 60% of the wetlands had at least 1 ditch and

almost 50% had a ditch intensity of 2 or more. In dairy and unimproved / rangeland

pastures, less than 30% of the wetlands had ditches and less than 30% had a ditch

intensity of 2 or more.

Land Use Differences

It was hypothesized in this study that wetlands surrounded by more intensely

fertilized and cattle-stocked areas would have higher surface soil TP (dairy > improved

pastures > unimproved pastures / rangeland). This ideas is supported by a study that

concluded that up to 80% of TP had the potential to leave heavily manured upland

regions, while less than 10% of the TP was likely to leave a low manure-impacted pasture

soil because of the mobility of P in manure (Graetz and Nair, 1999). Also, it is widely

accepted that the land use and land cover of a landscape have a major effect on water

quality (Herlihy et al., 1998; Cuffney et al., 2000; Berka et al., 2001). Sperry (2004)

found that ditched wetlands in improved pasture released 5 to 7 times more TP in surface

water runoff than wetlands in semi-native pastures. This study supports the hypothesis

that wetland soils in UNIMP areas have significantly less TP (median: 20.4 g m-2) than

DAIR (median: 35.3 g m-2) and IMP (median: 22.7 g m-2) land uses (while accounting for

differences due to vegetative community and wetland area) but there was not enough

evidence to show that DAIR and IMP were different. The lack of a significant difference









between wetland soils in IMP and DAIR land uses study may also be a result of the high

variability and lower number of samples in the DAIR treatment. The dairy land use class

grouped pastures with feed structures with non-irrigated fields near and beef pastures that

were similar to the improved pastures. The wide range of dates of sample collection was

probably a factor in the overall data variability as plant uptake of nutrients in wetlands

can be higher at the beginning of summer (when sampling began) than in the fall (when

sampling ended) (Tanner, 1999). Sperry (2004) found no differences in wetland soil TP

(g m-2) between improved and semi-native (UNIMP) pastures. This may have been due

to variability within a small sample size (12) and the fact that the IMP pastures in that

study were not fertilized after 1987.

Other Findings

Wetland size was a significant covariate for TP in centers and edges. Larger

wetlands stored less TP (g m-2) than smaller wetlands. This suggests that smaller

wetlands could become saturated with P and perhaps be a source for P in the future

(compare Reddy et al., 1995). Larger wetlands with larger pools of P binding sites and

longer hydraulic retention times that favor biogeochemical processes that bind P may

represent more long-term sinks for P.

In a study of sediment nutrient levels in 73 wetlands it was found that wetland size

was negatively correlated to sodium bicarbonate extractable P (Houlahan and Findlay,

2004). A possible explanation offered by Detenbeck et al. (1996) is that water and soils

in the center of large wetlands are better buffered against nutrient inputs than smaller

wetlands by virtue of the greater distance between the wetland center and input sources.

Land use would not explain this trend since there was no significant difference in wetland

size among land uses.









Forested / scrub-shrub systems had significantly more TP in wetland centers (while

accounting for land use and wetland size) than EM systems. Forested systems typically

have less organic matter than emergent marsh systems due to the recalcitrant nature of the

woody organic matter (Mitsch and Gosselink, 2000) so this result was surprising. Eight

out of the 12 FS wetlands were ditched and all had intense ditching (intensity of 2 or

more). Seven of them were associated with DAIR areas and one with UNIMP. The

forested systems had noticeable soil oxidation as seen from exposed tree roots and were

usually hummocky systems that allowed for aerobic soil processes to occur. This may

have had the effect of concentrating the P that was in surface soils since P has no gaseous

form as do carbon and nitrogen. It is possible that a combination of having more intense

ditching, receiving mostly DAIR runoff and experiencing soil oxidation caused the higher

P concentrations.

Storage

Dairies were found to be storing significantly more TP (median: 35.3 g m-2) than

UNIMP (median: 20.4 g m-2). Though DAIR wetlands comprised 18% of the sites, they

accounted for 29% of the 40,214 kg of P stored on all sampled wetland surface soils. The

sampled sites covered 154 ha and stored 261 kg ha-1 in the top 10 cm of soil. Reddy et al.

(1995) reported 750 kg ha-1 in streamside and dairy wetland soils (0-30 cm). Considering

that most of the P storage was found in the surface soils, the historically isolated wetlands

may be storing relatively low amounts of TP and may represent a potential sink for P

storage in the future.

Future Research

The cumulative impact of high nutrient loads in the past can leave a legacy that

may override the effects of current land-use practices related to stocking density and









fertilization (Steinman et al., 2003). In future research, it would be beneficial to create a

land cover map of the area using a satellite image classification based on apriori-

collected ground truth points of land uses where the fertilization and cattle stocking

history are known. These land cover classes may be more related to phosphorus

condition in uplands and wetlands and would have higher spatial and temporal

resolutions than the coarse-scale land-use layer used in this thesis. The land cover map

could be helpful in creating a sampling scheme according to the land cover classes such

that differences in those classes could be tested.

It is also recommended to test the effects of ditching on TP in wetlands. While

comparing wetlands of similar size and vegetation surrounded by similar land uses a

hypothesis to test would be that isolated wetlands store more TP than wetlands being

drained. Another possible hypothesis is that ditches that do not intersect wetland centers

(tangent ditches) conduct less TP from the wetland while still serving the function of

conducting water from it. A final recommendation is to characterize P fractions and

undertake P sorption studies using methods proven effective in non-calcareous, manure-

impacted wetland systems so that P dynamics can be further investigated in these

historically isolated wetland ecosystems.

Conclusions

Interacting factors affect the variability of P storage in historically isolated

wetlands and there is evidence to support the idea of hydrological restoration. This study

confirms previous findings that TP in DAIR wetlands store more TP than UNIMP, but

the comparison between DAIR and IMP wetlands was inconclusive. The high variability

in DAIR TP is probably due to a wide range of land uses within DAIR and may be a

reason why there was no significant difference between DAIR and IMP wetlands and









also why no differences in TP (g m-2) were found among DAIR hydrologic zones. The

development of a land cover map based on satellite imagery would assist in land use

classification for future studies.

The hypothesis that a TP gradient exists from wetland centers increasing outwardly

to upland soils, was accepted for IMP centers that had more TP than edges and uplands.

More intense ditching in IMP areas may be bringing more dissolved and particulate P to

the centers of this land-use type. Wetlands with larger ditches were shown to have

significantly more TP in center soils than those with smaller ditches.

It is hypothesized that mechanisms related to organic matter are responsible for the

majority of TP storage in these wetland systems including accretion and phosphate metal

bridging with humic substances. Through hydrological restoration, overall wetland size

would increase, providing additional storage capacity for P. Also, the amount of P per

square meter stored might also increase, due to increased hydrologic retention time.

Land managers should consider hydrological restoration of drained wetlands to increase

wetland size and hydraulic retention times. This would increase the number of P binding

sites on soil particles as well as increase the amount of center area soils which have

higher rates of P storage. More studies are needed to determine the P sorption capacity of

these soils. Increasing the size of smaller wetlands in more intensely managed pastures

may prevent them from becoming saturated with P and becoming sources of P to surface

waters.

Over 12,000 ha of isolated wetlands (drained and undrained) exist in the four

priority basins. About half are drained and represent an opportunity for increasing

storage of P in the Lake Okeechobee watershed through hydrological restoration. Land






75


managers should develop best management practices (BMPs) in IMP areas where the

largest proportion of ditched, isolated wetlands exist. They should consider restoring

hydrology of drained isolated wetlands, keeping ditches vegetated, and excluding cattle

from wetlands. Incentive programs offered by government agencies can help offset the

cost of BMP implementation while reducing the ecological and economic costs of

elevated P levels in Lake Okeechobee.














CHAPTER 4
UPSCALING TOTAL PHOSPHORUS TO UNSAMPLED WETLANDS

Introduction

Wetland ecosystems are an integration of physical and biological processes,

occurring at watershed- and field-scales. In the previous chapter, it was shown that

factors at multiple scales affect TP storage in historically isolated wetlands within the

four priority basins of the Lake Okeechobee watershed. These factors include land use,

ditch size, and wetland hydrologic zones. In this chapter, datasets available at the

watershed -scale (i.e., the four priority basins), were collected and analyzed to predict TP

storage in unsampled wetlands. Only in recent years, have the technologies of satellite

remote sensing and GIS, as well as natural resource datasets, become readily available to

study these systems at multiple spatial, spectral, and radiometric resolutions (Walsh et al.,

1998). The aim of this study was not to generate a mechanistic model that explains TP

storage factors, but to use patterns of available spatial datasets that characterize the

majority of wetland TP storage variability. These patterns of multi-scale variables may

represent a holistic view of TP conditions of wetlands and surrounding areas.

Environmental Variables

Ecological response variables can be predicted by environmental variables

(McKenzie and Austin, 1993; McKenzie and Ryan, 1999; Lapen et al., 2001). Land use,

for example, has been shown to account for a high amount of variability in stream water

quality (Hunsaker, et al., 1992; Roth et al., 1996; Herlihy et al., 1998; Behrendt et al.,

1999). In Chapter 3, it was shown that wetland soils in DAIR areas stored higher









amounts of TP than UNIMP areas. Land use may also help to predict TP stored in

unsampled wetlands. Soil type is an environmental variable that has been used to provide

input data to water quality models (Kuenstler et al., 1995; Wilson et al., 1996; Shaffer et

al., 1996). Soils adjacent to wetlands may be predictive of P in wetlands, because of

differing drainage and biogeochemical characteristics.

Spatial patterns of natural and anthropogenic features in a landscape strongly

influence the ecological characteristics (Risser et al., 1984; Walsh et al., 1998).

Landscape metrics are based on landscape geometry and spatial arrangement of patches

and features (Herzog and Lausch, 2001). For example, landscape characteristics that

describe the arrangement of human-altered land in a watershed can be correlated with

water biogeochemistry (Gergel et al., 2002; Yin et al., 2003). In a study of sediment

nutrient levels in 73 wetlands, it was found that wetland size was negatively correlated to

sodium-bicarbonate-extractable P (Houlahan and Findlay, 2004). In the same study, the

proportion of land within 1,250 m that was wetland, was also negatively correlated with

sodium-bicarbonate-extractable P, and nitrate levels were positively correlated with road

density within 500 m. Landscape metrics explained 65 to 86% of the total variation in

nitrogen yields to streams, and 73 to 79% of the variability in dissolved P (Jones et al.,

2001). Distances to features, such as roadways, dairies, waterways, and other wetlands

could be used in this study to partially quantify spatial landscape patterns of wetlands,

and may be predictive of their soil P condition.

Spectral Data

Remote sensing is the acquisition of data about an object or scene on earth by a

sensor that is far from the object (Colwell, 1983). Sensors may be hand-held or mounted

in vehicles, airplanes, or satellites. Remote sensing has emerged as a useful data source









for characterizing LULC (Burnett and Blaschke, 2003), and can provide valuable

quantitative data that is indicative of moisture, soils, and vegetation (Curran, 1985;

Jensen, 1996; Lillesand and Kiefer, 2000) that exist at a particular point in time. Satellite

remote sensing has shown its utility in wetland mapping (Ackleson and Klemas, 1987;

Mertes et al., 1995; Sader et al., 1995; Harvey and Hill, 2001; Schmidt and Skidmore,

2003) as well as land cover mapping (Jensen, 1996; San Miguel-Ayanz and Biging, 1997;

Lillesand and Kiefer, 2000; Pearlstine et al., 2002; Reese et al., 2002).

There are many studies quantifying biophysical measures such as biomass and soil

moisture (with remotely sensed spectral data), but few studies have linked

biogeochemistry with satellite spectral information (Numata et al., 2003). Green

vegetation, calculated from a Landsat7 ETM+ satellite image, was correlated with P

content in pastures in Rond6nia, Brazil (Numata et al., 2003). Asner et al. (1999) found

that spectral estimates of leaf area and nonphotosynthetic vegetation (NPV) of pastures

were correlated with soil P concentrations in the central Amazon region. These studies

used vegetation indices, which are quantitative measures of vegetative condition. These

indices are usually calculated from combinations of several spectral bands, whose values

are added, divided, or multiplied.

The pure spectral reflectance values for wetland features and surrounding areas are

a result of the interaction of rock, soil, vegetation, fauna, landform, and water (Schmidt

and Skidmore, 2003). They may be indicators for the properties of an earth feature (like

a wetland) as a whole (Schmidt and Skidmore, 2003). Polygon-based aggregation of

remotely sensed data has been shown to increase accuracy of land-cover biophysical

variable estimation at regional scales (Wicks et al., 2002). Spectral properties may be









indicative of wetland features (polygons): grazing intensity around it, amount of open

water, amount of exposed soil, amount of organic matter and vegetation density. These

features may correspond with TP magnitude stored in wetlands soils. For example, more

green vegetation within a wetland could indicate higher TP within the wetland.

Variability of vegetation around a wetland may be indicative of homogeneous pasture or

heterogeneous native land, which might be related to TP. Spectral reflectance data and

vegetation indices may be indicative of recent P conditions within, and around,

historically isolated wetlands of the study area.

Ecological Response Variables and Classification Trees

Austin et al. (1995) noted that the use of suitable environmental variables was more

important than the choice of prediction method, but classification and regression trees are

becoming increasingly used for this purpose (McKenzie and Ryan, 1999). Classification

and regression trees are ideally suited for analyzing complex ecological datasets, because

they provide robust analytical methods that can deal with nonlinearity, missing values,

interacting variables, and heterogeneous variances (De'ath and Fabricius, 2000). The

method is a form of binary, recursive, partitioning, that is well suited for determining

relationships within complex, nonparametric, interacting datasets that have a high number

of predictor variables (Breiman et al., 1984; Lewis, 2000; Steinberg and Colla, 1995).

Classification trees predict a pre-designated class for each case, whereas regression trees

predict a range of values for each case (Breiman et al., 1984).

McKenzie and Ryan (1999) used readily observable environmental features as a

basis for mapping soil properties. With digital terrain analysis and airborne gamma

radiometric data, they were able to use regression trees to account for 78% of the

variability of soil P in upland forest soils of southeastern Australia based on 165 sample









sites. In another study, the spatial distribution of alien plant stands (presence or absence)

was predicted with 73% accuracy using classification trees (based on the predictors of

temperature, number of growing days, altitude, geology, slope, roughness, precipitation,

distance to nearest road and distance to coastline) (Rouget et al., 2003). Spatial

distributions of 80 different tree species were predicted with regression tree analysis

using climate, soil factors, elevation, and land use as predictors (Iversen and Prasad,

1998). They reported that for tree species capable of occupying a wide range of habitats,

the classification accuracy was greater than 70%. The US Environmental Protection

Agency (Pan et al., 1999) found that broad spatial patterns of benthic diatom assemblages

in Mid-Atlantic streams could be predicted both by coarse-scale ecological factors (such

as land cover and land use), and by site-specific ecological factors (such as riparian

conditions), using a regression tree model.

Total P data collected in the field (described in Chapter 3) showed a non-normal

distribution, and had different degrees of variation within multiple categories (i.e., land

use, vegetative community). It is the aim of this chapter to use statistical relationships

between measured soil TP (mg kg-1) and ancillary spatial data available for the whole

study area, to predict TP in unsampled, historically isolated wetlands. Because each

wetland is an ecosystem, there are many interacting variables (both quantitative and

categorical) at multiple scales which may drive TP storage. Classification and regression

trees are analytical techniques that are not bound by restrictive assumptions of linear

statistical models (Iverson and Prasad, 1998), so are suited to the development of

predictive TP upscaling models in my study.









Hypotheses

* Satellite reflectance data from wetland areas and associated upland areas is
predictive of TP condition in historically isolated wetlands.

* Soils information related to the uplands is predictive of TP condition in historically
isolated wetlands.

* Land use information related to the uplands is predictive of TP condition in
historically isolated wetlands.

* Landscape metrics related to wetland location is predictive of TP condition in
historically isolated wetlands.

Objectives

* Develop a model that predicts TP storage in the surface soils of unsampled
wetlands using measures from the spatial sampling described in Chapter 3 in
combination with available spatial ancillary environmental datasets.

* Predict and map the TP stored in surface soils of unsampled historically isolated
wetlands.

Materials

Because we were interested in the TP stored in mostly disconnected historically

isolated wetlands, a polygon-based approach was employed in this GIS study. The NWI

(USFWS, 2002) dataset was used to define the unsampled wetlands. Other spatial layers

included the basin boundaries, waterways, land use, soils, major roads, DOQQ photos, a

2003 Landsat7 ETM+ satellite image, and the sampling data described in Chapter 3. All

layers were projected to HPGN (NAD83) Albers projection. Projection parameters and

metadata about the non-spectral data layers are listed in Appendix A. Appendix B

provides a glossary of GIS and remote sensing terms used in this text.

Land Use

The land use layer used in this chapter was from 2003 (an update to the 2001

land-use layer used in Chapter 3). It employed three levels of land use codes, according







82


to the Florida Land Use and Cover Classification System (FLUCCS) (FDOT, 1999). The

processes for classifying IMP and UNIMP land uses, and for integrating wetland land use

polygons into land use areas (as described in Chapter 3), were repeated for the 2003 land

use layer (Figure 4-1).


-

L~L




44


a. .

'^ '.-." -



"_. "il .-\



~- -- E. ,
h i-


-2


0 5 10


Sdairies
impro ed pastures
Sunimpro ed pastures / rangelands












n"-





.. N ...


20
SKilometers


Lgk e C'ikeeCliobee


Figure 4-1. Land uses (2003) of the priority basins.

National Wetland Inventory

The NWI dataset provided the basis of spatial wetland information in this study.

Associated with each wetland polygon was a hierarchical wetland classification system









with four levels: system, subsystem, class, and subclass. Class represents the appearance

of the wetland in terms of vegetation substrate (USFWS, 2002), and in this thesis is

referred to as the vegetative community. The NWI dataset also included modifiers that

described wetland hydrology. A problem with the dataset was that polygons exhibited

multidirectional shifts when compared to DOQQ photos (Figure 4-2) and the Landsat7

image. This is probably because the NWI layer was produced from stereophoto acetate

layers made by hand from unrectified paper images (before 1993) (J. Miner, NWI

employee, personal communication, July 21, 2004).




















SNW polygos 250 500 Meters
[ NWI polygons

Figure 4-2. National Wetland Inventory polygons. A) In the four basins. B) Overlaying
a 1 m DOQQ showing multidirectional shifts.

Soil Survey Geographic (SSURGO) Data Set

Soil spatial information (and associated database files) for each of the four counties

in the study area (Martin, Okeechobee, Highlands, and St. Lucie) were downloaded from

the Florida Geographic Data Library (FGDL) (Figure 4-3). This dataset is a digital




Full Text

PAGE 1

PREDICTING SOIL PHOSPHORUS STORAGE IN HISTORICALLY ISOLATED WETLANDS WITHIN THE LAKE OKEECHOBEE PRIORITY BASINS By KATHLEEN A. MCKEE A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE UNIVERSITY OF FLORIDA 2005

PAGE 2

Copyright 2005 by Kathleen A. McKee

PAGE 3

To my nieces and nephews Andrea, Matt, R obert, Jonathan and Andrew, for whom I hope to incite learning. “The learning process is something you can incite, literally in cite, like a riot.” --Audre Lorde

PAGE 4

iv ACKNOWLEDGMENTS This thesis would not be possible wit hout the gracious supp ort of at least 54 landowners from Okeechobee, Martin, Highlands and St. Lucie Counties, many of whom fostered a stimulating exchange between scie nce student and land st eward. Especially helpful were Pete Beaty, Louis Larson, J.C. Bass, and Norman Ephraim: expert cattle ranchers, and lovers of the land. Other land owners included Keith Rucks, Williamson Cattle Co., Triple Cross Horse and Cattle, Triple Diamond Ranch, Allen Smith, Allen Lewis, Edwin Walpole, the Duncans, Jerald Newcomer, Taylor Creek Ranch, B&E Ranch and Grove, Grassy Island Ranch, Pete Clemons, Patrick Luna, Ralph Palaez, James Fowler, Seller Prescott, Robert Ar nold, Lykes Brothers, Danny Fairclaw, Robert Edwards, Sacramento Farms, Haynes Williams, Nano Corona, Willoway Ranch, Richard Smith, David Durango, Richard Hales, Harv ey Cattle Co., Lois Johnson, the South Florida Water Management District, M Cr oss Ranch, Marion Wagner, Linda Harvey, Hamrick and Sons, Roy Hancock, Kirton Dudl ey, Bill Ritchie, and Larry Overton. The MacArthur dairy staff were es pecially open and helpful hosts. Other supportive dairies include Davie Dairy, Flying G Dairy, Larson Dairies and HW Rucks. The support of landowners was expertly ne gotiated by Mitch Flin chum (University of Florida Institute of Food and Agricultural Services, Extension) and Linda Crane (Florida Department of Agriculture and C onsumer Services). Ed Dunne and Charles Campbell (along with many other volunteers) worked from dark morning hours, filling ice chests, until dark evening hours, pulling soil cores. Erin Colburn at the South Florida

PAGE 5

v Water Management District provided inva luable, up-to-date GIS information and support. Biogeochemist Erin Bostic (o f the UF Wetland Biogeochemistry Lab) coordinated the processing of over 3000 samples, trained staff, and executed laboratory methods with quality contro l as her top priority. I thank Sabine Grunwald (my major prof essor) who gave me encouraging support and independence while keeping me moving on track. Committee member Sue Newman provided a strong scientific eye to this thesis. Special thanks go to Dr. Mark Clark for his multi-lateral support in many areas of my graduate experience, including field equipment, vegetation identification, wetla nd ecology, lab logistics, volun teers, landowners and data interpretation. He was always available wh en I needed a compassionate ear. Postdoctoral associates Ed Dunne, Greg Bruland, and Matt Cohen were invaluable guides of statistics, phosphorus, and the sc ientific process. I thank th e faculty and staff of the Soil and Water Science Department, who were very helpful, and interested in answering my many biogeochemistry and soil questions. La b-mates Rosanna Rivero, Sanjay Lamsal, Aarthy Sabesan, Adrien Mangeot and Vina y Ramasundaram were a huge source of support and friendship on a daily basis. Funding was provided by South Florida Water Management District, Florida Department of Agriculture and Consumer Services, and the Florida Department of Environmental Protection. Finally, my parents could not have been more supportive of my decision to leave a secure career, to try my hand at somethi ng new and ever more valuable to me.

PAGE 6

vi TABLE OF CONTENTS page ACKNOWLEDGMENTS.................................................................................................iv LIST OF TABLES...............................................................................................................x LIST OF FIGURES.........................................................................................................xiii ABSTRACT.....................................................................................................................xv i CHAPTER 1 INTRODUCTION........................................................................................................1 Need for Research.........................................................................................................2 Objectives..................................................................................................................... 4 2 SOIL PHOSPHORUS..................................................................................................6 Phosphorus in Wetlands........................................................................................7 Role of Wetlands for Retaining P..........................................................................9 3 PHOSPHORUS OBSERVATIONS IN NONRIPARIAN WETLANDS..................11 Introduction.................................................................................................................11 Land Use..............................................................................................................12 Hydrology............................................................................................................13 Phosphorus Gradient and Hydrologic Zones.......................................................13 Hypotheses..........................................................................................................15 Objectives............................................................................................................15 Materials and Methods...............................................................................................15 Study Area...........................................................................................................15 Spatial Data.........................................................................................................18 Generating Layers for Site Selection...................................................................20 Step 1: Creating riparian a nd nonriparian wetland layers............................22 Step 2: Creating nonriparian emergent marsh, forested and scrub shrub wetlands..................................................................................................24 Step 3: Creating land-use layers...................................................................25 Step 4: Create six treatment layers...............................................................27 Step 5: Stratified random sampling..............................................................28

PAGE 7

vii Field Sampling Methods and Site Descriptions..................................................29 Laboratory Analyses............................................................................................40 Data Management................................................................................................41 Statistical Analyses..............................................................................................42 Results........................................................................................................................ .45 Wetland Vegetative Communities and Surrounding Land Uses.........................45 Biogeochemical Measures in Hydrologic Zones.................................................45 Combined Effects of Land Use, Ve getative Community, and Wetland Size......46 Correlations between TP and Metals...................................................................47 Effects of Ditches................................................................................................47 Wetland Phosphorus Storage...............................................................................48 Discussion...................................................................................................................63 Zonal TP gradients..............................................................................................64 Hydrologic Connectivity.....................................................................................69 Land Use Differences..........................................................................................70 Other Findings.....................................................................................................71 Storage.................................................................................................................72 Future Research...................................................................................................72 Conclusions.................................................................................................................73 4 UPSCALING TOTAL PHOSPHORUS TO UNSAMPLED WETLANDS..............76 Introduction.................................................................................................................76 Environmental Variables.....................................................................................76 Spectral Data.......................................................................................................77 Ecological Response Variable s and Classification Trees....................................79 Hypotheses..........................................................................................................81 Objectives............................................................................................................81 Materials.....................................................................................................................8 1 Land Use..............................................................................................................81 National Wetland Inventory................................................................................82 Soil Survey Geographic (SSURGO) Data Set.....................................................83 Landsat7 ETM+ Spectral Data............................................................................86 Methods......................................................................................................................89 Upscaling Methodology......................................................................................89 Collection of Independent Variables...................................................................90 Field data......................................................................................................92 Water regime................................................................................................92 Distance to features......................................................................................92 Creating Buffers around Wetlands......................................................................95 Pre-Processing the Landsat7 ETM+ Image.........................................................99 Calculating Spectral Indices from the Landsat Image.......................................103 Classification Trees...........................................................................................106 Classifying unsampled wetlands................................................................109 Edge percent transfer function...................................................................110 Calculating Total Storage..................................................................................111 Results.......................................................................................................................1 13

PAGE 8

viii Important Predictors for TP...............................................................................113 Classification Trees for Upscaling TP Outside the Landsat Extent..................114 Upscaling Results..............................................................................................114 Discussion.................................................................................................................123 Prediction within the Landsat7 ETM+ Extent...................................................124 Prediction outside the Landsat7 ETM+ Extent.................................................126 Storage...............................................................................................................130 Future Research.................................................................................................131 Conclusions...............................................................................................................132 5 SYNTHESIS AND IMPLICATIONS......................................................................134 APPENDIX A SPATIAL DATA METADATA..............................................................................140 B GLOSSARY OF GIS TERMS.................................................................................143 C WETLAND VEGETATIVE COMMUNITY DISTRIBUTIONS...........................145 D SAMPLED SITES....................................................................................................146 E SAMPLING PERIOD RAINFALL..........................................................................149 F ADDITIONAL SOIL BIOGEOCHEMICAL DATA..............................................150 G LANDSAT7 ETM+ HEADER FILE.......................................................................163 H SOILS IN UPLANDS ADJACENT TO HISTORICALLY ISOLATED WETLANDS............................................................................................................165 I OUTPUT FILES OF FINAL CLASSIFICATION TREES.....................................168 Center Soils in Landsat Area....................................................................................168 Edge Soils in Landsat Area.......................................................................................178 Center Soils in NonLandsat Area.............................................................................188 Edge Soils in NonLandsat Area................................................................................196 J SCRIPTS FOR CLASSIFYING UNSAMPLED WETLANDS AND CALCULATING STORAGE..................................................................................205 Classifying Unsampled Wetlands.............................................................................205 Centers in Landsat Area....................................................................................205 Edges in Landsat Area.......................................................................................206 Centers in nonLandsat Area..............................................................................206 Edges in nonLandsat Area.................................................................................206

PAGE 9

ix Calculating Storage in Unsampled Wetlands...........................................................207 Edges.................................................................................................................207 Centers...............................................................................................................207 LIST OF REFERENCES.................................................................................................208 BIOGRAPHICAL SKETCH...........................................................................................224

PAGE 10

x LIST OF TABLES Table page 3-1 Frequency, proportion and area of wetlands sampled by land use........................32 3-2 Distribution of vegetation classes in hydrologic zones of sampled wetlands........32 3-3 Sampling scheme of historically isol ated wetlands by land-use type and vegetative community............................................................................................36 3-4 Distributions of National Wetland I nventory polygons by vegetative community and land use........................................................................................49 3-5 Physico-chemical propertie s among hydrologic zones..........................................50 3-6 Total P comparisons among hydrologic zones......................................................51 3-7 One M HCl extractable P comparisons..................................................................52 3-8 Metal comparisons among hydrologic zones.........................................................52 3-9 Effects of land use and vegetative community and wetland area on soil characteristics.........................................................................................................53 3-10 Estimated unweighted means and standard errors of pH.......................................53 3-11 Effects of land use and vegetative community and wetland area on center soil TP .................................................................................................................56 3-12 Effects of land use and vegetative community and wetland area on edge soil TP .................................................................................................................59 3-13 Spearman correlations for wetland cent er soil TP and biogeochemical parameters..............................................................................................................61 3-14 Spearman correlations for wetland e dge soil TP and biogeochemical parameters..............................................................................................................61 3-15 Total P stored in surface soils of sampled wetlands among land uses...................62 4-1 Landsat7 ETM+ spectral bands.............................................................................86

PAGE 11

xi 4-2 Non-tidal water regimes of wetlands in the National Wetland Inventory.............92 4-3 Linear transformation coefficients to produce three Tasseled Cap indices.........105 4-4 Predictors introduced into the cla ssification tree building process......................108 4-5 Mean TP and bulk density of sampled wetland soils by hydrologic zone and TP class.........................................................................................................112 4-6 Important variables for the center soils classification tree using spectral data....116 4-7 Prediction success table for the cen ter soils classification tree............................116 4-8 Important variables for the edge soils classification tree us ing spectral data......117 4-9 Prediction success table for the edge soils classification tree using spectral data.........................................................................................................117 4-10 Important variables for the center soils classification tree excluding spectral data.........................................................................................................118 4-11 Prediction success table for the center soils classification tree excluding spectral data.........................................................................................................118 4-12 Important variables for the edge soils classification tree excluding spectral data.........................................................................................................119 4-13 Prediction success table for the edge soils classification tree excluding spectral data.........................................................................................................119 4-14 Predicted TP storage in surface soils a nd descriptive statistics of unsampled historically isolated wetl ands by size and land use..............................................119 A-1 Projection information fo r GIS data layers..........................................................140 C-1 All National Wetland Inventor y polygons within the four priority basins by vegetative community..........................................................................................145 D-1 Sampled wetland locations and selected characteristics......................................146 E-1 Rainfall record by month of sampling period......................................................149 F-1 Comparisons of P extracti ons among hydrologic zones......................................151 F-2 Observed values and comparisons of P extractions among land uses.................152 F-3 Observed values and comparisons of metals among land uses............................154 F-4 Pearson correlations of selected soil biogeochemical parameters.......................156

PAGE 12

xii F-5 Pearson correlations of selected biogeo chemical parameters in wetland center soils in improved pastures....................................................................................157 F-6 Pearson correlations of selected biogeo chemical parameters in wetland center soils in dairies......................................................................................................158 F-7 Pearson correlations of selected biogeo chemical parameters in wetland center soils in unimproved past ures / rangelands...........................................................159 F-8 Pearson correlations of selected biogeo chemical parameters in wetland edge soils in improved pastures....................................................................................160 F-9 Pearson correlations of selected biogeo chemical parameters in wetland edge soils in dairies. ...................................................................................................161 F-10 Pearson correlations of selected biogeo chemical parameters in wetland edge soils in unimproved pastures / rangelands...........................................................162 H-1 Soil series in the components of prevalent map units of upland areas................166

PAGE 13

xiii LIST OF FIGURES Figure page 2-1 Selected P cycle components in wetlands..............................................................10 3-1 Lake Okeechobee with 41 watershe d basins and county lines..............................16 3-2 Four priority basins, counties and waterways........................................................19 3-3 Land use in the priority basins...............................................................................20 3-4 Procedure for creating spatial laye rs and selecting sample sites...........................21 3-5 Selecting riparian wetlands....................................................................................23 3-6 All NWI wetlands classified as riparian and nonriparian......................................24 3-7 Nonriparian wetlands by wetla nd vegetative community......................................25 3-8 Three steps to generalize pol ygons by vegetative community..............................26 3-9 Reclassification of land-use wetland polygons......................................................27 3-10 Land-use layer showing land-use areas of in terest in the priority basins after land-use reclassification.........................................................................................28 3-11 Sampled wetlands in the four prior ity basins of the Lake Okeechobee watershed...............................................................................................................30 3-12 Percentage of sites by land use per vegetative community....................................31 3-13 An example of zonation with in an isolated wetland..............................................33 3-14 An example of the “mixed transitional” vegetation class......................................34 3-16 Detail areas of DOQQ p hotos showing hydrologi cal connection types................37 3-17 Three intensities of man-made drainage ditches....................................................38 3-18 Ditch class by land use...........................................................................................39 3-19 Ditch number by land use......................................................................................39

PAGE 14

xiv 3-20 Ditch intensity by land use.....................................................................................40 3-21 Soil bulk density and percent organi c matter comparisons among vegetative communities...........................................................................................................54 3-22 Soil bulk density and percent organic matter by detailed land use showing mean values............................................................................................................55 3-23 Scatterplot of center soil total phosphorus ve rsus wetland size.............................56 3-24 Plots of total phosphorus in center soil s showing interaction of vegetation community and land use........................................................................................57 3-25 Total phosphorus comparisons of center soils showing means.............................58 3-26 Scatterplot of edge soil TP versus wetland size.....................................................59 3-27 Total phosphorus comparisons of edge soils showing means...............................60 3-28 Total P in center soils by ditch class......................................................................62 3-29 Land uses and TP stored in surface soils...............................................................63 3-30 Cartoon comparing relative magnitude s of wetland edge and center soil TP among land uses........................................................................................65 4-1 Land uses of the priority basins.............................................................................82 4-2 National Wetland Inventory polygons...................................................................83 4-3 Soil orders in the four priority basins....................................................................84 4-4 National wetland inventory wetland polygons shifted in multiple directions from the SSU RGO map unit polygons .................................................85 4-5 Landsat7 ETM+ image of path 16/41 from March 24, 2003.................................88 4-6 April 5, 2004 photo of a histor ically isolated wetland...........................................88 4-7 Steps for collecting sampled we tland independent variables................................91 4-8 Distance-to-high-inte nsity-area raster....................................................................93 4-9 Distance-to-major-roads raster...............................................................................94 4-10 Distance-to-waterways raster.................................................................................95 4-11 Landsat7 ETM+ false-color composite..................................................................97

PAGE 15

xv 4-12 Steps for creating buffers and associ ating them with the wetland ID....................98 4-13 The Landsat7 ETM+ subset image of study area.................................................101 4-14 Landsat7 ETM+ image and NWI polygons.........................................................102 4-15 Nonriparian polygons before and after manual alignment..................................103 4-16 Wetland area confidence inte rvals by edge percent.............................................110 4-17 Scatterplot of edge perc ent vs. wetland area........................................................111 4-18 Classification tree for sampled we tland center soils using Landsat....................115 4-19 Classification tree for sampled we tland edge soils using Landsat.......................116 4-20 Classification tree for samp led wetland center soils............................................117 4-21 Classification tree for sampled wetland edge soils..............................................118 4-22 Classification distributions of TP in wetland hydrologic zones..........................120 4-23 Unsampled historically isolated wetla nds indicating TP class in center and edge soils as predicted by tw o sets of classification trees...................................121 4-24 Predicted TP in the surface soils of histor ically isolated wetlands.....................122 4-25 Total P class distribution of sample d center soils among soil subgroups of major upland components....................................................................................127 4-26 Total P class distribution of samp led center soils among land uses....................128 4-27 Scatterplot of TP in sampled cen ter soils by wetland perimeter..........................129 5-1 Total P storage by land use: contrasting models..................................................139 E-1 May, 2003 precipitation map...............................................................................149

PAGE 16

xvi 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 PREDICTING SOIL PHOSPHORUS STORAGE IN HISTORICALLY ISOLATED WETLANDS WITHIN THE LAKE OKEECHOBEE PRIORITY BASINS By Kathleen A. McKee May 2005 Chair: Sabine Grunwald Cochair: Mark W. Clark Major Department: Soil and Water Science In South FloridaÂ’s Lake Okeechobee, th e problem of eutrophication is largely caused by phosphorus (P) pollution in runoff from dairies and beef-cattle pastures in four basins of the watershed. A recommendation for mitigating high P loads to surface waters is to restore the hydrology of isolated wetla nds which have been extensively ditched. A synoptic sampling of surface soils (0-10 cm) in 118 wetlands and surrounding uplands aimed to characterize historically isolated wetlands and relationships between total P (TP) and landscape-scale variables. A random sampling scheme, stratified by land use and wetland vegetative community, was used. It was hypothesized th at a decreasing TP gradient exists from wetland centers, to wetland edges, to surr ounding uplands, and that this storage varies with land use and ditchi ng intensity. Overall, wetland center soils stored significantly more TP (median: 24.9 g m-2) than edges (median: 16.5 g m-2) and uplands (median: 12.5 g m-2). Wetland dairy (median: 35.3g m-2) and improved pasture

PAGE 17

xvii (median: 22.7 g m-2) soils stored more TP than wetland soils in unimproved pastures and rangelands (median: 20.4 g m-2). Ditch intensity had a sign ificant effect on center soils, with more TP (g m-2) stored in wetlands with larger ditches. While there were no significant differences in the amounts of calci um, aluminum, iron or magnesium, organic matter content was significantly higher in ce nter soils (median: 20.4%) than in edges (median: 11.8%), and is a controlling factor for TP storage in these wetlands. To upscale these findings to the basin-scale, ancillary spatial data were used to predict TP stored in the all historically isolated wetlands in the priority basins. It was hypothesized that TP can be predicted using spectral data from a Landsat7 ETM+ image, upland soil taxonomic data, land use, landscap e metrics and wetland characteristics. Classification trees were ge nerated using the TP (mg kg-1) measures of the wetland center and edge soils to predict TP condition (hi gh or low) of unsampled wetland center and edge soils. Spectral data re flecting vegetation and soil moistu re were the most important predictors in the classification trees which ha d overall cross-validation accuracy rates of 76%. Sampled bulk density and TP (mg kg-1) means were used to predict TP storage in unsampled wetlands. It was predicted that 2,736,563 400,568 kg (292.3 kg ha-1) are being stored in the surface soils of all histori cally isolated wetlands in the four priority basins. By restoring the hydrology of these we tlands, more P would be stored in upland and wetland soils. Increased wetland area and hydraulic retention times would enhance organic matter accretion and other P-binding mechanisms. These findings support the idea of hydrologic restoration as a useful management pr actice to reduce P loads to surface waters. The landscapescale predictive model will help land managers target high P areas and estimate the effectiveness of future hydrological restoration efforts.

PAGE 18

1 CHAPTER 1 INTRODUCTION Natural eutrophication in freshwater la kes is the accumulation of sediments, nutrients, and organic matter from the surr ounding watershed, with time (Wetzel, 1983). Anthropogenic eutrophication is the acceleration of this proc ess due to human activities, and is a major problem in agricultural waters heds (Sharpley et al., 2000). Phosphorus (P) is often recognized as the limiting nutrient in freshwater aquatic systems, because of its key role in primary production (Syers et al ., 1973; Allen et al., 1982; Sonzogni et al., 1982; Wetzel, 1983). Phosphorus has been identi fied as the key element contributing to eutrophication in Lake Okeechobee in South Ce ntral Florida (Davis and Marshall, 1975; Federico et al., 1981). This ha s resulted in frequent algal bl ooms, detrimental changes in biological communities, and impaired use of water resources (South Florida Water Management District (SFWMD), 1993). Mo st of the P entering the lake from the watershed is from nonpoint sources (Federic o et al., 1981; Fluck et al., 1992; Florida Department of Environmental Protection (FDEP), 2001; Hiscock et al., 2003). Lake Okeechobee is a shallow, subtropical lake, with a surface area of 1,700 km2, and a drainage basin of about 12,000 km2 (SFWMD, 1993). In 1987, Florida adopted the Lake Okeechobee Surface Water Improvement and Management (SWIM) Plan (SFWMD, 1993). The major goal of the plan is to improve lake water quality through a watershed approach. The SWIM Plan identi fied four of the 41 basins as “priority basins,” because they were contributing most of the P loads to the lake (35%), while occupying 12% of the land area (FDEP, 2001). These basins are known as S-191 (Taylor

PAGE 19

2 Creek/Nubbin Slough), S-65D, S-154, and S-65E (t hree Kissimmee River basins). Land use in the priority basins is dominated by beef -cattle ranches and dair ies. Both of these land uses are thought to be re sponsible for large P exports to the lake via stormwater runoff from grazing pastures and cattle-feeding areas (MacGill et al., 1976; Allen et al., 1982; Anderson and Flaig, 1995). Dairies are co nsidered the most intensive land use for nutrient input in the basins (Reddy et al., 1995). The Clean Water Act (CWA) section 303 (d) lists Lake Okeechobee as an “impaired surface water” and requires the de velopment of pollutant total maximum daily loads (TMDLs). The TMDL goal for P to Lake Okeechobee is 140 t per year by 2015 (FDEP, 2001). Between 1995 and 2000, the lake received an average of 640 t of P per year (FDEP, 2001). In 2002, the P load measured into Lake Okeechobee was 543 t (SFWMD et al., 2004a). The SFWMD’s Works of the District permitting plan set Ploading targets for each basin in order to r each the TMDL goal, but these goals are still being exceeded in basins S-191 and S-154 (U S Army Corps of Engineers (USACE) and South Florida Water Management District (SFWMD), 2003). Phosphorus measured in basins S-65D and S-65E are thought to be belo w targets because of effective agricultural best management practices, but uncertainty rema ins about the future fa te and transport of P in all four of the basi ns (USACE and SFWMD, 2003). Need for Research Continued improvements are needed to reduce nonpoint sour ce P pollution from agriculture in these basins. It is known th at P retention by wetlands helps reduce P loads to receiving waters (Mitsch, 1992; Reddy et al., 1996a; Nairn and Mitsch, 1999; Braskerud, 2002; Coveney, 2002; Kuusemets and Mander, 2002; Shan et al., 2002). Wetland systems between agricultural land a nd water bodies can help improve water

PAGE 20

3 quality (Sharpley, 1995; Reddy et al., 1996b; Sharpley et al ., 2000; Tate et al., 2000). Boggess et al. (1995) concluded that 17% of total P inputs to the northern watershed of Lake Okeechobee (containing the priority basins ) was stored in wetland s, and that 60% of the mass transported from the uplands via su rface and subsurface water flow was stored in wetlands. They also concluded that ba sins with greater we tland areas could help reduce P loading from uplands to surface water systems. Hiscock et al. (2003) used a modeling approach to estimate that 32% of P runoff from uplands in the northern Okeechobee watershed was stored in wetlands, and that 8% of all P imported to the watershed was stored in wetlands. Reddy et al. (1996a) estimated that 70% of the P imported into the Lake Okeechobee watershed is stored in uplands; and an additional 18% is stored in wetlands and streams, with the remaining balance being in livestock and feed. While wetlands store P, much of what is not retained is exported downstream via drainage systems and stream s (Graetz and Nair, 1995). During the 1960s, 16,000 to 20,000 ha of floodpla in wetlands were drained for the development of agriculture (Lof tin et al., 1990). In the last 50 years, drainage ditches have been installed to drain lands for gr azing pasture (Flaig and Reddy, 1995). The Lake Okeechobee Protection Act of 2000 states that about 45% of the wetlands north of the lake have been ditched (SFWMD et al., 2004a ). Part of the Lake Okeechobee Protection Plan (LOPP), a response to the act, suggests that it may be necessary to “reclaim isolated wetlands on pasture lands” in order to redu ce P loads and attenuate peak water flows during storm events to the lake (SFWMD et al., 2004a). In our study, isolated wetlands were defi ned as depressional wetlands completely surrounded by upland, which may be hydrologi cally connected to other wetlands and

PAGE 21

4 waterbodies through ground-water flows or intermittent overflows (spillovers) (Tiner, 2003). Natural hydrologic influences on thes e wetlands are principally precipitation, groundwater, and local runoff. Man-made conne ctions to other surface water bodies are often associated with these wetlands, so they are referred to as hist orically isolated or nonriparian wetlands. Flaig and Havens (1995) estimated that 15% of wetlands in this area are isolated. According to the National Wetland Inventory (NWI) (U.S. Fish and Wildlife, 2002), wetlands cover 18% of the four priority basins, and 59% of those are hi storically isolated. While some studies have invest igated the role of riparian wetlands in the study area (Reddy et al., 1995; Reddy et al., 1996a), few st udies have addressed P retention in the historically isolated wetla nds. Moreover, little is known about the biogeochemical, vegetative, and hydrologic characteristics of these wetlands. To predict the success of reclaiming historically isolated wetlands, it is important to quantify the land area extent of these wetlands, determine how much tota l P (TP) they are currently storing, and investigate watershed and fieldscale factors that affect P st orage. This study focused on soil TP storage, as most of the long-term P storage in wetlands occurs in soils (Johnston, 1991; Reddy 1996a). Objectives Objectives of my study were as follows: Sample and characterize the historically is olated wetlands in dairies and beef-cattle pastures of the priority basins: land area, vegetation community, surrounding land use, hydrologic connectivity, soil physicoc hemical characteristics, and P content of surface soils. Use sampling data to predict the TP stored in surface soils of all historically isolated wetlands in dairies and beef-cat tle pastures of the priority basins.

PAGE 22

5 These objectives and related hypotheses are discussed in Chapters 3 and 4, respectively. Chapter 5 pr ovides a synthesis of both chapters and discusses the implications the results may have for land managers and researchers investigating the possibility of hydrologic restorati on of historically isolated we tlands in the priority basins of the Lake Okeechobee watershed.

PAGE 23

6 CHAPTER 2 SOIL PHOSPHORUS Soil phosphorus (P) exists in inorganic a nd organic forms that can either be in particulate or dissolved forms. Native sour ces of P include phospha te rock like apatite (inorganic) and dying cells (organic). Common alloch thonous sources of P in agriculturally dominated watersheds are commer cial fertilizer (ino rganic P) and cattle feed (organic and inorganic P) (Reddy et al ., 1999a). Graetz and Nair (1995) found that 18-20% of the P in upland surface soils of the northern Lake Okeechobee watershed was organic, compared to 25-50% in the subsur face spodic horizons, and suggested that this was due to leaching of organic P from surface horizons to subsurface layers. This form of P, which is not readily adso rbed to negatively charged soil particles, is susceptible to leaching from sandy upland soils (Reddy et al., 1996b). Organic P can be transported from upland soils of this region to a we tland system through lateral groundwater and surface flow (Graetz and Nair, 1995; Campbe ll et al., 1995). In most cases, organic forms of P are not directly available for bi ological uptake. En zymes called phosphatases are produced by many plants, algae, and microbi al species to mineralize organic P into inorganic phosphate anions (e.g., H2PO4 -), making it bioavailable (Troeh and Thompson, 1993; Brady and Weil, 1999). Dissolved inorganic P (DIP) is directly bi oavailable, so P conc entrations in this form are of greatest concern for water qual ity (Reddy et al., 1995). The adsorption of DIP to soil particles makes it unavailable for biological uptake and is of interest for mitigating excess P in aquatic systems. Organic matter is predominantly negatively

PAGE 24

7 charged so has little capacity to bind DIP. However, DIP can react with organometallic molecules: Fe+3 and Al+3 cations that are associated w ith organic matter molecules (Zhou et al., 1997). Phosphate ions can also be adsorbed to positively charged sites on clays and metal hydroxides (Brady and Wiel, 2002). Another adsorption process is ligand exchange, whereby a phosphate ion replaces a surface hydroxyl (OH-) on metal hydroxides (Rhue and Harris, 1999). After longe r periods of time, and with sufficient P concentrations, P can diffuse into amorphous and poorly crystalline metal oxides (Rhue and Harris, 1999). Once critical P concentr ations are exceeded, precipitation with Fe, Al, or Mn, (to form insoluble hydroxyl phosphates) can occur (Brady and Weil, 2002). Phosphorus in Wetlands Wetland soils can provide long-term storage of P with a turnover time of 96 years in contrast to herbaceous plants turnover ti me of two years (Johnston, 1991). If P inputs to wetland components are greater than out puts, a wetland is a sink for P (Johnston, 1991). Phosphorus retention in wetlands is not considered long term storage unless soil or litter is accumulating with time because the number of P adsorption sites on soil and organic matter is finite (J ohnston, 1991; Reddy et al., 1999b). The ability of wetlands to retain P depends on soil physico-chemical pr operties, the amount of P that has been loaded to the system, and hydraulic retenti on time (Reddy et al., 1998). These factors control P fluxes among components such as ve getation, soils, plant lit ter, and overlying water (Johnston, 1991). Some major aspects of the P cycle are shown in Figure 2-1. Dissolved inorganic P in the water colu mn and soil can be taken up by vegetation and other living organisms. Wh en plant leaves senesce and organisms die, organic P can accumulate in layers of detritus when mine ralization rates slow down due to anaerobic conditions in the flooded soils. This can be a major long term storag e component of P in

PAGE 25

8 wetland ecosystems (Reddy et al., 1993; Reddy et al., 1996b). If the P concentration of water loaded to a wetland is higher than the P concentration in interstitial soil water, P can diffuse from the water column to the soil porewater. If the P c oncentration of loading waters is lower, P will be released from th e soil porewater and the wetland may act as a source for P (Reddy et al., 1995). If similar gradients exist between the porewater and the soil particles, P can either be sorbed or de sorbed by soil particles. If P is adsorbed to soil particles for long periods of time, they ma y diffuse into the particle making it less bioavailable. This can be a long-term form of P storage until the gradient is reversed for a long period of time. At the soil-water interface, there can be an oxidized layer where P can precipitate with ferric iron (Fe+3) (Patrick and Khalid, 1974). Howe ver, deeper in the soil profile, low redox potential (below 120 mv) can cau se reducing conditions that reduce Fe+3 to ferrous form (Fe+2), releasing previously bound P (F aulkner and Richardson, 1989). Phosphorus can also precipitate with aluminum (Al) to form aluminum phosphate which is stable in acidic soils (Richardson, 1999). This reaction is not a ffected directly by redox conditions, but rather by pH (Reddy et al ., 1998; Kadlec and Knight, 1996; Mitsch and Gosselink, 2000). In alkaline wetland environments, P can bind with Ca and Mg. Changes in soil and surface water pH can cau se these compounds to become unstable, and release P (White and Taylor, 1977). Tw o examples of processes affecting pH are algal growth, which may increase water column pH, and decomposition of organic matter, which releases organic acids to lower the pH (Mitch and Gosselink, 2000; Troeh and Thompson, 1993). Liming and manure appli cation to pastures may also raise pH.

PAGE 26

9 Low redox conditions can slow anaerobic microbial decomposition processes (Lvesque and Mathur, 1979) lead ing to the accumulation of organic matter. In some wetlands, organic matter accumulation can be a major mechanism for long term P storage (Richardson, 1999; Reddy et al., 1996b). Aluminum and Fe hydroxides in amorphous and poorly crystalline forms (oxalateextractable) govern P sorption in many wetland soils and stream sediments (Reddy et al., 1995; Richardson, 1999; Patrick and Khalid, 1974) because of the high surface areas that provide many P binding sites (Rhue and Harri s, 1999). Flaig and Reddy (1995) described the different fractions of total P stored in wetland soils of differe nt land uses in the Okeechobee basins. Phosphorus associated with Feand Al-oxyhydroxides (amorphous forms) accounted for 17-43% of the TP in wetland soils. Calciumand magnesiumbound soil P accounted for 50% of TP in a da iry wetland, which received dairy waste for over 20 years. At nondairy sites within th e basin they found that Caand Mg-bound P was negligible. Role of Wetlands for Retaining P Most wetlands that are restored or creat ed have goals of improving water quality and habitat (Mitsh and Gosselink, 2000). Treatm ent wetlands have been constructed all over the world for the purpose of mitigati ng nonpoint source pollution in agricultural watersheds (Mitsch and Gosselink, 2000). In a review of treatment wetlands, Kadlec and Knight (1996) found that treatment wetlands (on average) removed 31% of the total phosphorus, and 41% of the inorganic phosphorus loaded to them. Other studies have shown that natural wetlands can remove P from downstream waters (Yin and Lan, 1994; Richardson, 2003; Nessel and Bayl ey, 1984), but can also be a source for P (Raisin and Mitchell, 1995; Reddy et al., 1995) (depending on the P concentration of loading waters).

PAGE 27

10 The two major compartments of P storag e in wetlands, are the accumulation of organic matter, and the adsorption sites on so il particles (Richardson, 1999). As wetlands in the Okeechobee basins continue to be drained, detrital layers will oxidize and release P to surface waters. As more P is loaded to wetlands through continuing agricultural practices, fewer binding sites for P will be available in wetland systems, and the wetlands can become a source for P when loaded with low P-concentration waters. In the subbasins of the Lake Okeechobee watershed, Re ddy et al. (1996a) conc luded that 45% of the P storage capacity of wetlands was still available. Wetlands are in need of further study to determine the mass of P that could poten tially be stored in them, and what kinds of land management practices may enhance P storage in wetland soils. Figure 2-1. Selected P cycle components in wetlands (not to scale). The dotted line varies vertically with hydroperi od. DIP = dissolved inorganic P.

PAGE 28

11 CHAPTER 3 PHOSPHORUS OBSERVATIONS IN NONRIPARIAN WETLANDS Introduction A recommendation for mitigating high P loads to surface waters of the four priority basins of the Lake Okeechobee watershed is to restore the hydrol ogy of historically isolated wetlands (SFWMD et al., 2004a; Bottche r et al., 1995). It is estimated that 45% of these wetlands have been ditched to re duce the wetland area, and to increase grazing pasture area (SFWMD et al., 2004a). Many studies of upland soils in the Okeechobee basins have been conducted (Mansell et al ., 1991; Graetz and Nair 1995; Nair et al., 1998; Graetz et al., 1999; Nair et al., 1999; Villapando and Graetz, 2001; Nair and Graetz, 2002; Pant and Reddy, 2003). Several st udies of riparian wetlands have been conducted (Scinto, 1990; Reddy et al., 1995; Reddy et al., 1996a; Reddy et al., 1996b; Reddy et al., 1998), but there is little informa tion about historically isolated wetlands in the Lake Okeechobee basins (Steinman et al., 2003; Sperry, 2004). These wetlands represent over half of the wetland area in the four priority basins (based on an analysis described in this chapter). As a result, they potentially represent a large storage compartment for P, which is currently unknown. Understanding the role of these wetlands in P storage at different scales, is important for assessing long-term efforts to reduce P loads to Lake Okeechobee. It is important to characterize the extent of these wetlands, and determine how much total phosphorus (TP) the soils are cu rrently storing. Factors th at influence P storage at multiple scales in historically isolated wetlands have not been addressed in this region.

PAGE 29

12 This chapter will investigate factors that a ffect TP storage in wetlands at fieldand watershed-scales. Land Use Landscape conditions at watershed-scale ca n influence nutrient loadings to surface waters (Hunsaker and Levine, 1995; Borsuk, et al ., 2002). It is widely accepted that the land use and land cover (LULC) of a landscap e have a major effect on water quality (Herlihy et al., 1998; Cuffney et al., 2000; Berka et al., 2001). Strong relationships exist between LULC and water quality as related to P (Hall and Schreier, 1996; Bolstad and Swank, 1997; Wear et al., 1998). Reddy et al. (1996a) found that TP stored in upland surface soils (ranging from 18-24 cm in depth) of the Lake Okeechobee watershed increased with intensity of land use: mean values of TP were 30 mg kg-1 in native unimpacted (rangeland) areas, followed by 46 mg kg-1 in forage areas (unimproved pastures), 84 mg kg-1 in improved pastures (fertilize d, artificially drained and possibly planted with forage grasses), and 1,792 mg kg-1 in intensive areas (dairy pastures). Total P concentrations in the A horiz on of manure-impacted upland so ils of the priority basins were 34 kg P ha-1 in nonimpacted (rangeland) soils, 165 kg P ha-1 in improved pasture soils and 1,680 kg P ha-1 in high manure-impacted dairy soils (Graetz and Nair, 1995). Up to 80% of TP had the potential to leav e heavily-manured upland regions, while less than 10% of the TP was likely to leave a low manure-impacted pa sture soil (Graetz and Nair, 1999). This suggests that more TP is tr ansported to wetlands and ditches of more intensely manure-impacted soils. Studies of historically isolated wetland soils in the Lake Okeechobee watershed found that TP in wetland soils was hi gher in areas with more intensively grazed and fertilized pa stures (Steinman et al., 2003; Sperry, 2004).

PAGE 30

13 Sperry (2004) found that ditched wetlands in improved pastures released 5 to 7 times more TP in surface water runoff than wetlands in semi-native pastures. Hydrology Studies have demonstrated that wetlands with surface connections to downstream waterways export nutrients (Nessel and Ba yley, 1984; Yin and Lan, 1994; Raisin and Mitchell, 1995; Richardson, 2003; Mwanuzi et al., 2003; Sperry, 2004). Surface connections may increase water velocities in wetlands to suspend and transport P-laden sediments (Phillips, 1989; Mitsch and Gosseli nk, 2000). Draining of wetlands can reduce their size and expose previously submer ged organic matter which may oxidize and release labile P to surface waters (Lves que and Mathur, 1979; Whigham and Jordan, 2003;). Campbell et al. (1995) concluded that surface drainage (ditches and canals) could be a significant mechanism for P transport to Lake Okeechobee. Phosphorus Gradient and Hydrologic Zones Properties of upland soils in the Lake Okeechobee watershed, which are mainly Spodosols, have important influences on the transport of P to wetland areas. The A horizon is sandy with some organic matter (2 -5%) and is generally 15 to 20 cm thick (Graetz and Nair, 1995; Lewis et al., 2001). The E horizon consists primarily of white sand as a result of the eluviation of clay a nd organic matter to the Bh (spodic) horizon (Graetz and Nair, 1995). The A horizon has very low P retention capacity and the E, or albic, horizon has almost none (Mansell et al., 1991). Phosphorus binds with organometallic complexes in the spodic horizon which creates a stable pool of relatively insoluble P (Yuan, 1966; Reddy et al., 1996a). While the spodic horizon is thought to have a high retention capacity for P, drainage water may not come into contact with this horizon, as water table levels ar e above it during much of the rainy season (Graetz et al.,

PAGE 31

14 1999; Blatie, 1980). The poor P retention and high water permeability of these sandy acidic surface soils (as well as the route of P transport vi a surface and subsurface flow along eluviated horizons) make th ese soils ineffective for P removal, which leads to a high P concentration in the drainage wate r (Campbell et al., 1995). Thus upland soils adjacent to wetlands should have less P than the more organic wetland soils that receive drainage waters. Phosphorus reactivity in wetland soils is known to depend on hydrologic conditions (Mitsch and Gosselink, 2000). Wetland cente rs typically have more undecomposed organic matter because of longer hydraulic retention times compared to wetland edges and this can be the dominant long-term P storage pool in wetlands (Richardson, 1999; Reddy et al., 1996b). Higher amounts of amor phous, and less crystalline forms of iron (Fe) and aluminum (Al) (which are thought to be the most important agents of P sorption in acidic soils) are also found in soils that are flooded for longer periods (Reddy et al., 1995; Richardson, 1985; Patrick and Khalid, 197 4). In a study of 12 improved pasture wetlands in a beef ranch in the Lake Okeec hobee watershed, surface soil (0-15 cm) TP in wetland centers was significantly higher (242 mg kg-1) than wetland edges (171 mg kg-1) (Sperry, 2004). When these numbers were adjusted for bulk density, there was no significant difference (Sperry, 2004). Characteriz ing differences of P storage in wetland hydrologic zones may be important for assessing the potential increase in P storage due to hydrologic restoration. The surface soil layers of the study area have been shown to contain the most TP and higher P sorption capacities (Scinto, 1990; Nair et al., 19 95; Rechcigl and Bottcher et al., 1995). My study surveyed the surface soils of historically isolated wetlands in order

PAGE 32

15 to characterize TP storage and investigate re lationships between TP and variables at the fieldand watershed-scales. Information colle cted in this chapter was used in Chapter 4 to predict TP storage in all historically is olated wetlands of the four priority basins. Hypotheses A zonal gradient exists within histori cally isolated wetlands, with TP storage decreasing from wetland centers, to wetland edges, to uplands. Hydrologic connectivity affect s TP storage, which decrea ses, as ditching intensity increases. Adjacent land use affects TP stored in wetlands, with highest TP stored by wetlands surrounded by dairies, followe d by improved pastures, followed by unimproved pastures and rangelands. Objectives Characterize extent, vege tative communities and surrounding land uses of historically isolated wetlands in dairies and beef-cattle pastures. Survey biogeochemical characteristics of historically isolated wetland surface soils. Investigate relative TP storage in the surface soils of two wetland hydrologic zones (center and edge) and surrounding upland. Investigate relationships between wetl and surface soil TP and the associated hydrologic connectivity, vegetative co mmunity, and surrounding land use. Materials and Methods Study Area The four priority basins are north of Lake Okeechobee and cover an area of 121,000 ha, with 101,200 ha (84%) in Okeechobee C ounty (Figure 3-1). The climate is subtropical, with a summer wet season and a winter dry season. Th e rainfall average, since 1948, is 127 cm per year in the southeas tern part of the study area (Southeastern Regional Climate Center, 2004). Most of the rainfall (75%) occurs during the wet season, between May and September (Shih, 1983). There is little topographic relief (0-

PAGE 33

16 2% slopes), and the water table is near or at the soil surface during much of the wet season (Blatie, 1980). It can drop to 2 m be low the surface during dry months (Knisel et al., 1985). The underlying geology is unconsolidated marine sediments, primarily sand and gravel, with clay lenses (P arker, 1955). A series of three confined aquifers occurs at depths greater than 40m: the Surfical Aquife r System, the Intermediate Aquifer System and the Floridan Aquifer (Miller, 1997). Ma jor streams that drain to Lake Okeechobee include the Kissimmee River, Fisheating Cr eek, and Taylor Creek (FDEP, 2001). Figure 3-1. Lake Okeechobee with 41 watershe d basins and county lines in Florida, USA. The study area consists of four priority basins colored orange. According to an analysis of the Flor ida National Wetland Inventory (NWI) (US Fish and Wildlife Service (USFWS), 2002) (descr ibed later in this chapter), 18% of the OKEECHOBEE STLUCIE HIGHLANDS GLADES INDIANRIVER MARTIN OSCEOLA POLKS-191 S-65D S-154 S-65E 020 10 Kilometers four priority basins counties Lake Okeechobee 0200 100 Kilometers other basins of the Lake Okeechobee watershed

PAGE 34

17 area consists of wetlands, but this was historic ally estimated to be 25% (McCaffery et al., 1976). Forty-one percent of the areal coverage of wetlands is associated with natural streams, and the rest are mostly small (typic ally less than 1 ha), depressional, nonriparian wetlands. They are usually dominated by em ergent macrophytes, forest, or scrub-shrub vegetation. It has been estimated that 45% of the wetlands in the study area are artificially drained ( SFWMD et al., 2004a). Upland soils of the Lake Okeechobee watershed are mainly Spodosols formed in sandy marine sediments (Graetz and Nair, 1995) which have a naturally low native P content (Hodges et al., 1967). Entisols ar e also common, having weakly formed Bh horizons (Lewis et al., 2001). According to soil data of the area (U.S. Department of Agriculture / Natural Resource Conservation Service (USDA/NRCS), 1995), historically isolated wetland soils are do minated by Aquepts and Aquent s and to a lesser extent, Aquolls and Aqualfs. The Spodosols are mo stly Alaquods. Immokalee (sandy, siliceous, hyperthermic Arenic Alaquod) and Myakka (sandy, siliceous, hyperthermic Aeric Alaquod) are the most common soil series occu rring in the upland ar eas. Alaquods are wet, sandy soils that have a fluctuating water table that moves iron out of the soil profile (Soil Survey Staff, 1999). They have hi gh humus content, and the spodic horizon is mostly an accumulation of Al and orga nic matter. Despite the high hydraulic conductivities of these soils (>16 cm hr-1), drainage is poor (L ewis et al., 2001). Drainage has been improved with extensive ditching of fields and wetlands to convey stormwater runoff towards Lake Okeechobee (Haan, 1995). Ditching densities in the landscape increase with land-use intensity from unimproved pastures to improved pastures to intensively mana ged pastures (dairies) to row crops (Heatwole, 1986).

PAGE 35

18 Based on 2003 land-use data (SFWMD, 2003), the four priority basins are 64.2% agriculture, 7.9% urban, 15.1% wetlands, 6.2% upland forests, 3.6% is transportation / utilities, and 1.5% water. Agriculture cons ists of 47.5% improved pastures, 7.2% dairies, and 6.2% unimproved pastures and rangeland. Ci trus, row crops, and sod farms make up 3.0%, 2.0%, and 0.5% respectively. Dairies and beef-cattle op erations of the study area contribute the majority of P to Lake Okeec hobee in equal amounts, but dairy inputs are much more concentrated (more P per unit ar ea) (MacGill et al., 1976; Allen et al., 1982; Anderson and Flaig, 1995; Reddy et al., 1995). Spatial Data Spatial data layers were assembled a nd analyzed in a geographic information system (GIS) to characterize wetlands and se lect sampling sites. The GIS software ArcGIS version 8.3 that consists of ArcCatalog, ArcMap and ArcToolbox modules developed by Environmental Systems Research Institute (ESRI Inc.; Redlands, CA) was used. The layers included watershed bounda ries, counties, waterways, land use, wetlands, and digital orthorectif ied quarter quadrangles (DO QQ) aerial photos (metadata is in Appendix A). Watershed boundaries de lineated the four priority basins and waterways consisted of canals, stream s and major tributar ies (Figure 3-2). Land use was based on 2001 conditions (Figur e 3-3) and employed three levels of land-use codes, according to the Florida Land Use and Cover Classification System (FLUCCS) (Florida Department of Trans portation (FDOT), 1999). The NWI provided the basis of spatial wetland information in this study. Associat ed with each NWI polygon, was a hierarchical wetl and classification system w ith four levels: system, subsystem, class and subclass. The class re presented the appearance of the wetland in

PAGE 36

19 terms of vegetation substrate (USFWS, 2002), and in this thesis it is referred to as the vegetative community. Lake OkeechobeeKi ssi m m e e R i ve r S-191 basin S-65D basin S-154 basin S-65E basinHIGHLANDS MARTIN STLUCIE OKEECHOBEE C 3 8T A Y L O R C R E E KL 6 2L 6 3 NL E T T U C E C R E E KL 6 2 01020 5Kilometers canals major tributaries streams countiesBASIN S-154 S-191 S-65D S-65E Figure 3-2. Four priority ba sins, counties and waterways. The 1-meter resolution 1999 DOQQ photos were used throughout this study for the verification of polygons of other data layers including wetlands, land use and hydrologic connectivity. All layers we re projected in High Precision Geodetic Network (HPGN), 1983 North American Datum (NAD 83), Albe rs Equal Area proj ection. Projection parameters are listed in Appendi x A. A glossary of GIS term s used in this text can be found in Appendix B. More information about GIS concepts and map projections can be found in Bolstad (2002).

PAGE 37

20 01020 5 Kilometers unimproved pastures woodland pastures dairies improved pastures wetlands rangeland basinsLake Okeechobee Figure 3-3. Land use (2001) in the priority ba sins. White areas represent other land uses (e.g., cropland, residential, upland fo rests, commercial) (SFWMD, 2003). Generating Layers for Site Selection The goal of the GIS analysis was to create 6 layers from which to select samples based on population proportions of representa tive wetland vegetative communities within the most common land uses. The land use trea tment had 3 levels: dairy (DAI), improved pasture (IMP) and unimproved pasture / range land combined (UNIMP). The vegetative community treatment had two levels: emergent marsh (EM) and forested / scrub-shrub

PAGE 38

21 combined (FOSS). These two treatments resulted in 6 cate gories. Figure 3-4 summarizes the steps to create these GIS laye rs which are describe d in detail in the following sections. Figure 3-4. Procedure for creating spatial la yers and selecting sample sites. EM = emergent marsh, FO = forested, SS = scrub-shrub, DAIR = dairy, IMP = improved pasture, UNIMP = unimproved pasture/rangeland. Basins DOQQs Clip to wetlands to four basins Exclude riparian wetlands Florida NWI wetlands Dissolve contiguous wetlands of same community type. Merge FO with SS Group wetlands by vegetative community and land use combinations Randomly select from each, proportional to population % EM-DAIR Exclude wetlands besides EM, FO, SS 2001 land use Hydrography Build 3 land use layers: IMP, UNIMP, DAIR EM-IMP EM-UNIMP FOSS-DAIR FOSS-IMP FOSS-UNIMP 145 sites Buffer streams / major tributaries (25m); Buffer canals ( 50m ) Ste p 1 Ste p 2 Ste p 4 Ste p 3 Ste p 5

PAGE 39

22 Step 1: Creating riparian and nonriparian wetland layers The NWI wetland layer was used to create two layers representing riparian and nonriparian wetlands. Within ArcMAP, NWI a nd the three waterway layers (canals, streams and major tributaries) were clipped to the four basin boundaries. Wetlands that were cut off at basin edges during the clippi ng were deleted to omit them from possible selection as sampling sites, but were included in calculations to quantify wetlands in the study area. To build the riparian wetland layer, th e three hydrography layers were assembled with the NWI wetland layer. Polygons that fell within 50 m of can als, 25 m of major tributaries or 25 m of streams, were selected and saved as a new layer. These distances were chosen based on trial and error attempts to capture the most riparian wetlands, and the least number of nonriparian wetlands (based on visual verification using the DOQQs). Wetlands that followed stream direction, t hose connected to waterways, and those forming chain patterns with in terconnecting natural streams, were considered riparian. Wetlands that were contiguously connected to the riparian wetlands were also selected (Figure 3-5A). These were then merged to th e first set of riparian wetlands to create a new set of riparian wetlands. The remain ing wetlands were saved as the nonriparian wetland layer. Because the stream and major tributary la yers included both natural and man-made hydrological features (i.e., sma ll canals and ditches), some wetlands were misclassified as riparian during the process described above Polygons adjacent to or cut through by man-made ditches were reassigned to the nonriparian wetland layer, based on visual interpretation of DOQQs (Figure 3-5B). On the other hand, some wetlands were

PAGE 40

23 misclassified as nonriparian, since some streams (evident on the DOQQs) were not represented by any of the hydrography layers. NWI polygons streams and major tributaries 25m buffer riparian wetlands 0500 250Meters 00.51 0.25 Kilometers riparian wetlands previously selected as riparian nonriparian wetlands major tributaries streams A B Figure 3-5. Selecting ripari an wetlands. A) NWI polygons intersecting stream and major tributary buffers (25 m wide) plus wetlands contiguous to those. B) Yellow polygons and green polygons repres ent wetlands classified based on distance to waterways. Or ange polygons were origina lly selected as riparian based on distance to waterways but then reclassified as nonriparian because the waterway was a manmade structure. Both layers were visually verified: wetlands following str eam direction, those connected to waterways, or those forming chain patterns with interconnecting natural streams, were considered riparian wetlands. If a polygon was cut from one layer, it was saved into the other one so th at both layers were mutually exclusive and comprehensive (Figure 3-6). Because EM, FO and SS wetlands represente d 88% of the historically isolated wetlands, polygons other than these were sel ected and deleted leav ing an “EM-FO-SS” nonriparian wetland layer (Figure 3-7). Th e deleted vegetative communities included aquatic bed, unconsolidated shor e, open water, and unconsolidated. Distributions of all NWI wetlands in the study area are listed by vegetative community in Appendix C.

PAGE 41

24 01020 5 KilometersLake Okeechobee nonriparian riparian basins Figure 3-6. All NWI wetlands classi fied as riparian and nonriparian. Step 2: Creating nonriparian emergent marsh, forested and scrub shrub wetlands Within the EM-FO-SS nonriparian wetland layer, there were neighboring polygons of the same type, but they had different hydr ological regimes according to data in the NWI (Figure 3-8A). For purposes of sel ecting a random sample based on vegetative community, these polygons were dissolved ba sed on the vegetative community. To do this, the polygon layers were firs t converted to a vegetative co mmunity raster with spatial resolution of 5 m (values were EM, SS or FO ) (Figure 3-8B). The raster was then converted to polygons (Figure 3-8C), e ssentially dissolving neighboring wetland

PAGE 42

25 polygons of the same vegetative community t ype together. Wetlands of type EM were exported into a new separate layer. There we re few FO and SS sites, so wetlands of these similar vegetative community types were exported into one FOSS layer. 01020 5 Kilometers emergent marsh forested scrub shrub basinsLake Okeechobee Figure 3-7. Nonriparian wetlands by wetland vegetative community. Step 3: Creating land-use layers To select nonriparian EM and FOSS wetla nds based on the land use they overlaid, three new land-use layers were needed: IMP, UNIMP and DAIR. The DAIR layer already existed as a separate layer. New IMP and UNIMP layers needed to be created but IMP and UNIMP polygons were interrupt ed by wetland land-use polygons (WETL)

PAGE 43

26 (Figure 3-9A). These land-use polygons of ten corresponded with NWI polygons but the NWI layer was more complete and accurate. In order to identify NWI wetlands that coincided to IMP or UNIMP land uses, the WETL polygons had to be incorporated into the IMP or UNIMP layers. EM EM EM EM EM EM EM EM FO EM CLASS1 EM FO SS A B EM EM EM EM EM FO C Figure 3-8. Three steps to generalize polygons by vegetative community (EM, FO or SS). A) Original nonriparian polygons. B) Conversion to 5 m raster cells. C) Conversion back to polygons. The 2001 land-use polygons were reclassifi ed as IMP, UNIMP or WETL according to FLUCCS codes. Polygons with FLUCCS le vel-2 code equal to 211 were classified as IMP. Those with FLUCCS level-2 code equal to 212 (unimproved pastures) or 213 (woodland pastures), or FLUCCS level-1 code equal to 300 (r angeland), were classified as UNIMP. Polygons with FLUCCS level-1 code equal to 600 were classified as WETL.

PAGE 44

27 The WETL polygons that were completely c ontained by the IMP land-use layer were selected, exported, and then merged with th e IMP land-use layer (Figure 3-9B). The WETL polygons occuring at the edges of the land-use polygons (that were at least 50% within the land-use polygon based on visual judgement) were selected, exported, and merged with the IMP land-use layer. These steps were repeated for the UNIMP layer. These steps were not necessary for the DAIR layer, because all polygons within it were already classified as DAIR. Figure 3-10 shows the three final land-use layers. NWI polygons improved land use land use wetlands 0500 250Meters Figure 3-9. Reclassification of land-use we tland polygons. A) Before reclassification, a NWI wetland (circled in yellow) did not intersect IMP land use. B) After reclassification, the NWI wetland (circled in yellow) intersected with IMP land use. Step 4: Create six treatment layers Six new NWI wetland layers were created by selecting EM or FOSS wetlands that were at least 50% contained with in each of the three land-use layers. It was ensured that the six new wetland layers (EM-DAIR, FOSS-DAIR, EM-UNIMP, FOSS-UNIMP, EMIMP, FOSS-IMP) were mutually exclusive. B A

PAGE 45

28 OKEECHOBEE STLUCIE GLADES HIGHLANDS MARTIN 01020 5 KilometersLake Okeechobee Counties 4 Priority Basins Dairies Wetlands Improved pasture Unimproved Pasture/Rangeland Figure 3-10. Land-use layer (2001) showing land-use areas of interest in the priority basins after land-use recl assification. White areas represent other land uses (e.g., cropland, residential, upland fo rests, commercial) (SFWMD, 2003). Step 5: Stratified random sampling Proportions represented by each of the six wetland strata (wetla nd type combined with land-use type) were determined based on population polygon counts. Based on a minimum sample size of 108, the number of wetlands to be sampled within each treatment was determined. This minimum number was a result of an initial sampling design based on three factors, each with three levels: land use, vegetative community and

PAGE 46

29 hydrologic-connection types. These 27 treatme nts would have a minimum of 4 samples each, thus the number 108. Since no a priori information was available for the wetland hydrological connections, this became a random factor in the current sampling scheme, and the number 108 remained as a commitme nt to the projectÂ’s funding agency. Random numbers for each of the 6 wetland layers were generated using SPSS software version 11.5 for Windows (SPSS, Inc, Chicago, IL). The quantity of sites selected was proportional to the populations of each strata in th e sampling scheme. These random numbers were used to se lect polygons based on polygon identification numbers (unique, GIS-assigned inte ger attributes). A minimum of four sites per category was adopted, with an added 25% in each category. This w ould provide extra sites if wetlands did not exist at the designated ge ographic coordinates or if wetlands were inaccessible. This produced a total nu mber of 145 possible sampling sites. Field Sampling Methods and Site Descriptions I sampled 118 nonriparian wetlands within the research land us es between May 19 and November 19, 2003 (Figure 3-11). Detail s about site locations, sampling dates and selected attributes are listed in Appendi x D. Rainfall during the sampling period is described in Appendix E. Land use and vegetative community. Land uses were assigned to sites based on a combination of the GIS land-use map and visu al observation at the sites. Improved pasture was defined as land which has been cl eared, tilled, reseeded with forage grasses, and periodically treated with brush control and fertiliz er application (FDOT, 1999). Unimproved pastures were defined as cleared la nd with major stands of trees and brush, where native grasses have been allowed to develop, and is typically not managed with brush control or fertilizer (FDOT, 1999). Rangeland was defined as land where the

PAGE 47

30 dominant vegetation is predominantly native gras ses, forbs, and shrubs, and is capable of being grazed (FDOT, 1999). Management may in clude brush control, but is not irrigated, cultivated, or fertilized. Da iry land was defined as any ar ea pertaining to a commercial dairy and has a wide range of usage intensit y (based on fertilization or stocking rates) including ungrazed fields, fertilized hayfield s, sprayfields, feeding pastures, and cowbarn areas. Any of these three land-use type s may have man-made drainage systems. Figure 3-11. Sampled wetlands in the four priority basins of the Lake Okeechobee watershed. EM = emergent marsh, FOSS = forested / scrub-shrub, UNIMP = unimproved pasture / rangeland, IMP = improved pasture, DAIR = dairy. A distinction between very improved (IM P1) and less improved (IMP2) pastures was made and assigned to sites, in order to compare TP content among these additional

PAGE 48

31 land-use classes. Less improve d pastures had less Bahia grass and more native grasses, or, had scrub present. A similar distinc tion was made for the unimproved pastures / rangeland class. Unimproved pastures were classes as UNI1, and rangelands (which had more scrub vegetation) were cl assed as UNI2. Within dairies, there was a variety of land uses. Some areas were grazed by beef cattle (DABE), and some were pastures where dairy cows sometimes received feed (DAPA). Some were in, or adjacent to, sprayfields (DASP) or hayfields (DAHA). Several were in areas that were not currently being managed with grazing or hay activities, a nd looked similar to unimproved pastures (DAUN) (Table 3-1). Many EM sites had large areas (more than 75% of the system) of open water. Due to this variability, a third class was defined as emergent marsh/open water (EW). No EW wetlands were sampled within the UNIMP areas and more than half the FOSS wetlands were sampled within dairies (Figure 3-12). Dairy Improved Unimproved Land Use Bars show percents EMEWFS V e g etation Communit y 0% 25% 50% 75%P e r c e n t Figure 3-12. Percentage of sites by land us e per vegetative community. EM= emergent marsh, EW=emergent marsh/open water, FS=forested/scrub-shrub.

PAGE 49

32 Table 3-1. Frequency, pr oportion and area of wetlands sampled by land use. Land Use Freq. Percent Total ha Dairy 21 17.8 32.86 Dairy Beef (DABE) 4 4.2 Dairy Hayfield (DAHA) 6 5.0 Dairy Pasture (DAPA) 4 3.4 Dairy Sprayfield (DASP) 1 .8 Dairy Unimproved (DAUN) 6 5.0 Improved pasture 85 72.0 97.14 Very improved (IMP1) 51 42.9 Less improved (IMP2) 34 28.6 Unimproved pasture / rangeland 12 10.2 24.4 Unimproved Pasture (UNI1) 7 5.9 Rangeland (UNI2) 5 4.2 Total 118 100.0 154.4 Hydrologic zone delineation and wetland size Vegetation and topography change were used to distinguish hydrologic zo nes. Functional vege tation classes (often found exclusively within a particular hydrol ogic zone) were defined and used to demarcate boundaries between upland, edge and center zones (Table 3-2). Table 3-2. Distribution of ve getation classes in hydrologic z ones of sampled wetlands. Center Edge Upland -------------------------------%---------------------------Aquatic vegetationa 4.2 --Open water 12.7 --Panicum hemitomum Shult. 11.9 -Trees 5.1 0.8 3.4 Shrubs 2.5 7.6 9.3 Juncus effuses L. 6.8 43.2 -Pontedaria cordata L. 25.4 --Polygonum spp. 17.8 --Other native herbaceous 5.9 9.3 4.2 Forage grassesb -8.5 63.6 Other grassesc 7.6 14.4 19.5 Mixed transitionald 0 16.1 -a For example, Nuphar luteum and Nymphea spp. b Often Paspalum notatum Fluegge (Bogdan) (Bahia grass); also Cynodon spp. (stargrass) and Hermarthria spp. (limpograss). c For example, Andropogon spp. and Panicum spp. d Refers to a mix of small herbaceo us plants which often contained Hydrocotyle spp.

PAGE 50

33 The transition between zones was often diffuse, due to overlap of species. Demarcation between upland and edge zones wa s often based on the wa terward extent of Bahia grass ( Paspalum notatum Fluegge (Bogdan). In most IMP emergent marshes, Soft Rush ( Juncus effusus L.) was the dominant species in th e edge zone, and could often be used to demarcate the outer or inner extent of that zone (Figure 3-13). A complex of emergent macrophytes, which often included Hydrocotyle spp., Centella spp and Eleocharis baldwinii (Torr.) Chapm., was often th e outermost wetland edge zone vegetative community, and was referred to as “mixed transitional” (Figure 3-14). Figure 3-13. An example of zonation within an isolated wetland. In this example, the center is the area containing surface water, Polygonum hydropiperoides Michx., and Pontedaria cordata L. The edge consists of Juncus effuses L. and “mixed transitional.” The edge ends where Paspalum notatum Fluegge (Bogdan) (Bahia grass) dominates. Although demarcation lines between zone s were not always easily defined, vegetation differences were easily identified, providing a high degree of confidence that samples were representative of the targeted hydrologic zone. The outer wetland edge was

PAGE 51

34 usually walked and recorded using a Pa thfinder Pro XR (Trimble Navigation LTD, Sunnyvale, CA) GPS receiver. The distribution of wetland si zes was right-skewed. The median area of sampled wetlands was 0.9 ha for EM, 1.5 ha for FS and 1.0 ha for EW. Figure 3-14. An example of the “mixed transitional” vegetation class showing Hydrocotyle umbellata L., Eleocharis baldwinii (Torr.) Chapm. and Diodia virginiana L Soils. Four 10 cm soil cores were extracted and composited for each hydrologic zone (Figure 3-15). If an ou tlet ditch was present, a threecore composite was collected from the base of the ditch beyond the wetla nd edge. Outlet ditches were identified by evidence of water flowing from the wetland or topography change that indicated outward flow. Upland cores were taken between three and twenty meters from the wetland edge. To take a soil core, thic k-walled (0.3 cm) polycarbonate tubes (7.6 cm internal diameter) sharpened at one end were hammere d in with rubber mallets and extracted by hand. The top 10 cm of the soil core was ex truded upward into a separate 10 cm section of tubing and then placed into a zip closur e plastic bag. The extruder was a PVC pipe fitted with a rubber stopper designed to prev ent water from draining from the sample.

PAGE 52

35 Living vegetation was cut off at the soil su rface with a knife. Th icker portions of flocculent layers were kept as part of the 10 cm depth. Litter on the cores was discarded and roots were left intact. If a large wetland had more than one center area, one of the areas and the adjacent edge and upland was sampled. On two occasions when the wetland was surrounded by wet prairie, the wet prairie was sampled as upl and. Samples were immediately stored on ice and transported to the laboratory within 80 hours, and usually within 36 hours. Figure 3-15. Zones of soil sample collection. Comparison to initial sampling scheme. Table 3-3 shows sampled wetlands compared to the initial sampling scheme. Forested / scrub-shrub sites were underrepresented in the IMP and UNIMP land-use types, in part, beca use upon visiting these sites, they were actually riparian or non-wetlands or the la nd use had changed. Unimproved pastures and rangelands were also under-represented because many had been converted to improved pastures or row crops.

PAGE 53

36 Table 3-3. Sampling scheme of historically isolated wetlands by land-use type (2001) and vegetative community. Proporti ons based on wetland polygon count. EM = emergent marsh, FOSS = forested / scrub shrub, EW = emergent marsh / open water. Land use Vegetative Community Polygon proportion Minimum samples Randomly selected** Sampled Dairy (FOSS) 1.0 1 4 6 6 Dairy (EM) 10.1 11 15 15 (6 EW) Improved pasture (FOSS) 5.0 5 7 4 Improved pasture (EM) 68.0 78 96 81 (15 EW) Unimproved pasture / rangeland (FOSS) 1.5 2 4 6 1 Unimproved pasture / rangeland (EM) 10.2 11 15 11 (0 EW) Total 108 145 118 Determined as percent of 108 minimu m samples; minimum per category is 4. ** After adding 25%. Hydrologic connectivity Hydrologic connectivity was characterized as isolated, flow-through, head-of-ditch, tangent, subsurf ace or end-of-ditch (Figure 3-16). Classes were defined based on effects of connectivity on nutrient export. Isolated wetlands had no ditch associated with them. Flow-thr ough wetlands had at least 2 ditches which influenced the center of the wetland. Head -of-ditch wetlands had at least one ditch carrying water away from the wetland. Tange nt wetlands were connected to a ditch which did not directly influence the center of the wetland. Subsurface site edges were within 30 m of a stream or major ditch (US DA/NRCS, 1982). End-of-ditch wetlands had one or more ditches bringing water to the wetland (some wetlands were being drained into others). Ditches were also classified according to intensity: 3 (major), 2 (intermediate) or 1 (minor) (Figure 3-17). Minor ditches were unmaintained, shallow (less than 15cm deep), and narrow (less than 1m). Intense ditches were wide (more than 1m) and deep (more than 40 cm). Intermediate ditches were in between. Less intense (smaller) ditches were usually more vegetated.

PAGE 54

37 A B C D E F Figure 3-16. Detail areas of DOQQ photos showing hydrologi cal connection types. A) isolated (a fence is on the left side ), B) head-of-ditch, C) flow-through, D) subsurface connection, E) tangent, and, F) end-of-ditch. (e)

PAGE 55

38 A B C Figure 3-17. Three intensities of man-made drainage ditches: A) minor B) intermediate C) major.

PAGE 56

39 About half of the sampled wetlands had no surface connection to waterways (they were classified as subsurface or isolated), a nd 32% of them were isolated (Figure 3-18). Improved pasture wetlands had more ditche s than wetlands in other land uses. The majority of wetlands in DAIR and UNIMP areas had one or no ditches (Figure 3-19). In the IMP areas, about half of the wetlands had intensity equal to two or greater (Figure 320). 1 7 2 1 9 10 1 1 10 5 15 17 11 13 8 2 22 21 4 27 1 3 3 2 13 10 3 3 9 23 6 38 1 2 1 2 6 6 2 6 12 2 20 1 3 2 6 2 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%Dairy Beef D ai r y Hay Dairy Pasture D airy S pray fiel d D airy U nimp More I mproved Le s s I mpr oved U n i m pro ved P a sture R angel a nd All Dairy Al l Im proved Al l U nim p/R ange Al l W et l and s Tangent Subsurface Isolated Head of Ditch Flowthrough End of Ditch Figure 3-18. Ditch class by land use. Numbers in the bars represent number of wetlands. 2 5 4 4 19 16 5 3 15 35 8 58 1 2 20 10 2 1 3 30 3 36 2 1 9 6 1 3 15 1 19 4 1 5 5 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%Dairy Beef Dai r y Hay Dairy Pa sture Dai r y Sp r ayfield Da ir y Un im p M o re Improved L ess Im p ro v ed U n improved pasture Ra ngeland A ll Da iry All Im p r ov e d All Un i mp / Range A ll We tla n d s 3 ditches 2 ditches 1 ditch 0 ditches Figure 3-19. Ditch number by land use. Numb ers in the bars represent number of wetlands.

PAGE 57

40 2 5 4 4 19 16 5 3 15 35 8 58 5 5 1 10 1 11 2 2 23 8 1 1 4 31 2 37 1 1 5 4 1 29 1 12 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100%Dai ry Be ef D a iry Ha y D a ir y Pa s tu r e Dai r y Spra y fi e l d D a iry Uni m p More Improv e d Less Improve d U n impr o v e d Pastur e R a n ge l an d All D a ir y All Imp r ove d Al l U n im p /Ran ge All W e tl a nd s Int = 3 Int = 2 Int = 1 Int = 0 Figure 3-20. Ditch intensity (3 being the largest ditches and 0 indicating lack of ditches) by land use. Numbers in the bars represent number of wetlands. Laboratory Analyses Composite soil samples were manually homoge nized. Roots larger than 2 mm in diameter, and live vegetation were removed. A subsample of the remaining composite soil sample was weighed, and pH was measured (usually within 48 hours) using a 1:1, soil to water, ratio (20g of so il to 20 mL of distilled, deioni zed water (DDI)). Moisture content was determined as the difference be tween the wet and dry weights of an ovendried (70C for three days) sample. Bulk density was calculated on a dry-weight basis using the known volume of soil cores. A subsample of each homogenized soil was placed in open tubs in a greenhouse to air-dry for approximately 3 weeks. The samples were then hand-mixed once a week to facili tate drying. After air-drying, the samples were brought back to the laboratory and placed in a 40C oven for 6 hours to stabilize the soil moisture content. Finally, samples were machine-ground and passed through a #100 mesh (0.15 mm openings) sieve. Total P was determined on a 70oC dried sample using the ignition method (Anderson, 1976). Soils were ashed at 550C for 4 hours in a muffle furnace. Loss on

PAGE 58

41 ignition (LOI) was determined as the differe nce between the initial soil weight and ash weight and is an estimate of organic matter content. Ashed samples were then digested with 6 M HCl and filtered using Whatman #41 filter paper. Digestate was kept at room temperature (20oC) until being analyzed calorimetri cally for soluble reactive P (SRP) using the automated ascorbic acid method (Method 365.1; US EPA, 1993). An estimate of total inorganic P (HCL-Pi) was determined after 0.5 g of soil was extracted with 25 mL of 1 M HCl (Reddy et al., 1998). Th is extraction was developed based on an empirical relationship between total inorganic P (determined from a P fractionation scheme) and HCl-Pi of organic soils collected from the Florida Everglades (Reddy et al., 1998). Soil solutions were filte red through 0.45 m membrane filters after centrifuging (6000 rpm 10 minutes). The extracts were stored at 4oC, and analyzed calorimetrically for soluble re active P (SRP) using the auto mated ascorbic acid method (Method 365.1; US EPA, 1993). These soil extracts were also analyzed on an Inductively Coupled Plasma Spectrophotometer (ICP) (US EPA, 1984) to determine total calcium (Ca) and magnesium (Mg) concentrations. To determine oxalate-extractable Al (Alox) and Fe (Feox), which represent amorphous forms, soils were extracted in the dark with 0.175 M ammonium oxalate and 0.1 M oxalic acid at a soil to solution rati o of 1:40 for 4 hours (McKeague and Day, 1966). Extracts were filtered through 0.45 m membrane filters, and analyzed using an ICP (US EPA, 1984). Data Management Polygons and points from the GPS data co llected in the field that delineated sampled wetlands were converted to shapefiles in ArcGIS. If GPS data was incomplete for a site, wetland shapefiles were created based on DOQQ interpretation and NWI

PAGE 59

42 polygons. Using Xtools (DataE ast LLC, Novosibirsk, Russia ), area, perimeter and XY coordinates were determined. This informa tion was imported into a MS Access database (Microsoft, Redmond, WA), and field notes about vegetation, land use, ditches, and edge percent were added for each wetland record. Da ta from the laboratory analyses were also stored in the relational databa se. Data queries of data we re generated for statistical analyses. Statistical Analyses Statistical analyses were performed using SPSS. Descriptive statistics of observed values were listed for physical properties, pH TP, and metals. Means of pH values were calculated from H+ ion concentrations. Total P, HCl-Pi and metals were bulk densityadjusted to calculate g m-2 in the top 10 cm of soil by multiplying the bulk density (g cm-3) by the respective sampleÂ’s mass-based measure (mg kg -1) and multiplying by 10. Parametric comparisons among means were made on normalized data. Most data were normalized by a natural log transformation. Bulk de nsity and mass-based TP (mg kg-1) were transformed by the square-root function. Data for each level to be compared was tested for normality using the Kolmogorov-Smirnov test at a 95% confidence level before a parametric test was run. Differences among hydrologic zones. Descriptive statistics were listed by hydrologic zone and comparisons among center, edge and upland zones were made using one-way analysis of variance (ANOVA). If va riances were equal according to LeveneÂ’s test (p < 0.05), least signif icant difference (LSD) post-hoc pairwise comparisons were used. Otherwise, Games-Howell comp arisons were used (SPSS, 2002). Combined effects. For determining the effects of categorical factors on bulk density, percent organic matter and TP, a ge neralized linear model using univariate

PAGE 60

43 factorial analyses of variance (ANOVA) and Type III sums of squares were computed with normalized dependent variables. Univar iate factorial ANOVAs test the effects of two or more categorical factors on one con tinuous dependent variable, as well as the interaction between the two factors (U nderwood, 1997; SPSS, 2002). The Type III sum of squares method uses unwei ghted, or estimated means, (not biased towards a group with the largest sample size) so can be used with different, and unproportional sample sizes (SPSS, 2002). Factorial ANOVAs were used to determ ine the effects of two fixed factors (vegetative community and land use) on pH, bul k density, percent organic matter, and TP (mg kg-1 and g m-2). Wetland area was a covariate (thi s was natural-log transformed). Mixed model factorial ANOVAs were used to determine the effects of hydrologic connectivity (a random factor), land use (a fi xed factor), and wetland area (a covariate) on TP (g m2) in wetland centers and edges and wetlands as a whole. The ditching variables tested were: a numbe r-of-ditches dummy, (value is 0 if number of ditches = 0, 1 otherwise) and a ditch-intensity dummy (value is 0 if intensity < 2, 1 otherwise). Dummy variables were used in order to have e nough degrees of freedom for the analyses. If the interaction of two factors was found to be significant, conclusions about one factor were made separately at each level of the other factor (Ott and Longnecker, 2001). Error bar graphs were used to illustrate interacting variables (O tt and Longnecker, 2001; Kinnear and Gray, 1999). Center and edge soil s were analyzed separately to maintain independence of samples. Homogeneity of va riances was checked with LeveneÂ’s test (p < 0.05), but if there were at least 6 cases within each group, hete roscedasticity was accepted (Underwood, 1997). The F-statistic and p-value were reported, and Bonferroni

PAGE 61

44 pairwise comparisons of estimated marginal (unweighted) means of factor levels were performed. Both the factorial ANOVAs and th e Bonferroni pairwise comparisons were tested at the 90% confidence level due to the coarse resolution, and high variability of the landscape-scale treatments being tested. Boxplots and error bars (show ing 95% confidence interval of the mean) were used to display observed distributions of data among treatments. The top of the box is at the 75th percentile, and the bottom of the box is at the 25th percen tile. The box represents the inter-quartile range and the horizontal line through the box represen ts the median. The ends of the whiskers represent the largest and smallest values that are not outliers. An outlier is represented by the symbol O, and is defined as a value that is smaller (or larger) than 1.5 times the inter-qua rtile range from the 25th (or 75th) percentile. An extreme value (represented by the symbol *) was defined as a value that is smaller (or larger) than 3 times the inter-quartile range from the 25th (or 75th) percentile. Correlations Spearman correlations (rs) between TP and metals (mg kg-1) and between TP (mg kg-1) and percent organic matter in cente r and edge soils were computed (for all wetlands and by land use). Spearma n correlations are computed from ranks, so they express the proportion of variability accounted for between non-normally distributed parameters (SPSS, 2002). Wetland TP storage. Storage of TP (g m-2) in the top 10 cm of each wetland as a whole was determined by multiplying the TP (g m-2) of each hydrologic zone by the respective percentage of each zone, and su mming the products. A univariate factorial ANOVA and Bonferroni pairwise comparisons of treatments were performed at the 90% confidence level with land use and vegetati on community as fixed factors and wetland

PAGE 62

45 area as a covariate. Another factorial ANOVA was performed to test if land use had an effect on wetland size. Total storage of TP (kg) was calculated by multiplying that number (g m-2) by the total area (m2) of the wetland and dividing by 1000 to convert g to kg. Maps of land use and TP (kg) within each sampled wetland were generated using ArcMap. Results Wetland Vegetative Communities and Surrounding Land Uses Fifty-nine percent of the we tlands in the study area were historically isolated, and 56% of those were emergent marsh, forested, or scrub-shrub wetlands (Table 3-4). Most historically isolated wetlands occurred within IMP areas. Biogeochemical Measures in Hydrologic Zones Physical properties and pH were signi ficantly different among hydrologic zones (Table 3-5). From centers to edges to upl ands, bulk density in creased while organic matter, moisture content and pH decreased. Overall, median TP (mg kg-1) was three times as high, and TP (g m-2) was 1.5 times as high in wetland centers than in edges (Table 3-6). Total P by volume was significantly higher in wetland centers compared to edges in IMP land-use areas, but not in DAIR or UNIMP areas. The amount of volume-based HCl-Pi was similar in all hydrologic zones (Table 3-7). In wetland center soils, medi an percent organic matter was 26.6% in DAI, 18.2% in IMP and 28.4% in UNIMP. For edges it was 15.6% in DAI, 11.4% in IMP, and 12.6% in UNIMP. There were no significant di fferences in metals between center and edge hydrologic zones (Table 3-8). Comparisons of other P extractions a nd metals among hydrologic zones and land uses are in Appendix F.

PAGE 63

46 Combined Effects of Land Use, Vegeta tive Community, and Wetland Size Bulk density, organic matter and pH. There was a significant effect of vegetative community on bulk density and perc ent organic matter in wetland centers and edges (Table 3-9 and Figure 3-21). Land use ha d no significant effect on bulk density or organic matter in either h ydrologic zone. Boxplots by detailed land use are shown in Figure 3-22. Land use and vegetative comm unity each had a significant effect on H+ concentration (Table 3-9). Estimated means and standard error per category were converted to pH (Table 3-10). Wetland size had no significant eff ect on bulk density, organic matter or pH. Center TP. The interaction of vegetati on community and land use had a significant effect on center TP (mg kg-1 and g m-2) in two factorial ANOVAs (Table 311). Wetland size had no significan t effect on mass-based TP (mg kg-1) but volume-based TP (g m-2) was lower in larger wetlands (T able 3-11 and Figure 3-23). The fixed factor interactions ar e illustrated by the dotted lines having opposite-signed slopes in Figure 3-24. When analyzed for the 3 levels of vegetation, land use only had a significant effect for EM wetlands, and when analyzed for each level of land use, vegetation community only had a significant e ffect within IMP wetlands (Table 3-11 and Figure 3-25). Edge TP. The interaction of vegetation comm unity and land use on wetland edge did not have a significant effect on soil TP. Wetland size had a significant effect: larger wetlands had less TP per unit of soil. (T able 3-12 and Figure 3-26). Vegetation community had a significant e ffect on mass-based TP (mg kg-1) but not on volume-based TP (g m-2) (Table 3-12 and Figure 3-27). Land use had a significant effect on both (Figure 3-27).

PAGE 64

47 Correlations between TP and Metals Overall, total P (mg kg-1) was most highly correlated with organic matter percentage in wetland centers and edges (T ables 3-13 and 3-14). In center soils, Alox was more highly correlated with TP, and in wetland edges, Mg and Feox were more highly correlated with TP. Effects of Ditches Total P storage (g m-2). The ditch number dummy did not have significant effect on TP g m-2) in centers, and neither dummy variable had a significant effect on edge soils or on the wetlands as a whole. The ditch intensity dummy variable had a significant effect on wetland center TP (g m-2) in a mixed model factoria l analysis (F = 8.482; p = 0.004) but there was no significant difference between the two estimated means in the Bonferroni comparisons at the 90% confidence level. The anal ysis was repeated for only IMP land-use areas to reduce variability. The ditch intensity dummy variable had a significant effect on center TP (g m-2) (F = 3.737; p = 0.057) and the two class estimated means of the two classes were significantly different (p = 0.057). The estimated mean TP for ditch intensity class 0 (ditch inte nsity of 1 or less) was 22.02 4.83 g m-2 and for class 1 (ditch intensity of 2 or more) was 29.50 7.27 g m-2. Ditch class and TP (mg kg-1). There were not enough samples per class to conclude that TP means were significantly di fferent among ditch clas ses but interestingly, TP (mg kg-1) in center soils varied similarly acco rding to ditch class, regardless of vegetative community or land use. The diffe rences in TP among ditch classes indicated that the TP in head-of-ditch, flowthrough, a nd end-of-ditch wetlands may be higher than in isolated, tangent, and subsurface wetlands (F igure 3-28). These relationships generally

PAGE 65

48 persisted among different land uses and vegetatio n types. This pattern did not exist for edge soils. Wetland Phosphorus Storage The TP (g m-2) stored in the top 10 cm of we tlands as a whole was significantly affected by land use (F = 4.381; p= 0.015) and wetland area (F = 4.598; p = 0.007) but not by vegetative community. Wetland sizes among land uses were not significantly different (p < 0.10). Total P st orage and size statistics for sa mpled wetlands are listed in Table 3-15. Dairy wetlands re presented 18% of the sampled sites (by count) and were storing 29% of the P. Improve d pasture wetlands represente d 72% of the sites and were storing 62% of the P. The unimproved past ure and rangeland wetlands represented 10% of the sites and stored 9% of the TP. A map of TP (g m-2) stored in the top 10 cm of sampled wetlands is shown in Figure 3-29.

PAGE 66

49 Table 3-4. Distributions of National We tland Inventory polygons within the four priority basins by vegetative community and land use (LU). Land-use data is from 2003. Research LU re fers to UNIMP (unimproved pasture/rangeland), IMP (improved pa stures) and DAIR (dairy). Vegetative communities are EM (emergent marsh), and FOSS (forested / scrub-shrub). Polygon count Polygon % Total ha Ha % All NWI 9,257-21,649 -Riparian 1,894 20.5 8,926 41.2 Nonripariana 7,363 79.5 12,723 58.8 EM+FOSS 6,471 87.9 b 12,121 56.0 b In research LU 5,095 78.7 b 9,887 81.6 b UNIMP 643 12.6 c 2,101 21.2 c EM 588 11.5 c 1,911 19.3 c FOSS 55 1.1 c 190 1.9 c IMP 3,895 76.4 c 6,792 68.7 c EM 3,571 70.0 c 6,228 63.0 c FOSS 324 6.4 c 564 5.7 c DAIR 557 10.9 c 994 10.1 c EM 495 9.7 c 878 8.9 c FOSS 62 1.2 c 116 1.2 c a Historically isolated wetlands. b Percentage of category above it. c Percentage of all EM+FOSS in research LU.

PAGE 67

50 Table 3-5. Physico-chemical properties among hydrologic zones based on one-way ANOVA of natural log transformed data and Games-Howell post-hoc comparisons. Mean values are followed by std. dev. and letters, which when the same, indicates the means are not different (p < 0.005). Ditch samples are not included in the comparisons. SE = one standard error of the mean. n Mean Median SE Min. Max. Bulk densitya ------------------------g cm-3 --------------------------Center 117 0.67 0.33 a 0.65 0.030.16 1.42 Edge 117 0.90 0.25 b 0.95 0.020.22 1.40 Upland 116 1.04 0.18 c 1.06 0.020.56 1.44 Ditch 60 0.79 0.27 x 0.80 0.030.10 1.35 Loss on ignition --------------------------(%) --------------------------Center 117 26.3 20.8 a 20.4 1.9 3.0 90.8 Edge 117 16.2 12.7 b 11.8 1.2 1.4 63.8 Upland 116 11.1 7.3 c 9.5 0.7 2.1 43.3 Ditch 60 14.9 17.0 x 9.3 1.5 1.3 93.1 Moisture content ------------------------(%) -------------------------Center 117 51.8 17.1 a 51.1 0.0219.6 82.6 Edge 117 39.5 11.5 b 36.0 0.0119.6 79.4 Upland 116 28.5 7.7 c 28.8 0.017.3 52.4 Ditch 60 36.2 14.2 s 34.0 0.360.1 79.4 pHb Center 117 5.0 4.8 a 5.4 0.074.0 7.5 Edge 117 4.9 4.9 b 5.2 0.054.3 6.7 Upland 116 4.7 4.7 b 4.8 0.063.7 7.9 Ditch 60 5.1 4.9 x 5.4 0.084.1 7.4 a Pairwise comparisons based on square-root transformed data. b Pairwise comparisons based on H+ concentrations.

PAGE 68

51 Table 3-6. Total P comparisons among hydr ologic zones based on one-way ANOVA and Games-Howell post-hoc comparisons (TP by volume was natural log transformed and TP by mass was square ro ot transformed). Mean values are followed by std. dev. and letters which, when the same, indicates the means are not different (p < 0.05). Ditc h samples are not included in the comparisons. SE = one standard error of the mean. n Mean Median SE Min. Max. All wetlands ------------------------------mg kg-1 ------------------------------------Center 117 671.5 622.9 a 560.5 56.9 35.9 4324.8 Edge 116 320.3 380.9 b 187.4 35.4 30.8 2606.2 Upland 117 199.4 212.6 c 125.7 19.7 39.3 1291.0 Ditch 62 304.0 316.8 x 183.9 38.2 6.2 1337.4 All wetlands ---------------------------g m-2 (0 10 cm) ----------------------------Center 116 34.4 28.0 a 24.9 2.6 4.5 188.3 Edge 116 23.0 21.3 b 16.5 23.0 3.9 152.4 Upland 117 19.4 19.1 b 12.5 19.4 4.2 96.3 Ditch 60 18.2 19.6 x 11.6 18.2 0.8 110.2 DAIR ------------------------------mg kg-1 ------------------------------------Center 21 1080.5 1083.6 a 1004.6 236.5 66.2 4324.8 Edge 21 638.3 659.9 ab 315.3 144.0 49.0 2606.2 Upland 21 322.9 346.1 b 189.5 75.5 74.6 1291.0 Ditch 7 478.4 508.3 x 257.7 192.1 51.8 1337.4 DAIRa ---------------------------g m-2 (0 10 cm) ----------------------------Center 21 45.6 40.6 a 36.5 8.9 8.1 188.3 Edge 21 36.7 36.0 a 23.5 7.9 6.6 153.4 Upland 21 27.8 24.5 a 18.1 27.8 8.0 96.3 Ditch 7 32.9 35.8 x 23.2 13.5 6.8 110.1 IMP ------------------------------mg kg-1 ------------------------------------Center 83 576.0 415.6 a 540.0 45.6 35.9 1655.2 Edge 83 254.4 245.2 b 172.6 26.9 37.8 1190.8 Upland 84 181.3 245.2 c 172.6 18.4 48.2 837.5 Ditch 48 253.3 246.8 x 169.1 35.6 18.5 1083.5 IMP ---------------------------g m-2 (0 10 cm) ----------------------------Center 83 33.3 24.6 a 22.3 2.7 4.5 108.8 Edge 83 20.1 15.1 b 15.4 1.7 3.9 85.2 Upland 84 18.4 18.2 b 12.3 2.0 4.6 91.0 Ditch 48 16.9 16.4 x 10.9 2.4 2.0 70.1 UNIMP ------------------------------mg kg-1 ------------------------------------Center 12 502.7 282.9 a 413.0 81.7 159.2 1000.8 Edge 12 219.4 200.0 b 169.2 57.7 30.8 777.2 Upland 12 110.4 54.7 b 99.5 15.8 39.3 212.5 Ditch 5 276.4 329.0 x 171.2 147.1 6.2 793.4 UNIMPa ---------------------------g m-2 (0 10 cm) ----------------------------Center 12 22.9 17.4 a 17.9 5.0 5.7 66.0 Edge 12 19.1 15.6 a 17.6 4.5 4.0 65.2 Upland 12 12.0 5.3 a 11.6 1.5 4.2 20.2 Ditch 5 9.4 9.1 x 6.6 4.1 0.8 21.6 a Pairwise comparisons based on LSD pos-hoc procedure (p < 0.05).

PAGE 69

52 Table 3-7. 1 M HCl extractable P comparisons based on one-way ANOVA and GamesHowell post-hoc comparisons of natural log transformed data. Mean values are followed by std. dev. and letters wh ich, when the same, indicates the means are not different (p < 0.05). D itch samples are no t included in the comparisons. Median percent of TP is based on g m-2. n Mean Median Mean Median Median % of TP ---------mg kg-1 ------------g m-2 (0-10 cm) -----% --Center 118 119.1 317.9 a 53.0 6.83 17.5 a 2.9 13 Edge 116 87.3 253.9 b 31.7 6.14 15.6 a2.9 18 Upland 116 41.0 46.7 b 25.5 3.95 4.17 a2.7 21 Ditch 60 43.2 49.8 x 25.6 2.74 2.87 x 1.8 20 Table 3-8. Metal comparisons among hydr ologic zones based on a one-way ANOVA and LSD post-hoc comparisons of na tural log transformed data. Mean values are followed by std. dev. and letters which, when the same, indicates the means are not different (p < 0.05). Ditch samples are not included in the comparisons. n Mean Median Mean Median -------------mg kg-1 --------------------g m-2 (0-10 cm) -----Oxalate Al Center 117 674.1 657.9 a 465.9 33.3 27.0 a 25.5 Edge 116 513.5 874.8 ab 274.1 38.5 44.4 a 24.9 Upland 116 537.8 926.4 b 236.4 49.7 74.6 a 23.9 Ditch 60 297.6 375.7 x 210.3 18.2 36.0 x 14.9 Oxalate Fe Center 117 911.8 1194.2 a 488.0 45.2 48.5 a 29.4 Edge 116 971.3 1851.0 ab 287.3 61.6 88.5 a 26.3 Upland 116 510.4 674.0 b 248.0 48.8 58.2 a 26.5 Ditch 60 475.4 691.0 x 248.1 31.5 36.0 x 17.6 HCl Ca Center 117 1383.3 1951.9 a 715.3 69.9 87.1 a 38.6 Edge 116 949.8 1403.9 a 524.5 73.2 82.3 a 38.4 Upland 116 858.8 959.2 a 505.2 80.9 76.7 b 54.7 Ditch 60 476.8 551.3 x 278.6 30.7 29.2 x 19.0 HCl Mg Center 117 674.1 657.9 a 465.9 48.2 83.7 a 22.9 Edge 116 513.5 874.8 a 274.1 54.6 88.7 a 21.8 Upland 116 391.6 926.4 b 155.8 36.3 53.4 b 16.0 Ditch 60 297.6 375.7 x 210.3 23.8 33.8 x 10.9

PAGE 70

53 Table 3-9. Effects of land us e (LU) and vegetative community (VC) (fixed factors) and wetland area (covariate) on soil characte ristics of wetland centers and edges based on a factorial ANOVA (p < 0. 10) and Bonferroni pairwise comparisons (p < 0.10). ns = not significant, DAI = dairy, IMP = improved pasture, UNI = unimproved pasture/ra ngeland, EM = emergent marsh, EW = emergent marsh/open water, FS = forested / scrub-shrub. Table 3-10. Estimated unweighted means and standard errors of pH among treatments based on a factorial ANOVA and Bonfe rroni pairwise comparisons (p < 0.01). Letters compare levels of land us e or vegetative community within a hydrologic zone. The same letter indi cate the means are not different. Fixed factors (vegetative community and land use) each had significant effects (p < 0.01) but thei r interaction and the covariate (wetland area) did not (p < 0.10). Center soil Edge soil Treatment and levels n -------------------pH----------------Land use Dairy 21 5.81 0.28 a 5.48 0.22 a Improved 85 5.35 0.24 b 5.07 0.20 b Unimproved pasture / rangeland 12 4.49 0.62 c 4 .65 0.51 b Vegetative community Emergent marsh 86 5.36 0.20 a 5.22 0.15 a Emergent marsh / open water 21 6.18 0.30 b 5.79 0.24 b Forested / scrub-shrub 11 4.67 0.50 c 4.73 0.39 c ----------------C enters-------------------------------Edges---------------Factor F p Level comparisons F p Level comparisons Bulk Density VC 6.626 0.008 EW > EM > FS 5.3030.006 EM, EW > FS LU ns ns VC*LU ns ns area ns ns Organic Matter VC 9.535 0.000 FS, EM > EW 6.6440.002 FS, EM > EW LU ns ns VC*LU ns ns area ns ns pH VC 12.946 0.000 EW > EM > FS 5.3490.006 EW > EM > FS LU 5.309 0.006 DAI > IMP > UNI 5.2690.007 DAI > IMP, UNI VC*LU ns ns area ns ns

PAGE 71

54 EMEWFS V e g etation Communit y 0.25 0.50 0.75 1.00 1.25B u l k D e n s i t y ( g c m 3 ) n=86n=21n=11 0.63 0.88 0.53 EMEWFSVegetation Community 0 25 50 75% O r g a n i c M a t t e r n=85n=21 n=12 27 14 44 A EMEWFSVegetative Community 0.25 0.50 0.75 1.00 1.25B u l k D e n s i t y ( g c m 3 ) n=85n=21n=11 0.94 0.92 0.49 EMEWFSVegetative Community 0 25 50 75% O r g n a i c M a t t e r n=85 n=21n=11 14 17 35 B Figure 3-21. Soil bulk density and percent organic matter comparisons among vegetative communities showing means. A) Wetland centers. B) Wetland edges. Error bars show 95% confidence interval of the mean. Data with the same letter above the boxplots are not different based on factorial ANOVAs (fixed factors were land use and vegetativ e community, covariate was wetland area) and Bonferroni pairwise compar isons (p < 0.10). Only vegetation community had a significant effect (p < 0.10). a b c a b a a a b a a b

PAGE 72

55 DABE DAHA DAPA DASP DAUN IMP1 IMP2 UNI1 UNI2 Land Use ( detail ) 0.00 0.50 1.00 1.50B u l k D e n s i t y ( g c m 3 ) n=4n=6n=4n=1n=6n=51n=34n=7n=5 0.52 0.53 1.020.18 0.67 0.73 0.66 0.49 0.48 DABE DAHA DAPA DASP DAUN IMP1 IMP2 UNI1 UNI2 Land Use (detail) 0 40 80 120% O r g a n i c M a t t e r n=5n=6n=4n=1n=6n=51n=33n=7n=5 46 31 1076 28 24 23 35 39 A DABE DAHA DAPA DASP DAUN IMP1 IMP2 UNI1 UNI2Land Use (detail) 0.00 0.50 1.00 1.50B u l k D e n s i t y ( g c m 3 ) n=4n=6 n=4 n=1n=6n=51n=33n=7n=5 0.63 0.60 0.770.22 0.88 0.92 0.96 0.91 1.08 Error Bars show 95.0% Cl of Mean DABE DAHA DAPA DASP DAUN IMP1 IMP2 UNI1 UNI2Land Use (detail) 0 25 50 75% O r g a n i c M a t t e r n=4n=6n=4n=1n=6n=51n=33n=7n=5 26 21 1764 16 17 12 17 11 B Figure 3-22. Soil bulk density and percent or ganic matter by detailed land use showing mean values. A) Wetland centers. B) Wetland edges. Error bars show 95% confidence interval of the mean. DAB E = dairy beef, DAHA = dairy hay, DAPA = dairy pasture, DASP = da iry sprayfield, DAUN = dairy unimproved field, IMP1 = more improve d pasture, IMP2 = less improved pasture, UNI1 = unimproved pasture, UNI2 = rangeland.

PAGE 73

56 Table 3-11. Effects of land use (LU) and vege tative community (VC) (fixed factors) and wetland area (covariate) on center soil TP (mg kg-1 and g m-2) based on factorial ANOVAs (p < 0.10) and Bonfe rroni pairwise comparisons (p < 0.10). ns = not significant, DAI = da iry, IMP = improved pasture, UNI = unimproved pasture/rangeland, EM = emergent marsh, EW = emergent marsh/open water, FS = forested / scrub-shrub. ------------------mg kg-1----------------------------g m-2 (0-10 cm)----------Factor F p Comparisons F p Comparisons VC*LU 2.873 0.040 2.231 0.089 VC 4.002a 0.022 FS > EM, EW 3.813a 0.024 FS > EM LU 4.173b 0.019 DAI > IMP, UNI 3.333b0.041 DAI > UNI Area ns 6.677 0.011 smaller > larger a Effect is only significant within IMP land-use areas. b Effect is only significant within EM wetland vegetative communities. Wetland size (ha)14 12 10 8 6 4 2 0TP (g m-2) in top 10cm of center soils200 180 160 140 120 100 80 60 40 20 0 Figure 3-23. Scatterplot of cen ter soil total phos phorus (g m-2) versus wetland size (ha).

PAGE 74

57 EMEWFSError Bars show Mean +/1.0 SE Dot/Lines show Medians dairy improved unimprovland use 0 500 1000 1500 2000T P i n c e n t e r s o i l s ( m g / k g ) i n t o p 1 0 c m n= 9 n= 65 n= 11 dairy improved unimprovland use n= 6 n= 15 dairy improved unimprovland use n= 6 n= 4n= 1 A EMEWFSError Bars show Mean +/1.0 SE Dot/Lines show Medians dairy improved unimprovland use 0.00 25.00 50.00 75.00T P i n c e n t e r s o i l s g / m 2 i n t o p 1 0 c m n= 9 n= 65 n= 11 dairy improved unimprovland use n= 6 n= 15 dairy improved unimprovland use n= 6 n= 4n= 1 B Figure 3-24. Plots of total phosphorus (TP) in center soil s showing interaction of vegetation community and land use. A) mg kg-1 B) g m-2. EW = emergent marsh / open water, FS = forested / sc rub-shrub, EM = emergent marsh. While EW and FS wetlands have lower TP within dairies compared to improved pastures, EM wetlands have higher TP within dairies.

PAGE 75

58 DairyImprovedUnimprovLand Use 0 1000 2000 3000 4000T P ( m g k g 1 ) i n t o p 1 0 c m n=9n=64n=11 1542 532 511 DairyImprovedUnimprov Land Use 0 50 100 150T P ( g m 2 ) i n t o p 1 0 c m n=9 n=64 n=11 58 29 24A EMEWFS Vegetative Community 0 500 1000 1500 2000 T P ( m g k g 1 ) i n t o p 1 0 c m n=64n=15n=4 532 539 1424 EMEWFSVegetative Community 0 50 100 150T P ( g m 2 ) i n t o p 1 0 c m n=64n=15n=4 29 43 60 Error Bars show 95.0% Cl of Mean B Figure 3-25. Total phosphorus comparisons of center soils showing means. A) Emergent marsh soils by land use. B) Impr oved pasture soils by vegetative community. Error bars show 95% conf idence interval of the mean. Data with the same letter above the boxplots are not different, based on factorial ANOVAs (fixed factors were vegetative community and land use, covariate was wetland area) and Bonferroni pa irwise comparisons (p < 0.10). The interaction of vegetation community and land use had a significant effect (p < 0.05). EM = emergent marsh, EW = emergent marsh / open water, FS = forested / scrub-shrub. a b b a a b a ab b a ab b

PAGE 76

59 Table 3-12. Effects of land use (LU) and vege tative community (VC) (fixed factors) and wetland area (covariate) on edge soil TP (mg kg-1 and g m-2) based on factorial ANOVAs (p < 0.10) and Bonfe rroni pairwise comparisons (p < 0.10). ns=not significant, DAI = da iry, IMP = improved pasture, UNI = unimproved pasture/rangeland, EM = emergent marsh, EW = emergent marsh/open water, FS = forested/scrub-shrub. --------------mg kg-1 -----------------------g m-2 (0-10 cm) ---------Factor F p Comparisons F P Comparisons VC*LU ns ns VC 2.788 0.066 FS > EM ns LU 2.961 0.056 DAI > IMP 2.845 0.063DAI > IMP area 6.995 0.009 smaller > larger 8.430 0.004smaller > larger Wetland size (ha)14 12 10 8 6 4 2 0TP (g m-2) in top 10cm in edge soils160 140 120 100 80 60 40 20 0 Figure 3-26. Scatterplot of edge soil TP (g m-2) versus wetland size (ha).

PAGE 77

60 DairyImprovedUnimprovLand Use 0 1000 2000 3000T P ( m g k g 1 ) i n t o p 1 0 c m n=21 n=83 n=12 638 254 219 DairyImprovedUnimprovLand Use 0.0 50.0 100.0 150.0T P ( g m 2 ) i n t o p 1 0 c m n=21 n=83 n=12 36.7 20.2 19.1A EMEWFS V e g etative Communit y 0 500 1000 1500 2000T P ( m g k g 1 ) i n t o p 1 0 c m n=84n=21n=11 270 346 656 EMEWFSVegetative Community 0 25 50 75 100T P ( g m 2 ) i n t o p 1 0 c m n=84n=21n=11 22 26 29B Figure 3-27. Total phosphorus comparisons of edge soils showing means. A) Land use. B) Vegetative community. Error bars s how 95% confidence interval of the mean. Data with the same letter a bove the boxplots are not different, based on a factorial ANOVA (fixed factors we re vegetative community and land use, covariate was wetland area) and B onferroni pairwise comparisons (p < 0.10). The interaction of vegeta tion community and land use had no significant effect. EM = emergent marsh, EW = emergent marsh / open water, FS = forested / scrub-shrub. a b ab a ab b a a a a b ab

PAGE 78

61 Table 3-13. Spearman correlations and significa nce (2-tailed) for wetland center soil TP and biogeochemical parameters (mg kg-1 except organic matter) by land use. IMP = improved pastures, DAIR = dairie s, UNIMP = unimproved pastures / rangeland. All IMP DAIR UNIMP n = 117 83 22 12 % organic matter 0.632 0.574 0.868 0.455 0.000 0.000 0.000 0.138 HCl-extr Ca 0.374 0.334 0.687 -0.217 0.000 0.002 0.000 0.499 HCl-extr Mg 0.476 0.479 0.415 0.273 0.000 0.000 0.055 0.391 Oxalate-extr Al 0.570 0.576 0.751 0.007 0.000 0.000 0.000 0.983 Oxalate-extr Fe 0.472 0.494 0.513 0.322 0.000 0.000 0.017 0.308 Table 3-14. Spearman correlations and significa nce (2-tailed) for wetland edge soil TP and biogeochemical parameters (mg kg-1 except organic matter) by land use. IMP = improved pastures, DAIR = dairie s, UNIMP = unimproved pastures / rangeland. All IMP DAIR UNIMP n = 117 83 22 12 % organic matter 0.748 0.717 0.786 0.727 0.000 0.000 0.000 0.007 HCl-extr Ca 0.179 0.118 0.223 0.042 0.055 0.291 0.330 0.897 HCl-extr Mg 0.657 0.577 0.617 0.783 0.000 0.000 0.003 0.003 Oxalate-extr Al 0.385 0.363 0.379 0.224 0.000 0.001 0.090 0.484 Oxalate-extr Fe 0.598 0.528 0.640 0.783 0.000 0.000 0.002 0.003

PAGE 79

62 0.00 500.00 1000.00 1500.00 2000.00T P (m g / k g ) i n c e n t e r s o i l s em marshem marsh / waterforest / scrub A Error Bars show Mean +/1.0 SE end of ditch flowthrough head of ditc isolated subsurface tangentditch class 500.00 1000.00 1500.00 2000.00T P ( m g / k g ) i n c e n t e r s o i l s dairyimprovedunimprov end of ditch flowthrough head of ditc isolated subsurface tangentditch class end of ditch flowthrough head of ditc isolated subsurface tangentditch class Circles show median Figure 3-28. Total P (mg kg-1) in center soils by ditch class. A) Varying similarly among vegetative communities. B) Varying similarly among and land uses. Table 3-15. Total P (TP) stored in surf ace soils of sampled wetlands (154 ha) among land uses. Mean values are followed by std. dev. and letters, which when the same, indicates the means are not different (p < 0.10). Comparisons are based on a factorial ANOVA (fixed factors were vegetative community and land use) and Bonferroni pairwise co mparisons of unweighted means. DAIR = dairies, IMP = improved pastur es, UNIMP = unimproved pastures / rangeland. Land Use (n) Mean TP Med. TP Mean size Med. size Mean TP Med. TP Total TP ---g m-2 (0-10 cm) ------------ha ----------------------------kg ----------------DAIR (21) 45.6 40.0 a 35.3 1.56 1.11 a 1.22628.5 522.1 437.3 11,701 IMP (84) 30.1 19.7 a 22.7 1.15 1.13 a 0.90325.6 424.0 212.9 24,946 UNIMP (12) 19.7 9.4 b 20.4 2.03 3.67 a 0.54188.1 148.1 125.6 3,566 40,214 B

PAGE 80

63 01020 5 Kilometers-Lake Okeechobee priority basins unimproved pastures / rangeland improved pasture dairiesTP (g/m2) in top 10 cm 5 14 14 22 22 36 36 58 58 185 Figure 3-29. Land uses (2003) and TP (g m-2) stored in surface soils (0-10 cm) of sampled wetlands within the four prio rity basins of the Lake Okeechobee watershed. Discussion This study aimed to assess general trends of TP storage within historically isolated wetlands of four sub-basins of the Lake Okeechobee watershed across a range of land uses (watershed-scale), for different ditching magnitudes (f ield-scale) and among hydrologic zones within wetlands (f ield-scale). The TP (g m-2) in surface soils (0-10 cm) reported in this study are comparable to thos e reported by studies of landscape wetlands

PAGE 81

64 and uplands in the study area (Sperry, 2004; Reddy et al., 1996a ; Nair and Graetz, 2002) and wetlands in agricultural areas of No rth Carolina (Bruland et al., 2003). Redox potential and pH are known to contro l P movement in wetlands (Richardson, 1999). Low redox potential slows microbial decomposition leading to organic matter accretion (Lvesque and Mathur, 1979) which may be the most important mechanism for P storage in more organic wetland soils (Axt and Walbridge, 1999; Richardson, 1999). In flooded soils containing Fe, low redox potentia l can cause ferric iron to be reduced to the soluble ferrous form releasing P that was bound to it (Faulkner and Richardson, 1989). Lower redox potentials that may exis t in the wetland centers can cause the transformation of crystalline forms of Fe and Al to more amorphous forms (oxalate extractable) which have more sites for P sorp tion and are more significant sorption agents for P than ferric iron (Reddy and Sm ith, 1987; Rhue and Harris, 1999). Zonal TP gradients Data showed that significant zonal TP gradients exist from wetland centers (median: 560.5 mg kg-1), to wetland edges (median: 187.4 mg kg-1) and surrounding uplands (median: 125.7 mg kg-1). The gradient of volume-based TP was only significant between the wetland center (median: 24.9 g m-2) and edge (16.5 g m-2) so a portion of the first hypothesis was accepted. These differences were significant in IMP areas but not in DAIR or UNIMP (Figure 3-30). Other studies have reported increased TP storage in more saturated soils within wetlands (Scint o, 1990; Sperry, 2004; Reddy et al., 1996a). The difference in IMP wetlands may be explaine d by ditches which are larger in size and number in this land use compared to the other land uses. Wetland centers act as hydrologic sinks accumulating water and other ma terial from surrounding areas. Ditches may transport dissolved and particulate P to the centers of thes e wetlands bypassing the

PAGE 82

65 buffering effects of the edges. The lack of significant differences in edges and centers within the other two land uses may be also a result of the hi gher TP variability combined with a smaller number of samples in those land uses. Figure 3-30. Cartoon comparing relative magn itudes (means and std. dev.) of wetland edge and center soil TP (g m-2) among land uses (only TP in IMP hydrologic zones are significantly different). Li nes connected to centers represent relative ditching. The relationships between P and other bi ogeochemical properties help explain the zonal TP gradient in the wetlands. Correla tions between TP and organic matter were strong in wetland centers (rs: 0.632) and in edges (rs: 0.748). Richardson (1999) pointed out that mineral sorption mechanisms are more important in mineral soils. In this study, the median organic matter in edges (12%) wa s significantly lower than that in centers (20%). These values are somewhat higher than the median organic matter found in 14 EM wetland soils (0-5 cm) of similar size in Pennsylvania which was 12% (Campbell et al., 2002) but similar to that re ported in a study of similar wetlands in the same study area (Sperry, 2004). Bruland and Ri chardson (2004) reported 9% or ganic matter in wet area soils (0-15 cm) of 11-year-old constructed wetl ands and about 6% in the edge areas. Unimproved Pastures / Rangeland Dairy Land Improved Pastures 36.7 36.0 g m-220.1 15.1 g m-2 19.1 15.6 g m -2 45.6 40.6 g m-233.3 24.6 g m-2 22.9 17.4 g m-2 n = 12 n = 21 n = 84

PAGE 83

66 More than half of the P in this study is estimated to be organic P. A 1 M HCl extraction was used to estimate inorganic P to be less than 20%, which would mean that more than 80% is organic P. This extrac tion is a conservative estimate of inorganic P (K.R. Reddy, personal comm unication, November 22, 2004) and was developed for organic wetland soils in calcareous systems that do not receive manure (Reddy et al., 1998). It is helpful to look at a study of IMP and UNIMP isolated wetland soils in the study area in which organic P of surface soil s (0-15 cm) was estimated to be 62% based on the difference of inorganic P (determined by a fractionation scheme) and total P (Sperry, 2004). Since the majority of the P stored in the wetlands is organic P, accretion mechanisms may be more important than so rption mechanisms in these historically isolated wetlands. The significantly higher TP in wetland centers corresponds with the significantly higher organic matte r content in wetland centers but the higher P in centers cannot be attributed to more metals since their quantities are not significantly different between hydrologic zones. Wh ile those metals may be playing a role in retaining P, sorption mechanisms with the organic matter-metal complexes may also be an important mechanism for P retention. Petrovic and Kastelan-Macan (1996) suggested that phosphate (H2PO4-) can be complexed with Ca, Mg, Fe and Al cations that are bound to negatively charged humic substances. This is sometimes referred to as bridging (Richardson, 1999). Other studies reported that Al complexation with organic matter played an important role in P sorption in surface wetlands soils (Axt and Walbridge, 1999; Haynes and Swift, 1989). More evidence for the important role of organic matter is shown by the higher correlation between TP a nd organic matter compared to metals.

PAGE 84

67 Scinto (1990) also found the highest correlation with TP to be organic matter in forested dairy wetlands within the study area. Though the correlati ons between TP and metals were not as high as between TP and organic matter, the trends are discussed below. Some studies have shown that amorphous Al and Fe hydroxides govern P sorption in acidic wetland soils and stream sedime nts (Patrick and Khalid, 1974; Reddy et al., 1995; Richardson, 1999; Axt and Walbridge, 199 9) whereas reactions with Ca and Mg are important binding mechanisms in more al kaline systems (Richardson, 1999). Total P in wetland edges was more correlated with Feox (rs: 0.598) than with Alox (rs: 0.385) and conversely the TP in wetland cente rs was more correlated with Alox (rs: 0.570) than with Feox (rs: 0.472) with the exception of UNIMP. In wetland edges, total Al may represent a significant potential pool of available P ad sorbing sites if the wetland were to be expanded in size by hydrological restoration. This would expose crystalline Al to flooding cycles and convert it to amorphous Al which has more sites for P adsorption than the crystalline form (Rhue and Harris, 1999). In the edges where soils have longer pe riods of oxidizing conditions than in centers, there will be more Fe in the oxidized form which binds P (Patrick and Khalid, 1974). In wetland centers, more amorphous Al with high P-binding capacity is expected than in edges. The results also show ed a correlation between TP and Mg (rs: 0.657) in wetland edges and a weaker correlation (rs: 0.476) in wetland centers. Because of the acidic nature of the soils in which Mg phosphate would not be stable, it is possible that the Mg is providing a bridge between phospha te and humic substances. Correlations in edge soils might be due to the influences of adjacent land uses whereas correlations in center soils might be more influenced by th e import and export of materials by ditches

PAGE 85

68 (confounded by the transport of material fr om the upland through the edge into the center). Higher pH along with higher ma nure input to DAIR wetlands may explain why there was a relatively strong co rrelation between TP and Ca (rs: 0.687) in the centers of those wetlands. Interestingly, results of this study indi cated that organic P was lower in IMP wetlands (18.2%) compared to DAIR (26.6%) or UNIMP (28.4%). Cattle may graze more wetland vegetation in IMP areas compared to other land uses since fewer cows are stocked in UNIMP areas (Steinman et al ., 2003) and cows in DAIR areas receive commercial feed (SFWMD et al., 2004a) and are excluded from many of the wetlands. Herbivory of aquatic vegetation may reduce th e amount of organic matter being accreted and the reduced vegetation combined with phys ical impact to soils by cattle may cause some sedimentation in the wetland. Studies have shown that vegetation around wetlands and vegetated ditches reduces sedimentation in wetlands suggesting that less particulate material (e.g., particulate P) is imported into the wetland (Fiener and Auerswald, 2003; Hook, 2003). A study by Clary (1999) indicated that the degree of sedimentation is related to the amount of grazing pr essure on a wetland (Clary, 1999). Total P (g m-2) stored in upland and wetland surfac e soils of unimproved pastures / rangelands was similar to that reported by other studies in the same land uses in the Lake Okeechobee watershed (Sperry, 2004; Re ddy et al., 1996a; Graet z and Nair, 1995; Scinto, 1990). Those studies reported higher va lues of TP in dairy soils than shown in this study (mean: 45.6 g m-2). While those studies focused sampling on high intensity dairy areas, this study showed a greater vari ability in sampled dairy areas that ranged

PAGE 86

69 from ungrazed fields to pastures where a la rge number of cows received supplemental feed. Hydrologic Connectivity Studies have demonstrated that wetlands with surface connections to downstream waterways can export nutrients (Nessel and Bayley, 1984; Yin and Lan, 1994; Raisin and Mitchell, 1995; Richardson, 2003; Mwanuzi et al., 2003; Sperry, 2004). This study hypothesized that wetlands with more ditchi ng would have less TP, but the inverse was true. Center soils of improved pasture wetlands with deeper and wider associated ditches had significantly more TP than wetlands with less intensive ditching while controlling for land use and wetland area. The number of ditches did not have a signific ant effect on TP. In this study, fertilizer and cattle density information could not be collected. More intense ditches typically have less vegetati on and faster water flow than less intense ditches. Vegetation in ditches filters particulate P and serves to slow down flow velocities (Fiener and Auerswal d, 2003). This enhances hydr aulic retention time in the ditches, and allows sediments to fall out of flowing waters to reduce the amount of particulate P being transported. More intens e ditches may bypass the buffering effects of edges bringing more P to wetland centers. It is also possible that ditch intensity is a proxy for fertilization and cattle stocking intens ity. A pasture with more intense ditching represents a significant economic investment which may be accompanied by additional fertilizer and more intense stocking of cattle. Interestingly, TP (mg kg-1) varied similarly by ditch cla ss regardless of land use or vegetative community. Flowthrough, end-of-dit ch and head-of-ditch wetlands generally had more TP than subsurface, tangent a nd isolated wetlands. Water flow may be transporting more particulat e and dissolved P into wetla nd centers of flow-through and

PAGE 87

70 end-of-ditch wetlands. Head-o f-ditch and isolated wetlands may have more vegetation to generate more organic matter that accretes a nd stores more TP. Isolated wetlands may be in areas with less intensive gr azing. Tangent wetlands have ditches that do not intersect the centers hence P being transported by d itches may bypass the wetland center. Based on these findings it can be speculated that wetla nds with less ditching are in pastures with less cattle activity. In improve d pastures, 60% of the wetlan ds had at least 1 ditch and almost 50% had a ditch intensity of 2 or more. In dairy and unimproved / rangeland pastures, less than 30% of the wetlands ha d ditches and less than 30% had a ditch intensity of 2 or more. Land Use Differences It was hypothesized in this study that wetlands surrounded by more intensely fertilized and cattle-stocked areas would have higher surface soil TP (dairy > improved pastures > unimproved pastures / rangeland) This ideas is supported by a study that concluded that up to 80% of TP had the potential to leave heavily manured upland regions, while less than 10% of the TP was likely to leave a low manure-impacted pasture soil because of the mobility of P in manur e (Graetz and Nair, 1999). Also, it is widely accepted that the land use and land cover of a landscape have a major effect on water quality (Herlihy et al., 1998; Cuffney et al ., 2000; Berka et al., 2001). Sperry (2004) found that ditched wetlands in improved pasture released 5 to 7 times more TP in surface water runoff than wetlands in semi-native pastures. This stud y supports the hypothesis that wetland soils in UNIMP areas have significantly less TP (median: 20.4 g m-2) than DAIR (median: 35.3 g m-2) and IMP (median: 22.7 g m-2) land uses (while accounting for differences due to vegetative community and wetland area) but there was not enough evidence to show that DAIR and IMP were diffe rent. The lack of a significant difference

PAGE 88

71 between wetland soils in IMP and DAIR land us es study may also be a result of the high variability and lower number of samples in the DAIR treatment. The dairy land use class grouped pastures with feed structures with non-ir rigated fields near a nd beef pastures that were similar to the improved pastures. The wi de range of dates of sample collection was probably a factor in the overall data variability as plant upta ke of nutrients in wetlands can be higher at the beginning of summer (whe n sampling began) than in the fall (when sampling ended) (Tanner, 1999). Sperry (2004) found no differences in wetland soil TP (g m-2) between improved and semi-native (UNIMP) pastures. This may have been due to variability within a small sample size (12) and the fact that the IMP pastures in that study were not fertilized after 1987. Other Findings Wetland size was a significant covariate for TP in centers and edges. Larger wetlands stored less TP (g m-2) than smaller wetlands. This suggests that smaller wetlands could become saturated with P and perhaps be a source for P in the future (compare Reddy et al., 1995). Larger wetlands with larger pools of P binding sites and longer hydraulic retention times that favor biogeochemical processes that bind P may represent more long-term sinks for P. In a study of sediment nutrient levels in 73 wetlands it was found that wetland size was negatively correlated to sodium bicar bonate extractable P (H oulahan and Findlay, 2004). A possible explanation o ffered by Detenbeck et al. (1996) is that water and soils in the center of large wetlands are better buffered against nutrient inputs than smaller wetlands by virtue of the greater distance betw een the wetland center and input sources. Land use would not explain this trend since there was no significant difference in wetland size among land uses.

PAGE 89

72 Forested / scrub-shrub systems had signifi cantly more TP in wetland centers (while accounting for land use and wetland size) than EM systems. Forested systems typically have less organic matter than emergent marsh sy stems due to the recalc itrant nature of the woody organic matter (Mitsch and Gosselink, 2000) so this result was surprising. Eight out of the 12 FS wetlands were ditched and al l had intense ditching (intensity of 2 or more). Seven of them were associated with DAIR areas and one with UNIMP. The forested systems had noticeable soil oxidation as seen from exposed tree roots and were usually hummocky systems that allowed for aerobic soil processes to occur. This may have had the effect of concentrating the P th at was in surface soils since P has no gaseous form as do carbon and nitrogen. It is possible that a combination of having more intense ditching, receiving mostly DAIR runoff and e xperiencing soil oxidation caused the higher P concentrations. Storage Dairies were found to be storing sign ificantly more TP (median: 35.3 g m-2) than UNIMP (median: 20.4 g m-2). Though DAIR wetlands comprised 18% of the sites, they accounted for 29% of the 40,214 kg of P stored on all sampled wetland surface soils. The sampled sites covered 154 ha and stored 261 kg ha-1 in the top 10 cm of soil. Reddy et al. (1995) reported 750 kg ha-1 in streamside and dairy wetland soils (0-30 cm). Considering that most of the P storage was found in the su rface soils, the historic ally isolated wetlands may be storing relatively low amounts of TP and may represent a potential sink for P storage in the future. Future Research The cumulative impact of high nutrient loads in the past can leave a legacy that may override the effects of current land-use practices related to stocking density and

PAGE 90

73 fertilization (Steinman et al., 2003) In future research, it woul d be beneficial to create a land cover map of the area using a sa tellite image classification based on a priori collected ground truth points of land uses where the fertilizati on and cattle stocking history are known. These land cover cla sses may be more related to phosphorus condition in uplands and wetlands and w ould have higher spatial and temporal resolutions than the coarse-scale land-use laye r used in this thesis. The land cover map could be helpful in creating a sampling sche me according to the land cover classes such that differences in those cl asses could be tested. It is also recommended to test the eff ects of ditching on TP in wetlands. While comparing wetlands of similar size and ve getation surrounded by similar land uses a hypothesis to test would be that isolated wetlands store mo re TP than wetlands being drained. Another possible hypothesis is that di tches that do not intersect wetland centers (tangent ditches) conduct less TP from the wetland while still serv ing the function of conducting water from it. A final recommenda tion is to characterize P fractions and undertake P sorption studies using methods pr oven effective in non-calcareous, manureimpacted wetland systems so that P dynamics can be further investigated in these historically isolated wetland ecosystems. Conclusions Interacting factors affect the variability of P storage in historically isolated wetlands and there is evidence to support the idea of hydrologi cal restoration. This study confirms previous findings that TP in DAI R wetlands store more TP than UNIMP, but the comparison between DAIR and IMP wetlands was inconclusive. The high variability in DAIR TP is probably due to a wide range of land uses within DAIR and may be a reason why there was no significant differe nce between DAIR and IMP wetlands and

PAGE 91

74 also why no differences in TP (g m-2) were found among DAIR hydrologic zones. The development of a land cover map based on sa tellite imagery would assist in land use classification for future studies. The hypothesis that a TP gradient exists from wetland centers increasing outwardly to upland soils, was accepted for IMP centers that had more TP than edges and uplands. More intense ditching in IMP areas may be bringing more di ssolved and particulate P to the centers of this land-use type. Wetlands with larger ditches were shown to have significantly more TP in center soils than those with smaller ditches. It is hypothesized that mechanisms relate d to organic matter are responsible for the majority of TP storage in these wetland sy stems including accretion and phosphate metal bridging with humic substances. Through hydr ological restoration, overall wetland size would increase, providing additional storage cap acity for P. Also, the amount of P per square meter stored might also increase, due to increased hydrol ogic retention time. Land managers should consider hydrological restoration of dr ained wetlands to increase wetland size and hydraulic retention times. This would increase the number of P binding sites on soil particles as well as increase the amount of center area soils which have higher rates of P storage. Mo re studies are needed to determine the P sorption capacity of these soils. Increasing the size of smaller we tlands in more intensely managed pastures may prevent them from becoming saturated w ith P and becoming sources of P to surface waters. Over 12,000 ha of isolated wetlands (drain ed and undrained) exist in the four priority basins. About half are drained and represent an opportunity for increasing storage of P in the Lake Okeechobee wate rshed through hydrologica l restoration. Land

PAGE 92

75 managers should develop best management practices (BMPs) in IMP areas where the largest proportion of ditched, isolated wetlands exist. They should consider restoring hydrology of drained isolated wetlands, keep ing ditches vegetated, and excluding cattle from wetlands. Incentive programs offered by government agencies can help offset the cost of BMP implementation while reduci ng the ecological and economic costs of elevated P levels in Lake Okeechobee.

PAGE 93

76 CHAPTER 4 UPSCALING TOTAL PHOSPHORUS TO UNSAMPLED WETLANDS Introduction Wetland ecosystems are an integration of physical and biological processes, occurring at watershedand field-scales. In the previous chapter, it was shown that factors at multiple scales affect TP storage in historically isolated wetlands within the four priority basins of the Lake Okeechobee watershed. Th ese factors include land use, ditch size, and wetland hydrologic zones. In this chapter, datase ts available at the watershed -scale (i.e., the four pr iority basins), were collected and analyzed to predict TP storage in unsampled wetlands. Only in recent years, have the tec hnologies of satellite remote sensing and GIS, as well as natural re source datasets, become readily available to study these systems at multiple spatial, spectral and radiometric resolutions (Walsh et al., 1998). The aim of this study was not to genera te a mechanistic model that explains TP storage factors, but to use patterns of ava ilable spatial datasets that characterize the majority of wetland TP storage variability. These patterns of multi-scale variables may represent a holistic view of TP conditions of wetlands and surrounding areas. Environmental Variables Ecological response variables can be pr edicted by environmental variables (McKenzie and Austin, 1993; McKenzie and Ry an, 1999; Lapen et al., 2001). Land use, for example, has been shown to account for a high amount of variability in stream water quality (Hunsaker, et al., 1992; Roth et al ., 1996; Herlihy et al., 1998; Behrendt et al., 1999). In Chapter 3, it was shown that we tland soils in DAIR areas stored higher

PAGE 94

77 amounts of TP than UNIMP areas. Land use may also help to predict TP stored in unsampled wetlands. Soil type is an environmen tal variable that has been used to provide input data to water qua lity models (Kuenstler et al., 1995; Wilson et al., 1996; Shaffer et al., 1996). Soils adjacent to wetlands may be predictive of P in wetlands, because of differing drainage and bioge ochemical characteristics. Spatial patterns of natural and anthr opogenic features in a landscape strongly influence the ecological characteristics (R isser et al., 1984; Walsh et al., 1998). Landscape metrics are based on landscape geom etry and spatial arrangement of patches and features (Herzog and Lausch, 2001). For example, landscape characteristics that describe the arrangement of human-altered land in a watershed can be correlated with water biogeochemistry (Gergel et al., 2002; Yin et al., 2003). In a study of sediment nutrient levels in 73 wetlands, it was found that wetland size was negatively correlated to sodium-bicarbonate-extractable P (Houlahan and Findlay, 2004). In the same study, the proportion of land within 1,250 m that was wetla nd, was also negatively correlated with sodium-bicarbonate-extractable P, and nitrate levels were positively correlated with road density within 500 m. Landscape metrics expl ained 65 to 86% of the total variation in nitrogen yields to streams, and 73 to 79% of the variability in dissolved P (Jones et al., 2001). Distances to features, such as roadwa ys, dairies, waterways, and other wetlands could be used in this study to partially qua ntify spatial landscape patterns of wetlands, and may be predictive of their soil P condition. Spectral Data Remote sensing is the acquisition of data about an object or scene on earth by a sensor that is far from the object (Colwell, 1983). Sensors may be hand-held or mounted in vehicles, airplanes, or satellites. Remote sensing has emerged as a useful data source

PAGE 95

78 for characterizing LULC (Burnett and Bl aschke, 2003), and can provide valuable quantitative data that is indicative of moisture, soils and vegetation (Curran, 1985; Jensen, 1996; Lillesand and Kiefer 2000) that exist at a particul ar point in time. Satellite remote sensing has shown its utility in wetland mapping (Ackleson and Klemas, 1987; Mertes et al., 1995; Sader et al., 1995; Harvey and Hill, 2001; Schmidt and Skidmore, 2003) as well as land cover mapping (Jensen, 1996; San Miguel-Aya nz and Biging, 1997; Lillesand and Kiefer, 2000; Pearlstine et al., 2002; Reese et al., 2002). There are many studies quantifying biophysical measures such as biomass and soil moisture (with remotely sensed spectr al data), but few studies have linked biogeochemistry with satel lite spectral information (N umata et al., 2003). Green vegetation, calculated from a Landsat7 ETM+ satellite image, was correlated with P content in pastures in Rondnia, Brazil (Num ata et al., 2003). Asner et al. (1999) found that spectral estimates of l eaf area and nonphotosynthetic ve getation (NPV) of pastures were correlated with soil P concentrations in the central Amazon region. These studies used vegetation indices, which are quantitati ve measures of vegetative condition. These indices are usually calculated from combinati ons of several spectral bands, whose values are added, divided, or multiplied. The pure spectral reflectance values for wetland features and surrounding areas are a result of the interaction of rock, soil, vegetation, fauna, landform, and water (Schmidt and Skidmore, 2003). They may be indicators for the properties of an earth feature (like a wetland) as a whole (Schmidt and Skidmo re, 2003). Polygon-based aggregation of remotely sensed data has been shown to increase accuracy of land-cover biophysical variable estimation at regiona l scales (Wicks et al., 2002). Spectral properties may be

PAGE 96

79 indicative of wetland features (polygons): grazing intens ity around it, amount of open water, amount of exposed soil, amount of or ganic matter and vegetation density. These features may correspond with TP magnitude stor ed in wetlands soils. For example, more green vegetation within a wetland could indicate higher TP within the wetland. Variability of vegetation around a wetland may be indicative of hom ogeneous pasture or heterogeneous native land, which might be re lated to TP. Spectral reflectance data and vegetation indices may be indicative of recent P conditions within, and around, historically isolated wetl ands of the study area. Ecological Response Variables and Classification Trees Austin et al. (1995) noted th at the use of suitable envir onmental variables was more important than the choice of prediction met hod, but classification a nd regression trees are becoming increasingly used for this purpose (McKenzie and Ryan, 1999). Classification and regression trees are ideally suited for an alyzing complex ecological datasets, because they provide robust analytical methods that can deal with nonlin earity, missing values, interacting variables, and heterogeneous va riances (DeÂ’ath and Fa bricius, 2000). The method is a form of binary, recursive, partitioning, that is well suited for determining relationships within complex, nonparametric, in teracting datasets that have a high number of predictor variable s (Breiman et al., 1984; Lewis, 2000; Steinberg and Colla, 1995). Classification trees predict a pre-designated class for each case, whereas regression trees predict a range of values for each case (Breiman et al., 1984). McKenzie and Ryan (1999) used readily observable environmental features as a basis for mapping soil properties. With di gital terrain analysis and airborne gamma radiometric data, they were able to use re gression trees to account for 78% of the variability of soil P in upland forest soils of southeastern Australia based on 165 sample

PAGE 97

80 sites. In another study, the spatial distribution of alien plant stands (presence or absence) was predicted with 73% accuracy using classification trees (b ased on the predictors of temperature, number of growing days, al titude, geology, slope, roughness, precipitation, distance to nearest road a nd distance to coastline) (R ouget et al., 2003). Spatial distributions of 80 different tree species were predicted wi th regression tree analysis using climate, soil factors, elevation, and land use as predictors (Iversen and Prasad, 1998). They reported that for tree species capab le of occupying a wide range of habitats, the classification accuracy was greater than 70%. The US Envi ronmental Protection Agency (Pan et al., 1999) found that broad spat ial patterns of benthic diatom assemblages in Mid-Atlantic streams could be predicted both by coarse-scale eco logical factors (such as land cover and land use), and by site-speci fic ecological factors (such as riparian conditions), using a regr ession tree model. Total P data collected in the field (des cribed in Chapter 3) showed a non-normal distribution, and had different degrees of variation within multiple categories (i.e., land use, vegetative community). It is the aim of this chapter to use statistical relationships between measured soil TP (mg kg-1) and ancillary spatial data available for the whole study area, to predict TP in unsampled, histor ically isolated wetlands. Because each wetland is an ecosystem, there are many inte racting variables (both quantitative and categorical) at multiple scales which may driv e TP storage. Classification and regression trees are analytical techni ques that are not bound by restri ctive assumptions of linear statistical models (Iverson and Prasad, 1998) so are suited to the development of predictive TP upscaling models in my study.

PAGE 98

81 Hypotheses Satellite reflectance data from wetland areas and associated upland areas is predictive of TP condition in historically isolated wetlands. Soils information related to the uplands is predictive of TP cond ition in historically isolated wetlands. Land use information related to the upla nds is predictive of TP condition in historically isolated wetlands. Landscape metrics related to wetland locat ion is predictive of TP condition in historically isolated wetlands. Objectives Develop a model that predicts TP storage in the surface soils of unsampled wetlands using measures from the spatia l sampling described in Chapter 3 in combination with available spatial ancillary environmental datasets. Predict and map the TP stored in surface soils of unsampl ed historically isolated wetlands. Materials Because we were interested in the TP stor ed in mostly disconnected historically isolated wetlands, a polygon-based approach was employed in this GIS study. The NWI (USFWS, 2002) dataset was used to define the unsampled wetlands. Other spatial layers included the basin boundaries, waterways, la nd use, soils, major roads, DOQQ photos, a 2003 Landsat7 ETM+ satellite image, and the sa mpling data described in Chapter 3. All layers were projected to HP GN (NAD83) Albers projection. Projection parameters and metadata about the non-spectral data layers are listed in Appendix A. Appendix B provides a glossary of GIS and remote sensing terms used in this text. Land Use The land use layer used in this chapte r was from 2003 (an update to the 2001 land-use layer used in Chapter 3). It employed three levels of land use codes, according

PAGE 99

82 to the Florida Land Use and Cover Classifi cation System (FLUCCS) (FDOT, 1999). The processes for classifying IMP and UNIMP land uses, and for integrating wetland land use polygons into land use areas (as described in Chapter 3), were repeated for the 2003 land use layer (Figure 4-1). 01020 5 Kilometers dairies improved pastures unimproved pastures / rangelandsLake Okeechobee Figure 4-1. Land uses (2003) of the priority basins. National Wetland Inventory The NWI dataset provided the basis of sp atial wetland information in this study. Associated with each wetland polygon was a hierarchical wetland cl assification system

PAGE 100

83 with four levels: system, subsystem, class, and subclass. Class re presents the appearance of the wetland in terms of ve getation substrate (USFWS, 2002) and in this thesis is referred to as the vegetative community. Th e NWI dataset also included modifiers that described wetland hydrology. A problem with the dataset was that polygons exhibited multidirectional shifts when compared to DOQQ photos (Figure 4-2) and the Landsat7 image. This is probably because the NWI layer was produced from stereophoto acetate layers made by hand from unrectified pape r images (before 1993) (J. Miner, NWI employee, personal communication, July 21, 2004). NWI polygons 0500 250Meters Figure 4-2. National Wetland Inventory polygons. A) In the four basins. B) Overlaying a 1 m DOQQ showing multidirectional shifts. Soil Survey Geographic (SSURGO) Data Set Soil spatial information (and associated databa se files) for each of the four counties in the study area (Martin, Okeechobee, Highlands and St. Lucie) were downloaded from the Florida Geographic Data Library (FGDL) (F igure 4-3). This dataset is a digital

PAGE 101

84 representation of the county soil survey maps published by the USDA/NRCS as the Soil Survey Geographic (SSURGO) da ta set (USDA/NRCS, 1995). Soil Orders Alfisols Entisols Histosols Inceptisols Mollisols Spodosols 01020 5 Kilometers Lake Okeechobee Figure 4-3. Soil orders in the four pr iority basins (Source: USDA/NRCS, 1995). The SSURGO dataset contains all levels of soil taxonomy. Th e spatial component of the dataset consists of map units. Each map unit represents 1 to 3 soil components, which usually correspond to soil series. Information about great groups and subgroups was updated manually in GIS layer attribute tables, since Haplaquods were changed to Alaquods, in Florida (G.W. Hurt, pers onal communication, September, 2004).

PAGE 102

85 Though the SSURGO and NWI datasets were both created on a 1:24,000 scale, there were many small NWI wetlands whose soils probably vary from the map unit polygon they coincide with (Figure 4-4). Most 1:24,000 scale maps have about a 2 ha minimum feature size (Soil Survey Staff, 1993) Considering that the median wetland size was about 1 ha, the soil map units that spatially coincided with wetland polygons may not describe the wetland soil type accura tely. For this reason, and because upland soils are expected to affect TP dynamics ar ound wetlands, soil information related to uplands adjacent to each wetland was collected for upscaling purposes. 11 7 14 14 2 14 2 11 11 2 14 14 11 11 14 11 2 2 2 11 11 3 11 4 11 2 7 11 2 2 11 7 11 3 4 2 14 3 14 3 3 2 3 4 7 3 5 7 2 7 11 7 5 7 3 3 11 11 7 3 3 4 5 11 18 11 7 7 3 3 3 14 18 15 3 6 7 2 18 3 2 7 18 5 3 3 7 18 3 3 3 7 3 7 7 3 3 3 3 3 3 3 3 3 3 3 3 18 14 7 7 9 3 3 14 3 18 7 7 18 18 7 11 7 99 3 2 012 0.5 Kilometers Figure 4-4. National wetland inventory wetland polygons (pi nk) shifted in multiple directions from the SSURGO map un it polygons (blue). 3 = Basinger and Placid soils; depressional, 14 = Myakka fine sand, 11 = Immokalee fine sand (Lewis et al., 2001). Wetland so ils in the red circle are probably Basinger and Placid soils; depressional not Myakka fine sand (G.W. Hurt, personal communication, February, 2004).

PAGE 103

86 Landsat7 ETM+ Spectral Data Different materials on earthÂ’s surface refl ect energy in different wavelengths. Images captured by satellites consist of same -size pixels that represent a quantity of energy reflected from the earth (Jensen, 1996). Spectral data from a Landsat7 Enhanced Thematic Mapper+ (ETM+) was used in this study as a representation of ecological response to nutrient condition in the histor ically isolated wetlands. Landsat7 ETM+ images have six spectral bands with 30 m spatial resolution (1 pixel represents 0.09 ha) (Table 4-1). Table 4-1. Landsat7 ETM+ spectral bands. Band Spectral range in micrometers Purpose 1 0.45-0.52 (blue-green) Designed for wa ter body penetration, soil and vegetation discrimination, and forest-type mapping.1 2 0.52-0.60 (green) Corresponds to th e green reflectance of healthy vegetation.2 3 0.63-0.69 (red) Operates in the chlorophyll absorption region.1 4 0.76-0.90 (near-infrared) This band abso rbs reflectance from healthy green vegetation and is used to estimate biomass.1 This band emphasizes soil-crop and land-water contrasts.2 5 1.55-1.75 (mid-infrared) This band is se nsitive to the amount of water in plants and soils.1,2 6 2.08-2.35 (mid-infrared) Useful for disc riminating mineral and rock types. Sensitive to vegetation and soil moisture.1 1 Lillesand and Kiefer, 2000. 2 Jensen, 1996. Each pixel consists of si x spectral data values, one for each band. These data values are referred to as the digital number (DN), and range from 0 to 255. They are a representation of the total solar radiance m easured by each sensor. More information about the Landsat7 satellite and ETM+ instrume nt can be found in the Landsat7 Science Data UserÂ’s Handbook (National Aeronautics and Space Administration, 2004). Spectral

PAGE 104

87 information from the image was processed using ERDAS Imagine software version 8.5 (Leica Geosystems, Atlanta, GA). The study area was overlapped by two Landsat 7 flight paths. A cloudless scene from path 16/41 (acquired on March 24, 2003) was purchased from the Earth Resources Observation Systems (EROS) Data Center (Fig ure 4-5). Detailed image metadata is in the header file listed in Appendix G. Th is scene included about 95% of the study area, and was temporally the closest cloudless scene to the sampling period available. It was expected that there would be slightly less vegetation at the time of image capture than during the sampling period, but ra infall records showed that rainfall between January and March, 2003 was 30 60% above the averag e (SFWMD, 2004b). Temperatures were normal (SFWMD, 2004c). Moreover, it has be en documented that springtime imagery in the Southeastern USA is optimal for wetland di scrimination (Jensen et al., 1984). Before plants have fully leafed out (e.g., when the image was captured), conditions are better for characterizing wet soils and standing water (Ozesmi and Bauer, 2002). On April 4, 2004, a year after the image da te, wetlands within the study area were visited to observe water levels and growth st atus of vegetation that would be comparable to the time of image capture (Figure 4-6). Water levels were lower than during the sampling period, but many wetland vegetation speci es were leafed out including some of the most common wetland species found in the sampled wetlands: Juncus effusus L. (was completely leafed out), Polyganum spp., and Pontedaria cordata L. (was about halfway leafed out). Less common species, such as Thalia geniculata L. Marantaceae and Eupatorium capillifolium, were incipient, with much standing dead plant material. Pasture grasses were generally brow ner than during the sampling period.

PAGE 105

88 Figure 4-5. Landsat7 ETM+ image of path 16/41 (183 km by 170 km) from March 24, 2003 and the study area. Figure 4-6. April 5, 2004 photo of a historic ally isolated wetland in the study area.

PAGE 106

89 Methods Upscaling Methodology Classification trees (classes assigned we re high TP and low TP) were used to upscale measured TP (mg kg-1) in the hydrologic zones (c enter and edge) of sampled wetlands to TP in unsampled wetlands. Mean bulk density values for sampled soils by TP class and hydrologic zone were used to calculate TP storage within unsampled wetlands. The general steps for the upscaling procedure were: 1. Sampled Wetlands a. Align sampled wetland polygons to match DOQQ photos. b. Create transition and upland polygons around each wetland (25 m and 75 m buffers). c. Collect independent variables rela ted to each wetland (Figure 4-7): i. spatial and spectral data for wetland polygons ii. spectral data for transition polygons iii. land use, soils and spectral data for upland polygons d. Convene the independent variables and measured TP values into two data sets: i. edge TP (mg kg-1) for each wetland and all independent variables ii. center TP (mg kg-1) for each wetland and all independent variables e. In each of the two datasets, classify each soil sample as being high or low TP (use mean TP as initial cutoff to separate classes). f. Use each dataset to build two classifi cation trees in an iterative process adjusting the high or low TP cutoff in order to make the most accurate tree: i. one that predicts if a sampled wetland center has high or low TP ii. one that predicts if a sampled wetland edge has high or low TP g. Repeat step 1f but exclude all spectral data to create two trees that predict edge and center TP classes for the un sampled wetlands in the southeastern part of the study area that are not covered by the Landsat7 ETM+ image.

PAGE 107

90 2. Unsampled Wetlands a. Identify all historically isolated wetlands. b. Align wetland polygons to match DOQQ photos. c. Create transition and upland polygons around each wetland (25 m and 75 m buffers respectively). d. Collect wetland independent variables th at were identified in decision nodes of the four classification trees. e. Classify each unsampled wetland base d on classificati on tree logic: i. low center / low edge ii. low center / high edge iii. high center / low edge iv. high center / high edge f. Use a transfer function based on sampled wetland size to estimate the areal proportion of each wetland that is edge or center. g. Use mean TP (mg kg-1), mean BD (g cm-3) and standard error (SE) measured in sampled centers and edges in each of the classes listed in step 2e to calculate soil TP storage (g m-2) in the top 10 cm of each hydrologic zone of the unsampled wetlands. h. Use proportions, estimated in step 2f, to calculate soil TP (g m-2 and kg) in each wetland. i. Create maps of TP storage (mean g m-2 and total kg) of each historically isolated wetland in the study area. Collection of Independent Variables Sample biogeochemical data for wetland cen ter and edge soils was stored in a MS Access relational database. Independent vari ables describing the sampled wetlands were redundantly stored for edge soils and center soils, for the purpo se of creating two separate classification trees. The collection of thes e independent variables is summarized in Figure 4-7.

PAGE 108

91 Figure 4-7. Steps for collecting samp led wetland independent variables. Field notes (vegetation, land use, edge %) a Wetland GPS data Match to corresponding NWI p ol yg ons Calculate wetland area and perimeter a Water regime a Create new shapefiles Wetland area and perimeter Create transition buffers (25 m) and upland b uffers ( 75 m ) around wetlands Find most common mapunit in 75 m u p land buffers Order, subgroup of major component NWI wetlands SSURGO soil data a Create distance rasters: to other wetlands, to major roads, to high intensity dairy areas and to waterways Roads Waterways Dairies Min. distances from wetlands tofeatures a Atmospherically correct Landsat7 image Georectify with DOQQs Calculate indices of Landsat7 spectral data Calculate zonal statistics of indices and 6 reflectance bands in wetland, transition, and upland polygons. Clip to 4 basins Zonal spectral statistics a a Edge and center database Land Use Create land use raster; calculate ma j or LU in u p land buffe r Major upland land use a

PAGE 109

92 Field data Information collected in the field about vegetative community, land use, and edge percent was entered for each wetland in the re lational database. Area and perimeter data were imported from the sampled wetland shapefile in ArcGIS. Water regime Polygons from the NWI dataset that coincided with sampled wetlands were selected and water regimes (Table 4-2) were recorded for each sampled wetland. Table 4-2. Non-tidal water regimes of wetlands in the National Wetland Inventory. A = Temporarily Flooded B = Saturated C = Seasonally Flooded D = Seasonally Flooded / Well Drained E = Seasonally Flooded / Saturated F = Semi-permanently Flooded G = Intermittently Exposed H = Permanently Flooded J = Intermittently Flooded K = Artificially Flooded W = Intermittently Flooded / Temporary Y = Saturated / Semipermanent / Seasonal Z = Intermittently Exposed / Permanent Distance to features Distance-to-feature landscape metrics were calculated for the sampled wetlands to be used as independent variables. These feat ures were: high intens ity dairy areas (HIAs) (Figure 4-8), major roads (Figure 4-9), wate rways (Figure 4-10) a nd wetlands (Figure 3-6). High intensity areas within dairies ar e the areas where feeding, milking, cow care, food storage, and waste management take pl ace. They are known to have very high TP concentrations (Graetz and Na ir, 1995), and proximity to t hose areas may be related to wetland soil TP. These areas along with major roads, ma y be indicative of more intensely developed areas. Areas further from dairies and major roads may be more

PAGE 110

93 remote, and may have less improved pastures. Proximity to other wetlands and to waterways may be predictive of wetland TP. If a sampled wetland is close to another wetland or waterway, it may be receiving less runoff, and perhaps less P. Proximity to another wetland or waterway may be an indi cator of the characteristics of a wetlandÂ’s upslope drainage area (UDA). Because th e study area has little topographical change, UDAs are difficult to delineate reliably, so they were not used in this study. sampled wetlands high intensity dairy areas dairies priority basinsdistance (m) to high intensity areas 29900 0 05101520 2.5 Kilometers Figure 4-8. Distance-to-high-in tensity-dairy-area raster. High intensity dairy areas were defined by first selecting all the barn area polygons of the dairy land use layer (using FLUCCS c odes in the attribute table and DOQQ aerial photos). Then, areas adjacent to those that were pastures, wetlands, ditches, lagoons,

PAGE 111

94 facilities, food storage areas, or fields, were se lected and exported as a separate HIA layer (red polygons in Figure 4-8). major roads sampled wetlands priority basinsdistance (m) to major road 30800 0 05101520 2.5 Kilometers Figure 4-9. Distance-to-major-road raster. A layer of waterways was created by merging the waterway layers: ditches, streams, canals and major tributaries (F igure 4-10). Upon comparing the layer representing ditches to DOQQ photos, it was observed that many ditches were not represented. Ditch lines (1,300) were digi tized for each DOQQ in the land uses of interest. A straight-line-distance raster with 15 m resolution was created. This defined the minimum distance of each cell to a feature. This was done for each of the four sets of features: HIAs, NWI wetlands, major roads, and waterways, to create four rasters. To

PAGE 112

95 determine how far a sampled wetland was fr om a feature, a minimum distance zonal statistic was determined for each sampled wetland. This generated four new data tables in ArcGIS which were then imported into the relational database. Metadata for wetlands, major roads, land use, and waterways is in Appendix A. Ditches, streams, canals, major tributariesdistance (m) to waterway 28000 0 05101520 2.5 Kilometers Figure 4-10. Distance-to-waterway raster. Creating Buffers around Wetlands Two buffers were created around each wetland in the GIS: a 25 m wide transition buffer around the wetland, and a 75 m wide upland buffer around the transition buffer. The upland buffer was used to characterize soils, land use, and spectral information. Only spectral information was collected within the transition buffer. The width of 75 m was chosen in order to include at least two Landsat7 pixels as a function of distance from

PAGE 113

96 the wetland. It can be seen in Figure 4-11 that the green vegetation in the areas surrounding historically isolated wetlands is variable, and may be related to nutrient condition. To create these buffers, four 25 m buffe rs were first created around each wetland polygon. The inner buffer was the transitional bu ffer. The outer 3 rings were dissolved to become the upland buffer. The two buffers were then associated with the identification number (ID) of each wetland. Detailed steps of how the buffers were dissolved and associated with the wetland ID are described below. A 100 m buffer was created for each we tland extending 100m outside as well as inside each wetland (Figure 4-12A). This crea ted an effective “umbrella” that overlaid the whole wetland, as well as all four of the 25 m buffers. Centroid points were created for the wetlands as well as for the 100 m buffers using XTools Pro (DataEast LLC, Moscow, Russia) for ArcMap. The wetland cen troids, which contained the wetland ID numbers, were joined to the 100 m buffer cen troids based on spatial location (points of each layer closest together were joined). Th is had the affect of associating the wetland ID with the 100 m centroid point s. Another spatial join be tween the 25 m buffers and the 100m buffers associated the wetland ID with the 25 m buffers (Figure 4-12B). The inner 25 m buffer was the transition buffer for each wetland. The three outer 25 m buffers were dissolved based on wetla nd ID number (Figure 4-12C) to create the 75 m upland buffer for each wetland. Both the trans ition and upland buffers had wetland ID associated with them, so that data collect ed for them could be associated to the appropriate wetland polygon.

PAGE 114

97 Figure 4-11. Landsat7 ETM+ false-color com posite of a portion of the study area (the stream/canal feature on the right-hand si de is Taylor Creek). Red represents band 4, green represents band 3 and blue represents band 2. Yellow polygons represent nonriparian wetlands within land uses of interest. Brighter reds indicate mo re vigorously growing vege tation. Soils with no or sparse vegetation will range from wh ite (sands) to greens or browns depending on moisture and organic matter content. Water bodies appear blue. Deep, clear water is dark blue to black in co lor, while sediment-laden or shallow waters appear lighter in colo r. Urban areas are blue-gray in color. (Source: Jensen, 1996). GIS Soil Characterization Soil map unit polygons in the SSURGO shapef ile that coincide with the wetland polygons, often reflected soil types of the upland area instead of the wetland soil. Though both the SSURGO and NWI datasets we re created at a scale of 1:24,000, there were more small polygons in the wetland shap efile generated from NWI, than in the SSURGO map unit shapefile (Figure 4-4). N onriparian wetlands were typically 0.7 to 1.3 ha in size, and the NWI dataset contai ned over 6,000 wetland polygons that were smaller than 1.3 ha. The SSURGO dataset had only 567 map unit polygons that were smaller than 1.3 ha. Therefore, assuming th e wetland soil type was distinct from the

PAGE 115

98 surrounding upland, it was not possible to a ccurately characteri ze the soil map units representative of these small wetlands. 0150300 75Meters historically isolated wetlands 100 m buffer (in and out) transition buffer upland buffer25 m buffers 3135 3143 3162 3158 3183 3175 A 3135 3143 3162 3158 3175 3183 31313135 3143 3162 3158 31753135 3135 3143 3135 3162 3158 3143 3175 3162 3158 3183 3143 3175 3162 3131 3183 3158 3175 3131 3183 3092 3135 3143 3162 3158 3175 3183 3131 3135 3143 3162 3158 3175 3135 3143 3162 3158 3175 3183 3131 3092 B C Figure 4-12. Steps for creating buffers and a ssociating them with the wetland ID. A) 100 m inner / outer buffer used as an “umbre lla” to spatially join the wetland ID to all other buffers. B) Wetlands a nd 25 m buffers around wetlands with wetland ID numbers. C) Final outer buffers (transition and upland) with wetland ID numbers. Since the transport of P in the landscape to the wetland was assumed to be affected by soils in the uplands, soils within the upland buffer areas of the wetlands were characterized. Because only 118 wetlands were sampled, not every possible map unit in the four counties was sampled. The higher le vel soil descriptors (order, great group and

PAGE 116

99 subgroup) of each major component of the map unit that dominated each upland buffer polygon, were identified for each wetland. To do this, the SSURGO shapefile was conve rted to a 10 m-resolution raster based on map unit (MUID) in ArcMAP. The ‘majority ’ zonal attribute was used to calculate which MUID value occurred most often within each upland buffer. The buffer attribute table containing the MUID and the wetland ID wa s imported into an relational database. In the SSURGO database, each map unit consists of 1 to 3 components. Components are generally phases of so il series (USDA/NRCS, 1995). A SSURGO components table for all four counties was im ported to the database and related to the buffer table by MUID. Each component record lists the series name, subgroup, suborder, order and component percentage that makes up the map unit. A query was generated in MS Access to list the order, great group a nd subgroup of the major component of the map unit that was most commonly found in the uplan d buffers of each wetland. (The series, great groups and orders associated with the upland buffer areas of the unsampled wetlands are summarized in Appendix H.) Pre-Processing the Landsat7 ETM+ Image Before spectral information in a satellite im age can be used in a geospatial analysis, the image must be pre-processed. This incl udes radiometric and geometric corrections. Radiometric correction is the process of re moving spectral influences that alter the measurement of earth object reflectance values (Wharton, 1989). Geometric correction is the process of rectifyi ng an image with another spatial layer so that they geographically coincide (Jensen, 1996). Radiometric correction. In an ideal model, satellite sensors would record only radiance reflected from objects on earth, but the sources of atmos pheric noise require

PAGE 117

100 atmospheric correction in order to obtain reflectance values. Energy recorded by a Landsat7 ETM+ sensor comes from three sources: the radian ce reflected by a ground object ( ), the radiance scattered diffusely by the atmosphere (which then further interacts with the target) (ET/ ), and the radiant energy scattered diffusely by the atmosphere (Lp) (Lillesand and Kiefer, 2000) These three forms of energy add up to total radiance (Ltot) and are recorded by the sensor (Equation 4-1) (Lillesand and Kiefer, 2000). The sensor records the radiance as an 8-bit-digital number (DN) whose values range from 0 to 255 (Jensen, 1996). Ltot = pL E T (Equation 4-1) Radiometric correction is classified into three parts: sensor errors, atmospheric effects, and sun angle (Wharton, 1989). Sens or errors were corrected by the vendor before delivery of the Landsat7 ETM+ scene (National Aeronautics and Space Administration, 2004). Many different atmo spheric correction methods have been explored in order to obtain surface reflectance values (L iang et al., 2001). A most common method is dark object subtraction ( DOS) (Teillet and Fedosej evs, 1995; Liang et al., 2001; Butson and Fernandes, 2004). The co ncept of DOS, is that reflectance values (and corresponding DN values) for very dark obj ects are close to zero in all the spectral bands when there are no atmospheric effect s (Liang et al., 1997). Since atmospheric interference is wavelength-dependent, the mi nimum pixel DN values will vary from band to band. Thus, during the DOS, the lowest pixel value (from a la rge dark lake for example) in a band is subtracted from every pixe l in the scene. This is repeated for every band, and is based on the assumption that at mospheric conditions w ithin the scene are constant. Atmospheric errors for this anal ysis were corrected using a DOS called the

PAGE 118

101 COST method (Chavez, 1996) in ERDAS Imagin e. Sun-angle effects were corrected during the DOS process when radiance was converted to reflectance (Kaufman and Sendra, 1988). Reprojecting and subsetting the image. Using ERDAS Imagine, the Landsat7 ETM+ image was reprojected from Universal Transverse Mercator (UTM) projection to High Precision GPS Network (HPGN), Nort h American Datum 1983 (NAD83) Albers equal-area projection. Projection parameters are listed in Appendix A. A shapefile of the four-basin study area within the image extent wa s created. This was used to subset the image (Figure 4-13). Figure 4-13. The Landsat7 ETM+ subset image of study area. Image georectification and NWI polygon alignment. Since the data of interest (historically isolated wetlands ) is polygon-based, it was important to line up wetland polygons with the image. Attempts to georectify the image to DOQQs were unsuccessful. Attempts at an image-to-m ap rectification involved creating over 300 ground control points between the image and D OQQ photos distributed in all areas of the

PAGE 119

102 study area (Jensen, 1996). The rectification RMS error was 0.39 but resulted in minimal and heterogeneous improvement of wetland alig nment (Figure 4-14). There were multidirectional shifts of polygon positi ons relative to other layers. This was thought to be an artifact of the NWI layer creation met hods before 1993 (J. Miner, NWI employee, personal communication, July 21, 2004). Since the Landsat7 image was rectified with DOQQ images, the wetland polygons were manua lly adjusted to ma tch the 1 m DOQQ photos (Figure 4-15). This would ensure the best possible alig nment between wetland polygons and the Landsat7 image. A B Figure 4-14. Landsat7 ETM+ image (bright green areas represent wetlands) and NWI polygons (black outlines). A) Before r ectification. B) Af ter rectification.

PAGE 120

103 manually adjusted wetlands nonriparian wetlands 0350 175 Meters Figure 4-15. Nonriparian polygons before (pi nk) and after (blue) manual alignment to match DOQQ photos. Calculating Spectral Indices from the Landsat Image Spectral indices are used to identify sma ll differences between various rock types and vegetation classes that may not be observed in the display of the original color bands (ERDAS, 1999). A vegetative index is a valu e that is calculated from remotely-sensed data and is used to quantify vegetative cover. Also, a vegetation index can be viewed as an indicator for the properties of the landscap e as a whole (Schmidt and Skidmore, 2003). These properties may affect th e retention of P in that la ndscape. Though many vegetative indices exist, the most widely used index is the Normalized Difference Vegetative Index (NDVI) (Rondeaux et al., 1996; Lillesand and Kiefer, 2000). The NDVI is calculated as a ratio between the red and near infrared (N IR) portions of the spectrum (Equation 4-2). These two spectral bands are most affected by the absorption of chlorophyll in leafy green vegetation, and by the de nsity of green vegetation (L illesand and Kiefer, 2000).

PAGE 121

104 NDVI d NIR d NIR Re Re (Equation 4-2) Because the Landsat7 ETM+ scene was from the end of the dry period, it was important to use indices that consider ed nonphotosynthetic vegetation and soils. Characterization of vegetation reflectance can be improved with soil-adjusted soil indices (Huete, 1988). The Modified Soil-Adjuste d Vegetation Index (MSAVI) is a vegetation index that accounts for soil background reflectance so that it is more representative of just green healthy vegetation (Rondeaux et al., 1996). This should be an important consideration in this study since wetlands often have pa rtial vegetation canopies and because the sandy soils of South Florida are quite bright. This could also be important in our study since there was a large variability in the amount of soil organic matter found in the sampled wetlands soils. Qi et al. (1994) developed severa l MSAVIs. The one used in this analysis (MSAVI2) (ERDAS, 2004a), did not require precalculation of other vegetation indices, or the sl ope of the soil line (Equation 4-3) (Qi et al., 1994). 2 8 1 2 2 1 22MSAVI d NIR NIR NIR ) Re ( ) ( ) ( (Equation 4-3) The Tasseled Cap (TC) is a linear tr ansformation of the six Landsat7 ETM+ spectral bands to create three mutually inde pendent vectors that represent brightness, greenness (Kauth and Thomas, 1976), and wetne ss (Crist and Cicone 1984). These linear transformations were developed to correlate to physical characteristics of agricultural fields (Crist and Cic one, 1984). Brightness (TC1), is a we ighted sum of all bands defined in the direction of the principal variation in soil reflectance (Li llesand and Kiefer, 2000). Greenness (TC2), is orthogonal to brightness, an d is a contrast between near-infrared and visible reflectance (Lillesa nd and Kiefer, 2000). Thus, it measures the presence and density of green vegetation (Lillesand and Ki efer, 2000). Wetness (TC3), is a contrast

PAGE 122

105 between shortwave-infrared (SWIR) and vi sible/near-infrared (VNIR) reflectance, providing a measure of soil and canopy moistu re (Crist and Kauth, 1986). Coefficients for the indices (ERDAS, 2004b) ar e listed in Table 4-3. Table 4-3. Linear transfor mation coefficients for the 6 spectral bands of Landsat7 ETM+ to produce three Tasseled Ca p (TC) indices (ERDAS, 2004b). TC index ----------------------------------Band---------------------------------1 2 3 4 5 6 Brightness (TC1) 0.3561 0.3972 0.3904 0.6966 0.2286 0.1596 Greenness (TC2) -0.3344 -0.3544 -0.4556 0.6966 -0.0242 -0.263 Wetness (TC3) 0.2626 0.2141 0.0926 0.0656 -0.7629 -0.5388 Pixels are either pure or mixed. If, for example, a pixel represents an area on the ground that consists of part past ure and part wetland, it is a mi xed pixel. If it consists only of deep water, it is a pure pixel. Tran sition zones between different land covers will consist of mixed pixels. Because the median size of the historically isolated wetlands in the study area small (less than 1 ha), most wetlands will contain a high number of mixed pixels. These pixels may quantify a combin ation of a wetlandÂ’s vegetation, soils and moisture. Thus, the reflectance values with in each band, in add ition to the vegetation indices, may be important indicators of we tland TP condition. More information about remote sensing concepts and methods can be found in Jensen (1996) and Lillesand and Kiefer (2000). Zonal statistics (minimum, maximum, mean, range and standard deviation) of the 6 reflectance bands and the indices were calculat ed for the wetlands, transition buffers and upland buffers. These statistics were stored in the attribute tables for the wetland shapefiles and then imported into the relational database.

PAGE 123

106 Classification Trees Urban (2002) describes a clas sification tree as a model th at partitions the dataset recursively into subsets that are increasi ngly homogeneous with respect to the defined groups. It generates an inverted decision tree providing a dichotomous key to classify unknown samples into groups. The top node of the tree is a parent nod e and is split into two child nodes. Each child node can beco me a parent node by be ing partitioned into two nodes. The algorithm chooses the split that partitions the data into two parts such that it minimizes the sum of the squared devia tions from the mean in the separate parts (Bishop et al., 2001). Terminal nodes classify (classification trees) or predict (regression trees) the value of the dependent variable. CART version 5.0 software (Salford System s, San Diego, CA) was used to create classification and regression trees. Regression trees to predict TP values in unsampled wetlands were attempted as well as classificati on trees with 5 and 3 cl asses of TP, but the small sample size proved to preclude acceptabl e cross-validation accuracies (they were below 50%). Thus 2-level classification trees were built. When creating classification trees for binary dependent variables (2 cl asses), it is recommended that at least 200 learning samples be used and about 100 more for each additional leve l of the dependent variable (i.e., 300 samples is recommended to predict high, medium and low TP classes) (Steinberg and Colla, 1995). Classification trees were built to class edge and center soils as having high or low TP (2 classes). First, sampled edge and cen ter soils were classed as having high or low TP (mg kg-1). (Bulk-density-adjusted TP number s were not used because bulk density numbers were thought to intr oduce more error into the pr edictions.) One TP (mg kg-1) value was missing from the edge soil data, a nd two values were missing from center soils

PAGE 124

107 data. To replace those values, linear regres sion using environmental and laboratory data was investigated. Regression yiel ded poor overall results (adjusted R2 < 0.5 in center soils). Since all missing values were for IMP EM wetlands (the group of wetlands with the lowest variability compared to other land uses and vegetative communities), they were replaced by the median values for each zone of sampled IMP EM wetlands (400.2 mg kg-1 for centers, and 173.5 mg kg -1in edges). Initially, the cutoff between high and low TP for each hydrologic zone was the mean of each group (sampled center soils and sa mpled edge soils). More trees were run with slightly different cutoff values until the tree with the lowest relative cost was generated. The best classification trees (u sing all the independent variables listed in Table 4-4) resulted when the high / low TP cutoff was 250 mg kg-1 for edges, and 620 mg kg-1 for centers. As different trees were bei ng generated, variables were removed from the analysis and sometimes added back. Cr oss-validation accuracy and relative cost values changed with the variables that were introduced to the model building. The best tree was created when fewer than all of the origin al variables were used to build the tree. Cross-validation is the process of us ing a portion of the data for testing classification accuracy in an iterative manner. Cross-validation to test tree accuracy is used when there are fewer than the recomme nded number of cases for building trees to predict a certain number of dependent variable s (i.e., number of cla sses) (Steinberg and Colla, 1995). First, a tree is generated from the entire dataset, and then if a 10% crossvalidation is used, the data is divided into 10 equal sets (random with equal distributions of the dependent variable). Trees are then generated 10 different times with 90% of the data. Classification errors are then averaged from the multiple tree generations.

PAGE 125

108 Table 4-4. Predictors introduced into th e classification tree building process. Categorical data (8 predictors) Predictor Levels of Sampled Wetlands Soil order of map unit major component Entisol, Histosol, Alfisol, Mollisol, in upland buffer Spodosol Great group of map unit major component Alaquod, Ochraqualf, Argiaquoll, in upland buffer Endoaqualf, Medisaprist, Psammaquent Soil subgroup of map unit major component Aeric Alaquod, Arenic Alaquod, in upland buffer Arenic Ochraqualf, Arenic Argiaquoll, Typic Argiaquoll, Typic Endoaqualf, Terric Medisaprist, Spodic Psammaquent Land use IMPa, UNIMPb, DAIRc Wetland vegetative community EMd, EWe, FOSSf Wetland hydrological regime Fg, Ah, Bi, Cj, Hk Basin S-191, S-154, S-65E, S-65D Quantitative spatial data (31 predictors) Distance from wetland center statistics (min, max, range, mean, std. dev.): road, major road, high intensity dairy area, ditch or stream, other wetland In upland 75 m buffer, number of soil map units and land use types Wetland area and perimeter No. of ditches Ditch intensity Spectral data (165 predictors) In wetland, 25 m buffer and upl and 75 m buffer (min., max., range, mean, std. dev.): Tasseled Cap 1, 2, 3 Reflectance in 6 bands NDVI MSAVI a Improved pasture b Unimproved pasture or rangeland c Dairy d Emergent marsh e Emergent marsh / open water f Forest / scrub-shrub g Semi-permanently flooded h Temporarily flooded i Saturated j Seasonally flooded k Permanently flooded

PAGE 126

109 The best classification trees have the lowest relative cost. Relative cost takes into account classification success rates as well as tree complexity. A tree with high classification success has terminal nodes that are very pure because cases of different classes would very seldom be put into the same terminal node. Models that are less complex (more parsimonious) are estimated with greater precision (Breiman et al., 1984). More cost is associated with bigger trees (more terminal nodes), which, although they may fit the learning dataset with higher accuracy than smaller trees, are overfit to the particular dataset (Steinberg and Colla, 1995). A large complex tree may fit the idiosyncrasies and noise in the learning data set which are not likely to occur in another set of similar samples (Lewis, 2000). For mo re information on classification trees, refer to Breiman et al. (1984) and Steinberg and Colla (1995). In my study, a 10% cross-validation was us ed, and the best trees were those with the lowest relative cost and highest cross-va lidation accuracy rates fo r both classes (high and low TP). A tree with an overall relative cost of 0.40 and a cross-validation accuracy rate of 0.52 for high TP and 0.89 for low TP was not considered as good as a tree with relative cost of 0.43, and cr oss-validation accuracy rates of 0.71 for both classes. Since 5% of unsampled wetlands were not in the Landsat7 ETM+ image extent, two trees were created using spectral data independent variables, and two trees were created without it. Both trees used the same cutoff value between high and low TP. Classifying unsampled wetlands Unsampled wetlands were separated into two layers: those within the extent of the Landsat7 ETM+ image extent, and those that were not. Visual Basic (Microsoft, Redmond, WA) scripts were written (using the fi eld calculator in ArcMap) to assign each

PAGE 127

110 wetland with an edge class (high or low TP) and a center class (high or low TP). The scripts were based on the logic of the classification trees and are listed in Appendix J. Edge percent transfer function In order to calculate TP storage in unsa mpled wetlands, a func tion relating percent edge in sampled wetlands to a parameter that was available in watershed-scale data layers was needed. It is well known that smaller sh apes within an ecological context have more edge (Sharpe et al., 1981). Edge percent wa s estimated in the field in units of 10. A relationship between wetland area (ha) and edge percent can be seen in Figures 4-16 and 4-17. However, this relationship is not linear, and showed a low adjusted R2 = 0.041. Almost all wetlands sampled that were sma ller than 2.6 ha, had an edge proportion greater than 20% (Figure 4-17). 1 2 3 5 8 15 11 26 43 4 N =Ed g e % 90.00 80.00 70.00 60.00 50.00 40.00 30.00 20.00 10.00 .00Area (ha) 95% Confidence Int.4 3 2 1 0 -1 Figure 4-16. Wetland area confidence intervals by edge percent. Wetlands were divided into two classes: great er than or smaller than 2.6 ha. There was a significant difference in proportion that was edge between th e two classes using a factorial ANOVA with Type III sum of squares (p < 0.005). This relationship was used to calculate TP in unsampled wetlands. Wetlands larger than 2.6 ha had a mean

PAGE 128

111 proportion of 15.8% edge. Smaller ones had a mean edge proportion of 27.9%. These proportions were attributed to unsampled we tlands based on their size in order to calculate predicted TP storage. 0.0 1.2 2.4 3.6 4.8 6.0 7.2 8.4 9.6 10.8 12.0 13.2Wetland area ( ha ) 0.00 10.00 20.00 30.00 40.00 50.00 60.00 70.00 80.00 90.00W e t l a n d e d g e % Figure 4-17. Scatterplot of edge percent vs. wetland area. Most wetlands < 2.6 ha (vertical line) had > 20% edge. Calculating Total Storage Using the measured mean TP (mg kg -1) and BD (g cm-3) within each class (high TP and low TP) of center and edge soils (Table 4-5), TP storage (g m-2) in the top 10 cm of soil was computed as the product of BD (g cm-3) and TP (mg kg-2) divided by 10. To verify that the mean bulk densities woul d be appropriate to use for upscaling, comparisons were made between high and lo w TP classes in each hydrologic zone using factorial ANOVA with Type III su m of squares. Mean bulk densities of the two center classes were significantly diffe rent: F = 223.12; p = 0.000, and mean bulk densities of the two edge classes were significantly differe nt: F = 201.91; p = 0.000. Also, TP (mg kg-1) and BD were significantly inversely co rrelated (Pearson r = -0.671; p = 0.000).

PAGE 129

112 Table 4-5. Mean and std. error of TP a nd bulk density (BD) of sampled wetland soils by hydrologic zone and TP class. ---------------Center --------------------------Edge------------high TP low TP high TP low TP ----------------------------------mg kg-1 ------------------------------------TP (mg kg-1) 1157.89 85.99 270.67 19.27 645.90 74.223 135.44 6.271 BD (g cm-3) 0.50 0.0400 0.79 0.0397 0.66 0.0343 1.02 0.0189 As numbers with associated errors are multiplied, errors must be appropriately propagated (Taylor, 1997). The standard erro r (SE) of BD TP was calculated as the square root of the proportional errors that have been squared and summed (Equation 4-4) (Taylor, 1997). The SE of the summed TP st ored in all sampled wetlands is the square root of the sum of the squares of the sta ndard errors weighted by the wetland areal percent that is within that hydrologic zone plus the covariance of the samples in the particular edge TP class and center TP class (Equations 4-5 and 4-6) (Taylor, 1997; K.M. Portier, personal communicat ion, November 24, 2004). ** 2 2BD SE TP SE BD TP SEBD TP BD TP (Equation 4-4) where BD TPSE* = standard error of the TP*BD mean TPSE = standard error of the TP mean BDSE = standard error of the BD mean TP = total phosphorus (mg kg-1) BD = bulk density (g cm-3) cov * *eTPcTP e c eTP e cTP c sumTPw w SE w SE w SE 2 2 (Equation 4-5) where sumTPSE = standard error of mean of edge TP + center TP (g m-2) cw = weight (percent of wetland that is center) cTPSE = standard error of center TP (g m-2) ew = weight (percent of wetland that is edge) eTPSE = standard error of edge TP (g m-2) eTPcTPcov = covariance of center and edge TP (g m-2)

PAGE 130

113 n x TP x TPn i TP c e TP e eTPcTPc 1) )( ( cov (Equation 4-6) where eTPcTPcov = covariance of center and edge TP (g m-2) eTP = edge TP (g m-2) e TPx = mean of edge TP (g m-2) cTP = center TP (g m-2) c TPx = mean of center TP (g m-2) n = number of samples Results Important Predictors for TP The most important predictors for both edge and center TP were the Landsat7 ETM+ spectral data independent variables. Reflectance data for individual bands was most explanatory for center soils (Figure 4-18), and both reflecta nce and tasseled cap statistics were most explanatory for edge soils (Figure 4-19). For wetland center soils within the Lands at area, the relative cost of the classification tree was 0.488. The test data classification success rate for the high TP class (65 cases) was 71.1%, and for the low TP class (52 cases) was 80.0%. The overall cross-validation success rate for centers was a bout 75.6%. Important variables are listed in Table 4-6. A prediction success table is li sted in Table 4-7. For wetland edges within the Landsat area, the relative cost of the cl assification tree was 0.484. The test data classification success rate for the high TP cl ass (42 cases) was 70.7%, and for the low TP class (75 cases) was 81.0%. The overall cr oss-validation success rate for edges was about 75.8%. Important variables are listed in Table 4-8. A prediction success table is listed in Table 4-9.

PAGE 131

114 Classification Trees for Upscaling TP Outside the Landsat Extent The classification tree built for soil centers using only non-spectral data depended mostly on soil information and wetland size. The tree had an ove rall cross-validation accuracy of 62.3% (Figure 4-20). The most im portant variables are listed in Table 4-10. A prediction success table is listed in Table 4-11. The best tree for wetland edge soils depended mostly on spatial location and had an overall cross-valid ation accuracy of 73.3% (Figure 4-21). Priority basin and distance to high intensity area were most important, followed by vegetation community and land use (Table 4-12). A prediction success table is listed in Table 4-13. Classi fication tree outputs from the CART software are listed in Appendix I. Upscaling Results When classifying wetlands outside th e Landsat7 ETM+ area, 52 out of the 422 wetlands in that area had soil subgroups th at were not represented by the sampled wetlands and those soil subgroups were not part of the center classi fication tree (Figure 420). For these wetlands, the classification tree was traversed to the right branch at the first decision node so that a decision be tween high and low could be made based on wetland perimeter. The distribution of high / lo w TP classifications of unsampled wetlands among land use classes compare to those of sampled wetla nds in Figure 4-22. The variations in the distributions among land uses between sample d and unsampled may be partially due to the large variability in fiel d conditions within each land use discussed in Chapter 3. Total kg of P predicted in DAIR area wetlands was 12.8% of the total; in IMP it was 69.7%; and in UNIMP it was 17.4% (Tab le 4-14). The propor tions of wetland area by land use were 9.6% (DAIR), 67.7% (IMP) and 22.7% (UNIMP) respectively. The

PAGE 132

115 population of wetlands in the priority basins was predicted to be storing 292.3 kg ha-1 in the surface soils. Predicted high / low TP classes per hydrologic zone for unsampled wetlands were mapped (Figure 4-23). Basin S-191 had the highest proportion of high center / high edge wetlands. In the other th ree basins, more than half of the wetlands had low center / low edge wetlands. The mean predicted TP (g m-2) in the surface soils of the whole wetland is mapped in Figure 4-24. Figure 4-18. Classification tree for sampled wetland center soils using Landsat7 data. White boxes are decision nodes based on Landsat7 ETM+ spectral data, and colored boxes are terminal nodes cla ssifying a percentage of learning samples (actual classes are listed) as having high (orange) or low (blue) TP (mg kg-1). Overall cross-validati on accuracy is 75.6%. Wetland Refl. Band 6 Std. Dev .<= Wetland Refl. Band 4 Std. Dev. 11 Low (81%) 2 High 17.261 32 High (85%) 7 Low 12.719 Upland 75 m Refl. Band 2 Mean Wetland Refl. Band 5 Range <= 11 High (100%) 33 Low (84%) 5 High 6 Low (83%) 1 High > > > > 10.903 <= <= 49.500 Wetland Refl. Band 2 Std. Dev. > <= 8 Low (87%) 1 High 1.580

PAGE 133

116 Table 4-6. Important variables for the center soils classification tree using spectral data. Variables used in decision nodes are designated with *. Predictor Band Statistic Relative Importance *Wetland reflectance 6 std. dev. 100.00 |||||||||||||||||||||||||||||||||||||||||| Wetland reflectance 5 std. dev. 48.06 |||||||||||||||||||| Wetland tasseled cap 3 -std. dev. 44.16 ||||||||||| ||||||| *Upland buffer reflectance 2 mean 43.12 |||||||||||||||||| *Wetland reflectance 4 std. dev. 37.45 ||||||||||||||| *Wetland reflectance 5 range 33.33 ||||||||||||| Wetland reflectance 6 range 33.02 ||||||||||||| Wetland reflectance 3 std. dev. 32.05 ||||||||||||| *Wetland reflectance 2 std. dev. 24.00 ||||||||| Wetland reflectance 4 range 13.32 ||||| Upland buffer reflectance 1 mean 12.81 ||||| Upland buffer reflectance 3 mean 12.52 |||| Upland buffer reflectance 6 mean 6.98 || Table 4-7. Prediction success table for the center soils classification tree. Predicted Predicted Actual Class Total Cases % correct Class low Class high Learning data (overall success rate 86.0%) low 65 89.2 58 7 high 52 82.7 9 43 Test data (overall success rate 75.6%) low 65 80.0 52 13 high 52 71.1 15 37 Figure 4-19. Classification tree for sampled wetland edge soils using Landsat7 data. White boxes are decision nodes based on Landsat7 ETM+ spectral data, and colored boxes are terminal nodes cla ssifying a percentage of learning samples (actual classes are listed) as having high (orange) or low (blue) TP (mg kg-1). Overall cross-validation accuracy is 75.8%. Wetland Reflectance Band 4 Max <= Upland buffer Tasseled Cap 1 Mean 41 Low (88% ) 3 High 37 High (76%) 21 Low 129.913<= > >124.500<= 13 Low (78%) 2 High

PAGE 134

117 Table 4-8. Important variables for the edge soils classification tree using spectral data. Variables used in decision nodes are designated with *. Predictor Band Statistic Relative Importance *Wetland reflectance 4 max. 100.00 |||||||||||||||||||||||||||||||||||||||||| Transition buffer reflectance 4 mean 77.13 |||||||||||||||||||||||||||||||| Wetland tasseled cap 2 -m ean 57.39 ||||| ||||||||||||| |||||| Wetland tasseled cap 2 -max. 50.12 | |||||||||||||| |||||| Wetland reflectance 4 mean 46.91 ||||||||||||||||||| *Upland buffer tasseled cap 1 -mean 38.99 |||||||||||||||| Transition buffer tasseled cap 2 -mean 35.53 |||||||||||||| Upland buffer reflectance 3 min. 11.81 |||| Transition buffer reflectance 3 min. 8.41 ||| Upland buffer tasseled cap 1 -max. 4.93 | Table 4-9. Prediction success table for the e dge soils classification tree using spectral data. Predicted Predicted Actual Class Total Cases % correct Class low Class high Learning data (overall success rate 80.1%) low 7572.05421 high 4288.1537 Test data (overall success rate 75.8%) low 7570.75322 high 4281.0834 Figure 4-20. Classification tr ee for 117 sampled wetland center soils. White boxes are decision nodes based on non-spectral da ta, and colored boxes are terminal nodes classifying a percentage of learni ng samples (actual classes are listed) as having high (orange) or low (blue) TP (mg kg-1). Blue nodes classify cases as low, orange as high. Overall cross-validation accuracy is 62.3%. Upland buffer Soil Subgroup Wetland Perimeter 44 Low (66.7% ) 22 High 26 High (72.6% ) 11 Low 323.24 <= > 10 Low (69.1%) 4 Hi g h Arenic Alaquod, Typic Argiaquoll, Arenic Argiaquoll, Typic Endoaqualf, Spodic Psammaquent Aeric Alaquod, Arenic Ochraqualf Terric Medisaprist

PAGE 135

118 Table 4-10. Important variables for the cente r soils classification tree excluding spectral data. Variables used in decision nodes are designated with *. Predictor Relative Importance *Upland buffer majority subgroup 100.00 |||||||||||||| |||||||||||||||| |||||||||||| Wetland size 50.11| |||||||||||||||||||| *Wetland perimeter 32.50||||||||||||| Vegetation type + land use 20.71|||||||| Upland buffer majority soil order 13.96||||| Basin 4.26| Min. distance to major road 0.6 Table 4-11. Prediction success table for th e center soils classi fication tree excluding spectral data. Predicted Predicted Actual Class Total Cases % correctClass lowClass high Learning data (overall success rate 66.6%) low 65 83.15411 high 52 50.02626 Test data (overall success rate 62.3%) low 65 78.55114 high 52 45.22824 Figure 4-21. Classification tr ee for 117 sampled wetland edge soils. White boxes are decision nodes based on non-spectral da ta, and colored boxes are terminal nodes classifying a percentage of learni ng samples (actual classes are listed) as having high (orange) or low (blue) TP (mg kg-1). Overall cross-validation accuracy is 73.3%. Priority Basin Distance to High Intensity Dairy Area 46 Low (81.1%) 6 High 10 High (71.8%) 7 Low 1,849.35 m <= > 22 High (92.9%) 3 Low S-154, S-65D S-191, S-65E Distance to Major Road 19 Low (72.7%) 4 High > <=1,601.84 m

PAGE 136

119 Table 4-12. Important variables for the edge soils classification tree excluding spectral data. Variables used in decision nodes are designated with *. Predictor Relative Importance *Priority basin 100.00 |||| |||||||||||||||| ||||||||||| *Min. distance to high intensity dairy area 36.52 ||||||||||||||| Vegetative community + land use 15.46 |||||| Upland buffer major component of most common map unit 15.17 |||||| *Min. distance to major road 13.45 ||||| Upland buffer subgroup 12.03 |||| Upland buffer major land use 9.69 ||| Wetland perimeter 1.75 Table 4-13. Prediction success table for th e edge soils classifi cation tree excluding spectral data. Predicted Predicted Actual Class Total Cases % correctClass lowClass high Learning data (overall success rate 81.5%) low 75 86.76510 high 42 76.21032 Test data (overall success rate 73.3%) low 75 80.06015 high 42 66.71428 Table 4-14. Predicted TP and SE (kg) storage in surface soils (0-10 cm) and descriptive statistics of unsampled historically isolated wetlands by size and land use. Land Use Total TP (kg) in soil (0-10 cm) Area (ha) mean SE Area (ha) median Area (ha) total DAIR 350,967 62,131 2.11 6.09 0.69 899 IMP 1,908,698 289,410 1.96 4.81 0.71 6,337 UNIMP 476,897 49,025 3.93 33.54 0.81 2,127 Total 2,736,563 400,568 9,363

PAGE 137

120 h, hh, ll, hl, ledge, center class 250 500 750 1000 1250C o u n t 143 40 122 121DAIIMPUNI h, hh, ll, hl, ledge, center class 540 426 722 1389 h, hh, ll, hl, ledge, center class 40 51 83 367 A h, hh, ll, hl, ledge, center class 0 10 20 30C o u n t 11 2 3 5 D I U h, hh, ll, hl, ledge, center class 13 21 12 38 h, hh, ll, hl, ledge, center class 2316 B Figure 4-22. Classification di stributions of TP (mg kg-1) in wetland hydrologic zones. Each bar represents a different combination of high (h) or low (l) TP in edges and centers by land use. A) Pr edicted TP in all wetlands. B) TP measured in sampled wetlands. DAI = dairies, IMP = improved pastures, UNI = unimproved pastures / rangeland.

PAGE 138

121 nonLandsat area dairies improved pastures unimproved pasture / rangelandTP class by zone low center, low edge low center, high edge high center, low edge high center, high edge 01020 5 KilometersLake Okeechobee Figure 4-23. Unsampled historically is olated wetlands indicating TP (mg kg-1) class in center and edge soils as pr edicted by two sets of classification trees (inside and outside of Landsat7 ETM+ image area ). High center TP is defined as greater than 620 mg kg-1 and high edge TP as greater than 250 mg kg-1. Inset bar graph shows wetla nd counts within each basin by zonal TP classes. Basin S-65E Basin S-154 Basin S-191 Basin S-65D S-154S-191S-65DS-65EBasin 0 500 1000 1500C o u n t

PAGE 139

122 dairies improved pastures unimproved pasture / rangeland priority basinsMean g/m^2 in top 10 cm 19 20 21 27 28 51 52 55Lake Okeechobee A Figure 4-24. Predicted TP (g m-2) in the surface soils (0-10 cm) of historically isolated wetlands in the priority basins of the Lake Okeechobee watershed. A) Southeast portion. B) Northeast portion.

PAGE 140

123 01020 5 Kilometers B Figure 4-24. Continued. Discussion Prediction of TP (mg kg-1) magnitude in surface soils of historically isolated wetlands in the study area was successful using satellite imagery and cl assification trees. Other studies have used other statistical techniques to relate environmental variables to soil nutrient conditions (Numata et al., 2003; Ba llester et al., 2003; Asner et al., 1999).

PAGE 141

124 Since soil and water reflectance characteristics are important indicators of TP condition and land use intensity in wetlands, using a sp ringtime image may have had an advantage over a summertime image, where vegetation might obscure wet soils and standing water (Ozesmi and Bauer, 2002). Studies have used remotely sensed data in classification and regression trees to predict fore st cover types and soil properti es (Xian et al., 2002; Bishop and McBratney, 2001). Accounting for the differences between the wetland edge and center soils made the model more robust. The overall cross-valida tion accuracies of th e trees using Landsat7 ETM+ spectral data were quite high. Other studies using classifi cation and regression trees with environmental variables reported similar accuracies (Rouget et al., 2003; Pan et al., 1999). Prediction within the Landsat7 ETM+ Extent Total P condition in center soils depended on the variability of spectral data in the wetlands and central tendency of spectral data in the uplands. More variability in bands 4 and 2 in the wetlands (indicative of green vege tation), was predictive of higher center TP for most of the wetlands. In Chapter 3, the centers of wetlands with more intense ditching were shown to have significantly hi gher TP. Ditches ma y prevent the uniform distribution of nutrients a nd water causing the centers to have more heterogeneous vegetation patterns. Higher TP areas ma y also experience more grazing and the associated wetlands may have areas that cattle prefer making the vegetation more heterogeneous. In some of the unimpr oved emergent marsh wetlands visited, the vegetation cover was observed to be low re lative to more improve d areas, so spectral variability of bands registering vegetati on in those wetlands would be low.

PAGE 142

125 Lower mean green reflectance (band 2) in uplands was predictive of lower center TP for most of the wetlands. It is hypothesi zed that more fertiliz ed and grazed pastures are more homogeneous and green and this is supported by field observations. More variability in bands 5 and 6 in the wetland (soil and vegetation moisture) was predictive of lower center TP for many of the wetlands. This could be indicative of more soil disturbance by cattle, causing variability in soil moisture, and it could be indicative of the variability in non -photosynthetic vegetation (N PV). Non-photosynthetic vegetation represents the pr evious yearÂ’s green vegeta tion (Asner et al., 1998), and reflects in the shortwave infr ared region of the electroma gnetic spectrum (Asner et al., 1999) (recorded by bands 5 and 6 of Landsat7). The image used in this study was from the end of the dry season, when standing d ead wetland vegetation was greater than the new growth for certain species. High variability in these bands may be a signature that represents a mix of species typical of lower nutrient wetlands. Less wetland vegetation was predictive of lower edge TP, and higher upland soil brightness was predictive of higher edge P. Nutrients in agricultural runoff from uplands are known to enhance plant growth in rece iving wetlands, resulting in a higher total percent cover of species (G ustafson and Wang, 2002). Hi gher upland soil brightness (represented by TC1) is indicative of lower organic matter and lower moisture (Curran, 1985). More intense grazing may expose more soil by the shortening and removal of grass. More intensely managed pastures will be better drained than less improved pastures, and have lower moistures and lower organic matter due to oxidation in the drier conditions.

PAGE 143

126 The vegetation indices NDVI and MSAVI were not predictive in the classification trees. These indices only use the red (band 2) and near-infrared bands (band 4) where the contrast between vegetation and soil is at a maximum (Lillesand and Kiefer, 2000). Based on the results of the trees, soil and ve getation moisture were important factors in wetland soil TP prediction, and th ese are not registered in thes e vegetation indices. Asner et al. (1999) reported that NDVI may not inte grate senesced plant materials during drier periods, so may not be good indicators of the amount of total vegetation present. They also found that P availability in pastures was not correlated to NDVI during drier periods, but was correlated with leaf area and NPV. The spectral reflectance curves for soil and litter have similar shapes in the visible and near infrared wavelength ranges (.4 to 1.1 micrometers) (Nagler et al., 2000; Aase a nd Tanaka, 1991). The MSAVI was developed to account for soil backgrounds (Qi et al., 1994) but van Leeuwen and Heute (1996) reported that soil adjusted vegetation indice s (SAVIs) did not respond to emerging green vegetation combined with leaf litter. Prediction outside the Landsat7 ETM+ Extent The soil subgroup of the major component of the most common map unit in the upland buffers was the most important vari able, followed by wetland size in center soil classification trees for wetlands outside of the Landsat7 ETM+ extent. Other studies have shown predictive relationships between soil properties and environmental variables (McKenzie and Austin, 1993; Odeh et al., 1994). Most of the soils surrounding sampled wetlands were Aeric Alaquods (mostly Mya kka series), Arenic Alaquods (mostly Immokalee and some Waveland series), and Spodic Psammaquents (mostly Basinger and some Valkaria series) (Figure 4-25).

PAGE 144

127 Low HighTP Class A e r i c A l a q u o d A r e n i c A l a q u o d A r e n i c A r g i a q u o l l A r e n i c O c h r a q u a l f S p o d i c P s a m m a q u e n t T e r r i c M e d i s a p r i s t T y p i c A r g i a q u o l l T y p i c E n d o a q u a l fSoil Subgroup 10 20 30 40C o u n t n=28 n=20 n=14 n=28 n=1 n=2 n=2n=6 n=13 n=1n=1n=1 Figure 4-25. Total P class di stribution of sampled center soils among soil subgroups of major upland components. The majority of the sampled wetlands w ith upland Immokalee soils had lower TP than the 620 mg kg-1 cutoff, whereas most of the wetla nds with upland Myakka soils had higher TP. Immokalee soils have faster permeability and deeper Bh horizons than Myakka soils (Lewis, 2000) thus more P may infiltrate the soil and less may enter the wetlands from surface runoff. All wetlands wi th upland Valkaria soils had low TP and wetlands with upland Basinger soils were even ly split. These Entisols occur in sloughs and depressions and poorly defined draina geways (Lewis, 2000) which may preclude them from being optimal grazing areas compared to Alaquods. Soil subgroup was a better splitter in the tree logic than land use because there were larger differences in the nu mber of high and low cases among soil types compared to land uses (Figure 4-26). Among land uses, low and high TP were more balanced so it was not

PAGE 145

128 a definitive splitter. Based on the analysis in Chapter 3, this may be caused by interacting factors such as wetland area a nd vegetation community, as well as the high TP variability measured in each land use. Though the center soil tree using soil subgroup had an overall cross-validation accuracy of 62.3%, the classifi cation success rate of high TP center soils was low (45.2%). This shows that spectral da ta variables are more reliable for predicting TP condition. low highTP Class dairyimprovedunimprov Land Use 0 25 50 75C o u n t n=13 n=8 n=34 n=50 n=5 n=7 Figure 4-26. Total P class distribution of sampled center soils among land uses. Wetland size was also an important variable with larger wetlands being classified as having higher TP (mg kg-1). This contradicts a conclusi on from Chapter 3 that larger wetlands have less TP, but this is because of an unbalanced sampling design according to wetland size. A disproportionate amount of small wetlands were sampled, which decreased the average and median within each class (Figure 4-27). It is probable that smaller wetlands in improved pastures were pr eviously drained completely, so that they are no longer wetlands. This is supported by the fact that in unimproved areas, the sampled wetland median size is lower than in dairies and improved pastures (Table 3-9).

PAGE 146

129 Also, field observations were made of severa l wetlands that had been mowed; possibly in an attempt to convert them to pasture land use. For edge soils outside the Landsat7 ETM + image area, locati on within the study area (defined by sub-basin) and landscape metr ics were the most important variables. Wetlands closer to dairy HIAs were predicte d to have higher TP. The decision node for distance from major roads only applies to cases that are farther away from dairies; those that are further from major roads, had higher TP This may be a coincidence in the data for those 40 cases. 25050075010001250Perimeter (m) 0 1000 2000 3000 4000C e n t e r S o i l T P ( m g / k g ) Figure 4-27. Scatterplot of TP (mg kg-1) in sampled center soils by wetland perimeter. Includes wetlands with major comp onent subgroups = Aeric Haplaquod, Arenic Ochraqualf and Terric Medisapris t in the uplands. Position of cutoff and medians is approximate. It is logical that the spectral data was th e most important predictor of TP since it had the highest spatial and tempor al resolution of any of the layers used in the analysis. The Landsat7 ETM+ image had 30 m pixels (0 .09 ha each), providing information at a 1:5,000 scale (Soil Survey Staff, 1993). The land use data was developed at a 1:24,000 High / low TP cutoff (620 mg kg-1) High TP median perimeter (394 m) Low TP median p erimeter ( 357 m )

PAGE 147

130 scale and was known to be erroneous base d on comparisons to DOQQ images and field observations. Soil data, which was shown to be an important predictor, was created at a similar scale as land use, but is temporally more accurate than land use which changes much more quickly. Land use is also a mu ch coarser categorical variable than soil taxonomy, as it groups many levels of land use in to a single class (discussed in Chapter 3). Attempting to explain the relationshi ps between landscape variables and P condition in wetlands is diffi cult, and biogeochemical differences between and within ecosystems are spectrally subtle (Asner et al., 2000). Little literature is available that correlates wetland nutrient condi tion in wetlands with satell ite reflectance data. The spectral data statistics of wetlands and uplands used in this study represent complex ecological systems with many interacting fact ors which are difficult to mechanistically describe. Storage Total P Storage maps showed that ba sin S-191 had the highest TP in wetland surface soils. This agrees with reports that basin S-191 had the highest P concentrations measured in surface waters by the SFWMD comp ared to the other three basins (USACE and SFWMD, 2003). These basins contain the majority of dair ies in the four basins as well as several retired dairies. A surprising result is the high TP measured and predicted in wetlands in the northeastern portion of basin S-65D. This area has no dairy land use, a lot of wetlands, and a large area of unimprove d pastures and rangela nds. It is possible that there is row crop activity, but this is an area that deserves further investigation.

PAGE 148

131 Future Research Since spatial patterns of na tural and anthropogenic feat ures influence ecological characteristics of a landscape (Walsh et al ., 1998; Risser et al., 1984), the pattern of historically isolated and sometimes connected wetlands could be ch aracterized and made part of a model to better dete rmine effects on TP storage. Yin et al. (2003) found that patterns of land use within la rger rather than smaller buffer areas around sample sites were more predictive of water quality parame ters. Future research should include multiple buffer sizes around wetlands in order to de tect regional as well as local effects. At the regional scale, watersheds are often nested components of larger drainage basins, suggesting nutrient export coefficients will change from one sub-watershed to the next (Whickham et al., 2003) so that the upla nd area surrounding a wetland as well as the number of wetlands connected to this wetland may be predictive of TP stored in the wetland. Landscape indicators incorporating hist orical land use may hold promise for predicting and assessing the status of all wetl and systems (Gergel et al., 2002). Rate of fertilization, land and history could be impor tant at the watershed scale. These were found to be important in fiel d studies by Mander and Kuusem ets (2000) and Steinman et al. (2003). Asner et al. (1999) found that spectral estimates of leaf area and NPV of pastures were correlated with soil P concentrations in the central Amazon. Future studies might include information about NPV in pastures and wetlands as well as improved registration techniques to align digital spectr al information to the study area. In order to obtain better TP predictions within wetlands, more sample collection is recommended. Using classification trees w ith categorical indepe ndent variables and

PAGE 149

132 small datasets tends to find relationships in the data that are coincidental with that data. The method should be used with caution a nd datasets of over 200 data points was recommended for classifying samples into two classes (Steinberg and Colla, 1995). It is also recommended that an independent dataset of wetland soils be collected to serve as a validation dataset for determining classification tree accuracy. Conclusions Remotely sensed satellite spectral data and classification trees were useful for predicting phosphorus conditions in historica lly isolated wetlands within pasture and dairy land uses in the Lake Okeechobee priori ty basins. Spectral data provide high spatial resolution information that produced more accurate TP predictions compared to categorical, coarser-scale datasets such as soils, land use and NWI, or quantitative information about wetland size and distances to landscape features. Accounting for the differences in TP storage between wetland ce nters and edges in th e classification trees provided more robust predictions. The high spatia l variability of TP measured in wetland soils reflects the complexity of wetland ecosystems that precludes the prediction of more explicit TP values at watershed scale. Basin S-191 had the most dairy land use and the highest proportion of predicted high TP wetland soils (as expected). Surpri singly high TP values were predicted in the northeastern part of basin S-65D. This basin lacks any obvious reasons for these elevated TP values so meri ts further investigation. Information from the synoptic sampling of th ese historically isolated wetlands and this upscaling analysis can be used by land managers to estimate how much additional P may be stored after hydrologic restoration of th ese historically isolated wetlands. Future work should include the quantification of NPV using spectral data, improved

PAGE 150

133 georectification techniques, and the addition of more la ndscape metrics to quantify wetland spatial patterns and their hydraulic relationships. Additionally, sampling an independent, separate set of we tland soils to validate the predictions of the classification tree models, would enhance the r obustness of prediction models. Though predictive modeling techniques are le ss reliable than field observations, they provide a toolset for holistic modeling of TP in wetland soils at watershed-scale, while taking into account the variability of wetland and upland characteristics. Fieldbased observations are labor intensive, costly and they constrain the number of sampled wetlands. Statistical models provide a cost -effective method to predict wetland soil TP throughout a watershed by combining field obs ervations, remotely sensed datasets, and ancillary spatial datasets.

PAGE 151

134 CHAPTER 5 SYNTHESIS AND IMPLICATIONS Multiple interrelated factors influence P stor age in historically isolated wetlands of the Lake Okeechobee priority basi ns at fieldand watershed-scales. Land use, hydrologic connectivity, wetland hydrologic conditions, and wetland size he lp to explain soil TP storage variability. Unimproved pasture / ra ngeland land use provide d the lowest P input into wetland ecosystems, with higher P inputs from dairies and improved pastures. This suggests that agricultural BMPs within dairies and improved pa stures will con tinue to be important for mitigating P inputs to surface waters (USACE and SFWMD, 2003). More drainage has been installed within these two land uses, and since it was shown that larger ditches are related to more TP stored with in wetland centers, they represent opportunities for hydrological restorati on of inflowing ditches to isolated wetlands. According to Reddy et al. (1996a), uplands of dairies and pastures in the Lake Okeechobee watershed have 75% of their P st orage capacity remaining, compared with 45% remaining in wetlands and streams. A recommendation from Flaig and Reddy (1995) for reducing P loads in surface waters of the watershed was to retain more P in upland soils. Hydrologic rest oration should aim to convert small wetlands into larger ones, extending the total wetland area within th e priority basins. While the restoration of ditches is expected to keep some P within upland soils, the expans ion of wetlands will convert some upland soils to wetland soils. Since upland soils have more P retention capacity remaining than wetlands, their conversion to wetlands will provide the opportunity for P in surface waters to reside over these soils for longer periods, allowing

PAGE 152

135 for more P-binding processes to occur. In creased hydraulic retention times in wetlands are another expected effect of hydrological restoration. By increasing their size, wetlands would acc rete more organic matter thus storing more organic P in undecomposed vegetation (Richardson, 1999) as well as phosphate ions bound to organometallic complexes (Petrovic and Kastelan-Macan, 1996; Richardson, 1999). Aluminum and Fe in more flooded wetland edges would become more amorphous over time binding more i norganic P (Patrick and Khalid, 1974; Richardson, 1985; Reddy et al., 1995; Rhue and Harris, 1999). Because organic matter is an important mechanism for P storage in historically isolated wetlands of the priority basins, it is worth considering mana gement practices that enhance vegetative growth. Cows and calves ma y enjoy the cooling benefits of wetlands, but the physical impacts may compromise ve getation (Clary, 1999; Harris and Asner, 2003), as well as grind and resuspend organic matter that may release P to surface waters. Vegetation around wetlands reduces sedimenta tion within them, s uggesting that more particulate P would remain stored in the uplands (Fiener and Auerswald, 2003; Hook, 2003). Perhaps the economic investment put into the maintenance of ditches could be shifted to enhance wetland func tions. This could include pr oviding other cooling options for cattle (mesh structures or trees for shade) or installing fencing around wetlands, popular to cattle, to exclude them. Smaller historically isolated wetlands store more P per unit area than larger wetlands, hence, it is possible that they have less P storage capacity remaining compared to larger sites which have more P binding site s overall. Smaller wetlands with high soil TP concentrations represent a risk of bei ng a P source if P concentrations of incoming

PAGE 153

136 waters are relatively low (Reddy et al., 1996a ). This can happen during a rain event causing a shift in P concentrations so that P-laden soils release P to surface waters. If an wetland enlarged from hydrological restorat ion receives the same P loads as before restoration, the P stored per unit area may d ecrease as the P coming into the system is distributed among more available P-binding site s in the larger wetland. Larger wetlands pose a smaller risk of becoming P sources. Another benefit to hydrological restoration is the potential reduced water flows to the lake. This is an important considerat ion, since flood control and the attenuation of peak water flows to Lake Okeechobee are go als of the Lake Okeechobee Protection Plan (SFWMD et al., 2004a). It is recognized that the hydrologic rest oration of historica lly isolated wetlands comes with associated costs: losing pasture ar ea and implementing infrastructure to keep cattle out of wetlands. Best management pr actices are currently being implemented in dairies and cattle ranches to reduce the amount of P going into surface waters from point and non-point sources. These are part of land management programs that involve economic subsidies for landowners who implem ent these BMPs. This research, as well as other studies related to the hydrologic rest oration of historically isolated wetlands, will be part of new BMP proposals designed to help landowners reduce P loads going to Lake Okeechobee while being economically feasible and incentive-based (SFWMD et al., 2004a). The Wetlands Reserve Program (WRP) is an example of a cost-sharing program that is available to landowners for wetland e nhancement. It is offered by NRCS, and is authorized by the 2002 Farm Bill. This is a voluntary program that provides technical

PAGE 154

137 and financial assistance to landowners who rest ore, protect, and enhance wetlands within agricultural land uses. There are 3,900 historically isolated wetlands within IM P areas that cover almost 6,000 ha. These wetlands contain about two th irds of the TP stored in historically isolated wetlands of the dairies and beef-cattle pastures of the priority basins. Future studies should focus on these improved pastur e wetlands, because they account for 70% of the pasture and dairy wetlands in the prior ity basins, and have the most opportunity for hydrological restoration. Many best manageme nt practices have been put into place in dairies to mitigate P losses to waterways, but comparatively few have been implemented on improved pastures (SFWMD et al., 2004a). A predictive model was developed in Ch apter 4 that accounts for TP and BD differences between wetland center and edge soils. Classification trees of this model used watershed-scale spatial and spectral da ta to predict if a wetland had relatively high or low TP, and mean sampled TP and BD values per class were used to determine TP storage. This model will be referred to as Model1. For comparison, another model that simply multiplies the total area of land use by the average TP and BD within that land use is considered (Model2). Both models were us ed to calculate total kg of P stored in the historically isolated wetlands of the priority basins (Figure 5-1). Though both methods yielded similar overall results for TP stor age in the four basins, Model1 is preferred over Model2. Model1 considers diffe rences between wetland edges and centers (TP and areal coverage). Mo reover, it was shown in Chapter 3 that the difference between DAIR and IMP wetland TP storage was not significant, so a model relying on land use types is not robust. In Ch apter 4, the prediction of TP condition (high

PAGE 155

138 or low) was shown to statistically rely on spectral information which has high spatial and temporal resolution. Finally, a hypothetical scenario presented belo w shows that the two models contrast significantly when considering a single-land use type. A hypothetical scenario of the restoration of 100 artificially drained IMP wetlands that are one ha in size, with 25% of the area being edge, and both hydrologic zones having “high TP,” is considered here. If th ese wetlands increased in size by 5%, storage could change from 540,750 kg in those 100 ha, to 567,790 kg (according to Model1). This amounts to an additional 2,700 kg TP that c ould be stored in the top 10 cm of soil. Using Model2, the additional storage calcu lated is only 985 kg P. This scenario demonstrates that the two models vary in thei r predictions, despite the similar totals that they predicted for the population of histor ically isolated wetla nds in the study area. Most notable, is the difference in uncertainty between the two models. The uncertainty of Model 1 is much smaller wh en compared to Model2, due to the large variability of TP within each land-use class. Model 1 showed that spectral reflectance of wetland and surrounding upland areas was more important for predic ting wetland soil TP than soil taxonomy, individual wetland charac teristics, landscape metrics, and land use layers. It is robust, and met hods used to develop it can be applied in other watersheds. Moreover, Landsat7 ETM+ images are relatively inexpensive to acquire, and documentation for processing the imag es is readily available. Site-specific investigations can only provide estimates for a number of limited sites whereas a landscape-scale predic tive model, like the one developed in this thesis, can address the land resource management issues of P storage at watershed scale. In addition to developing the TP prediction model, this thesis summarized important information

PAGE 156

139 about the wetlands of the four priority basins that was not available before. This includes spatial distribution of wetlands, wetland size distribution, wetland vegetation community distribution, spatial differe nces of TP storage within wetlands (edge and center hydrologic zones), and types of metals and phosphorus in the soils with comparisons among land uses. It is hoped that this information will assist scientists and land managers to investigate the effects of ditches an d hydrological restoration on TP storage and retention in future field-scale studies, and to investigate wetl and areas that were predicted to be storing high amounts of TP in surface soils. Figure 5-1. Total P storage by land use: contrasting models. A) Calculations using land use means and total ha (Model2). B) Calculations using classification trees for edge and center soils (Model1). Total using land use means (Ch. 3) = 2,736,400 kg (928,473 to 4,544,327 kg)A Unimproved Pastures / Rangeland 197 94 kg ha-1 2,127 ha = 419,019 kg (219,081 to 618,957 kg) Dairy Land 456 400 kg ha-1 899 ha = 409,944 kg (50,344 to 769,544) Improved Pastures 301 197 kg ha-1 6,337 ha = 1,907,437 kg (659,048 to 3,155,826) Total using upscaling model means (Ch. 4) = 2,736,562 kg (2,335,995 to 3,137,131)B Dairy Land 350,967 62,131 kg (288,836 to 413,098) Improved Pastures 1,908,698 289,410 kg (1,619,288 to 2,198,108) Unimproved Pastures / Rangeland 476,897 49,025 kg (427,872 to 525,922)

PAGE 157

140 APPENDIX A SPATIAL DATA METADATA Table A-1. Projection inform ation for GIS data layers. Projection Albers Conical Equal-Area Units Meters Datum HPGN (NAD83) Spheroid GRS1980 1st standard parallel 24 00 0.000 2nd standard parallel 31 30 0.000 Central meridian -84 00 0.000 Latitude of projection's origin24 00 0.000 False easting (m) 400000.00000 False northing (m) 0.00000 Watershed boundaries. This layer delineates the four basins of the study area as determined by the Florida Department of Environmental Protection (FDEP) and was received from the South Florida Water Mana gement District (SFWMD) in September, 2002. County boundaries. This layer contains Florid a county boundaries created by the US Census Bureau in 1990 from TIGER/Li ne files at a 1:100,000 scale. It was downloaded from the Florida Geogra phic Data Library (FGDL) website ( www.fgdl.org ) in September, 2002. Hydrography: main tributaries, streams, canals. These layers were created by the U.S. Geological Survey (USGS) from 1:24,000-scale digital lin e graph (DLG) files which originated from aerial photos and 1:20,000-, 1:24,000-, and 1:25,000-scale 7.5minute topographic quadrangle maps. The data were last updated in 1994-1995. The

PAGE 158

141 layers (hysurcrc.shp, hysurwtl.shp and midhydro. shp) contained data for the four priority basins and were received from the SFWMD in September, 2002. Ditches and streams. This layer was created by the USGS based on 1:24,000 DLGs. The hy24l*.* line shapefiles delinea ted many ditches and streams for each county. These were downloaded from the FGDL website ( www.fgdl.org ) in June, 2004. Land use. Mock, Roos & Associates, Inc., subcontracted by the South Florida Water Management District (SFWMD), crea ted a land use polygon coverage using the SFWMD 1:24,000 1995 land use coverage as a base layer. Changes reflected 2001 conditions based on input from SFWMD Ok eechobee Service Center staff, field observations and examination of 1:12,000 colo r infrared orthophotogr aphy. The land use layer employs three levels of land use codes according to th e Florida Land Use and Cover Classification System (FLUCCS) developed by the Florida De partment of Transportation (FDOT, 1999). The Florida Department of Ag riculture and Consumer Services (FDACS) with Mock-Roos Consulting created a separate dairy land use layer in 2002 with four levels of FLUCCS codes identifying pastures barn areas, ponds, fields etc. These two layers were received from the SFWMD in September, 2002. In October 2003, an update of non-dair y land uses was supplied by the SFWMD which had noticeable improvements to impr oved pastures based on verification with DOQQs and sampling field notes. National Wetland Inventory (NWI). This 1:24,000 polygon layer provided the basis of spatial wetland information in this study. It was generate d from digital line graph (DLG) files between 1988 and 1993 by the US Fish and Wildlife Service (USFWS, 2002). The files were based on 1:58,000 colo r infrared aerial photography or 1:24,000

PAGE 159

142 natural color aerial photography capture d between 1972 and 1984. The NWI polygon shapefiles for each of the four counties we re downloaded in September, 2002 from the Florida Geographic Data Library (FGDL) website ( www.fgdl.org ). Digital Orthophoto Quarter Quadrangles (DOQQs). Forty-five 1m resolution DOQQs corresponding to the study area were downloaded from the FDEP Land Boundary Information System (LABINS) website ( data.labins.org ). Boundaries of these aerial, geo-rectified, color-infrared images correspond to USGS topographic quadrangles. The images were produced by the USGS in 1999. Soil Survey Geographic (SSURGO) data set. This dataset (version 3.0, July 2000, scale 1:24,000) is a digital represen tation of the County Soil Survey maps published by the US Department of Ag riculture (USDA) National Resource and Conservation Service (NRCS) as the Soil Su rvey Geographic (SSU RGO) data set. Source data is from 1990. Detailed inform ation about the SSURGO dataset can be found in USDA/NRCS (1995). Soil polygons (sso ils*.*) and the soil component tables (comp*.*) for each of the four counties were downloaded from the FGDL website ( www.fgdl.org ). Roads. A 1:24,000 shapefile of major roads (m ajrds*.*) was developed by Florida Department of Transportation. A more de tailed 1:24,000 layer of roads (rds24*.*) was developed by the Earth Science informati on center of the U.S. Geological Survey. Layers for the four counties were download ed from the FGDL website in September, 2002.

PAGE 160

143 APPENDIX B GLOSSARY OF GIS TERMS Albers conical equal-area a map projection that relate s a global map onto a conical surface that intercepts the globa l sphere two times at 2 sta ndard parallels thus reducing distortion compared to a cone that w ould be tangential at one parallel. Arc: A two dimensional feature representing a line such as a stream or road. Attributes: A set of characteristics of a feature described by numbers, characters etc. typically stored in tabular format in a table and linked to the feature by a unique identifier (e.g., the attributes of a land use polygon might include land use type and area). Attributes may also be referred to as fields or columns. Buffer: A zone of a specified distance around a se t of features. Buffers are useful for proximity analysis (e.g., find all wetlands w ithin 50 meters of a canal). Buffers are created in ArcMap using the buffering tool. Clip: The extraction of a subset of one layer using one or more of the polygons in another layer as a "cookie cutter". This is particular ly useful for creating a new layer that is a subset of the features in a nother larger layer. The Geopr ocessing wizard in ArcMAP is used to clip. Extent: Geographical area covered by a spatial data layer. Feature: The conceptual representation of a geographic feature. When referring to geographic features, feature cl asses include points, lines (arcs) and areas (polygons). Field calculator: A feature in ArcMAP that performs a mathematical calculation to set a field (attribute) value for a si ngle record or all records. HPGN: High Precision Geodetic Netw ork, also referred to as High Accuracy Reference Networks (HARNs). NAD 83 coordinates based on the HPGN/HARN surveys changed approximately 0.2 to 1.0 meter relative to the original NAD 83 (1986) adjustment. Intersect: An operation in the ArcMap Geoprocessi ng wizard that integrates two spatial data sets (shapefiles) preserving features th at fall within the area common to both input data sets. Join: The relating of two or more attribute tables on the basis of a common spatial location (spatial join) or attribute.

PAGE 161

144 Layer: A collection of similar geogr aphic features (such as we tlands, streams, land uses or counties) in a particular area or place re ferenced together for display on a map. Merge: An operation in the ArcMap Geoprocessi ng wizard that appends the features of two layers into one single layer, where featur es in each layer are mutally exclusive and of the same feature type (e.g., polygons or arcs). NAD83: North American Datum of 1983. Polygon: A two-dimensional feature representing an area (e.g., a county or wetland). Raster: A representation of spatia l areas by regularly sized ce lls in a matrix or grid pattern. They may be continuous throughout an extent or may be intermittent, representing local areas or sh apes. All cells of a raster hold a common type of value (text, integer, real etc.) and cell each holds a single value. Shapefile: ArcView’s GIS format for storing geom etry and attribute information about a set of features. Table: A set of data elements that has a hor izontal dimension (rows – records) and a vertical dimension (columns – attributes of each row) in a relational database system. Related to a feature layer is a tabl e and each row identifies one feature. Union: An operation in the ArcMap Geoprocessing wizard that combines all the features of two layers so that all features of both layers are retained. (Sources: Kennedy, 2001; ESRI, 2002; American Congress on Surveying and Mapping et al., 1994)

PAGE 162

145 APPENDIX C WETLAND VEGETATIVE COMMUNITY DISTRIBUTIONS Table C-1. All National Wetla nd Inventory polygons within the four priority basins by vegetative community (VC). VCa Polygon count Total Percent Mean Medianb Max Minb ----------------------------------ha ------------------------------------------AB 364 316.151.46 0.87 1.56 0.38 14.85 0.05 EM 7006 15,748.0772.74 2.21 8.23 0.73 320.43 0.04 FO 811 3,259.2215.05 3.94 10.10 1.34 179.39 0.07 OW 93 25.950.12 0.28 0.33 0.16 2.08 0.05 SS 441 1,551.237.17 3.44 6.17 1.21 55.06 0.08 UB 537 745.043.44 0.64 2.13 0.25 342.02 0.05 US 5 3.790.02 0.76 0.34 0.76 1.31 0.24 Total 9,257 21,649.45 a AB = aquatic bed, EM = emergent marsh, FO = forested, OW = open water, SS = scrub/shrub, UB = unconsolidated bottom, US = unconsolidated shore. b Calculated after wetlands that were partially cu t off by the clipping of NWI to the study area were deleted.

PAGE 163

146 APPENDIX D SAMPLED SITES Table D-1. Sampled wetland locations and selected characteristics ID Date sampled X Y No. ditches Ditch intensity Ditch class Veg. type Surr. land use Edge % 1 5/20/2003 701985.203 375244.349 1 2 H EM IMP1 40 2 5/19/2003 701103.971 382144.05 1 1 H EM IMP1 10 3 8/14/2003 719861.555 367546.736 0 0 I SS DAHA 20 4 8/14/2003 719809.932 368167.055 0 0 I EM DAHA 20 6 8/21/2003 725679.48 365906.194 2 3 F SS DASP 30 7 9/25/2003 728886.131 355088.813 3 2 F EM IMP1 50 8 9/25/2003 697036.839 380308.068 0 0 S EM UNI1 70 10 9/11/2003 705026.998 403184.624 0 0 S EM IMP2 10 65 7/11/2003 701185.571 384599.864 1 2 H FO UNI1 20 66 9/11/2003 704756.884 400639.033 1 2 H EM UNI2 10 72 7/15/2003 718419.362 383402.432 0 0 S FS DAHA 60 88 9/11/2003 703886.123 398957.214 2 3 H EM UNI2 20 101 7/16/2003 710171.955 381235.777 0 0 S FS DABE 90 133 7/30/2003 711100.895 379197.7 1 3 H FO DAHA 50 263 7/23/2003 712936.913 369211.907 1 2 H FO IMP1 20 273 8/6/2003 717288.042 368929.111 0 0 I FO IMP2 20 283 8/20/2003 717113.979 367915.545 1 2 H FO IMP1 10 342 8/20/2003 730159.079 358578.865 2 2 F SS IMP1 10 443 9/26/2003 701291.568 392462.592 1 2 H EM IMP1 10 478 8/7/2003 704818.207 392465.171 2 3 E EM IMP2 50 615 8/8/2003 689329.872 390243.608 0 0 S EM IMP2 20 696 9/26/2003 699594.315 388611.733 0 0 S EM IMP2 10 750 9/10/2003 694564.276 387753.832 0 0 S EW IMP2 60 756 9/10/2003 697301.327 387670.624 0 0 S EM IMP1 10 763 9/10/2003 692561.075 387576.183 2 1 F EM IMP2 20 805 9/10/2003 690001.384 386879.607 2 1 F EM IMP2 20 813 7/11/2003 710887.372 386323.94 0 0 I EW DAPA 40 818 9/10/2003 691898.021 386621.105 0 0 I EM IMP2 10 855 7/11/2003 710358.389 385915.182 0 0 I EW DAPA 40 962 8/6/2003 697513.213 384075.808 0 0 I EM IMP1 70 1017 7/15/2003 710423.401 383039.417 2 2 F FS DABE 40 1058 7/10/2003 701281.077 383159.998 1 2 H EM IMP2 30 1097 8/6/2003 697037.637 382208.782 3 2 E EW IMP1 30 1128 9/11/2003 688735.58 381663.489 1 2 E EM IMP1 10 1188 9/26/2003 708734.588 381072.005 0 0 I EM IMP1 10 1195 7/25/2003 707075.766 380989.959 1 3 E EM IMP1 10 1227 7/25/2003 704135.306 380701.364 1 2 H EM IMP1 10

PAGE 164

147 147 Table D-1. Continued. ID Date Sampled X Y No. Ditches Ditch Intensity Ditch Class Veg. Type Surr. Land Use Edge % 1315 7/30/2003 717391.067 379694.274 1 3 T EM IMP2 20 1424 7/25/2003 715458.304 378804.393 0 0 S EM IMP1 30 1506 9/10/2003 692306.15 378147.584 2 2 F EM IMP1 40 1609 9/24/2003 687186.084 377111.395 1 1 H EM IMP2 10 1692 9/24/2003 689081.467 376465.048 1 2 H EM IMP2 60 1780 7/24/2003 714901.059 375823.162 0 0 I EW IMP1 20 1791 7/25/2003 718847.126 375798.971 0 0 S EM IMP2 10 1809 8/6/2003 709777.182 375644.141 0 0 I EM UNI2 10 1815 9/18/2003 696730.691 375526.966 0 0 S EM IMP1 20 1843 7/23/2003 711864.361 375516.148 3 2 F EW IMP1 50 1852 7/24/2003 713827.557 375507.125 0 0 S EM IMP2 10 1873 8/1/2003 703677.622 375118.3 0 0 I EW IMP1 40 2005 7/17/2003 721090.043 374076.295 0 0 I EM IMP1 80 2045 8/8/2003 697269.236 373752.067 2 2 F EW IMP1 40 2097 7/18/2003 714115.291 373737.09 2 2 E EM IMP1 10 2120 7/17/2003 720931.616 373259.332 1 2 E EM IMP1 70 2146 7/17/2003 720558.818 373073.053 0 0 S EM IMP1 10 2174 7/23/2003 711054.455 372682.596 0 0 I EM UNI2 10 2189 8/9/2003 706651.266 372663.907 1 2 H EM DAUN 10 2295 11/19/2003 700304.021 370279.311 1 3 H EM IMP1 10 2301 8/10/2003 707296.766 372202.764 0 0 I EW DAPA 10 2309 7/18/2003 713082.634 372045.205 2 1 T EM IMP1 30 2314 7/31/2003 702778.244 372091.236 0 0 S EW DAUN 10 2420 9/24/2003 691286.668 371299.847 0 0 I EM IMP1 40 2441 7/31/2003 703623.08 371203.806 0 0 I EM DAUN 10 2445 7/17/2003 722341.183 371114.259 1 3 H EM IMP2 10 2452 7/24/2003 716063.818 370752.26 1 2 E EM IMP2 50 2460 7/31/2003 704338.539 371039.365 0 0 I EM DAUN 10 2470 8/6/2003 692058.615 370881.038 0 0 S EM DAUN 50 2535 7/16/2003 721809.274 370538.737 0 0 I EM IMP1 50 2573 7/24/2003 716022.36 370410.359 0 0 I EW IMP2 10 2578 9/18/2003 698480.134 369743.31 3 3 F EW IMP1 40 2601 7/17/2003 721219.662 369978.371 1 1 H EM UNI1 10 2609 8/1/2003 705105.848 369938.621 1 2 H EM IMP2 20 2620 7/16/2003 723710.956 369820.769 1 2 T EM IMP1 50 2699 7/22/2003 711135.657 369269.136 0 0 I EM UNI1 10 2725 8/10/2003 708359.682 368903.798 0 0 I EM DABE 10 2834 8/14/2003 701933.667 368213.832 2 2 H EM IMP2 10 2869 7/30/2003 723802.81 367860.359 1 2 H EM IMP1 40 3071 9/24/2003 691274.961 366569.273 2 2 F EM IMP1 30 3086 9/18/2003 702933.996 366399.242 3 1 F EW IMP2 20 3125 7/30/2003 723021.571 366067.685 1 2 H EW IMP1 30 3151 7/22/2003 716747.094 365959 0 0 I EM IMP1 10 3203 9/24/2003 691809.248 365641.84 0 0 I EM IMP1 20 3249 9/24/2003 693021.708 365142.315 1 1 H EW IMP1 40

PAGE 165

148 148 Table D-1. Continued. ID Date Sampled X Y No. Ditches Ditch Intensity Ditch Class Veg. Type Surr. Land Use Edge % 3250 9/18/2003 709440.366 365231.33 0 0 I EM IMP2 30 3270 7/22/2003 718258.601 364961.751 1 1 T EM IMP2 30 3339 9/18/2003 705787.645 364329.181 2 1 F EM IMP1 30 3387 8/21/2003 723435.659 364028.969 1 2 H EM IMP1 0 3409 9/18/2003 704009.383 363888.707 0 0 S EW IMP1 0 3592 8/21/2003 722235.04 361906.978 0 0 I EM IMP2 10 3622 8/21/2003 721778.365 361889.212 0 0 I EM IMP1 40 3640 8/14/2003 719512.092 361591.7 0 0 I EM IMP1 60 3840 8/20/2003 732765.977 357807.392 1 2 H EM IMP2 10 3887 11/18/2003 737244.895 356976.927 0 0 I EM UNI2 10 3929 8/20/2003 729574.754 356605.211 1 2 H EM IMP2 20 4006 9/25/2003 729977 354970.224 0 0 I EM IMP2 20 5002 9/26/2003 708539.585 381920.77 0 0 I EW IMP1 10 5003 9/26/2003 700696.5 392789.823 2 2 F EM IMP2 40 5004 9/27/2003 691796.699 370090.319 1 2 T EM DAUN 40 5005 9/27/2003 702592.829 365408.97 2 3 F EW IMP2 10 5006 9/27/2003 702880.783 365806.298 1 2 H EM IMP1 0 5007 9/27/2003 701583.545 366719.164 0 0 S EW IMP1 30 5008 9/27/2003 725573.952 367739.665 2 2 E EW DABE 20 5020 11/17/2003 700766.032 390058.596 0 0 I EM IMP2 10 5021 11/17/2003 700838.036 390485.245 1 2 E EM IMP1 20 5022 11/17/2003 701488.126 390343.839 2 2 F EM IMP1 10 5023 11/17/2003 712680.987 384496.149 0 0 S EM DAHA 10 5024 11/17/2003 713414.679 384108.059 0 0 I EM DAHA 40 5025 11/18/2003 737292.604 356051.683 0 0 I EM UNI1 10 5026 11/18/2003 723237.077 359090.049 1 1 T EM IMP1 20 5027 11/18/2003 723360.156 359270.879 0 0 I EM IMP1 20 5028 11/18/2003 699616.51 376794.254 0 0 I EM UNI1 80 5029 11/18/2003 700042.559 377012.651 0 0 S EM UNI1 60 5030 11/19/2003 706118.486 390999.356 0 0 I EM IMP2 10 5031 11/19/2003 705913.263 391920.222 0 0 I EM IMP2 20 5032 11/19/2003 706748.834 390167.622 2 3 E EM IMP1 20 5033 11/19/2003 707166.848 390187.997 1 3 F EM IMP1 20 5034 11/19/2003 714296.509 383529.956 0 0 S EW DAPA 20 5037 11/19/2003 700734.298 370184.864 0 0 I EM IMP2 0 5038 5/19/2003 701473.603 375298.838 1 2 H EM IMP1 20 H = head-of-ditch E = end-of-ditch F = flowthrough I = isolated T = tangent S = subsurface EM = emergent marsh EW = emergent marsh / open water FO = forested SS = scrub shrub FS = forested scrub mix IMP1 = more improved pasture IMP2 = less improved pasture UNIMP1 = unimproved pasture UNIMP2 = rangeland DAPA = dairy pasture DAHA = dairy hay field DABE = dairy beef pasture DASP = dairy sprayfield DAUN = dairy unimproved

PAGE 166

149 APPENDIX E SAMPLING PERIOD RAINFALL Figure E-1. May, 2003 precipitati on map, showing inches of rainfall, and comparison to average (four priority basin study area is outlined in red) (Source: SFWMD, 2004b). Table E-1. Rainfall record by month of sa mpling period and percent of historical average by area designated in Figure E-1 (SFWMD, 2004b). Lower Kissimmee ---------cm --------Lake Okeechobee ---------cm --------No. sites sampled November, 2003 4.50 (84%) 3.38 (59%) 18 September, 2003 80.01 (124%) 66.17 (103%) 35 August, 2003 64.62 (129%) 48.39 (98%) 25 July, 2003 101.35 (116%) 29.62 (90%) 37 May, 2003 53.72 (126%) 52.45 (122%) 3

PAGE 167

150 APPENDIX F ADDITIONAL SOIL BIOGEOCHEMICAL DATA Mehlich I, or double-acid extractabl e, phosphorus (P) (Mehlich, 1953) is commonly used as a soil extractant for P in the southeastern US as an indicator of potentially leachable P (Graetz et al., 1999). The Mehlich 1 extracts P from aluminum, iron, and calcium phosphates and is best suit ed to acid soils (pH < 6.5) with low cation exchange capacities (< 10 cmol/kg) and low organic matter contents (< 5%) (Pierzynski, 2000). It was determined using a 1:4 soil to solution (0.025 M H2SO4 and 0.05 M HCl) ratio. Air-dried (40oC) samples were weighed (5 g) and shaken for 5 minutes before being filtered through a Whatman #41 filter pa per. Soil extracts where stored at 4oC and analyzed calorimetrically for soluble reactive P (SRP) using the automated ascorbic acid method (Method 365.1; US EPA, 1993). A one hour water extraction on air-dried samples using a 1:10, soil to DDI water ratio was used to determine the quantity of water-extractable P. Extracts were filtered through Whatman #41 filter paper an d acidified to pH 2 with 11 N H2SO4 for preservation purposes. All wate r extracts were stored at 4oC and analyzed as described above (Method 365.1; US EPA, 1993). To determine oxalate extractable P, Fe and Al, soils were extracted with 0.175 M ammonium oxalate and 0.1 M oxalic acid at a soil to so lution ratio of 1:40 for 4 hours (McKeague and Day, 1966). After extraction, extract were filtered through 0.45 m membrane filters and analyzed using an I nductively Coupled Plasma Spectrophotometer (US EPA, 1984).

PAGE 168

151 Table F-1. Comparisons of P extractions among hydrologic zones. Statistical analyses are by one-way ANOVA and LSD pairwi se comparisons. Mean values and std. dev. followed by the same le tter are not different (p < 0.05). Ditch samples were not included in comparison analyses. n Mean Median Min. Max. Mehlich1-extr P -------------------------mg kg-1 --------------------------Center 118 36.9 75.9 a 23.24.1 792.2 Edge 116 29.7 95.0 b 14.81.2 1005.8 Upland 116 20.6 31.7 b 12.81.5 270.0 Ditch 60 19.8 18.5 11.91.1 80.1 HCl-extr P -----------------------mg kg-1 --------------------------Center 118 119.1 317.9 a 53.0 5.2 3233.0 Edge 116 87.3 253.9 b 31.7 2.2 2078.3 Upland 116 41.0 46.7 b 25.5 3.6 348.3 Ditch 60 43.2 49.8 25.6 1.6 277.9 Oxalate-extr P -------------------------mg kg-1 --------------------------Center 117 155.9 361.8 a 66.411.1 3596.0 Edge 117 109.8 131.5 b 67.65.4 788.5 Upland 115 146.1 301.2 c 66.15.2 2712.0 Ditch 61 102.0 108.7 58.258.2 535.5 Water-extr P -------------------------mg kg-1 --------------------------Center 116 23.3 26.4 a 13.5 1.8 173.6 Edge 116 12.7 12.3 b 9.5 0.8 79.9 Upland 117 12.7 12.3 b 8.0 0.6 68.3 Ditch 60 11.7 10.9 7.7 0.2 46.9 Mehlich1-extr P ----------------g m-2 (0 10 cm) ----------------------Center 118 2.2 3.8 a 1.20.2 34.5 Edge 116 2.2 5.6 a 1.30.2 58.8 Upland 116 2.1 3.0 a 1.30.2 26.2 Ditch 60 1.3 1.2 0.90.1 6.1 HCl-extr P ----------------g m-2 (0 – 10 cm) ---------------------Center 118 6.83 17.5 a 2.9 0.2 140.7 Edge 116 6.14 15.6 a 2.9 0.3 121.5 Upland 116 3.95 4.17 a 2.7 0.4 33.8 Ditch 60 2.74 2.87 1.8 0.2 17.7 Oxalate-extr Pa ----------------g m-2 (0 10 cm) ----------------------Center 118 9.9 16.1 a 6.10.9 156.5 Edge 116 10.7 19.0 a 5.40.9 177.5 Upland 116 6.9 5.6 a 4.90.9 24.7 Ditch 60 5.6 5.8 4.20.6 34.5 Water-extr P ----------------g m-2 (0 10 cm) ----------------------Center 116 1.1 1.1 a 0.8 0.1 7.6 Edge 116 1.0 0.8 a 0.8 0.1 4.7 Upland 117 1.0 0.8 a 0.8 0.1 4.9 Ditch 60 0.8 0.7 0.6 0.0 3.8 a Pairwise comparisons based on Games-Howell procedure (p < 0.05)

PAGE 169

152 Table F-2. Observed values and comparisons of P extractions (mg kg -1) among land uses within hydrologic zones. Statistical analyses are by one factor ANOVA and Bonferroni pairwise comparisons based on estimated marginal means. Mean values and std. dev. followed by the same letter are not different (p < 0.01). n Mean Median Min. Max. ---------------------------mg kg-1 ---------------------------Centers Water-extr Pa Dairy 22 32.0 42.7 a 14.84.8 173.7 Improved 83 21.0 20.2 a 13.61.8 101.7 Unimproved 12 22.6 25.0 a 11.72.5 84.4 Mehlich1-extr P Dairy 2281.4 165.1 a 31.0 7.0 792.2 Improved 8327.8 23.4 b 22.8 4.1 126.1 Unimproved 1219.6 18.5 b 15.7 5.9 74.6 HCl-extr Pb Dairy 22326.40 688.8 a 87.224.1 3233.0 Improved 8372.40 88.1 b 50.05.2 510.3 Unimproved 1269.36 76.0 b 36.012.6 273.2 Oxalate-extr P Dairy 22441.7 775.8 a 214.837.9 3596.0 Improved 83142.1 149.5 b 86.712.3 788.5 Unimproved 12154.2 152.5 ab 97.627.4 535.5 Edges Water-extr P Dairy 2120.1 19.6 a 17.7 1.5 79.9 Improved 8211.9 9.8 a 9.8 0.8 54.4 Unimproved 126.1 4.3 b 5.6 0.8 14.7 Mehlich1-extr P Dairy 2188.7 215.7 a 32.0 2.8 1005.8 Improved 8216.7 11.5 b 14.2 2.1 67.0 Unimproved 1216.2 26.6 b 7.1 1.2 95.8 HCl-extr P Dairy 21222.7 437.1 a 90.9 11.7 2078.2 Improved 8240.4 35.4 b 29.4 4.6 177.2 Unimproved 12175.5 507.6 b 14.3 2.2 783.7 Oxalate-extr P Dairy 21320.1 338.1 a 175.4 18.2 1442.7 Improved 8288.0 96.8 b 53.4 10.0 646.7 Unimproved 12290.6 765.5 b 40.7 6.6 2712.0 Uplands Water-extr Pa Dairy 2113.8 15.1 a 10.0 1.8 68.3 Improved 839.5 7.2 a 7.9 0.8 37.4 Unimproved 127.1 5.4 a 4.3 0.6 19.9 Mehlich1-extr Pa Dairy 2129.1 29.7 a 21.3 6.1 130.9 Improved 8320.1 33.9 b 12.8 1.5 270.0 Unimproved 129.7 7.8 b 6.6 2.1 27.9

PAGE 170

153 Table F-2. Continued n Mean Median Min. Max. ---------------------------mg kg-1 ---------------------------HCl-extr Pa Dairy 21 63.3 48.1 a 48.9 11.1 180.3 Improved 83 38.8 47.4 b 23.9 3.6 348.3 Unimproved 12 17.1 13.1 c 14.0 5.9 52.4 Oxalate-extr P Dairy 21 116.4 85.9 a 78.8 14.9 313.0 Improved 83 65.7 60.8 b 42.8 11.1 284.3 Unimproved 12 37.3 26.0 b 29.4 8.3 104.6 a (p < 0.10) b (p < 0.05)

PAGE 171

154 Table F-3. Observed values a nd comparisons of metals (mg kg -1) among land uses within hydrologic zones. Statistical analyses are by one-factor ANOVA and Bonferroni pairwise comparisons based on estimated marginal means. Mean values and std. dev. followed by the same letter are not different (p < 0.10). n Mean Median Min. Max. ----------------------------mg kg-1 ----------------------------Centers Oxalate-extractable Al Dairy 21 942.1 1067.8 a 467.943.5 3924.0 Improved 84 594.1 526.3 a 411.548.1 3047.4 Unimproved 12 765.0 454.2 a 791.1117.9 1443.3 Oxalate-extractable Fe Dairy 21 1208.9 1589.7 a 586.1157.8 5992.0 Improved 84 804.8 1068.9 a 460.134.2 7141.2 Unimproved 12 1141.3 1228.4 a 539.6145.8 4162.7 HCl-extractable Ca Dairy 21 2118.1 3721.5 a 493.133.6 15421.6 Improved 84 1250.2 1316.7 a 749.554.5 5256.4 Unimproved 12 1028.8 781.2 a 763.9357.6 3059.9 HCl-extractable Mg Dairy 21 1501.8 1631.9 a 804.6136.5 6079.9 Improved 84 702.6 1059.7 b 365.734.1 6717.5 Unimproved 12 1039.0 1219.9 ab 523.440.5 4032.7 HCl-extractable Al Dairy 21 804.5 956.9 a 481.3122.9 3735.3 Improved 84 486.3 491.7 a 318.320.4 2789.6 Unimproved 12 531.9 376.0 a 431.8113.7 1111.0 HCl-extractable Fe Dairy 21 2075.6 2244.3 a 1225.0341.0 9926.0 Improved 84 1463.1 1708.6 a 788.134.24 8630.5 Unimproved 12 1331.8 1751.4 a 407.5117.2 4910.0 Edges Oxalate-extractable Al Dairy 21 722.4 729.9 a 596.4101.2 2852.1 Improved 83 455.4 952.3 b 237.177.6 8558.9 Unimproved 12 550.0 402.1 ab 542.868.0 1271.5 Oxalate-extractable Fe Dairy 21 1613.0 1935.3 a 415.8112.1 5844.9 Improved 83 679.9 1248.1 b 273.058.5 8063.9 Unimproved 12 1863.6 3885.9 ab 283.882.6 13720.0 HCl-extractable Ca Dairy 21 1452.9 2114.8 a 587.341.1 8999.0 Improved 83 875.7 1252.3 a 452.884.5 7757.9 Unimproved 12 582.7 337.1 a 613.295.5 1247.0 HCl-extractable Mg Dairy 21 1691.5 1992.1 a 666.2109.0 5755.9 Improved 83 575.6 1252.6 b 198.434.7 7883.9 Unimproved 12 1904.0 4029.7 b 190.59.3 14145.0

PAGE 172

155 Table F-3. Continued. HCl-extractable Al Dairy 21 474.9 384.2 a 311.826.8 1529.6 Improved 83 285.8 390.6 b 183.718.5 2900.0 Unimproved 12 445.5 498.3 ab 165.973.9 1430.5 HCl-extractable Fe Dairy 21 1795.8 1147.4 a 1147.4125.4 6595.9 Improved 83 936.6 1163.5 b 468.452.5 7249.5 Unimproved 12 1047.5 1741.3 ab 297.392.1 5965.0 Uplands Oxalate-extractable Al Dairy 21 636.1 939.2 a 363.787.3 4523.6 Improved 83 543.9 985.3 a 214.640.9 4709.2 Unimproved 12 323.8 248.4 a 232.755.4 957.0 Oxalate-extractable Fe Dairy 21 734.1 761.7 a 381.990.6 2539.4 Improved 83 474.6 685.8 a 215.753.7 3443.1 Unimproved 12 367.0 245.2 a 311.261.8 858.2 HCl-extractable Ca Dairy 21 1157.5 1423.2 a 444.586.0 4519.8 Improved 83 844.0 860.8 a 525.6100.7 4108.6 Unimproved 12 438.6 217.1 a 463.4155.6 815.8 HCl-extractable Mg Dairy 21 735.4 804.7 a 416.774.9 2649.1 Improved 83 319.2 576.9 b 137.76.9 3354.7 Unimproved 12 290.9 369.9 b 95.215.4 1211.9 HCl-extractable Al Dairy 21 238.7 239.7 a 152.027.0 813.3 Improved 83 162.1 188.8 a 118.214.6 1443.3 Unimproved 12 144.8 84.7 a 123.843.8 338.7 HCl-extractable Fe Dairy 21 1279.0 1137.9 a 1009.0144.4 4483.3 Improved 83 1116.9 1511.9 a 636.164.4 8441.8 Unimproved 12 719.0 993.8 b 318.762.6 3505.0

PAGE 173

156Table F-4. Pearson correlations of selected soil biogeochemical parameters in wetland edge soils. To normalize data, TP was square-root transformed a nd other variables were na tural-log transformed. TP Waterextr P Percent LOI HCl-extr Fe HCl-extr Al HCl-extr Ca HCl-extr Mg Oxalateextr Al Oxalate -extr Fe Water-extr P Pearson Correlation 0.514 1 Sig. (2-tailed) 0.000 n 116 117 Percent loss on ignition Pearson Correlation 0.631 0.315 1 Sig. (2-tailed) 0.000 0.001 n 116 117 117 HCl-extr Fe Pearson Correlation 0.316 0.209 0.484 1 Sig. (2-tailed) 0.001 0.024 0.000 n 115 116 116 116 HCl-extr Al Pearson Correlation 0.291 0.214 0.302 0.194 1 Sig. (2-tailed) 0.002 0.021 0.001 0.037 n 115 116 116 116 116 HCl-extr Ca Pearson Correlation 0.447 0.316 0.354 -0.066 0.385 1 Sig. (2-tailed) 0.000 0.001 0.000 0.483 0.000 n 115 116 116 116 116 116 HCl-extr Mg Pearson Correlation 0.367 -0.059 0.507 0.524 0.218 0.127 1 Sig. (2-tailed) 0.000 0.528 0.000 0.000 0.019 0.173 n 115 116 116 116 116 116 116 Oxalate-extr Al Pearson Correlation 0.304 -0.071 0.407 0.213 0.174 0.581 0.423 1 Sig. (2-tailed) 0.001 0.447 0.000 0.022 0.063 0.000 0.000 n 116 117 117 116 116 116 116 117 Oxalate-extr Fe Pearson Correlation 0.347 -0.090 0.523 0.508 0.329 0.161 0.976 0.439 1 Sig. (2-tailed) 0.000 0.334 0.000 0.000 0.000 0.084 0.000 0.000 n 116 117 117 116 116 116 116 117 117

PAGE 174

157Table F-5. Pearson correlations of selected biogeochemical parameters in wetla nd center soils in improved pastures. To normal ize data, TP was square-root transformed and othe r variables were natu ral-log transformed. TP Waterextr P Percent LOI HClextr Fe HClextr Al HCl-extr Ca HCl-extr Mg Oxalateextr Al Oxalateextr Fe Water-extr P Pearson Correlation 0.569 1 Sig. (2-tailed) 0.000 n 83 84 Percent loss on ignition Pearson Correlation 0.582 0.588 1 Sig. (2-tailed) 0.000 0.000 n 83 84 84 HCl-extr Fe Pearson Correlation 0.456 0.596 0.372 1 Sig. (2-tailed) 0.000 0.000 0.000 n 83 84 84 84 HCl-extr Al Pearson Correlation 0.491 0.363 0.647 0.278 1 Sig. (2-tailed) 0.000 0.001 0.000 0.010 n 83 84 84 84 84 HCl-extr Ca Pearson Correlation 0.271 0.112 0.417 -0.239 0.564 1 Sig. (2-tailed) 0.013 0.310 0.000 0.029 0.000 n 83 84 84 84 84 84 HCl-extr Mg Pearson Correlation 0.406 0.158 0.505 0.346 0.132 -0.063 1 Sig. (2-tailed) 0.000 0.150 0.000 0.001 0.232 0.571 n 83 84 84 84 84 84 84 Oxalate-extr Al Pearson Correlation 0.545 0.322 0.662 0.310 0.849 0.520 0.200 1 Sig. (2-tailed) 0.000 0.003 0.000 0.004 0.000 0.000 0.069 n 83 84 84 84 84 84 84 84 Oxalate-extr Fe Pearson Correlation 0.428 0.108 0.547 0.298 0.221 0.002 0.963 0.301 1 Sig. (2-tailed) 0.000 0.330 0.000 0.006 0.044 0.983 0.000 0.005 n 83 84 84 84 84 84 84 84 84

PAGE 175

158Table F-6. Pearson correlations of selected biogeochemical parameters in wetla nd center soils in dairies. To normalize data, TP was square-root transformed and other vari ables were natural-log transformed. TP Waterextr P Percent LOI HCl-extr Fe HClextr Al HCl-extr Ca HClextr Mg Oxalateextr Al Oxalateextr Fe Water-extr P Pearson Correlation 0.632 1 Sig. (2-tailed) 0.002 n 22 22 Percent loss on ignition Pearson Correlation 0.621 0.367 1 Sig. (2-tailed) 0.002 0.093 n 22 22 22 HCl-extr Fe Pearson Correlation -0.102 -0.039 0.174 1 Sig. (2-tailed) 0.653 0.865 0.438 n 22 22 22 22 HCl-extr Al Pearson Correlation 0.586 0.689 0.198 -0.051 1 Sig. (2-tailed) 0.004 0.000 0.378 0.820 n 22 22 22 22 22 HCl-extr Ca Pearson Correlation 0.648 0.764 0.108 -0.286 0.851 1 Sig. (2-tailed) 0.001 0.000 0.632 0.198 0.000 n 22 22 22 22 22 22 HCl-extr Mg Pearson Correlation 0.084 -0.011 0.036 0.767 0.209 0.061 1 Sig. (2-tailed) 0.709 0.963 0.875 0.000 0.351 0.786 n 22 22 22 22 22 22 22 Oxalate-extr Al Pearson Correlation 0.685 0.593 0.442 -0.131 0.811 0.794 0.228 1 Sig. (2-tailed) 0.001 0.005 0.045 0.571 0.000 0.000 0.321 n 21 21 21 21 21 21 21 21 Oxalate-extr Fe Pearson Correlation 0.239 -0.049 0.529 0.170 -0.160 -0.117 0.319 0.295 1 Sig. (2-tailed) 0.298 0.834 0.014 0.461 0.489 0.612 0.159 0.194 n 21 21 21 21 21 21 21 21 21

PAGE 176

159Table F-7. Pearson correlations of selected biogeochemical parameters in wetla nd center soils in unimp roved pastures / rangela nds. To normalize data, TP was square-root transformed a nd other variables were na tural-log transformed. TP Water-extr P Percent LOI HCl-extr Fe HClextr Al HCl-extr Ca HClextr Mg Oxalateextr Al Oxalateextr Fe Water-extr P Pearson Correlation 0.610 1 Sig. (2-tailed) 0.035 n 12 12 Percent loss on ignition Pearson Correlation 0.419 0.864 1 Sig. (2-tailed) 0.175 0.000 n 12 12 12 HCl-extr Fe Pearson Correlation 0.281 0.568 0.364 1 Sig. (2-tailed) 0.376 0.054 0.244 n 12 12 12 12 HCl-extr Al Pearson Correlation -0.135 0.203 0.487 0.106 1 Sig. (2-tailed) 0.675 0.527 0.109 0.742 n 12 12 12 12 12 HCl-extr Ca Pearson Correlation -0.309 -0.159 -0.063 -0.236 -0.070 1 Sig. (2-tailed) 0.328 0.621 0.847 0.461 0.830 n 12 12 12 12 12 12 HCl-extr Mg Pearson Correlation 0.189 0.553 0.448 0.333 -0.236 -0.303 1 Sig. (2-tailed) 0.555 0.062 0.144 0.290 0.460 0.338 n 12 12 12 12 12 12 12 Oxalate-extr Al Pearson Correlation -0.138 0.163 0.418 0.117 0.677 0.032 0.000 1 Sig. (2-tailed) 0.669 0.614 0.176 0.718 0.016 0.921 1.000 n 12 12 12 12 12 12 12 12 Oxalate-extr Fe Pearson Correlation 0.262 0.582 0.434 0.259 -0.286 -0.344 0.980 -0.100 1 Sig. (2-tailed) 0.411 0.047 0.159 0.415 0.367 0.274 0.000 0.757 n 12 12 12 12 12 12 12 12 12

PAGE 177

160Table F-8. Pearson correlations of selected biogeochemical parameters in wetla nd edge soils in improved pastures. To normaliz e data, TP was square-root transformed and othe r variables were natu ral-log transformed. TP Waterextr P Percent LOI HCl-extr Fe HClextr Al HCl-extr Ca HClextr Mg Oxalateextr Al Oxalateextr Fe Water-extr P Pearson Correlation 0.235 1 Sig. (2-tailed) 0.032 n 83 84 Percent loss on ignition Pearson Correlation 0.744 0.364 1 Sig. (2-tailed) 0.000 0.001 n 83 84 84 HCl-extr Fe Pearson Correlation 0.283 0.275 0.327 1 Sig. (2-tailed) 0.010 0.012 0.003 n 82 83 83 83 HCl-extr Al Pearson Correlation 0.114 0.022 0.249 0.030 1 Sig. (2-tailed) 0.308 0.846 0.023 0.785 n 82 83 83 83 83 HCl-extr Ca Pearson Correlation 0.376 0.049 0.557 0.007 0.324 1 Sig. (2-tailed) 0.001 0.657 0.000 0.950 0.003 n 82 83 83 83 83 83 HCl-extr Mg Pearson Correlation 0.582 -0.114 0.533 0.227 -0.009 0.346 1 Sig. (2-tailed) 0.000 0.307 0.000 0.039 0.937 0.001 n 82 83 83 83 83 83 83 Oxalate-extr Al Pearson Correlation 0.435 -0.118 0.437 0.204 0.109 0.686 0.636 1 Sig. (2-tailed) 0.000 0.286 0.000 0.064 0.326 0.000 0.000 n 83 84 84 83 83 83 83 84 Oxalate-extr Fe Pearson Correlation 0.548 -0.161 0.553 0.195 0.266 0.486 0.939 0.666 1 Sig. (2-tailed) 0.000 0.143 0.000 0.077 0.015 0.000 0.000 0.000 n 83 84 84 83 83 83 83 84 84

PAGE 178

161Table F-9. Pearson correlations of selected biogeochemical parameters in wetland edge soils in dairies. To normalize data, TP was square-root transformed and other vari ables were natural-log transformed. TP Waterextr P Percent LOI HCl-extr Fe HClextr Al HCl-extr Ca HClextr Mg Oxalateextr Al Oxalateextr Fe Water-extr P Pearson Correlation 0.622 1 Sig. (2-tailed) 0.003 n 21 21 Percent loss on ignition Pearson Correlation 0.564 0.129 1 Sig. (2-tailed) 0.008 0.577 n 21 21 21 HCl-extr Fe Pearson Correlation 0.259 0.015 0.781 1 Sig. (2-tailed) 0.256 0.948 0.000 n 21 21 21 21 HCl-extr Al Pearson Correlation 0.651 0.633 0.356 0.288 1 Sig. (2-tailed) 0.001 0.002 0.113 0.205 n 21 21 21 21 21 HCl-extr Ca Pearson Correlation 0.502 0.559 -0.077 -0.302 0.624 1 Sig. (2-tailed) 0.020 0.008 0.741 0.184 0.003 n 21 21 21 21 21 21 HCl-extr Mg Pearson Correlation 0.299 -0.268 0.624 0.692 0.202 -0.140 1 Sig. (2-tailed) 0.188 0.239 0.003 0.001 0.381 0.545 n 21 21 21 21 21 21 21 Oxalate-extr Al Pearson Correlation 0.184 -0.113 0.246 0.205 0.216 0.379 0.270 1 Sig. (2-tailed) 0.425 0.625 0.282 0.372 0.347 0.090 0.236 n 21 21 21 21 21 21 21 21 Oxalate-extr Fe Pearson Correlation 0.290 -0.289 0.653 0.704 0.141 -0.238 0.981 0.235 1 Sig. (2-tailed) 0.202 0.204 0.001 0.000 0.542 0.299 0.000 0.305 n 21 21 21 21 21 21 21 21 21

PAGE 179

162Table F-10. Pearson correlations of selected biogeochemical parameters in wetla nd edge soils in unimproved pastures / rangelan ds. To normalize data, TP was square-root transformed a nd other variables were na tural-log transformed. TP Waterextr P Percent LOI HCl-extr Fe HCl-extr Al HCl-extr Ca HClextr Mg Oxalateextr Al Oxalateextr Fe Water-extr P Pearson Correlation 0.816 1 Sig. (2-tailed) 0.001 n 12 12 Percent loss on ignition Pearson Correlation 0.422 0.595 1 Sig. (2-tailed) 0.172 0.041 n 12 12 12 HCl-extr Fe Pearson Correlation 0.218 0.338 0.337 1 Sig. (2-tailed) 0.497 0.282 0.284 n 12 12 12 12 HCl-extr Al Pearson Correlation -0.070 0.190 0.392 0.499 1 Sig. (2-tailed) 0.828 0.554 0.207 0.098 n 12 12 12 12 12 HCl-extr Ca Pearson Correlation -0.208 0.146 0.309 -0.007 0.427 1 Sig. (2-tailed) 0.517 0.650 0.329 0.983 0.166 n 12 12 12 12 12 12 HCl-extr Mg Pearson Correlation 0.279 0.511 0.624 0.879 0.530 0.051 1 Sig. (2-tailed) 0.380 0.089 0.030 0.000 0.076 0.875 n 12 12 12 12 12 12 12 Oxalate-extr Al Pearson Correlation -0.086 0.030 0.526 0.277 0.748 0.618 0.329 1 Sig. (2-tailed) 0.790 0.927 0.079 0.384 0.005 0.032 0.297 n 12 12 12 12 12 12 12 12 Oxalate-extr Fe Pearson Correlation 0.264 0.508 0.621 0.870 0.537 0.050 0.999 0.321 1 Sig. (2-tailed) 0.408 0.092 0.031 0.000 0.072 0.876 0.000 0.308 n 12 12 12 12 12 12 12 12 12

PAGE 180

163 APPENDIX G LANDSAT7 ETM+ HEADER FILE NDF_REVISION=2.00; DATA_SET_TYPE=EDC_ETM+; PRODUCT_NUMBER=080040210017900001; PIXEL_FORMAT=BYTE; PIXEL_ORDER=NOT_INVERTED; BITS_PER_PIXEL=8; PIXELS_PER_LINE=7468; LINES_PER_DATA_FILE=7048; DATA_ORIENTATION=UPPER_LEFT/RIGHT; NUMBER_OF_DATA_FILES=6; DATA_FILE_INTERLEAVING=BSQ; TAPE_SPANNING_FLAG=1/1; START_LINE_NUMBER=1; START_DATA_FILE=1; LINES_PER_VOLUME=42288; BLOCKING_FACTOR=1; RECORD_SIZE=7468; UPPER_LEFT_CORNER=0824436.2753W,0282216.3674N,329160.000,3139560.000; UPPER_RIGHT_CORNER=0802726.3195W,0282252.6217N,553170.000,3139560.000; LOWER_RIGHT_CORNER=0802759.6308W,0262821.9011N,553170.000,2928150.000; LOWER_LEFT_CORNER=0824249.3197W,0262748.4721N,329160.000,2928150.000; REFERENCE_POINT=SCENE_CENTER; REFERENCE_POSITION=0813542.9425W,0272536.6592N,441165.000,3033855.000,3 734.50,3524.50; REFERENCE_OFFSET=-70.40,-24.28; ORIENTATION=0.000000; MAP_PROJECTION_NAME=UTM; USGS_PROJECTION_NUMBER=1; USGS_MAP_ZONE=17; USGS_PROJECTION_PARAMETERS=6378137.000000000000000,6356752.314140000400 000,0.000000000000000,0.000000000000000,0.000000000000000,0.00000000000 0000,0.000000000000000,0.000000000000000,0.000000000000000,0.0000000000 00000,0.000000000000000,0.000000000000000,0.000000000000000,0.000000000 000000,0.000000000000000; HORIZONTAL_DATUM=NAD83; EARTH_ELLIPSOID_SEMI-MAJOR_AXIS=6378137.000; EARTH_ELLIPSOID_SEMI-MINOR_AXIS=6356752.314; EARTH_ELLIPSOID_ORIGIN_OFFSET=0.000,0.000,0.000; EARTH_ELLIPSOID_ROTATION_OFFSET=0.000000,0.000000,0.000000; PRODUCT_SIZE=FULL_SCENE; PIXEL_SPACING=30.0000,30.0000; PIXEL_SPACING_UNITS=METERS; RESAMPLING=NN; PROCESSING_DATE/TIME=2004-02-11T09:56:13; PROCESSING_SOFTWARE=NLAPS_4_2_04e13; NUMBER_OF_BANDS_IN_VOLUME=6;

PAGE 181

164 WRS=016/041; ACQUISITION_DATE/TIME=2003-03-24T15:44:27Z; SATELLITE=LANDSAT_7; SATELLITE_INSTRUMENT=ETM+; PROCESSING_LEVEL=08; SUN_ELEVATION=53.38; SUN_AZIMUTH=130.28; BAND1_NAME=ETM+_BAND_1; BAND1_FILENAME=LE7016041000308350.I1; BAND1_WAVELENGTHS=0.45,0.52; BAND1_RADIOMETRIC_GAINS/BIAS=0.7756863,-6.1999969; BAND2_NAME=ETM+_BAND_2; BAND2_FILENAME=LE7016041000308350.I2; BAND2_WAVELENGTHS=0.52,0.60; BAND2_RADIOMETRIC_GAINS/BIAS=0.7956862,-6.3999939; BAND3_NAME=ETM+_BAND_3; BAND3_FILENAME=LE7016041000308350.I3; BAND3_WAVELENGTHS=0.63,0.69; BAND3_RADIOMETRIC_GAINS/BIAS=0.6192157,-5.0000000; BAND4_NAME=ETM+_BAND_4; BAND4_FILENAME=LE7016041000308350.I4; BAND4_WAVELENGTHS=0.76,0.90; BAND4_RADIOMETRIC_GAINS/BIAS=0.9654902,-5.1000061; BAND5_NAME=ETM+_BAND_5; BAND5_FILENAME=LE7016041000308350.I5; BAND5_WAVELENGTHS=1.55,1.75; BAND5_RADIOMETRIC_GAINS/BIAS=0.1257255,-0.9999981; BAND6_NAME=ETM+_BAND_7; BAND6_FILENAME=LE7016041000308350.I7; BAND6_WAVELENGTHS=2.08,2.35; BAND6_RADIOMETRIC_GAINS/BIAS=0.0437255,-0.3500004; END_OF_HDR;

PAGE 182

165 APPENDIX H SOILS IN UPLANDS ADJACENT TO HI STORICALLY ISOLATED WETLANDS The Soil Survey Geographic (SSURGO) da ta set (version 3.0, July 2000, scale 1:24,000) is a digital repres entation of the county soil su rvey maps published by the USDA/NRCS. Original source data is from 1990 (USDA/NRCS, 1995). Subgroup and great group information was outdated (e.g. Haplaquod is now termed Alaquod) based on the Okeechobee County Soil Survey (Lewis et al., 2001; G.W. Hurt, personal communication, September 20, 2004). The information was manually corrected. The total number of compone nts per series in all the major map units found within 75 m upland buffer areas created around historically isolated wetlands (within vegetative communities and land uses of interest in this thesis), and the average percent of each series was calculated. To calculate the percentage of the major map units each series represented, the number of components for each series was divided by the total number of components (4,548). This was multiplie d by the average component percentage (which was based on the number of com ponents) and was then normalized to 4,048 buffers; thus giving the proportions of series and associated highe r taxonomic divisions for all the upland areas (Table H-1).

PAGE 183

166 Table H-1. Soil series in the components of prevalent map units of upland areas around 4,048 nonriparian emergent marsh, forest ed or scrub-shrub wetlands within improved pastures, dairies, unimproved pastures or rangeland areas of the four priority basins of the Lake Ok eechobee watershed. Great groups and orders are summarized. Series name Total no. of components Average % of mapunit Normalized % of all major map units Soil subgroup Drainage classa Myakka 1430 98 35.89 Aeric Alaquod p Immokalee 1151 97 28.59 Arenic Alaquod p Valkaria 372 100 9.53 Spodic Psammaquent p Basinger 402 90 9.27 Spodic Psammaquent p Floridana 171 43 1.88 Arenic Argiaquoll vp Bradenton 72 100 1.84 Typic Endoqualf p Manatee 103 68 1.79 Typic Argiaquoll vp Placid 219 32 1.79 Typic Humaquept vp Waveland 147 43 1.62 Arenic Alaquod p Wabasso 57 98 1.43 Alfic Alaquod p Riviera 114 40 1.17 Arenic Glossaqualf p Pinellas 38 80 0.78 Arenic Ochraqualf p Oldsmar 43 70 0.77 Arenic Alaquod p Okeelanta 61 48 0.75 Terric Medisaprist vp Pineda 26 98 0.65 Arenic Glossaqualf p Parkwood 18 100 0.46 Mollic Endoaqualf p Pomello 21 81 0.44 Arenic Alaquod semi-p Lawnwood 23 52 0.31 Aeric Alaquod p Malabar 18 60 0.28 Grossarenic Ochraqualf p Tequesta 32 26 0.21 Arenic Glossaqualf vp Holopaw 7 90 0.16 Grossarenic Ochraqualf p Ft. Drum 3 100 0.08 Aeric Endoaquept p St. Johns 3 100 0.08 Typic Alaquod p Pendarvis 4 60 0.06 Arenic Alorthod p Salerno 4 62 0.06 Grossarenic Alaquod p Gator 4 50 0.05 Terric Medisaprist vp Punta 3 40 0.03 Grossarenic Alaquod p Samsula 1 100 0.03 Terric Medispaprist vp Alaquodsb 69.20 Psammaquentsb 18.80 Argiaquolls 3.67 Endoaqualfs 2.30 Glossaqualfs 2.03 Humaquepts 1.79 Ochraqualfs 1.22 Medisaprists 0.83 Endoaquepts 0.08 Alorthods 0.06 Spodosols 69.26 Entisols 18.80 Alfisols 5.55 Inceptisols 1.87 Histosols 0.83

PAGE 184

167 Table H-1. Continued. a p = poor, vp = very poor, semi-p = semi-poor b Alaquods are wet Spodosols characterized by a fluctuating water table and a spodic horizon consisting of mostly organic matter an d aluminum with a very low iron content (Soil Survey Staff, 1996). Psammaquents are sandy soils, wet during most of the year, have grayish or bluish colors and sometim es a weak B horizon that represents an incipient spodic horizon (Soil Survey Staff, 1996). (Sources: Carter et al., 1989; Lewis et al., 2001; McCollum and Cruz, 1979; USDA/NRCS, 1995)

PAGE 185

168 APPENDIX I OUTPUT FILES OF FINAL CLASSIFICATION TREES Note: In the following outputs from CART ve rsion 5.0 (Salford Systems, San Diego, CA), variables are coded according to a naming convention described below: Beginning letters of the variable name SW = sampled wetland S2 = sampled 25m transition buffer SB = sampled 75m upland buffer UW = unsampled wetland, U2 = unsampled 25m transition buffer UU = unsampled 75m upland buffer Middle letters of the variable name REF* = reflectance (where is a band, 1 through 6) TC1 = tasseled cap band 1 (brightness), TC2 = tasseled cap band 2 (greenness), TC3 = tasseled cap band 3 (wetness), Ending letters of the variable name MI = minimum MA = maximum ST = std. dev. ME = mean RA = range Center Soils in Landsat Area ====================== Target Frequency Table ====================== Variable: TP_CLAS N Classes: 2 Data Value N Wgt Count -------------------------------------------1 65 65 2 52 52 Total 117 117 Missing Value Prevalence Learn ---------------SBREF6ME 0.0598 SWTC3ST 0.0598 SWREF1ST 0.0598

PAGE 186

169 SWREF2RA 0.0598 SWREF2ST 0.0598 SWREF3ST 0.0598 SWREF4RA 0.0598 SWREF4ST 0.0598 SWREF5RA 0.0598 SWREF5ST 0.0598 SBREF5ME 0.0598 SBREF4ME 0.0598 SBREF3ME 0.0598 SBREF2ME 0.0598 SBREF1ME 0.0598 SWREF6RA 0.0598 SWREF6ST 0.0598 PRIORS SET EQUAL CURRENT MEMORY REQUIREMENTS TOTAL: 14053. DATA: 2106. ANALYSIS: 14053. AVAILABLE: 13500000. SURPLUS: 13485947. The data are being read ... 117 Observations in the learning sample. FILE: G:\GIS_Okeechobee\CARTupscaling\NEW files thesis\final center tree\CART centers thesis.txt CART is running. ============= TREE SEQUENCE ============= Dependent variable: TP_CLAS Terminal Cross-Validated Resubstitution Complexity Tree Nodes Relative Cost Relative Cost Parameter -----------------------------------------------------------------1 9 0.542 +/0.083 0.223 0.000000 2 8 0.523 +/0.082 0.231 0.003867 3 7 0.488 +/0.080 0.250 0.009625 4** 6 0.488 +/0.080 0.281 0.015395 5 3 0.538 +/0.078 0.458 0.029497 6 2 0.623 +/0.076 0.588 0.065395 7 1 1.000 +/0.000 1.000 0.205779 Initial misclassification cost = 0.500 Initial class assignment = 1 ================ NODE INFORMATION ================ ********************************************* Node 1: SWREF6ST * N: 117 ********************************************* ********************************* ================================= Node 2 = Terminal Node 6 = N: 79 = N: 38 = ********************************* =================================

PAGE 187

170 Node 1 was split on SWREF6ST A case goes left if SWREF6ST <= 10.903 Improvement = 0.078979 Complexity Threshold = 0.205769 Node Cases Wgt Counts Cost Class 1 117 117.00 0.500 1 2 79 79.00 0.353 2 -6 38 38.00 0.159 1 Weighted Counts Class Top Left Right 1 65.00 32.00 33.00 2 52.00 47.00 5.00 Within Node Probabilities Class Top Left Right 1 0.500 0.353 0.841 2 0.500 0.647 0.159 Surrogate Split Assoc Improvement 1 SWTC3ST s 21.198 0.771 0.042 2 SWREF3ST s 3.821 0.701 0.037 3 SWREF5ST s 22.899 0.618 0.046 4 SWREF6RA s 35.500 0.605 0.039 5 SWREF5RA s 80.500 0.535 0.025 Competitor Split Improvement 1 SWTC3ST 22.484 0.051 2 SWREF5ST 22.899 0.046 3 SWREF6RA 36.500 0.044 4 SWREF5RA 48.500 0.038 5 SWREF3ST 3.821 0.037 ********************************************* Node 2: SBREF2ME * N: 79 ********************************************* ================================= ********************************* = Terminal Node 1 = Node 3 = N: 13 = N: 66 ================================= ********************************* Node 2 was split on SBREF2ME A case goes left if SBREF2ME <= 17.261 Improvement = 0.034792 Complexity Threshold = 0.065385 Node Cases Wgt Counts Cost Class 2 79 79.00 0.353 2 -1 13 13.00 0.185 1 3 66 66.00 0.272 2 Weighted Counts Class Top Left Right 1 32.00 11.00 21.00 2 47.00 2.00 45.00 Within Node Probabilities Class Top Left Right 1 0.353 0.815 0.272 2 0.647 0.185 0.728

PAGE 188

171 Surrogate Split Assoc Improvement 1 SBREF3ME s 13.841 0.444 0.023 2 SBREF5ME s 83.150 0.426 0.011 3 SBREF1ME s 4.475 0.426 0.021 4 SBREF6ME s 34.281 0.352 0.016 5 SBREF4ME s 95.613 0.148 0.006 Competitor Split Improvement 1 SBREF3ME 14.871 0.031 2 SWREF4ST 12.719 0.024 3 SBREF1ME 4.475 0.021 4 SBREF5ME 76.509 0.018 5 SBREF6ME 34.281 0.016 ********************************************* Node 3: SWREF4ST * N: 66 ********************************************* ********************************* ================================= Node 4 = Terminal Node 5 = N: 27 = N: 39 = ********************************* ================================= Node 3 was split on SWREF4ST A case goes left if SWREF4ST <= 12.719 Improvement = 0.029085 Complexity Threshold = 0.029487 Node Cases Wgt Counts Cost Class 3 66 66.00 0.272 2 4 27 27.00 0.463 2 -5 39 39.00 0.149 2 Weighted Counts Class Top Left Right 1 21.00 14.00 7.00 2 45.00 13.00 32.00 Within Node Probabilities Class Top Left Right 1 0.272 0.463 0.149 2 0.728 0.537 0.851 Surrogate Split Assoc Improvement 1 SWREF4RA s 37.500 0.694 0.015 2 SWREF5RA s 62.500 0.388 0.003 3 SBREF1ME r 6.236 0.331 0.005 4 SWREF5ST s 19.208 0.322 0.002 5 SWREF2ST s 2.558 0.314 0.010 Competitor Split Improvement 1 SWREF2RA 9.500 0.018 2 SWREF4RA 36.000 0.017 3 SBREF5ME 106.032 0.013 4 SWREF5RA 66.500 0.012 5 SWREF2ST 2.429 0.011 ********************************************* Node 4: SWREF5RA * N: 27

PAGE 189

172 ********************************************* ********************************* ================================= Node 5 = Terminal Node 4 = N: 18 = N: 9 = ********************************* ================================= Node 4 was split on SWREF5RA A case goes left if SWREF5RA <= 49.500 Improvement = 0.012405 Complexity Threshold = 0.044231 Node Cases Wgt Counts Cost Class 4 27 27.00 0.463 2 5 18 18.00 0.286 2 -4 9 9.00 0.135 1 Weighted Counts Class Top Left Right 1 14.00 6.00 8.00 2 13.00 12.00 1.00 Within Node Probabilities Class Top Left Right 1 0.463 0.286 0.865 2 0.537 0.714 0.135 Surrogate Split Assoc Improvement 1 SWREF5ST s 16.010 0.784 0.012 2 SWTC3ST s 17.132 0.676 0.005 3 SWREF6RA s 28.500 0.676 0.005 4 SWREF6ST s 9.315 0.568 0.003 5 SWREF4ST s 9.641 0.459 0.002 Competitor Split Improvement 1 SWREF5ST 16.010 0.012 2 SWTC3ST 13.290 0.010 3 SWREF2ST 1.529 0.009 4 SWREF2RA 9.500 0.007 5 SWREF6ST 7.192 0.007 ********************************************* Node 5: SWREF2ST * N: 18 ********************************************* ================================= ================================= = Terminal Node 2 = = Terminal Node 3 = = N: 7 = = N: 11 = ================================= ================================= Node 5 was split on SWREF2ST A case goes left if SWREF2ST <= 1.580 Improvement = 0.016170 Complexity Threshold = 0.036538 Node Cases Wgt Counts Cost Class 5 18 18.00 0.286 2 -2 7 7.00 0.172 1 -3 11 11.00 0.000 2 Weighted Counts Class Top Left Right 1 6.00 6.00 0.00

PAGE 190

173 2 12.00 1.00 11.00 Within Node Probabilities Class Top Left Right 1 0.286 0.828 0.000 2 0.714 0.172 1.000 Surrogate Split Assoc Improvement 1 SWREF2RA s 7.000 0.483 0.008 2 SBREF5ME r 102.882 0.241 0.001 3 SWREF4RA s 23.500 0.207 0.002 4 SWREF4ST s 7.360 0.207 0.002 5 SWREF1ST s 1.446 0.172 0.006 Competitor Split Improvement 1 SWREF1ST 1.566 0.008 2 SWREF2RA 7.000 0.008 3 SWREF3ST 1.900 0.003 4 SWREF4RA 36.000 0.003 5 SWREF4ST 9.999 0.003 ========================= TERMINAL NODE INFORMATION ========================= [Breiman adjusted cost, lambda = 0.051] Node N Prob Cost Class ----------------------------------------------------------------------------1 13 0.1038 0.1852 1 Parent C.T. = 0.065 [0.5060] 11 0.8148 1 2 0.1852 2 2 7 0.0558 0.1724 1 Parent C.T. = 0.037 [0.6380] 6 0.8276 1 1 0.1724 2 3 11 0.1058 0.0000 2 Parent C.T. = 0.037 [0.3169] 0 0.0000 1 11 1.0000 2 4 9 0.0712 0.1351 1 Parent C.T. = 0.044 [0.5420] 8 0.8649 1 1 0.1351 2 5 39 0.3615 0.1489 2 Parent C.T. = 0.029 [0.2692] 7 0.1489 1

PAGE 191

174 32 0.8511 2 6 38 0.3019 0.1592 1 Parent C.T. = 0.206 [0.2999] 33 0.8408 1 5 0.1592 2 Node Learn 1 13.00 11.00 2.00 2 7.00 6.00 1.00 3 11.00 0.00 11.00 4 9.00 8.00 1.00 5 39.00 7.00 32.00 6 38.00 33.00 5.00 ========================== MISCLASSIFICATION BY CLASS ========================== (Cross Validation) Prior Wgt Class Prob Wgt Count Count Misclass Misclass Cost ---------------------------------------------------------------------------1 0.500 65.00 65 7.00 7 0.108 (65.00 65 13.00 13 0.200) 2 0.500 52.00 52 9.00 9 0.173 (52.00 52 15.00 15 0.288) ---------------------------------------------------------------------------Total 1.000 117.00 117 16.00 16 (117.00 117 28.00 28) =================== VARIABLE IMPORTANCE =================== Relative Number Of Importance Categories Penalty --------------------------------------------------------------------------SWREF6ST 100.000 SWREF5ST 73.543 SWTC3ST 57.123 SWREF6RA 53.010 SWREF5RA 49.274 SWREF3ST 45.016 SBREF2ME 42.439 SWREF4ST 39.833 SWREF2ST 32.120 SBREF1ME 30.899 SBREF3ME 27.733 SWREF4RA 20.610 SBREF6ME 19.533 SBREF5ME 15.061 SWREF2RA 9.661 SWREF1ST 7.548 SBREF4ME 6.965 =============== OPTION SETTINGS ===============

PAGE 192

175 Construction Rule Gini (priors altered by costs) Estimation Method 10-fold cross-validation Misclassification Costs Unit Tree Selection 0.000 se rule Linear Combinations No Initial value of the complexity parameter = 0.000 Minimum size below which node will not be split = 10 Node size above which sub-sampling will be used = 117 Maximum number of surrogates used for missing values = 5 Number of surrogate splits printed = 5 Number of competing splits printed = 5 Maximum number of trees printed in the tree sequence = 10 Max. number of cases allowed in the learning sample = 117 Maximum number of cases allowed in the test sample = 0 Max # of nonterminal nodes in the largest tree grown = 117 (Actual # of nonterminal nodes in largest tree grown = 12) Max. no. of categorical splits including surrogates = 1 Max. number of linear combination splits in a tree = 0 (Actual number cat. + linear combination splits = 0) Maximum depth of largest tree grown = 11 (Actual depth of largest tree grown = 6) Exponent for center weighting in split criterion = 0.220 Maximum size of memory available = 13500000 (Actual size of memory used in run = 55799) =========================== CV-tree Competitor List =========================== (Type, Predictor, Split if continuous, Improvement) Top Split Competitors Main 1 2 3 4 -----------------------------------------------------------------------CV 1 | Numeric Numeric Numeric Numeric Numeric | SWREF6ST SWTC3ST SWREF5ST SWREF6RA SWREF5RA | 10.781 22.484 24.660 36.500 48.500 | 0.081 0.058 0.053 0.047 0.040 CV 2 | Numeric Numeric Numeric Numeric Numeric | SWREF6ST SWTC3ST SWREF5ST SWREF3ST SWREF6RA | 10.903 22.484 21.855 3.821 36.500 | 0.076 0.050 0.045 0.041 0.038 CV 3 | Numeric Numeric Numeric Numeric Numeric | SWREF6ST SWTC3ST SWREF5ST SWREF6RA SWREF3ST | 10.903 21.445 23.812 36.500 3.821 | 0.081 0.054 0.047 0.045 0.041 CV 4 | Numeric Numeric Numeric Numeric Numeric | SWREF6ST SWTC3ST SWREF3ST SWREF5ST SWREF6RA | 10.903 22.484 3.821 24.068 35.500 | 0.105 0.066 0.056 0.056 0.052 CV 5 | Numeric Numeric Numeric Numeric Numeric | SWREF6ST SWTC3ST SWREF5ST SWREF6RA SWREF5RA | 10.903 17.944 22.899 35.500 48.500 | 0.073 0.055 0.043 0.037 0.034 CV 6 | Numeric Numeric Numeric Numeric Numeric | SWREF6ST SWTC3ST SWREF5ST SWREF6RA SWREF5RA | 10.903 22.484 22.899 36.500 48.500 | 0.088 0.065 0.058 0.054 0.047 CV 7 | Numeric Numeric Numeric Numeric Numeric | SWREF6ST SWTC3ST SWREF5ST SWREF5RA SWREF6RA | 10.903 13.554 13.583 48.500 36.500

PAGE 193

176 | 0.062 0.058 0.046 0.045 0.043 CV 8 | Numeric Numeric Numeric Numeric Numeric | SWREF6ST SWTC3ST SWREF5ST SWREF6RA SWREF3ST | 10.974 13.554 22.899 36.500 3.428 | 0.079 0.052 0.043 0.041 0.038 CV 9 | Numeric Numeric Numeric Numeric Numeric | SWREF6ST SWTC3ST SWREF5ST SWREF5RA SWREF6RA | 10.903 13.233 24.068 48.500 38.500 | 0.067 0.041 0.040 0.036 0.036 CV 10 | Numeric Numeric Numeric Numeric Numeric | SWREF6ST SWTC3ST SWREF6RA SWREF5ST SWREF3ST | 10.903 21.445 36.500 22.899 3.821 | 0.086 0.051 0.051 0.048 0.042 FINAL | Numeric Numeric Numeric Numeric Numeric | SWREF6ST SWTC3ST SWREF5ST SWREF6RA SWREF5RA | 10.903 22.484 22.899 36.500 48.500 | 0.079 0.051 0.046 0.044 0.038 Left Split Competitors Main 1 2 3 4 -----------------------------------------------------------------------CV 1 | Numeric Numeric Numeric Numeric Numeric | SBREF2ME SBREF3ME SWREF4ST SBREF1ME SBREF5ME | 17.261 14.871 12.719 4.475 76.509 | 0.035 0.030 0.024 0.024 0.021 CV 2 | Numeric Numeric Numeric Numeric Numeric | SBREF2ME SBREF3ME SBREF1ME SWREF2RA SWREF4ST | 17.261 14.871 5.094 9.500 12.709 | 0.035 0.029 0.022 0.022 0.022 CV 3 | Numeric Numeric Numeric Numeric Numeric | SBREF3ME SBREF2ME SWREF4ST SWREF2RA SWREF4RA | 14.871 17.304 12.719 8.500 36.000 | 0.032 0.030 0.022 0.018 0.017 CV 4 | Numeric Numeric Numeric Numeric Numeric | SBREF2ME SBREF3ME SBREF1ME SWREF4ST SBREF5ME | 17.261 14.871 5.094 12.719 76.509 | 0.039 0.033 0.028 0.025 0.023 CV 5 | Numeric Numeric Numeric Numeric Numeric | SBREF2ME SBREF3ME SWREF4ST SBREF1ME SWREF4RA | 17.127 14.872 12.719 4.181 36.000 | 0.030 0.030 0.026 0.021 0.017 CV 6 | Numeric Numeric Numeric Numeric Numeric | SBREF2ME SBREF3ME SWREF4ST SBREF1ME SBREF5ME | 17.261 14.871 12.719 4.181 76.509 | 0.034 0.029 0.022 0.020 0.020 CV 7 | Numeric Numeric Numeric Numeric Numeric | SBREF2ME SBREF3ME SWTC3ST SWREF4ST SBREF1ME | 17.261 14.871 13.554 12.719 4.475 | 0.032 0.028 0.022 0.019 0.017 CV 8 | Numeric Numeric Numeric Numeric Numeric | SBREF2ME SBREF3ME SWREF4ST SWREF4RA SBREF1ME | 17.261 14.871 12.736 36.000 4.475 | 0.040 0.035 0.030 0.023 0.023 CV 9 | Numeric Numeric Numeric Numeric Numeric | SBREF2ME SWREF4ST SBREF1ME SBREF3ME SWREF4RA | 17.261 12.719 4.481 14.871 36.000 | 0.036 0.030 0.030 0.027 0.020 CV 10 | Numeric Numeric Numeric Numeric Numeric | SBREF2ME SBREF3ME SBREF1ME SWREF4ST SBREF5ME | 17.261 14.871 5.094 12.719 76.509 | 0.042 0.036 0.025 0.023 0.021 FINAL | Numeric Numeric Numeric Numeric Numeric | SBREF2ME SBREF3ME SWREF4ST SBREF1ME SBREF5ME

PAGE 194

177 | 17.261 14.871 12.719 4.475 76.509 | 0.035 0.031 0.024 0.021 0.018 Right Split Competitors Main 1 2 3 4 -----------------------------------------------------------------------CV 1 | Numeric Numeric Numeric Numeric Numeric | SBREF5ME SBREF2ME SBREF6ME SWREF5RA SWREF5ST | 116.117 21.085 55.528 107.000 33.875 | 0.015 0.014 0.011 0.010 0.007 CV 2 | Numeric Numeric Numeric Numeric Numeric | SBREF5ME SBREF2ME SBREF6ME SWREF5RA SWREF5ST | 116.643 21.085 55.528 104.500 33.875 | 0.015 0.012 0.011 0.010 0.010 CV 3 | Numeric Numeric Numeric Numeric Numeric | SBREF5ME SBREF2ME SBREF6ME SWREF5RA SWREF5ST | 116.643 21.085 55.528 107.000 33.875 | 0.015 0.013 0.011 0.010 0.007 CV 4 | Numeric Numeric Numeric Numeric Numeric | SBREF2ME SBREF5ME SWREF4RA SWREF4ST SWREF5RA | 21.085 116.643 113.500 36.686 107.500 | 0.006 0.005 0.005 0.005 0.005 CV 5 | Numeric Numeric Numeric Numeric Numeric | SBREF5ME SBREF2ME SBREF6ME SWREF5RA SBREF1ME | 116.643 21.085 55.528 107.000 7.043 | 0.017 0.015 0.014 0.010 0.008 CV 6 | Numeric Numeric Numeric Numeric Numeric | SBREF5ME SBREF6ME SBREF2ME SBREF1ME SBREF3ME | 116.643 55.528 21.085 7.584 24.825 | 0.017 0.015 0.010 0.007 0.007 CV 7 | Numeric Numeric Numeric Numeric Numeric | SBREF5ME SBREF2ME SWREF5RA SBREF6ME SWREF5ST | 116.643 21.085 107.000 55.528 33.875 | 0.014 0.011 0.010 0.010 0.006 CV 8 | Numeric Numeric Numeric Numeric Numeric | SBREF5ME SBREF2ME SWREF5RA SBREF6ME SWREF5ST | 116.643 21.188 107.000 56.149 33.471 | 0.011 0.011 0.011 0.008 0.007 CV 9 | Numeric Numeric Numeric Numeric Numeric | SBREF5ME SBREF2ME SBREF6ME SWREF5RA SWREF5ST | 116.643 20.965 55.528 107.000 33.875 | 0.017 0.014 0.012 0.010 0.007 CV 10 | Numeric Numeric Numeric Numeric Numeric | SWREF5RA SBREF2ME SBREF5ME SWREF5ST SBREF6ME | 107.000 21.085 116.671 33.875 55.528 | 0.011 0.010 0.009 0.007 0.007 FINAL | | | ========================= Gains and ROC for TP_CLAS ========================= TP_CLAS = { 1 } Class Class P of Cum P of Cum N P of Class P of N Pop Cum P Lift Lift Node in Bin Bin in Bin Class in Bin in Bin of Pop Ratio Index -----------------------------------------------------------------------------4 8.00 0.8889 0.1231 0.1231 9.00 0.0769 0.0769 1.60 1.60 6 33.00 0.8684 0.5077 0.6308 38.00 0.3248 0.4017 1.57 1.56 2 6.00 0.8571 0.0923 0.7231 7.00 0.0598 0.4615 1.57 1.54

PAGE 195

178 1 11.00 0.8462 0.1692 0.8923 13.00 0.1111 0.5726 1.56 1.52 5 7.00 0.1795 0.1077 1.0000 39.00 0.3333 0.9060 1.10 0.32 3 0.00 0.0000 0.0000 1.0000 11.00 0.0940 1.0000 1.00 0.00 -----------------------------------------------------------------------------65.00 117.00 Score Threshold Sensitivity Specificity ---------------------------------------0.8888889 0.0000 1.00000 0.8684211 0.1231 0.9808 0.8571429 0.6308 0.8846 0.8461538 0.7231 0.8654 0.1794872 0.8923 0.8269 0.0000000 1.00000 0.2115 0.0000000 1.00000 0.0000 TP_CLAS = { 2 } Class Class P of Cum P of Cum N P of Class P of N Pop Cum P Lift Lift Node in Bin Bin in Bin Class in Bin in Bin of Pop Ratio Index -----------------------------------------------------------------------------3 11.00 1.0000 0.2115 0.2115 11.00 0.0940 0.0940 2.25 2.25 5 32.00 0.8205 0.6154 0.8269 39.00 0.3333 0.4274 1.94 1.85 1 2.00 0.1538 0.0385 0.8654 13.00 0.1111 0.5385 1.61 0.35 2 1.00 0.1429 0.0192 0.8846 7.00 0.0598 0.5983 1.48 0.32 6 5.00 0.1316 0.0962 0.9808 38.00 0.3248 0.9231 1.06 0.30 4 1.00 0.1111 0.0192 1.0000 9.00 0.0769 1.0000 1.00 0.25 -----------------------------------------------------------------------------52.00 117.00 Score Threshold Sensitivity Specificity ---------------------------------------1.0000000000 0.0000 1.00000 0.8205128 0.2115 1.00000 0.1538462 0.8269 0.8923 0.1428571 0.8654 0.7231 0.1315789 0.8846 0.6308 0.1111111 0.9808 0.1231 0.1111111 1.00000 0.0000 C:\Documents and Settings\kamckee\Local Settings\Temp\s1i050: 8.26 kb Grove file created containing 1 Tree. G:\GIS_Okeechobee\CARTupscaling\NEW files thesis\final center tree\CART centers thesis.txt: 117 records. Edge Soils in Landsat Area ====================== Target Frequency Table ====================== Variable: ETP_CLAS N Classes: 2

PAGE 196

179 Data Value N Wgt Count -------------------------------------------1 75 75 2 42 42 Total 117 117 Missing Value Prevalence Learn ---------------S2TC3ST 0.0598 SWREF1RA 0.0598 SWREF1ST 0.0598 SWREF2MA 0.0598 SWREF2MI 0.0598 SWREF2ME 0.0598 SWREF2RA 0.0598 SWREF2ST 0.0598 SWREF4MA 0.0598 SWREF4MI 0.0598 SWREF4ME 0.0598 SWREF4RA 0.0598 SWREF4ST 0.0598 SWREF5MA 0.0598 SWREF5MI 0.0598 SWREF5ME 0.0598 SWREF5RA 0.0598 SWREF5ST 0.0598 SWREF6MA 0.0598 SWREF6MI 0.0598 SWREF1ME 0.0598 SWREF1MI 0.0598 SWTC2MA 0.0598 SWTC2MI 0.0598 SWTC2ME 0.0598 SWTC2RA 0.0598 SWTC2ST 0.0598 SBTC1MA 0.0598 SBTC1MI 0.0598 SBTC1ME 0.0598 SBTC1RA 0.0598 SBTC1ST 0.0598 SBTC2MA 0.0598 SBTC2MI 0.0598 SBTC2ME 0.0598 SBTC2RA 0.0598 SBTC2ST 0.0598 SBTC3MA 0.0598 SBTC3MI 0.0598 SBTC3ME 0.0598 SWREF1MA 0.0598 SWREF6ME 0.0598 SBREF2MA 0.0598 SBREF2MI 0.0598 S2REF3MA 0.0598 S2REF3MI 0.0598 S2REF3ME 0.0598 S2REF3RA 0.0598 S2REF3ST 0.0598 S2REF4MA 0.0598 S2REF4MI 0.0598 S2REF4ME 0.0598 S2REF4RA 0.0598 S2REF4ST 0.0598

PAGE 197

180 S2TC2MA 0.0598 S2TC2MI 0.0598 S2TC2ME 0.0598 S2TC2RA 0.0598 S2TC2ST 0.0598 S2TC3MA 0.0598 S2TC3MI 0.0598 S2TC3ME 0.0598 S2TC3RA 0.0598 S2REF2ST 0.0598 S2REF2RA 0.0598 SBREF2ME 0.0598 SBREF3MA 0.0598 SBREF3MI 0.0598 SBREF3ME 0.0598 SBREF3RA 0.0598 SBREF3ST 0.0598 SBREF4MA 0.0598 SBREF4MI 0.0598 SBREF4ME 0.0598 SBREF4RA 0.0598 SBREF4ST 0.0598 S2REF2ME 0.0598 S2REF2MI 0.0598 S2REF2MA 0.0598 S2REF1ST 0.0598 S2REF1RA 0.0598 S2REF1MA 0.0598 S2REF1MI 0.0598 S2REF1ME 0.0598 PRIORS SET EQUAL CURRENT MEMORY REQUIREMENTS TOTAL: 32277. DATA: 9945. ANALYSIS: 32277. AVAILABLE: 13500000. SURPLUS: 13467723. The data are being read ... 117 Observations in the learning sample. FILE: G:\GIS_Okeechobee\CARTupscaling\NEW files thesis\final edge tree\CART edges thesis.txt CART is running. ============= TREE SEQUENCE ============= Dependent variable: ETP_CLAS Terminal Cross-Validated Resubstitution Complexity Tree Nodes Relative Cost Relative Cost Parameter -----------------------------------------------------------------1 10 0.650 +/0.091 0.147 0.000000 2 9 0.587 +/0.088 0.157 0.005259 3 8 0.590 +/0.088 0.178 0.010486 4 7 0.590 +/0.088 0.202 0.011915 5 6 0.587 +/0.088 0.236 0.017153 6 5 0.569 +/0.086 0.290 0.026677 7** 3 0.484 +/0.080 0.399 0.027391

PAGE 198

181 8 2 0.617 +/0.083 0.525 0.062867 9 1 1.000 +/0.000 1.000 0.237629 Initial misclassification cost = 0.500 Initial class assignment = 1 ================ NODE INFORMATION ================ ********************************************* Node 1: SWREF4MA * N: 117 ********************************************* ================================= ********************************* = Terminal Node 1 = Node 2 = N: 44 = N: 73 ================================= ********************************* Node 1 was split on SWREF4MA A case goes left if SWREF4MA <= 124.500 Improvement = 0.115099 Complexity Threshold = 0.237619 Node Cases Wgt Counts Cost Class 1 117 117.00 0.500 1 -1 44 44.00 0.116 1 2 73 73.00 0.328 2 Weighted Counts Class Top Left Right 1 75.00 41.00 34.00 2 42.00 3.00 39.00 Within Node Probabilities Class Top Left Right 1 0.500 0.884 0.328 2 0.500 0.116 0.672 Surrogate Split Assoc Improvement 1 S2REF4ME s 111.721 0.892 0.100 2 SWTC2ME s 40.561 0.746 0.089 3 SWTC2MA s 58.520 0.690 0.084 4 S2TC2ME s 46.759 0.669 0.061 5 SWREF4ME s 107.734 0.656 0.082 Competitor Split Improvement 1 S2REF4MA 135.000 0.115 2 SWTC2MA 63.848 0.105 3 SBTC1ME 130.088 0.102 4 S2REF4ME 111.721 0.100 5 SBTC2MA 70.361 0.098 ********************************************* Node 2: SBTC1ME * N: 73 ********************************************* ================================= ================================= = Terminal Node 2 = = Terminal Node 3 = = N: 15 = = N: 58 =

PAGE 199

182 ================================= ================================= Node 2 was split on SBTC1ME A case goes left if SBTC1ME <= 129.913 Improvement = 0.044871 Complexity Threshold = 0.062857 Node Cases Wgt Counts Cost Class 2 73 73.00 0.328 2 -2 15 15.00 0.216 1 -3 58 58.00 0.241 2 Weighted Counts Class Top Left Right 1 34.00 13.00 21.00 2 39.00 2.00 37.00 Within Node Probabilities Class Top Left Right 1 0.328 0.784 0.241 2 0.672 0.216 0.759 Surrogate Split Assoc Improvement 1 SBREF3MI s 8.500 0.422 0.032 2 S2REF3MI s 8.500 0.362 0.027 3 SBREF2ME s 16.131 0.349 0.008 4 SBTC1MA s 142.858 0.349 0.016 5 SBREF2MI s 9.500 0.302 0.006 Competitor Split Improvement 1 SBREF3MI 8.500 0.032 2 SBREF4ME 132.344 0.030 3 SWREF5ST 21.889 0.029 4 S2TC2ME 67.340 0.028 5 SBTC1RA 31.761 0.028 ========================= TERMINAL NODE INFORMATION ========================= [Breiman adjusted cost, lambda = 0.033] Node N Prob Cost Class ----------------------------------------------------------------------------1 44 0.3090 0.1156 1 Parent C.T. = 0.238 [0.2102] 41 0.8844 1 3 0.1156 2 2 15 0.1105 0.2155 1 Parent C.T. = 0.063 [0.4407] 13 0.7845 1 2 0.2155 2 3 58 0.5805 0.2412 2 Parent C.T. = 0.063 [0.2940] 21 0.2412 1

PAGE 200

183 37 0.7588 2 Node Learn 1 44.00 41.00 3.00 2 15.00 13.00 2.00 3 58.00 21.00 37.00 ========================== MISCLASSIFICATION BY CLASS ========================== (Cross Validation) Prior Wgt Class Prob Wgt Count Count Misclass Misclass Cost ---------------------------------------------------------------------------1 0.500 75.00 75 21.00 21 0.280 (75.00 75 22.00 22 0.293) 2 0.500 42.00 42 5.00 5 0.119 (42.00 42 8.00 8 0.190) ---------------------------------------------------------------------------Total 1.000 117.00 117 26.00 26 (117.00 117 30.00 30) =================== VARIABLE IMPORTANCE =================== Relative Number Of Importance Categories Penalty --------------------------------------------------------------------------SWREF4MA 100.000 S2REF4ME 86.453 SWTC2ME 76.954 SWTC2MA 72.611 SWREF4ME 71.466 S2TC2ME 53.126 SBTC1ME 38.985 SBREF3MI 27.960 S2REF3MI 23.218 SBTC1MA 14.124 SBREF2ME 6.732 SBREF2MI 5.491 SWREF2MI 0.000 SWREF2ME 0.000 SWREF2RA 0.000 SWREF2ST 0.000 SWREF5ST 0.000 SWREF4MI 0.000 SWREF4RA 0.000 SWREF4ST 0.000 SWREF5MA 0.000 SWREF5RA 0.000 SWREF5MI 0.000 SWREF5ME 0.000 SWREF2MA 0.000 SWREF1ST 0.000 SWREF1RA 0.000 SWTC2MI 0.000 SWTC2RA 0.000 SWTC2ST 0.000

PAGE 201

184 SBTC1MI 0.000 SBTC1RA 0.000 SBTC1ST 0.000 SBTC2MA 0.000 SBTC2MI 0.000 SBTC2ME 0.000 SBTC2RA 0.000 SBTC2ST 0.000 SBTC3MA 0.000 SBTC3MI 0.000 SBTC3ME 0.000 SWREF1MA 0.000 SWREF1MI 0.000 SWREF1ME 0.000 SWREF6MA 0.000 SWREF6MI 0.000 S2REF2ST 0.000 S2REF3MA 0.000 S2REF3ME 0.000 S2REF3RA 0.000 S2REF3ST 0.000 S2REF4MA 0.000 S2REF4MI 0.000 S2REF4RA 0.000 S2REF4ST 0.000 S2TC2MA 0.000 S2TC2MI 0.000 S2TC2RA 0.000 S2TC2ST 0.000 S2TC3MA 0.000 S2TC3MI 0.000 S2TC3ME 0.000 S2TC3RA 0.000 S2REF2RA 0.000 S2REF2ME 0.000 S2REF2MI 0.000 SWREF6ME 0.000 S2TC3ST 0.000 SBREF3MA 0.000 SBREF3ME 0.000 SBREF3RA 0.000 SBREF3ST 0.000 SBREF4MA 0.000 SBREF4MI 0.000 SBREF4ME 0.000 SBREF4RA 0.000 S2REF2MA 0.000 S2REF1ST 0.000 S2REF1RA 0.000 S2REF1ME 0.000 S2REF1MI 0.000 SBREF4ST 0.000 S2REF1MA 0.000 SBREF2MA 0.000 =============== OPTION SETTINGS =============== Construction Rule Gini (priors altered by costs) Estimation Method 10-fold cross-validation Misclassification Costs Unit Tree Selection 0.000 se rule

PAGE 202

185 Linear Combinations No Initial value of the complexity parameter = 0.000 Minimum size below which node will not be split = 10 Node size above which sub-sampling will be used = 117 Maximum number of surrogates used for missing values = 5 Number of surrogate splits printed = 5 Number of competing splits printed = 5 Maximum number of trees printed in the tree sequence = 10 Max. number of cases allowed in the learning sample = 117 Maximum number of cases allowed in the test sample = 0 Max # of nonterminal nodes in the largest tree grown = 117 (Actual # of nonterminal nodes in largest tree grown = 12) Max. no. of categorical splits including surrogates = 1 Max. number of linear combination splits in a tree = 0 (Actual number cat. + linear combination splits = 0) Maximum depth of largest tree grown = 11 (Actual depth of largest tree grown = 6) Exponent for center weighting in split criterion = 0.140 Maximum size of memory available = 13500000 (Actual size of memory used in run = 77465) =========================== CV-tree Competitor List =========================== (Type, Predictor, Split if continuous, Improvement) Top Split Competitors Main 1 2 3 4 -----------------------------------------------------------------------CV 1 | Numeric Numeric Numeric Numeric Numeric | SBTC1ME SWREF4MA S2REF4MA S2REF4ME SWTC2MA | 130.088 124.500 135.000 118.513 63.848 | 0.108 0.108 0.103 0.097 0.094 CV 2 | Numeric Numeric Numeric Numeric Numeric | SWREF4MA S2REF4MA S2REF4ME SWTC2MA SBTC1ME | 124.500 135.000 111.721 63.848 130.088 | 0.123 0.111 0.105 0.102 0.097 CV 3 | Numeric Numeric Numeric Numeric Numeric | SWREF4MA S2REF4MA S2REF4ME SBREF4ME SWTC2MA | 124.500 135.000 118.513 119.726 63.848 | 0.147 0.144 0.144 0.133 0.124 CV 4 | Numeric Numeric Numeric Numeric Numeric | SWREF4MA SBTC1ME SWTC2MA SBTC2MA S2REF4MA | 124.500 130.088 63.848 70.361 135.000 | 0.105 0.102 0.100 0.094 0.093 CV 5 | Numeric Numeric Numeric Numeric Numeric | S2REF4MA SBTC1ME SWREF4MA SWTC2MA SWREF4ME | 135.000 130.088 124.500 63.848 110.737 | 0.102 0.099 0.099 0.093 0.090 CV 6 | Numeric Numeric Numeric Numeric Numeric | S2REF4MA SWREF4MA SBTC2MA SWTC2MA SBTC1ME | 135.000 130.000 69.808 63.848 130.088 | 0.130 0.116 0.110 0.109 0.106 CV 7 | Numeric Numeric Numeric Numeric Numeric | SWREF4MA SWREF4ME S2REF4MA S2REF4ME SBTC2MA | 124.500 110.737 135.000 118.513 70.361 | 0.118 0.110 0.110 0.110 0.103 CV 8 | Numeric Numeric Numeric Numeric Numeric | S2REF4MA SWREF4MA SWTC2MA SBREF4ME SBREF4MA | 135.000 130.000 63.848 129.044 142.500 | 0.132 0.125 0.117 0.115 0.101

PAGE 203

186 CV 9 | Numeric Numeric Numeric Numeric Numeric | SWREF4MA S2REF4MA SBTC1ME SWTC2MA SBTC2MA | 124.500 135.000 130.088 63.638 70.361 | 0.117 0.111 0.108 0.100 0.099 CV 10 | Numeric Numeric Numeric Numeric Numeric | S2REF4MA SWREF4MA SWTC2MA SBTC1ME S2REF4ME | 135.000 124.500 63.848 130.088 111.721 | 0.118 0.117 0.113 0.103 0.100 FINAL | Numeric Numeric Numeric Numeric Numeric | SWREF4MA S2REF4MA SWTC2MA SBTC1ME S2REF4ME | 124.500 135.000 63.848 130.088 111.721 | 0.115 0.115 0.105 0.102 0.100 Left Split Competitors Main 1 2 3 4 -----------------------------------------------------------------------CV 1 | Numeric Numeric Numeric Numeric Numeric | SWREF2ST SWREF1ST SWREF2RA S2REF2RA S2REF2ST | 1.472 1.083 6.000 7.000 1.345 | 0.017 0.013 0.013 0.013 0.013 CV 2 | Numeric Numeric Numeric Numeric Numeric | SBTC2RA SBREF4RA SBTC2MA SBREF4MA SBREF4ST | 54.695 97.000 71.267 144.000 19.409 | 0.014 0.014 0.011 0.011 0.011 CV 3 | Numeric Numeric Numeric Numeric Numeric | SBTC1ST SBTC3ME SBTC3MA SBREF4ST SBTC1RA | 4.623 -108.008 -71.865 5.269 31.526 | 0.008 0.008 0.006 0.006 0.004 CV 4 | Numeric Numeric Numeric Numeric Numeric | SBTC2RA SBREF4RA SBREF4ST SBREF4MA SBTC2ST | 54.695 97.000 19.409 144.000 17.955 | 0.013 0.013 0.013 0.011 0.010 CV 5 | Numeric Numeric Numeric Numeric Numeric | SWREF6MI SWREF5MI SWREF4RA S2TC3MA SWREF2ST | 24.500 53.500 36.000 -46.535 2.449 | 0.049 0.045 0.036 0.036 0.031 CV 6 | Numeric Numeric Numeric Numeric Numeric | SWREF6MI SWREF2ST SWREF4RA SWREF5MI SWREF2RA | 25.000 2.449 36.000 53.500 9.500 | 0.031 0.026 0.026 0.026 0.024 CV 7 | Numeric Numeric Numeric Numeric Numeric | SBTC2RA SBREF4RA SBREF4MA SBREF4ST SBREF4MI | 54.136 97.000 144.000 19.409 53.000 | 0.015 0.015 0.013 0.013 0.011 CV 8 | Numeric Numeric Numeric Numeric Numeric | SWREF6MI SWREF5MI SWREF4RA S2TC3MA S2TC3RA | 24.500 53.500 36.000 -46.535 42.453 | 0.029 0.025 0.023 0.021 0.020 CV 9 | Numeric Numeric Numeric Numeric Numeric | SBREF4MA SBTC2RA SBREF4RA SBTC2MA SBREF4ST | 144.000 54.695 97.000 71.267 19.409 | 0.017 0.013 0.013 0.011 0.011 CV 10 | Numeric Numeric Numeric Numeric Numeric | SWREF6MI SWREF5MI SWREF4RA S2TC3MA SWREF2ST | 24.500 53.500 36.000 -46.535 2.449 | 0.048 0.042 0.037 0.037 0.032 FINAL | | | Right Split Competitors Main 1 2 3 4 ------------------------------------------------------------------------

PAGE 204

187 CV 1 | Numeric Numeric Numeric Numeric Numeric | SWTC2MA SWREF5MA S2REF3ME SWREF4MA SWTC2ST | 63.848 109.500 17.386 124.500 7.912 | 0.066 0.059 0.059 0.057 0.053 CV 2 | Numeric Numeric Numeric Numeric Numeric | SBTC1ME SBREF3MI S2REF3MI SBREF4ME SWREF5ST | 129.913 8.500 8.500 132.344 21.889 | 0.034 0.032 0.032 0.026 0.025 CV 3 | Numeric Numeric Numeric Numeric Numeric | SBTC1ME SBREF3MI SBREF3ST SBREF4ME SBTC1ST | 129.913 8.500 3.714 132.290 5.169 | 0.042 0.039 0.031 0.027 0.026 CV 4 | Numeric Numeric Numeric Numeric Numeric | SBTC1ME SWREF5ST SBREF3MI SWREF5RA S2TC2RA | 129.913 21.889 8.500 68.000 30.375 | 0.042 0.036 0.035 0.029 0.028 CV 5 | Numeric Numeric Numeric Numeric Numeric | SBTC1ME SBREF3MI S2TC2ME SBTC1RA SWREF5MI | 129.913 8.500 67.340 31.824 52.000 | 0.034 0.028 0.025 0.023 0.023 CV 6 | Numeric Numeric Numeric Numeric Numeric | SBREF3MI SBTC1ME SBREF4ME S2REF1MI S2REF2ME | 8.500 129.913 133.112 1.500 15.537 | 0.037 0.034 0.029 0.023 0.023 CV 7 | Numeric Numeric Numeric Numeric Numeric | SBTC1ME SWREF5ST SBREF3MI S2TC2ME SBREF4ME | 129.876 21.889 8.500 67.340 132.344 | 0.045 0.041 0.037 0.035 0.033 CV 8 | Numeric Numeric Numeric Numeric Numeric | S2TC2ME SWTC2ST SBTC1RA SWREF5MA SBREF4ME | 67.340 7.948 31.761 107.500 133.112 | 0.035 0.034 0.029 0.028 0.025 CV 9 | Numeric Numeric Numeric Numeric Numeric | SBTC1ME SBREF3MI S2REF3MI SBTC1RA S2REF1MI | 129.913 8.500 8.500 31.761 1.500 | 0.046 0.031 0.031 0.027 0.025 CV 10 | Numeric Numeric Numeric Numeric Numeric | SBREF3MI SBTC1ME SBREF4ME SBTC1RA S2REF2MI | 8.500 129.913 133.112 31.761 10.500 | 0.028 0.026 0.023 0.023 0.023 FINAL | Numeric Numeric Numeric Numeric Numeric | SBTC1ME SBREF3MI SBREF4ME SWREF5ST S2TC2ME | 129.913 8.500 132.344 21.889 67.340 | 0.045 0.032 0.030 0.029 0.028 ========================== Gains and ROC for ETP_CLAS ========================== ETP_CLAS = { 1 } Class Class P of Cum P of Cum N P of Class P of N Pop Cum P Lift Lift Node in Bin Bin in Bin Class in Bin in Bin of Pop Ratio Index -----------------------------------------------------------------------------1 41.00 0.9318 0.5467 0.5467 44.00 0.3761 0.3761 1.45 1.45 2 13.00 0.8667 0.1733 0.7200 15.00 0.1282 0.5043 1.43 1.35 3 21.00 0.3621 0.2800 1.0000 58.00 0.4957 1.0000 1.00 0.56 -----------------------------------------------------------------------------75.00 117.00

PAGE 205

188 Score Threshold Sensitivity Specificity ---------------------------------------0.9318182 0.0000 1.00000 0.8666667 0.5467 0.9286 0.3620690 0.7200 0.8810 0.3620690 1.00000 0.0000 ETP_CLAS = { 2 } Class Class P of Cum P of Cum N P of Class P of N Pop Cum P Lift Lift Node in Bin Bin in Bin Class in Bin in Bin of Pop Ratio Index -----------------------------------------------------------------------------3 37.00 0.6379 0.8810 0.8810 58.00 0.4957 0.4957 1.78 1.78 2 2.00 0.1333 0.0476 0.9286 15.00 0.1282 0.6239 1.49 0.37 1 3.00 0.0682 0.0714 1.0000 44.00 0.3761 1.0000 1.00 0.19 -----------------------------------------------------------------------------42.00 117.00 Score Threshold Sensitivity Specificity ---------------------------------------0.6379310 0.0000 1.00000 0.1333333 0.8810 0.7200 0.0681818 0.9286 0.5467 0.0681818 1.00000 0.0000 C:\Documents and Settings\kamckee\Local Settings\Temp\s1i0595: 10.5 kb Grove file created containing 1 Tree. G:\GIS_Okeechobee\CARTupscaling\NEW files thesis\final edge tree\CART edges thesis.txt: 117 records. Center Soils in NonLandsat Area ====================== Target Frequency Table ====================== Variable: TP_CLAS N Classes: 2 Data Value N Wgt Count -------------------------------------------1 65 65 2 52 52 Total 117 117 Missing Value Prevalence Learn --------------SBLUVA$ 0.0085 SBLUMJ$ 0.0085 OPENWAT 0.0085 Mix Priors

PAGE 206

189 Target Class Prior ------------------------------------------------1 0.528 2 0.472 CURRENT MEMORY REQUIREMENTS TOTAL: 26762. DATA: 1521. ANALYSIS: 26762. AVAILABLE: 13500000. SURPLUS: 13473238. The data are being read ... 117 Observations in the learning sample. FILE: G:\GIS_Okeechobee\CARTupscaling\NEW files thesis\final center tree\CART centers thesis_new.txt CART is running. ============= TREE SEQUENCE ============= Dependent variable: TP_CLAS Terminal Cross-Validated Resubstitution Complexity Tree Nodes Relative Cost Relative Cost Parameter -----------------------------------------------------------------1 18 0.887 +/0.097 0.259 0.000000 2 15 0.909 +/0.097 0.293 0.005434 3 14 0.926 +/0.097 0.307 0.006207 4 12 0.924 +/0.097 0.345 0.009091 5 10 0.945 +/0.097 0.386 0.009572 6 6 0.911 +/0.097 0.539 0.018172 7 5 0.812 +/0.092 0.589 0.023407 8** 3 0.779 +/0.090 0.689 0.023675 9 2 0.859 +/0.095 0.784 0.044882 10 1 1.000 +/0.000 1.000 0.101933 Initial misclassification cost = 0.472 Initial class assignment = 1 ================ NODE INFORMATION ================ ********************************************* Node 1: SBSUBGR$ * N: 117 ********************************************* ================================= ********************************* = Terminal Node 1 = Node 2 = N: 66 = N: 51 ================================= ********************************* Node 1 was split on SBSUBGR$ A case goes left if SBSUBGR$ = ("Arenic Haplaquod","Argiaquoll", "Endoaqualf","Spodic Psammaquent") Improvement = 0.007409 Complexity Threshold = 0.101923 Node Cases Wgt Counts Cost Class 1 117 117.00 0.472 1

PAGE 207

190 -1 66 66.00 0.359 1 2 51 51.00 0.385 2 Weighted Counts Class Top Left Right 1 65.00 44.00 21.00 2 52.00 22.00 30.00 Within Node Probabilities Class Top Left Right 1 0.528 0.641 0.385 2 0.472 0.359 0.615 Surrogate Split Assoc Improvement 1 OPENWAT r 9.000 0.264 0.004 2 SWDIMJMI s 3503.355 0.194 0.000 3 BASIN$ s "S-154","S-191","S-65E" 0.189 0.002 4 VEGLU$ s "emdair","emimp","emunimp" 0.178 0.008 5 SWORDER$ s "E","M" 0.143 0.007 Competitor Split Improvement 1 VEGLU$ "emdair","fossdai","fossimp" 0.006 2 OPENWAT 2.000 0.004 3 BASIN$ "S-154","S-65D","S-65E" 0.004 4 SWPERIM 283.244 0.004 5 SBLUMJ$ "1","2","3","4" 0.003 ********************************************* Node 2: SWPERIM * N: 51 ********************************************* ================================= ================================= = Terminal Node 2 = = Terminal Node 3 = = N: 14 = = N: 37 = ================================= ================================= Node 2 was split on SWPERIM A case goes left if SWPERIM <= 323.239 Improvement = 0.002408 Complexity Threshold = 0.044872 Node Cases Wgt Counts Cost Class 2 51 51.00 0.385 2 -2 14 14.00 0.309 1 -3 37 37.00 0.274 2 Weighted Counts Class Top Left Right 1 21.00 10.00 11.00 2 30.00 4.00 26.00 Within Node Probabilities Class Top Left Right 1 0.385 0.691 0.274 2 0.615 0.309 0.726 Surrogate Split Assoc Improvement 1 SWHECT s 0.569 0.707 0.005 2 SWDIMJMI s 501.316 0.130 0.000 3 SWDIRDMI s 7.500 0.077 0.000 4 SWORDER$ s "E","M" 0.069 0.001 5 VEGLU$ s "fossuni" 0.069 0.001

PAGE 208

191 Competitor Split Improvement 1 SWHECT 0.882 0.002 2 SWDIDIMI 18.107 0.001 3 SWDIRDMI 1355.090 0.001 4 SBLUMJ$ "1","2","4" 0.001 5 VEGLU$ "emdair","fossdai","fossimp" 0.001 ========================= TERMINAL NODE INFORMATION ========================= [Breiman adjusted cost, lambda = 0.021] Node N Prob Cost Class ----------------------------------------------------------------------------1 66 0.5571 0.3586 1 Parent C.T. = 0.102 [0.4153] 44 0.6414 1 22 0.3586 2 2 14 0.1175 0.3091 1 Parent C.T. = 0.045 [0.5453] 10 0.6909 1 4 0.3091 2 3 37 0.3254 0.2745 2 Parent C.T. = 0.045 [0.3689] 11 0.2745 1 26 0.7255 2 Node Learn 1 66.00 44.00 22.00 2 14.00 10.00 4.00 3 37.00 11.00 26.00 ========================== MISCLASSIFICATION BY CLASS ========================== (Cross Validation) Prior Wgt Class Prob Wgt Count Count Misclass Misclass Cost ---------------------------------------------------------------------------1 0.528 65.00 65 11.00 11 0.169 (65.00 65 14.00 14 0.215) 2 0.472 52.00 52 26.00 26 0.500 (52.00 52 28.00 28 0.538) ---------------------------------------------------------------------------Total 1.000 117.00 117 37.00 37 (117.00 117 42.00 42) ===================

PAGE 209

192 VARIABLE IMPORTANCE =================== Relative Number Of Importance Categories Penalty --------------------------------------------------------------------------VEGLU$ 100.000 6 SWORDER$ 84.199 5 SBSUBGR$ 82.849 7 SWHECT 58.700 OPENWAT 47.475 SWPERIM 26.923 BASIN$ 18.701 4 SWDIMJMI 3.287 SWDIRDMI 2.767 SBLUMJ$ 0.000 5 SBLUVA$ 0.000 4 SWDIDIMI 0.000 =============== OPTION SETTINGS =============== Construction Rule Ordered Twoing Estimation Method 10-fold cross-validation Misclassification Costs Unit Tree Selection 0.000 se rule Linear Combinations No Initial value of the complexity parameter = 0.000 Minimum size below which node will not be split = 10 Node size above which sub-sampling will be used = 117 Maximum number of surrogates used for missing values = 5 Number of surrogate splits printed = 5 Number of competing splits printed = 5 Maximum number of trees printed in the tree sequence = 10 Max. number of cases allowed in the learning sample = 117 Maximum number of cases allowed in the test sample = 0 Max # of nonterminal nodes in the largest tree grown = 117 (Actual # of nonterminal nodes in largest tree grown = 19) Max. no. of categorical splits including surrogates = 1000 Max. number of linear combination splits in a tree = 0 (Actual number cat. + linear combination splits = 32) Maximum depth of largest tree grown = 11 (Actual depth of largest tree grown = 7) Exponent for center weighting in split criterion = 0.560 Maximum size of memory available = 13500000 (Actual size of memory used in run = 75390) =========================== CV-tree Competitor List =========================== (Type, Predictor, Split if continuous, Improvement) Top Split Competitors Main 1 2 3 4 -----------------------------------------------------------------------CV 1 | Categorical Categorical Categorical Numeric Numeric | SBSUBGR$ VEGLU$ BASIN$ OPENWAT SWDIRDMI | 2.000 265.402 | 0.009 0.006 0.006 0.004 0.003 CV 2 | Categorical Numeric Numeric Categorical Numeric

PAGE 210

193 | SBSUBGR$ OPENWAT SWPERIM VEGLU$ SWHECT | 2.000 283.244 0.389 | 0.010 0.006 0.006 0.005 0.004 CV 3 | Categorical Categorical Numeric Numeric Categorical | SBSUBGR$ VEGLU$ OPENWAT SWDIRDMI BASIN$ | 2.000 265.402 | 0.006 0.006 0.004 0.003 0.003 CV 4 | Categorical Categorical Numeric Numeric Categorical | SBSUBGR$ VEGLU$ SWPERIM OPENWAT SBLUMJ$ | 283.244 2.500 | 0.012 0.007 0.005 0.004 0.004 CV 5 | Categorical Categorical Categorical Numeric Categorical | VEGLU$ SBSUBGR$ BASIN$ OPENWAT SBLUMJ$ | 2.000 | 0.005 0.004 0.004 0.003 0.003 CV 6 | Categorical Categorical Numeric Numeric Categorical | SBSUBGR$ VEGLU$ OPENWAT SWDIDIMI BASIN$ | 2.000 90.621 | 0.006 0.005 0.005 0.004 0.004 CV 7 | Categorical Categorical Numeric Categorical Numeric | SBSUBGR$ VEGLU$ SWPERIM BASIN$ SWHECT | 283.244 0.382 | 0.006 0.004 0.004 0.003 0.003 CV 8 | Categorical Numeric Categorical Categorical Numeric | SBSUBGR$ SWPERIM VEGLU$ BASIN$ SWHECT | 307.131 0.382 | 0.007 0.006 0.006 0.005 0.004 CV 9 | Categorical Categorical Categorical Numeric Numeric | SBSUBGR$ VEGLU$ BASIN$ SWPERIM OPENWAT | 307.131 2.000 | 0.010 0.007 0.005 0.004 0.004 CV 10 | Categorical Categorical Numeric Numeric Numeric | SBSUBGR$ VEGLU$ SWPERIM SWDIRDMI OPENWAT | 307.131 265.402 2.000 | 0.006 0.006 0.004 0.003 0.003 FINAL | Categorical Categorical Numeric Categorical Numeric | SBSUBGR$ VEGLU$ OPENWAT BASIN$ SWPERIM | 2.000 283.244 | 0.007 0.006 0.004 0.004 0.004 Left Split Competitors Main 1 2 3 4 -----------------------------------------------------------------------CV 1 | Categorical Numeric Numeric Numeric Categorical | BASIN$ SWHECT SWDIRDMI SWPERIM SBLUVA$ | 0.668 236.673 404.048 | 0.003 0.003 0.002 0.002 0.001 CV 2 | Numeric Numeric Numeric Numeric Numeric | SWPERIM SWHECT SWDIRDMI SWDIDIMI OPENWAT | 323.239 0.882 1355.090 18.107 2.000 | 0.002 0.002 0.002 0.001 0.001 CV 3 | Categorical Numeric Categorical Numeric Categorical | BASIN$ SWHECT SBLUVA$ SWPERIM VEGLU$ | 0.668 404.048 | 0.003 0.002 0.002 0.002 0.001 CV 4 | Categorical Categorical Numeric Numeric Numeric | BASIN$ SBLUVA$ SWHECT SWPERIM OPENWAT | 0.668 391.795 16.000 | 0.002 0.002 0.002 0.002 0.001 CV 5 | Categorical Numeric Numeric Numeric Numeric | SBLUVA$ SWDIMJMI SWHECT SWPERIM SWDIRDMI | 855.000 1.187 834.413 485.411 | 0.001 0.001 0.000 0.000 0.000

PAGE 211

194 CV 6 | Categorical Numeric Categorical Numeric Numeric | BASIN$ SWHECT SBLUVA$ SWPERIM SWDIRDMI | 0.668 404.048 275.643 | 0.003 0.002 0.001 0.001 0.001 CV 7 | Categorical Categorical Numeric Numeric Numeric | SBLUVA$ BASIN$ SWHECT SWPERIM OPENWAT | 0.668 408.721 13.500 | 0.003 0.002 0.002 0.001 0.001 CV 8 | Categorical Categorical Numeric Numeric Numeric | BASIN$ SBLUMJ$ SWPERIM SWHECT SWDIDIMI | 200.321 2.175 90.621 | 0.003 0.001 0.001 0.001 0.001 CV 9 | Categorical Numeric Numeric Numeric Numeric | BASIN$ SWPERIM SWHECT SWDIDIMI SWDIRDMI | 404.048 0.668 134.790 285.220 | 0.004 0.002 0.002 0.001 0.001 CV 10 | Numeric Numeric Categorical Numeric Numeric | SWHECT SWPERIM BASIN$ OPENWAT SWDIRDMI | 0.668 404.048 13.500 275.643 | 0.002 0.002 0.002 0.002 0.001 FINAL | | | Right Split Competitors Main 1 2 3 4 -----------------------------------------------------------------------CV 1 | Numeric Numeric Numeric Numeric Categorical | SWPERIM SWDIRDMI SWHECT OPENWAT SBLUMJ$ | 323.239 1561.530 0.882 2.000 | 0.002 0.002 0.001 0.001 0.001 CV 2 | Categorical Numeric Numeric Numeric Categorical | BASIN$ OPENWAT SWDIDIMI SWPERIM SBLUVA$ | 16.000 90.621 404.048 | 0.002 0.001 0.001 0.001 0.001 CV 3 | Numeric Numeric Numeric Categorical Numeric | SWPERIM SWHECT SWDIDIMI VEGLU$ SWDIRDMI | 323.239 0.896 18.107 713.775 | 0.003 0.003 0.001 0.001 0.001 CV 4 | Numeric Numeric Numeric Categorical Categorical | SWHECT SWPERIM SWDIRDMI SBLUMJ$ VEGLU$ | 0.882 323.239 1355.090 | 0.004 0.004 0.002 0.002 0.002 CV 5 | Numeric Numeric Categorical Numeric Categorical | SWDIRDMI SWDIMJMI SBSUBGR$ SWDIDIMI BASIN$ | 226.737 694.587 90.621 | 0.002 0.002 0.002 0.001 0.001 CV 6 | Numeric Numeric Numeric Numeric Numeric | SWHECT SWPERIM SWDIRDMI SWDIDIMI OPENWAT | 0.882 323.239 1355.090 18.107 2.000 | 0.002 0.002 0.002 0.001 0.001 CV 7 | Numeric Numeric Numeric Categorical Categorical | SWHECT SWPERIM SWDIDIMI SBLUMJ$ SBLUVA$ | 0.882 349.375 18.107 | 0.002 0.002 0.002 0.001 0.001 CV 8 | Numeric Numeric Numeric Numeric Categorical | SWHECT SWPERIM SWDIRDMI SWDIDIMI SBLUMJ$ | 0.882 347.957 1355.090 18.107 | 0.002 0.002 0.001 0.001 0.001 CV 9 | Numeric Numeric Numeric Categorical Numeric | SWPERIM SWHECT SWDIRDMI VEGLU$ SWDIDIMI | 323.239 0.882 1355.090 18.107 | 0.004 0.003 0.001 0.001 0.001

PAGE 212

195 CV 10 | Numeric Numeric Numeric Categorical Categorical | SWPERIM SWHECT SWDIDIMI SBLUMJ$ VEGLU$ | 323.239 0.876 18.107 | 0.003 0.002 0.001 0.001 0.001 FINAL | Numeric Numeric Numeric Numeric Categorical | SWPERIM SWHECT SWDIDIMI SWDIRDMI SBLUMJ$ | 323.239 0.882 18.107 1355.090 | 0.002 0.002 0.001 0.001 0.001 ========================= Gains and ROC for TP_CLAS ========================= TP_CLAS = { 1 } Class Class P of Cum P of Cum N P of Class P of N Pop Cum P Lift Lift Node in Bin Bin in Bin Class in Bin in Bin of Pop Ratio Index -----------------------------------------------------------------------------2 10.00 0.7143 0.1538 0.1538 14.00 0.1197 0.1197 1.29 1.29 1 44.00 0.6667 0.6769 0.8308 66.00 0.5641 0.6838 1.22 1.20 3 11.00 0.2973 0.1692 1.0000 37.00 0.3162 1.0000 1.00 0.54 -----------------------------------------------------------------------------65.00 117.00 Score Threshold Sensitivity Specificity ---------------------------------------0.7142857 0.0000 1.00000 0.6666667 0.1538 0.9231 0.2972973 0.8308 0.5000 0.2972973 1.00000 0.0000 TP_CLAS = { 2 } Class Class P of Cum P of Cum N P of Class P of N Pop Cum P Lift Lift Node in Bin Bin in Bin Class in Bin in Bin of Pop Ratio Index -----------------------------------------------------------------------------3 26.00 0.7027 0.5000 0.5000 37.00 0.3162 0.3162 1.58 1.58 1 22.00 0.3333 0.4231 0.9231 66.00 0.5641 0.8803 1.05 0.75 2 4.00 0.2857 0.0769 1.0000 14.00 0.1197 1.0000 1.00 0.64 -----------------------------------------------------------------------------52.00 117.00 Score Threshold Sensitivity Specificity ---------------------------------------0.7027027 0.0000 1.00000 0.3333333 0.5000 0.8308 0.2857143 0.9231 0.1538 0.2857143 1.00000 0.0000 C:\Documents and Settings\kamckee\Local Settings\Temp\s1cg500: 18.5 kb Grove file created containing 1 Tree. G:\GIS_Okeechobee\CARTupscaling\NEW files thesis\final center tree\CART centers thesis_new.txt: 117 records.

PAGE 213

196 Edge Soils in NonLandsat Area ====================== Target Frequency Table ====================== Variable: ETP_CLAS$ N Classes: 2 Data Value N Wgt Count ------------------------------------------------------------1 75 75 2 42 42 Total 117 117 Missing Value Prevalence Learn --------------NWIMOD$ 0.0342 SBLUVA$ 0.0085 SBLUMJ$ 0.0085 PRIORS SET EQUAL CURRENT MEMORY REQUIREMENTS TOTAL: 35104. DATA: 1989. ANALYSIS: 35104. AVAILABLE: 13500000. SURPLUS: 13464896. The data are being read ... 117 Observations in the learning sample. FILE: G:\GIS_Okeechobee\CARTupscaling\NEW files thesis\final edge tree\CART edges thesisx.txt CART is running. ============= TREE SEQUENCE ============= Dependent variable: ETP_CLAS$ Terminal Cross-Validated Resubstitution Complexity Tree Nodes Relative Cost Relative Cost Parameter -----------------------------------------------------------------1 12 0.765 +/0.094 0.210 0.000000 2 11 0.744 +/0.093 0.221 0.005259 3 9 0.730 +/0.093 0.250 0.007153 4 5 0.650 +/0.091 0.342 0.011558 5** 4 0.533 +/0.086 0.371 0.014772 6 2 0.577 +/0.083 0.530 0.039534 7 1 1.000 +/0.000 1.000 0.235248 Initial misclassification cost = 0.500 Initial class assignment = 1 ================ NODE INFORMATION

PAGE 214

197 ================ ********************************************* Node 1: BASIN$ * N: 117 ********************************************* ================================= ********************************* = Terminal Node 1 = Node 2 = N: 52 = N: 65 ================================= ********************************* Node 1 was split on BASIN$ A case goes left if BASIN$ = ("S-154","S-65D") Improvement = 0.069877 Complexity Threshold = 0.235238 Node Cases Wgt Counts Cost Class 1 117 117.00 0.500 1 -1 52 52.00 0.189 1 2 65 65.00 0.311 2 Weighted Counts Class Top Left Right 1 75.00 46.00 29.00 2 42.00 6.00 36.00 Within Node Probabilities Class Top Left Right 1 0.500 0.811 0.311 2 0.500 0.189 0.689 Surrogate Split Assoc Improvement 1 SBCOMPO$ s "BASINGER","MANATEE","VALKARIA" 0.184 0.018 2 SWHECT r 1.209 0.178 0.002 3 SBSUBG2$ s "Psammaquent" 0.166 0.016 4 SWPERIM r 516.805 0.162 0.008 5 SBSUBGR$ s "Spodic Psammaquent" 0.149 0.014 Competitor Split Improvement 1 VEGLU$ "emimp","emunimp" 0.045 2 SWDIHIMI 1398.275 0.041 3 GENVE$ "E" 0.026 4 VTYP$ "EM","EW" 0.026 5 SBLUMJ$ "1","3","4" 0.025 ********************************************* Node 2: SWDIHIMI * N: 65 ********************************************* ================================= ********************************* = Terminal Node 2 = Node 3 = N: 25 = N: 40 ================================= ********************************* Node 2 was split on SWDIHIMI A case goes left if SWDIHIMI <= 1849.350 Improvement = 0.025521 Complexity Threshold = 0.039524 Node Cases Wgt Counts Cost Class 2 65 65.00 0.311 2 -2 25 25.00 0.071 2

PAGE 215

198 3 40 40.00 0.490 1 Weighted Counts Class Top Left Right 1 29.00 3.00 26.00 2 36.00 22.00 14.00 Within Node Probabilities Class Top Left Right 1 0.311 0.071 0.510 2 0.689 0.929 0.490 Surrogate Split Assoc Improvement 1 VEGLU$ s "emdair","fossdai","fossimp", 0.633 0.017 "fossuni" 2 SBLUMJ$ s "2","5" 0.549 0.012 3 SBSUBGR$ s "Aeric Haplaquod", 0.346 0.016 "Arenic Ochraqualf","Argiaquoll", "Medisaprist" 4 VTYP$ s "EW","FO","FS","SS" 0.323 0.002 5 SBCOMPO$ s "FLORIDANA","MANATEE","MYAKKA", 0.304 0.018 "OKEELANTA","PINELLAS","VALKARIA" Competitor Split Improvement 1 SBCOMPO$ "BASINGER","BRADENTON","IMMOKALEE","MANATEE", 0.018 "VALKARIA","WAVELAND" 2 VEGLU$ "emimp","emunimp","fossuni" 0.017 3 SBSUBGR$ "Arenic Haplaquod","Endoaqualf", 0.016 "Spodic Psammaquent" 4 SWPERIM 323.239 0.014 5 SBLUMJ$ "1","3","4" 0.012 ********************************************* Node 3: SWDIMJMI * N: 40 ********************************************* ================================= ================================= = Terminal Node 3 = = Terminal Node 4 = = N: 23 = = N: 17 = ================================= ================================= Node 3 was split on SWDIMJMI A case goes left if SWDIMJMI <= 1601.840 Improvement = 0.009398 Complexity Threshold = 0.072381 Node Cases Wgt Counts Cost Class 3 40 40.00 0.490 1 -3 23 23.00 0.273 1 -4 17 17.00 0.282 2 Weighted Counts Class Top Left Right 1 26.00 19.00 7.00 2 14.00 4.00 10.00 Within Node Probabilities Class Top Left Right 1 0.510 0.727 0.282 2 0.490 0.273 0.718 Surrogate Split Assoc Improvement

PAGE 216

199 1 SWDIRDMI s 569.230 0.592 0.002 2 SBCOMPO$ s "IMMOKALEE","MANATEE", 0.359 0.005 "OKEELANTA","PINELLAS","VALKARIA", "WAVELAND" 3 SBSUBGR$ s "Arenic Haplaquod", 0.287 0.003 "Arenic Ochraqualf","Medisaprist" 4 SWHECT s 1.239 0.256 0.000 5 SWPERIM s 352.160 0.247 0.000 Competitor Split Improvement 1 SBCOMPO$ "BRADENTON","IMMOKALEE","MANATEE","OKEELANTA", 0.007 "PINELLAS","VALKARIA","WAVELAND" 2 SWDIRDMI 321.302 0.006 3 SWDIWEMI 102.812 0.006 4 SBLUMJ$ "2","3","4","5" 0.005 5 SBSUBGR$ "Arenic Haplaquod","Arenic Ochraqualf", 0.004 "Endoaqualf","Medisaprist" ========================= TERMINAL NODE INFORMATION ========================= [Breiman adjusted cost, lambda = 0.046] Node N Prob Cost Class ----------------------------------------------------------------------------1 52 0.3781 0.1889 1 Parent C.T. = 0.235 [0.3037] 46 0.8111 1 6 0.1889 2 2 25 0.2819 0.0709 2 Parent C.T. = 0.040 [0.2194] 3 0.0709 1 22 0.9291 2 3 23 0.1743 0.2732 1 Parent C.T. = 0.072 [0.4943] 19 0.7268 1 4 0.2732 2 4 17 0.1657 0.2816 2 Parent C.T. = 0.072 [0.5117] 7 0.2816 1 10 0.7184 2 Node Learn 1 52.00 46.00 6.00 2 25.00 3.00 22.00 3 23.00 19.00 4.00 4 17.00 7.00 10.00 ==========================

PAGE 217

200 MISCLASSIFICATION BY CLASS ========================== (Cross Validation) Prior Wgt Class Prob Wgt Count Count Misclass Misclass Cost ---------------------------------------------------------------------------1 0.500 75.00 75 10.00 10 0.133 (75.00 75 15.00 15 0.200) 2 0.500 42.00 42 10.00 10 0.238 (42.00 42 14.00 14 0.333) ---------------------------------------------------------------------------Total 1.000 117.00 117 20.00 20 (117.00 117 29.00 29) =================== VARIABLE IMPORTANCE =================== Relative Number Of Importance Categories Penalty --------------------------------------------------------------------------BASIN$ 100.000 4 SBCOMPO$ 58.503 10 SBSUBGR$ 46.830 7 SWDIHIMI 36.523 VEGLU$ 24.407 6 SBSUBG2$ 22.476 6 SBLUMJ$ 17.645 5 SWDIMJMI 13.449 SWPERIM 10.795 VTYP$ 2.916 5 SWDIRDMI 2.879 SWHECT 2.794 SWDIWEMI 0.000 GENVE$ 0.000 2 NWIMOD$ 0.000 6 SBLUVA$ 0.000 4 =============== OPTION SETTINGS =============== Construction Rule Gini (priors altered by costs) Estimation Method 10-fold cross-validation Misclassification Costs Unit Tree Selection 0.000 se rule Linear Combinations Minimum node size is 3 Variable deletion parameter is 0.200 Initial value of the complexity parameter = 0.000 Minimum size below which node will not be split = 10 Node size above which sub-sampling will be used = 117 Maximum number of surrogates used for missing values = 5 Number of surrogate splits printed = 5 Number of competing splits printed = 5 Maximum number of trees printed in the tree sequence = 10 Max. number of cases allowed in the learning sample = 117 Maximum number of cases allowed in the test sample = 0 Max # of nonterminal nodes in the largest tree grown = 117 (Actual # of nonterminal nodes in largest tree grown = 14)

PAGE 218

201 Max. no. of categorical splits including surrogates = 1000 Max. number of linear combination splits in a tree = 100 (Actual number cat. + linear combination splits = 36) Maximum depth of largest tree grown = 11 (Actual depth of largest tree grown = 6) Exponent for center weighting in split criterion = 0.360 Maximum size of memory available = 13500000 (Actual size of memory used in run = 86929) =========================== CV-tree Competitor List =========================== (Type, Predictor, Split if continuous, Improvement) Top Split Competitors Main 1 2 3 4 -----------------------------------------------------------------------CV 1 | Categorical Categorical Numeric Categorical Categorical | BASIN$ VEGLU$ SWDIHIMI GENVE$ VTYP$ | 1398.275 | 0.073 0.053 0.052 0.030 0.030 CV 2 | Categorical Categorical Categorical Categorical Numeric | BASIN$ VEGLU$ GENVE$ VTYP$ SWDIHIMI | 1398.275 | 0.061 0.048 0.035 0.035 0.030 CV 3 | Linear Comb Categorical Categorical Numeric Categorical | BASIN$ VEGLU$ SWDIHIMI SBCOMPO$ | 1398.275 | 0.062 0.061 0.045 0.044 0.028 CV 4 | Categorical Categorical Numeric Categorical Categorical | BASIN$ VEGLU$ SWDIHIMI GENVE$ VTYP$ | 1421.730 | 0.073 0.053 0.046 0.030 0.030 CV 5 | Categorical Numeric Categorical Categorical Categorical | BASIN$ SWDIHIMI VEGLU$ SBCOMPO$ GENVE$ | 1398.275 | 0.082 0.044 0.041 0.031 0.025 CV 6 | Categorical Categorical Numeric Categorical Categorical | BASIN$ VEGLU$ SWDIHIMI SBLUMJ$ GENVE$ | 1398.275 | 0.079 0.050 0.040 0.031 0.026 CV 7 | Categorical Categorical Numeric Numeric Categorical | BASIN$ VEGLU$ SWDIHIMI SWDIMJMI SBLUMJ$ | 2749.765 732.627 | 0.073 0.040 0.038 0.024 0.024 CV 8 | Categorical Categorical Numeric Categorical Numeric | BASIN$ VEGLU$ SWDIHIMI SBCOMPO$ SWDIMJMI | 1359.685 715.701 | 0.063 0.039 0.039 0.025 0.023 CV 9 | Categorical Numeric Categorical Categorical Categorical | BASIN$ SWDIHIMI VEGLU$ SBLUMJ$ SBCOMPO$ | 1398.275 | 0.070 0.045 0.044 0.027 0.024 CV 10 | Categorical Categorical Numeric Categorical Categorical | BASIN$ VEGLU$ SWDIHIMI GENVE$ VTYP$ | 1398.275 | 0.066 0.042 0.038 0.030 0.030 FINAL | Categorical Categorical Numeric Categorical Categorical | BASIN$ VEGLU$ SWDIHIMI GENVE$ VTYP$ | 1398.275 | 0.070 0.045 0.041 0.026 0.026

PAGE 219

202 Left Split Competitors Main 1 2 3 4 -----------------------------------------------------------------------CV 1 | Linear Comb Numeric Numeric Categorical Numeric | SWDIRDMI SWDIMJMI NWIMOD$ SWDIWEMI | 684.937 1363.185 249.269 | 0.008 0.005 0.005 0.003 0.003 CV 2 | Categorical Categorical Categorical Numeric Numeric | SBSUBG2$ SBCOMPO$ SBSUBGR$ SWDIWEMI SWDIMJMI | 249.035 715.701 | 0.005 0.005 0.004 0.004 0.003 CV 3 | Categorical Numeric Categorical Numeric Numeric | BASIN$ SWDIMJMI SBCOMPO$ SWDIHIMI SWDIWEMI | 1250.385 1740.210 128.598 | 0.019 0.011 0.008 0.007 0.007 CV 4 | Categorical Categorical Categorical Categorical Numeric | SBLUVA$ GENVE$ VTYP$ VEGLU$ SWDIRDMI | 1818.715 | 0.003 0.003 0.003 0.003 0.003 CV 5 | Linear Comb Categorical Numeric Categorical Categorical | NWIMOD$ SWDIWEMI SBSUBGR$ SBCOMPO$ | 249.269 | 0.005 0.004 0.004 0.004 0.004 CV 6 | Categorical Categorical Categorical Categorical Categorical | SBLUMJ$ SBLUVA$ GENVE$ VTYP$ VEGLU$ | 0.004 0.003 0.003 0.003 0.003 CV 7 | Linear Comb Categorical Categorical Categorical Categorical | SBSUBG2$ SBCOMPO$ SBSUBGR$ BASIN$ | 0.005 0.004 0.004 0.003 0.003 CV 8 | Linear Comb Categorical Categorical Categorical Categorical | SBSUBG2$ SBCOMPO$ SBSUBGR$ NWIMOD$ | 0.007 0.004 0.004 0.004 0.004 CV 9 | Categorical Categorical Categorical Numeric Categorical | SBSUBG2$ SBCOMPO$ SBSUBGR$ SWDIMJMI NWIMOD$ | 715.701 | 0.004 0.004 0.004 0.004 0.003 CV 10 | Categorical Categorical Categorical Categorical Categorical | SBSUBG2$ SBCOMPO$ BASIN$ SBSUBGR$ NWIMOD$ | 0.004 0.004 0.004 0.004 0.003 FINAL | | | Right Split Competitors Main 1 2 3 4 -----------------------------------------------------------------------CV 1 | Linear Comb Numeric Categorical Categorical Numeric | SWDIHIMI VEGLU$ SBCOMPO$ SWPERIM | 1421.730 323.239 | 0.041 0.029 0.020 0.019 0.018 CV 2 | Linear Comb Numeric Categorical Categorical Numeric | SWDIHIMI VEGLU$ SBCOMPO$ SWPERIM | 1883.650 323.239 | 0.042 0.021 0.018 0.017 0.015 CV 3 | Categorical Categorical Categorical Categorical Categorical | BASIN$ SBSUBG2$ SBCOMPO$ VTYP$ VEGLU$ | 0.005 0.003 0.003 0.003 0.003 CV 4 | Numeric Categorical Categorical Numeric Categorical | SWDIHIMI VEGLU$ SBCOMPO$ SWPERIM SBLUMJ$ | 1849.350 323.239 | 0.027 0.020 0.016 0.015 0.014 CV 5 | Numeric Categorical Categorical Numeric Categorical | SWDIHIMI SBCOMPO$ VEGLU$ SWPERIM SBSUBGR$

PAGE 220

203 | 1421.730 323.239 | 0.028 0.019 0.019 0.019 0.017 CV 6 | Numeric Categorical Categorical Categorical Numeric | SWDIHIMI SBCOMPO$ VEGLU$ SBSUBGR$ SWPERIM | 1849.350 323.239 | 0.030 0.020 0.018 0.017 0.016 CV 7 | Numeric Categorical Numeric Categorical Categorical | SWDIHIMI SBCOMPO$ SWDIMJMI VEGLU$ SBSUBGR$ | 1849.350 738.256 | 0.020 0.014 0.012 0.012 0.011 CV 8 | Numeric Numeric Categorical Numeric Categorical | SWDIHIMI SWHECT SBCOMPO$ SWPERIM VEGLU$ | 1849.350 0.418 313.185 | 0.025 0.018 0.018 0.017 0.015 CV 9 | Numeric Categorical Categorical Categorical Categorical | SWDIHIMI SBCOMPO$ SBSUBGR$ VEGLU$ SBLUMJ$ | 1849.350 | 0.024 0.019 0.017 0.013 0.010 CV 10 | Numeric Categorical Categorical Numeric Categorical | SWDIHIMI SBSUBGR$ SBCOMPO$ SWPERIM VEGLU$ | 1849.350 317.568 | 0.025 0.021 0.021 0.019 0.016 FINAL | Numeric Categorical Categorical Categorical Numeric | SWDIHIMI SBCOMPO$ VEGLU$ SBSUBGR$ SWPERIM | 1849.350 323.239 | 0.026 0.018 0.017 0.016 0.014 =========================== Gains and ROC for ETP_CLAS$ =========================== ETP_CLAS$ = { 1 } Class Class P of Cum P of Cum N P of Class P of N Pop Cum P Lift Lift Node in Bin Bin in Bin Class in Bin in Bin of Pop Ratio Index -----------------------------------------------------------------------------1 46.00 0.8846 0.6133 0.6133 52.00 0.4444 0.4444 1.38 1.38 3 19.00 0.8261 0.2533 0.8667 23.00 0.1966 0.6410 1.35 1.29 4 7.00 0.4118 0.0933 0.9600 17.00 0.1453 0.7863 1.22 0.64 2 3.00 0.1200 0.0400 1.0000 25.00 0.2137 1.0000 1.00 0.19 -----------------------------------------------------------------------------75.00 117.00 Score Threshold Sensitivity Specificity ---------------------------------------0.8846154 0.0000 1.00000 0.8260870 0.6133 0.8571 0.4117647 0.8667 0.7619 0.1200000 0.9600 0.5238 0.1200000 1.00000 0.0000 ETP_CLAS$ = { 2 } Class Class P of Cum P of Cum N P of Class P of N Pop Cum P Lift Lift Node in Bin Bin in Bin Class in Bin in Bin of Pop Ratio Index -----------------------------------------------------------------------------2 22.00 0.8800 0.5238 0.5238 25.00 0.2137 0.2137 2.45 2.45 4 10.00 0.5882 0.2381 0.7619 17.00 0.1453 0.3590 2.12 1.64 3 4.00 0.1739 0.0952 0.8571 23.00 0.1966 0.5556 1.54 0.48

PAGE 221

204 1 6.00 0.1154 0.1429 1.0000 52.00 0.4444 1.0000 1.00 0.32 -----------------------------------------------------------------------------42.00 117.00 Score Threshold Sensitivity Specificity ---------------------------------------0.8800000 0.0000 1.00000 0.5882353 0.5238 0.9600 0.1739130 0.7619 0.8667 0.1153846 0.8571 0.6133 0.1153846 1.00000 0.0000 C:\Documents and Settings\kamckee\Local Settings\Temp\s1mo167: 15.9 kb Grove file created containing 1 Tree. G:\GIS_Okeechobee\CARTupscaling\NEW files thesis\final edge tree\CART edges thesisx.txt: 117 records.

PAGE 222

205 APPENDIX J SCRIPTS FOR CLASSIFYING UNSAMP LED WETLANDS AND CALCULATING STORAGE Classifying Unsampled Wetlands Note: The following MS Visual Basic (Mic rosoft, Redmond, Washi ngton) scripts were used in ArcMapÂ’s attribute calculator to classify unsampled wetlands. The variable y represents the TP class of th e zone where 1 = low TP and 2 = high TP. Variables were coded according to a naming convention described below. Beginning letters of the variable name UW = unsampled wetland, UU = unsampled 75m upland buffer Middle letters of the variable name REF* = reflectance (where is a band, 1 through 6) TC1 = tasseled cap band 1 (brightness), Ending letters of the variable name MA = maximum ST = standard deviation ME = mean RA = range Centers in Landsat Area Dim y as double a = [UWREF6ST] b = [UUREF2ME] c = [UWREF4ST] d = [UWREF5RA] e = [UWREF2ST] if a > 10.9 then y = 1 if (a <= 10.9 and b <= 17.261) then y = 1 if (a <= 10.9 and b > 17.261 and c > 12.719) then y = 2 if (a <= 10.9 and b > 17.261 and c <= 12.719 and d > 49.5) then y = 1 if (a <= 10.9 and b > 17.261 and c <= 12.719 and d <= 49.5 and e > 1.580) then y = 2 if (a <= 10.9 and b > 17.261 and c <= 12.719 and d <= 49.5 and e <= 1.580) then y = 1

PAGE 223

206 Edges in Landsat Area Dim y as double a = [UWREF4MA] b = [UUTC1ME] if a <= 124.5 then y = 1 if (a > 124.5 and b <= 129.913) then y = 1 if (a > 124.5 and b > 129.913) then y = 2 Centers in nonLandsat Area Note: Two scripts were used in sequence to classify centers that were outside the Landsat area. In the first script, z repres ents a field that holds the classification based only on soil. It is re ferred to in the second script. Dim z as double a = [subgroup] z = 0 if a = "Arenic Alaquod" then z = 1 if a = "Arenic Argiaquoll" then z = 1 if a = "Spodic Psammaquent" then z = 1 Dim y as double z = [lowsoil] b = [perimeter] if z = 1 then y = 1 if (z = 0 and b <= 323.25) then y = 1 if (z = 0 and b > 323.25) then y = 2 Edges in nonLandsat Area Dim y as double a = [dist_hia] b = [dist_mjrd] if a <= 1849.35 then y = 2 if (a > 1849.35 and b > 1601.84) then y = 2 if (a > 1849.35 and b <= 1601.84) then y = 1

PAGE 224

207 Calculating Storage in Unsampled Wetlands Note: The following scripts were used in Ar cMapÂ’s attribute calculator to calculate storage in the hydrologic z ones of unsampled wetlands. The variable y represents TP (g m-2) in the top 10 cm of soil. Edges Dim y as double a = [pct_edg] b = [EDG_CLAS] c = [HECTARES] if b = 2 then y = 645.9 0.66 a c 10000 / 10 if b = 1 then y = 135.44 1.02 a c 10000 / 10 Centers Dim y as double a = [pct_ctr] b = [CTR_CLAS] c = [HECTARES] if b = 2 then y = 1157.89 0.5 a c 10000 / 10 if b = 1 then y = 270.67 0.79 a c 10000 / 10

PAGE 225

208 LIST OF REFERENCES Ackleson, S.G. and V. Klemas. 1987. Remote sensing of submerged aquatic vegetation in Lower Chesapeake Bay: a comparison of Landsat MSS to TM imagery. Rem. Sens. Environ. 22:235-248. Allen, L.H., Jr., J.M. Ruddell, G.J. Rutter, and P. Yates. 1982. Land effects on Taylor Creek water quality. p. 67-77. In E.G. Kruse et al. (ed.) Environmentally Sound Water and Soil Management. American Soci ety of Civil Engineers, New York. American Congress on Surveying and Mapping (ACSM), American Society for Photogrammetry and Remote Sensing (ASPR S), and American Society of Civil Engineers (ASCE). 1994. Glossary of th e Mapping Sciences. American Society of Civil Engineers, Bethesda, Maryland. Anderson, J.M. 1976. An ignition method for de termination of total phosphorus in lake sediments. Water Res. 10:329-331. Anderson, D.L. and E.G. Flaig. 1995. Agri cultural best management practices and surface water improvement and management. Water Sci. and Tech. 31:109-121. Asner, G.P. 1998. Biophysical and biochemical sources of variability in canopy reflectance. Rem. Sens. Environ. 64:234-253. Asner, G.P., A.R. Townsend, and M.M. Bustamante. 1999. Spectrometry of pasture condition and biogeochemistry in the Central Amazon. Geophysical Research Letters 26(17):2769-2772. Asner, G.P., C.A. Wessman, C.A. Bateson, and J.L. Privette. 2000. Impact of tissue, canopy, and landscape factors on the hyperspect ral reflectance variability of arid ecosystems. Rem. Sens. Environ. 74:69-84. Austin, M.P., J.A. Meyers, L. Belbin, and M.D. Doherty. 1995. Simulated data case study, subproject 5, modelling of landscap e patterns and proce sses using biological data. Division of Wildlife And Ecology, Commonwealth Scientific and Industrial Research Organization (CSIRO), Canberra, Australia. Axt, J.R. and M.R. Walbridge. 1999. Phosphate removal capacity of palustrine forested wetlands and adjacent uplands in Virginia. Soil Sci. Soc. Am. J. 63:1019-1031.

PAGE 226

209 Behrendt, H., M. Ley, R. Korol, M. Str onska-Kedzia, and W. Pagenkopf. 1999. Point and diffuse nutrient emissions and transports in the Odra Basin and its main tributaries. Acta Hydrobiol. Hydrochim. 27:274–281. Berka, C.H. Schreier and K. Hall. 2001. Linking water quality with agricultural intensification in a rural watershed. Water, Air and Soil Pollution 127:389-401. Bishop, T.F.A, and A.B. McBratney. 2001. A comparison of prediction methods for the creation of field-extent soil pr operty maps. Geoderma 103:149-160. Blatie, D. 1980. Land into water, water into land. Florida State University Press, Tallahassee, FL. Boggess, C.F., E.G. Flaig, and R.C. Fluck. 1995. Phosphorus budget-basin relationships for Lake Okeechobee tributary basins Ecological Engineering 5:143-162. Bolstad, P. 2002. GIS Fundamentals: A Fi rst Text on Geographic Information Systems. Eider Press, White Bear Lake, Minnesota. Bolstad, P. and W. Swank. 1997. Cumulative impacts of landuse on water quality in a southern Appalachian watershed. J. Am Water Resour. Assoc. 33(3):519-533. Borsuk, M.E., C.A. Stow, and K.H. Reckhow 2002. Integrative environmental prediction using Bayesian networks: A synthesi s of models describing estuarine eutrophication. IEMSS Conference proceedings, Lugano, Switzerland. Bottcher, A.B., T.K. Tremwel, and K.L. Campbell. 1995. Best management practices for water quality improvement in the La ke Okeechobee watershed. Ecological Engineering 5:341-356. Brady, N.C. and R.R. Weil. 2002. Elements of the nature and properties of soils. Prentice Hall, Upper Saddle River, New Jersey. Braskerud, B.C. 2002. Factor s affecting phosphorus reten tion in small constructed wetlands treating agricultur al non-point source pollution. Ecological Engineering 19:41-61. Breiman, L., J.H. Friedman, R.A. Olshen and C.J. Stone. 1984. Classification and regression trees. Wadsworth, Belmont, California. Bruland, G.L., M.F. Hanchey, and C.J. Ri chardson. 2003. Effects of agriculture and wetland restoration on hydrology, soils, and water quality of a Carolina bay complex. Wetlands Ecol. and Management 11:141-156. Bruland, G.L., and C.J. Richardson. 2004. H ydrologic and topsoil additions affect soil properties of Virginia created wetl ands. Soil. Soc. Am. J. 68:2069-2077.

PAGE 227

210 Burnett, C. and T. Blaschke. 2003. A mu lti-scale segmentation/ object relationship modeling methodology for landscape anal ysis. Ecological Mo deling 168:233-249. Butson, C.R. and R.A. Fernandes. 2004. A consistency analysis of surface reflectance and leaf area index retrieval from ove rlapping clear-sky Landsat ETM+ imagery. Rem. Sens. Environ. 89:369-380. Campbell, D.A., C.A. Cole, and R.P. Brooks 2002. A comparison of created and natural wetlands in Pennsylvania, USA. Wetla nds Ecology and Management 10:41-49. Campbell, K.L., J.C. Capece, and T.K. Tremwell. 1995. Surface / subsurface hydrology and phosphorus transport in the Kissimm ee River Basin, Florida. Ecological Engineering 5:301-330. Carter, L.J., D. Lewis, L. Cr ockett, and J. Vega. 1989. Soil Survey of Highlands County, Florida. USDA/NRCS in cooperation with th e University of Florida, Institute of Food and Agricultural Sciences, Agricult ural Experimental Stations and Soil Science Department; Florida Department of Agriculture and Consumer Services; and the Florida Department of Transportation. Chavez, P.S. 1996. Image-based atmospheric corrections revisited and revised. Photogrammetric Engineering a nd Remote Sensing 62(9):1025-1036. Clary, W.P. 1999. Stream channel and vegetatio n responses to late spring cattle grazing. J. Range Manage. 52:218–227. Colwell, R.N. (ed.) 1983. Manual of re mote sensing. American Society of Photogrammetry. Falls Church, Virginia. Crist, E.P. and R.C. Cicone. 1984. A physi cally based transfor mation of Thematic Mapper data: the TM tasseled cap. IEEE Transaction on Geosciences and Remote Sensing 22:256-263. Crist, E.P. and R.J. Kauth. 1986. The ta sseled cap de-mystified. Photogrammetric Engineering and Remote Sensing 52:81-86. Cuffney, T.F., M.R. Meador, S.D. Porter, and M.E. Gurtz. 2000. Responses of physical, chemical, and biological indicators of water to a gradient of agricultural land use in the Yakima River, Washington. Envi ronmental Monitoring and Assessment 64:259-270. Curran, P.J. 1985. Principles of Remote Sensing. Longman Publishing, London. Davis, F.E. and M.M. Marshall. 1975. Chemi cal and biological inve stigations of Lake Okeechobee January 1973 June 1974. Tec hnical Publication 75-1. South Florida Water Management District, West Palm Beach, Florida.

PAGE 228

211 De’ath, G. and K.E. Fabricius. 2000. Classi fication and regression trees: a powerful yet simple technique for ecological data analysis. Ecology 81:3178-3193. Detenbeck N.E., D.L. Taylor, A. Lima, a nd C. Hagley. 1996. Temporal and spatial variability in water quality of wetlands in the Minneapolis/St. Paul, MN metropolitan area: Implications for monitoring strategies and designs. Environmental Monitoring and Assessment 40:11–40. ERDAS Inc. 1999. ERDAS field guide. ERDAS Inc, Atlanta, Georgia. ERDAS Inc. 2004a. Modified Soil-Adjuste d Vegetation Index (MSAVI). [Online] Availble at: http://gis.leicageosystems.com/Support/Downloads/d wnld_detail.aspx?download_ID=8 (accessed Nov 1, 2004; verified 20 Dec 2004). ERDAS Inc. Atlanta, Georgia. ERDAS Inc. 2004b. Tasseled Cap Transformation for Landsat 7 ETM+ Sensor [Online] Available at: http://gis.leicageosystems.com/Support/Downloads/d wnld_detail.aspx?download_ID=9 (accessed Nov 1, 2004; verified 20 Dec 2004). ERDAS Inc. Atlanta, Georgia. ESRI Inc. 2002. ArcMap Help system. Versi on 8.3. ESRI Inc., Redlands, California. Faulkner, S.P. and C.J. Richardson. 1989. P hysical and chemical characteristics of freshwater wetland soils. pp. 41-72. In: D.A. Hamner, (ed.) Constructed Wetlands for Wastewater Treatment. Lewis Publishers, Chelsea, Michigan. Federico, A.C., K.G. Dickson, C.R. Krat zer, and F.E. Davis. 1981. Lake Okeechobee water quality studies and eu trophication assessment. Re port, South Florida Water Management District, West Palm Beach, Florida. Fiener, P. and K. Auerswald. 2003. Effec tiveness of grassed waterways in reducing runoff and sediment delivery from agri cultural watersheds. J. Environ. Qual. 32:927-936. Flaig, E.G. and K.E. Havens. 1995. Historical trends in the Lake Okeechobee ecosystems. I. Land use and nutrient load ings. Arch. Hydrobiol. (Suppl.) 107:1-24. Flaig, E.G. and K.R. Reddy. 1995. Fate of phosphorus in the La ke Okeechobee basin, Florida, USA: overview and recommendations. Ecological Engineering 5:127-143. Florida Department of Environmental Prot ection. 2001. Total Maximum Daily Load for Total Phosphorus – Lake Okeec hobee. [Online] Available at: http://www.dep.state.fl.us/ water/wqssp/lakeo_tmdl.htm (modified 01 Jul 2004; accessed 21 Aug 2004; verified 20 D ec 2004). Florida Department of Environmental Protection, Tallahassee, Florida.

PAGE 229

212 Florida Department of Environmental Protection. 2003. 1999 Digital Orthographic Quarter-Quad. [Online] Available at: http://data.labins.org/ 2003/MappingData/DOQQ/doqq.cfm (modified 05 Oct 2002; accessed 01 Nov 2002; verified 09 N ov 2004). Florida Department of Environmental Protection, Tallahassee, Florida. Florida Department of Transportation. 1999. Florida land use, cover and forms classification system. [Online] Available at: http://www.dot.state.fl.us/surveyingandma pping/fluccmanual.pdf (accessed 11 Sept 2004; verified 20 Dec 2004). Florida Departme nt of Transportation, Tallahassee, Florida. Fluck, R.C., C.A. Fonyo, and E.G. Flaig. 1992. Land-use-based phosphorus balances for Lake Okeechobee, Florida, drainage basi ns. J. Appl. Eng. Agric. 8:813-820. Gergel, S.A., M.G. Turner, J.R. Miller, J. M. Melack, and E.H. Stanley. 2002. Landscape indicators of human impacts to riverine systems. Aquatic Sciences 64:118-128. Graetz, D.A. and V.D. Nair. 1995. Fate of phosphorus in Florida Spodosols contaminated with cattle manure. Ecological Engineering 5:163-181. Graetz, D.A. and V.D. Nair. 1999. Inorganic fo rms of phosphorus in soils and sediments. pp. 171-186. In K.R. Reddy, G.A. OConnor, and C.L. Schelske (eds.) Phosphorus biogeochemistry in subtropical ecosyst ems. Lewis Publishers, Boca Raton. Graetz, D.A., V.D. Nair, K.M. Portier, and R.L. Voss. 1999. Phosphorus accumulation in manure-impacted Spodosols of Florida. Agriculture, Ecosystems and Environment 75:31-40. Gustafson, S. and D. Wang. 2002. Eff ects of agricultural runoff on vegetation composition of a priority c onservation wetland, Vermont, USA. J. Environ. Qual. 31:350-257. Haan, C.T. 1995. Fate and transport of phosphorus in the Lake Okeechobee Basin, Florida. Ecological Engineering 5:331-339. Hall, K. and H. Schreier. 1996. Urbanization an d agricultural intensif ication in the lower Fraser River Valley: impact s on water use and quality. GeoJournal 40:135-146. Harris, A.T. and G.P. Asner. 2003. Grazing gr adient detection with airborne imaging spectroscopy on a semi-arid rangeland. J ournal of Arid Environments 55:391-404. Harvey, K.R. and G.J.E. Hill. 2001. Vegetati on mapping of a tropical freshwater swamp in the Northern Territory, Australia: a comparison of aerial photography, Landsat TM and SPOT satellite imagery. In t. J. Rem. Sensing 22(15):2911-2925. Haynes, R.J. and R.S. Swift. 1989. The e ffects of pH and drying on adsorption of phosphate by aluminum-organic matter associations. J. Soil Sci. 40:773.

PAGE 230

213 Heatwole, C.D. 1986. Field and basin-scale water quality models for evaluating agricultural nonpoint pollution abatement programs in a south Florida flatwoods watershed. Ph.D. diss.Univ. of Florida, Gainesville. Herlihy, A.T., J.L. Stoddard, and C.B. J ohnson. 1998. The relationship between stream chemistry and watershed land use data in the mid-Atlantic region. U.S. Water Air and Soil Pollution 105:377-386. Herzog, F. and A. Lausch. 2001. Supplementing land-use statistics with landscape metrics: some methodological considerat ions. Environmental Monitoring and Assessment 72:37-50. Hiscock, J.G., C.S. Thourot, and J. Zang. 2003. Phosphorus budget-land use relationships for the northern Lake Okeechobee watershe d, Florida. Ecological Engineering 21:63-74. Hodges, J.R., G.K. Kirk, R.L. Shirley, F.M. Peacock, J.F. Easley, H.L. Breland, and F.G. Martin. 1967. Phosphorus fertilization of pangola grass pastures, direct and residual effects. Research Report, Range Experiment Station, University of Florida, Ona, Florida. Hook, P.B. 2003. Sediment retention in range land riparian buffers. J. Environ. Qual. 32:1130-1137. Houlahan, J.E. and C.S. Findlay. 2004. Estimating the ‘critical’ distance at which adjacent land-use degrades wetland water and sediment quality. Landscape Ecology 19:677-690. Huete, A.R. 1988. A soil-adjusted vegetati on index (SAVI). Rem. Sens. Environ. 25:295309. Hunsaker, C.T. and D.A. Levine. 1995. Hierar chical approaches to the study of water quality in rivers. Bioscience 45:193-203. Hunsaker, C.T., D.A. Levine, S.P Timmins B.L. Jackson, and R.V. O’Neill. 1992. Landscape characterization for assess ing regional water quality. p. 997–1006. In Ecological Indicators Edited by D.H. McKe nzie, D.E Hyatt, and V.J. McDonald (eds.) Elsevier Appl. Sci., New York, New York. Iverson, L.R. and A.M. Prasad. 1998. Predicting abundance for 80 tree species following climate change in the eastern United States. Ecological Monogra phs 68: 465-485. Jensen, J.R. 1996. Introductory digital image processing, a remote se nsing perspective. Prentice-Hall, Upper Saddle River, New Jersey. Jensen, J.R, E.J. Christensen, and R. Shar itz. 1984. Nontidal wetland mapping in South Carolina using airborne multi-spectral scan ner data. Rem. Sens. Environ.16:1-12.

PAGE 231

214 Johnston, C.A. 1991. Sediment and nutrient retention by freshwater wetlands: effects on surface water quality. Critical Reviews in Environmental Control 21(5,6): 491-565. Jones, K.B., A.C. Neale, M.S. Nash, R.D. Van Remortel, J.D. Wickham, K.H. Ritters, and R.V. O’Neill. 2001. Predicting nutrient and sediment loadings to streams from landscape metrics: A multiple watershed study from the United States Mid-Atlantic Region. Landscape Ecology 16:301–312. Kadlec, R.H. and R.L. Knight. 1996. Treatment wetlands. Lewis Publishers, Boca Raton, Florida. Kaufman, Y.J. and C. Sendra. 1988. Automatic atmospheric correction. Int. J. Remote Sensing. 9:1357–1381. Kauth, R.J. and G.S. Thomas. 1976. The ta sseled cap a graphic description of the spectral-temporal development of agricultural crops as seen by Landsat. Proceedings of the Symposium on Machine Processing of Remotely Sensed Data. Purdue University, West Lafayette, Indiana, p. 4B41-51. Kennedy, H. (ed.) 2001. ESRI Press Dictionary of GIS Terminology. [Online] Available at http://www.esri.com/library /glossary/glossary.html (accessed 08 Feb 2004; verified 11 Nov 2004). ESRI Pr ess, Redlands, California. Kinnear, P.R. and C.D. Gray. 1999. SPSS fo r Windows made simple. Psychology Press Ltd., East Sussex, U.K. Knisel, W.G., P. Yates, J.M. Sheridan, T. K. Woody, III, L.H. Allen, Jr., and L.E. Asmusssen. 1985. Hydrology and hydrogeology of Upper Taylor Creek watershed, Okeechobee County, Florida: data and analysis. ARS-25. USDA-Agricultural Research Service, Washington, D.C. Kuenstler, W.F., D. Ernstrom, and E. Seeley 1995. Nutrient management-building tools to predict potent ial nutrient losses. p. 459-467. In C. Heatwole (ed.) Water quality modeling. American Society of Agricultu ral Engineers (ASAE) Publication 5-95. ASAE, St. Joseph, Michigan. Kuusemets, V. and U. Mander. 2002. Nutr ient flows and management of a small watershed. Landscape Ecology 17:59-68. Lapen, D.R., G.G. Topp, E.G. Gregorich, H.N. Hayhoe, and W.E. Curnoe. 2001. Divisive field-scale associations between corn yi elds, management, and soil information. Soil and Tillage Research 58(3-4):193-206. Lvesque, M.P., and S.P. Mathur. 1979. A co mparison of various means of measuring the degree of decomposition of virgin peat mate rials in the context of their relative biodegradability: Canadian Jour nal of Soil Science 59:397–400.

PAGE 232

215 Lewis, D.L. K.J. Liudahl, C.V. Noble, and L.J.Carter. 2001. Soil Survey of Okeechobee County, Florida. USDA/NRCS in cooperati on with the University of Florida, Institute of Food and Agricultural Sciences, Agricultural Experimental Stations and Soil Science Department and Florida Depa rtment of Agriculture and Consumer Services, Gainesville, Florida. Lewis, R.J. 2000. An introduction to classifi cation and regression tr ee (CART) analysis [Online]. Available at http://www.saem.org/download/lewis1.pdf (accessed 3 June 2004; verified 24 Aug. 2004). Society for Academic Emergency Medicine, Lansing, Michigan. Liang, S., H. Fallah-Adl, S. Kalluri, J. Ja Ja, Y.J. Kaufman, and J. Townshend. 1997. Development of an operational atmospheri c correction algorithm for TM imagery. J. Geophys. Res. 102:17,173-17,186. Liang, S., H. Fang, M. Chen. 2001. Atmospheric correction of Landsat ETM+ land surface imagery: I. Methods. IEEE Tran sactions on Geosciences and Remote Sensing. 39:2490-2498. Lillesand, T. and R. Kiefer. 2000. Remote se nsing and image interpretation. John Wiley & Sons. New York, New York. Loftin, M.K., L.A. Toth, and J.T.B. Obeysekera. 1990. Kissimmee river restoration: alternative plan evaluation and prelimin ary design report. South Florida Water Management District, West Palm Beach, Florida. MacGill, R.A., S.E. Gatewood, C. Hutchins on, and D.D. Walker. 1976. Final report on the special project to prevent the eutr ophication of Lake Okeechobee. Rep. DSPBCP-36-76. Div. of State Pla nning, Tallahassee, Florida. Mander, U., A. Kull, and V. Kuusemets. 2000. Nutrient flows and land use change in a rural catchment: a modeling approach. Landscape Ecology 15:187-199. Mansell, R.S., Bloom, and S.A. Burgoa. 1991. Phosphorus transport with water flow in acid, sandy soils. p. 271-314. In: M.Y. Corapcioglu (ed.) Transport processes in porous media. Kluwer Academic Publishers, Boston, Massachusetts. McCaffery, P.M., W.M. Hinkley, R. MacG ill and G.D. Cherr. 1976. Report of investigations in the Kissimmee-Lake Okeechobee Watershed. Fla. Depart. Environ. Reg. Tech. Series, Vol. 2, No. 2. Tallahassee, Florida. McCollum, S.H and O.E. Cruz. 1979. Soil Survey of St. Martin County Area, Florida. USDA/NRCS in cooperation with the Universi ty of Florida, Institute of Food and Agricultural Sciences, Agricultural Experimental Stations and Soil Science Department; the Martin County Board of County Commissioners, and Florida Department of Agriculture and Consum er Services, Gainesville, Florida.

PAGE 233

216 McKeague, J.A. and J.H. Day. 1966. Dithionite and oxalate-extractable Fe and Al as aids in differentiating various classes of so ils. Canadian Journal of Soil Science. 46:1322. McKenzie, N. J. and M. P. Austin. 1993. A qua ntitative Australian approach to medium and small scale surveys based on soil stra tigraphy and environmental correlation. Geoderma 57:329-355. McKenzie, N.J. and P.J. Ryan. 1999. Spa tial prediction of soil properties using environmental correlation. Geoderma 89:67-94. Mehlich, A. 1953. Determinations of P, Ca, Mg, K, Na and NH4 by North Carolina soil testing laboratories. Mimeo. North Caroli na Department of Agriculture, Raleigh, North Carolina. Mertes, L.A.K., D.L. Daniel, J.M. Melack, B. Nelson, L.A. Marinelli, and B.R. Forsberg. 1995. Spatial patterns of hydrology, geomorphology and vegetation on the floodplain of the Amazon River in Brazil from a remote sensing perspective. Geomorphology 13:215-232. Miller, J.A. 1997. Hydrogeology of Florida. p. 69-88. In A.F. Randazzo and D.S. Jones (eds.) The geology of Florida. Univer sity Press, Gainesville, Florida. Mitsch, W.J. 1992. Basin design and the role of created, restored, and natural riparian wetlands in controlling non-poi nt source pollution. Ecological Engineering, 1:2747. Mitsch, W.J. and J.G. Gosselink. 2000. We tlands. John Wiley and Sons, New York, New York. Nagler, P.L., C.S.T. Daughtry, and S.N. Goward. 2000. Plant litter and soil reflectance. Rem. Sens. Environ. 71:207-215. Nair, V.D. and D.A. Graetz. 2002. Phosphor us saturation in Spodosols impacted by manure. J. Environ. Qual. 31:1279–1285 Nair, V.D., D.A. Graetz, and K.M. Portier. 1995. Forms of phosphorus in soil profiles from dairies of South Florida. So il Sci. Soc. Am. J. 59:1244-1259. Nair, V.D., D.A. Graetz, and K.R. Reddy. 1998. Dairy manure influences on phosphorus retention capacity of Spodosols. J. Environ. Qual. 27:522–527. Nair, V.D., R.R. Villapando, and D.A. Graetz. 1999. Phosphorus retention capacity of the spodic horizon under varying environmenta l conditions. J. Environ. Qual. 28:1308– 1313. Nairn, R.W., and W.J. Mitsch. 1999. P hosphorus removal in created wetland ponds receiving river overflow. Ec ological Engineering 14:107-126.

PAGE 234

217 National Aeronautics and Space Administrati on. 2004. Landsat 7 science data user’s handbook. Report 430-15-01-003-0. [Online]. Available at http://ltpwww.gsfc.nasa.gov/IAS/handbook/handbook_toc.html (modified 24 Aug 2004; accessed 2 Oct 2004; verified 12 Nov 2004). National Aeronautics and Space Administration, Washington D.C. Nessel, J. K. and S. E. Bayley. 1984. Dist ribution and dynamics of organic matter and phosphorus in a sewage enrich ed cypress swamp. p. 262–278. In K.C. Ewel and H.T. Odum (eds.) Cypress Swamps. Universi ty Presses of Florida, Gainesville, FL, USA. Numata, I., J.V. Soares, D.A. Roberts, F. C. Leonidas, O.A. Chadwick, G.T. Batista. 2003. Relationships among soil fertility dynamics and remotely sensed measures across pasture chronosequences in Rondni a, Brazil. Rem. Sens. Environ. 87:446455. Ott, R.L. and M. Longnecker. 2001. An Intr oduction to Statistical Methods and Data Analysis. Duxbury, Pacific Grove, California. Ozesmi, S.L. and M.E. Bauer. 2002. Satel lite remote sensing of wetlands. Wetlands Ecology and Management 10:381-4002. Pan, Y., R.J. Stevenson, B.H. Hill, P.R. Kaufmann, and A.T. Herlihy. 1999. Spatial patterns and ecological determinants of be nthic algal assemblages in Mid-Atlantic Highland streams, U.S.A. J ournal of Phycology 35:460-468. Pant, H.K. and K.R. Reddy. 2003. Potential loading of phosphorus in a wetland constructed in agricultural land. Water Research 37:965-972. Parker, G.G. 1955. Water resources in southeas tern Florida with spec ial reference to the geology and ground water of the Miami area. Water supply paper. USGS, Tallahassee, Florida. Patrick, W.H., Jr. and R.A. Khalid. 1974. Phosphate release and sorption by soils and sediments: effect of aerobic and an aerobic conditions. Science. 186:53-55. Pearlstine, L.G., S.E. Smith, L.A. Brandt, C. R. Allen, W.M. Kitchens, and J. Stenberg. 2002. Assessing state-wide biodiversity in th e Florida gap analysis project. Journal of Environmental Management. 66:127-144. Petrovic, M. and M. Kast elan-Macan. 1996. The uptake of inorganic phosphorus by insoluble metal-humic complexes. Wa ter Science and Technology 34:253-258. Phillips, J.D. 1989. Fluvial sediment storage in wetlands. Water Resour. Bull. 25(4):867– 873.

PAGE 235

218 Pierzynski, G.M. (ed.) 2000. Methods of phosphorus analysis for soils, sediments, residuals and waters. Southern cooperat ive series bulletin no. 396. North Carolina University, Raleigh, North Carolina. Qi, J., A. Chehbouni, A.R. Huete, and Y.H. Kerr. 1994. Modified Soil Adjusted Vegetation Index (MSAVI). Rem. Sens. Environ. 48:119–126. Raisin, G.W. and G.S. Mitchell. 1995. The us e of wetlands for the control of non-point source pollution. Water Sci. and Tech. 32(3):177-186. Rechcigl, J.E. and A.B. Bottcher. 1995. Fate of phosphorus on Bahiagrass (Paspalum notatum) pastures. Ecological Engineering 5:247-259. Reddy, K.R., R.D. Delaune, W.F. DeBusk, and M.S. Koch. 1993. Long-term nutrient accumulation rates in the Everglades. Soil Sci. Soc. Am. J. 57:1147-1155. Reddy, K.R., O.A Diaz, L.J Scinto, and M. Agami. 1995. Phosphorus dynamics in selected wetlands and streams of the Lake Okeechobee Basin. Ecological Engineering 5:183-207. Reddy, K.R., E.G. Flaig, and D.A. Graetz. 1996a. Phosphorus storage capacity of uplands, wetlands and streams of the Lake Okeechobee Watershed, Florida. Agric. Ecosyst. Environ.59:203-216. Reddy, K.R., E.G. Flaig, L.J. Scinto, O. Diaz, and T.A. DeBusk. 1996b. Phosphorus assimilatin in a stream system of the La ke Okeechobee Basin. Journal of American Water Resources Association 32 (5):901-913. Reddy, K.R., R.H. Kadlec, E.G. Flaig, a nd P.M. Gale. 1999a. Phosphorus retention in streams and wetlands: a review. Critical Reviews in Environmental Science and Technology 29:83-146. Reddy, K.R., E. Lowe, and T. Fontaine. 1999b. Phosphorus in Florida’s ecosystems: analysis of current issues. p.111-142. In K.R. Reddy, G.A. O’Connor, and C.L. Schelske (eds.). Phosphorus biogeochemi stry in subtropical systems. Lewis Publishers, Boca Raton, Florida. Reddy, K.R., G.A. O Connor, and P.M. Gale 1998. Phosphorus sorption cap acities of wetland soils and stream sediments imp acted by dairy effluent. Journal of Environmental Quality. 27:438-447. Reddy, K. R. and W.H. Smith (eds.) 1987. Aquatic plants for water treatment and resource recovery, Magnolia Publishers, Orlando, Florida. Reese, H.M., T.M. Lillesand, D.E. Nagel, J.S. Steward, R.A. Goldmann, T.E. Simmons, J.W. Chipman, and P.A. Tessar. 2002. Statewide land cover derived from multiseasonal Landsat TM data: a retros pective of the WISCLAND project. Rem. Sens. Environ. 82:224-237.

PAGE 236

219 Rhue, R.D. and R.G. Harris. 1999. Phosphorus sorption/desorption reactions in soils and sediments. pp.187-206. In K.R. Reddy, G.A. O’Connor, and C.L. Schelske (eds.) Phosphorus biogeochemistry in subtropi cal ecosystems. Lewis Publishers, Boca Raton, Florida. Richardson, C.J. 1985. Mechanisms cont rolling phosphorus retention capacity in freshwater wetlands. Science 228:1424–1427. Richardson, C.J. 1999. The role of wetlands in storage, release and cycling of Phosphorus on the landscape: a 25-year retrospective. p. 47-68. In K.R. Reddy, G.A. O’Connor, and C.L. Schelske (eds.) Phosphorus bi ogeochemistry in subtropical systems. Lewis Publishers, Boca Raton, Florida. Richardson, C.J. 2003. Pocosins: hydrologically isolated or integrated wetlands on the landscape? Wetlands 23(3):563-576. Risser, P.G., J.R. Karr, and R.T.T. Form an. 1984. Landscape ecology: directions and approaches. Special Publication 2. Illinois Natural History Survey, Urbana, Illinois. Rondeaux, G., M. Steven, and F. Baret. 1996. Optimization of soiladjusted vegetation indices. Rem. Sens. Environ. 55:95-107. Roth, N.E., J.D. Allan, and D.L. Erickson. 1996. Landscape influences on stream biotic integrity assessed at multiple scales. Landscape Ecology 11:141–156. Rouget, M., D.M. Richardson, R.M. Cowling, J.W. Lloyd, and A.T. Lombard. 2003. Current patterns of habitat transformati on and future threats to biodiversity in terrestrial ecosystems of the Cape Flor istic Region, South Africa. Biological Conservation 112:63-85. Sader, A.S., A. Douglas, and W. Liou. 1995. Accuracy of Landsat-TM and GIS rulebased methods for forest wetland classificat ion in Maine. Remote Sens. Environ. 53:133-144. San Miguel-Ayanz, J. and G.S. Biging. 1997. Comparison of single-stage and multi-stage classification approaches for cover type mapping with TM and SPOT data. Remote Sens. Environ. 59:92-104. Schmidt, K.S. and A.K. Skidmore. 2003. Spect ral discrimination of vegetation types in a coastal wetland. Rem. Sens. Environ. 5814:1-17. Scinto, L.J. 1990. Seasonal variation in so il phosphorus distributi on in two wetlands of south Forida. Master’s Thesis. Univ. of Florida, Gainesville, Florida.

PAGE 237

220 Shaffer, M.J., M.D. Hall, B.K. Wylie, and D.G. Wagner. 1996. NLEAP/GIS approach for identifying and mitigating regional ni trate-nitrogen leaching. p. 283-294. In D.L. Corwin and K. Loague (ed.) Applications of GIS to the modeling on non-point source pollutants in the vados e zone. Soil Sci. Soc. Am. Spec. Pub. 48. Soi Sci. Soc. Am., Madison, WI. Shan, B., C. Yin, and G. Li. 2002. Transport and retention of phosphor us pollutants in the landscape with a traditional, multipond sy stem. Water, Air and Soil Pollution 139:15-34. Sharpley, A.N. 1995. Fate and transport of nutrients – phosphorus. Agricultural Research Service Working Paper No. 8. U.S. Depart ment of Agriculture, Washington, D.C. Sharpley, A.N., B. Foy, and P. Withers. 2000. Practical and innovative measures for the control of agricultural phosphor us losses to water: an overview. J. Environ. Qual. 29:1-9 Shih, G. 1983. Data analysis to detect ra infall changes in south Florida. Technical Memorandum. South Florida Water Mana gement District, West Palm Beach, Florida. Soil Survey Staff. 1996. Keys to soil taxonomy. U.S. Gov. Print. Office, Washington, D.C. Soil Survey Staff. 1993. Soil survey manual. So il Conservation Service. U.S. Department of Agriculture Handbook 18. U.S. Gov. Print. Office, Washington, D.C. Sonzogni, W.C., S.C. Chapra, D.E. Armstrong, and T.J. Logan. 1982. Bioavailability of phosphorus inputs to lakes. Journal of Environmental Quality 11:555-563. South Florida Water Management Distri ct. 1993. Surface Water Improvement and Management (SWIM) Plan Update for Lake Okeechobee, Volume 1, Planning Document. South Florida Water Management District, West Palm Beach, Florida. South Florida Water Manageme nt District. 2003. 2003 Land use GIS layer. South Florida Water Management District, West Palm Beach, Florida. South Florida Water Management District Florida Department of Environmental Protection, and Florida Department of Agriculture and Consumer Services. 2004a. Lake Okeechobee Protection Plan. South Flor ida Water Management District, West Palm Beach, Florida. South Florida Water Management District. 2004b. District-w ide rainfall maps. [Online]. Available at http://www.sfwmd.gov/curre/rainmaps/period.html (accessed 1 May 2004; verified 12 Nov 2004). South Florid a Water Management District, West Palm Beach, Florida.

PAGE 238

221 South Florida Water Manageme nt District. 2004c. DBHYDRO. [Online]. Available at http://www.sfwmd.gov/org/ema/dbhydro/ (accessed 11 May 2004; verified 12 Nov 2004). South Florida Water Management Di strict, West Palm Beach, Florida. Southeastern Regional Climate Center. 2004. Historical Climate Su mmaries for Florida [Online]. Available at http://water.dnr.state.sc.us/climate/sercc/climateinfo/historical/historical_fl.html (modified 13 Aug 2004; accessed 12 Nov 2004). SE Regional Climate Center, Columbia, South Carolina. Sperry, C.M. 2004. Soil phosphorus in isolat ed wetlands of subt ropical beef cattle pastures. Master’s thesis. Univ. of Florida, Gainesville, Florida. SPSS. 2002. SPSS for Windows. Release 11.5.0. SPSS, Inc., Chicago, Illinois. Steinberg, D. and P. Colla. 1995. CART: Treestructured non-parametric data analysis. Salford Systems, San Diego, CA. Steinman, A.D., J. Conklin, P.J. Bohlen, a nd D.G. Uzarski. 2003. Influence of cattle grazing and pasture land use on macroinvert ebrate communities in freshwater wetlands. Wetlands 23(4):877-889. Syers, J.K., R.F. Harris, and D.E. Arms trong. 1973. Phosphate chemistry in lake sediments. Journal of Environmental Quality 2:1-14. Tate, K.W., G.A. Nader, D.J. Lewis, E.R. Atwill, and J.M. Connor. 2000. Evaluation of buffers to improve the quality of runoff fr om irrigated pastures. Journal of Soil and Water Conservation, 55(4):473-478. Taylor, J. 1997. An introduction to error anal ysis. University Science Books, Sausalito, California. Teillet, P.M. and G. Fedosejevs. 1995. On the dark target approach to atmospheric correction of remotely sensed data Can. J. Remote Sensing. 21:374–387. Tiner, R.W. 2003. Geographically isolated wetlands of the United States. Wetlands 23(3):494-516. Troeh, F.R. and L.M. Thompson. 1993. Soils an d soil fertility. Oxford University Press, New York, New York. Underwood, A.J.. 1997. Experiments in ecology : their logical design and interpretation using analysis of variance. Cambridge University Press, Cambridge, U.K. U.S. Army Corps of Engin eers and South Florida Water Management District. 2003. Central and Southern Florida Project, Co mprehensive Everglades Restoration Plan, Watershed Assessment Report, Lake Okeec hobee Watershed Project. June 2003.

PAGE 239

222 U.S. Department of Agriculture – Natura l Resources Conservation Service, Florida Engineering and Soils Staffs. 1982. Fl orida Drainage Guide. USDA, NRCS, Gainesville, Florida. U.S. Department of Agriculture – Natura l Resource Conservation Service. 1995. Soil survey geographic (SSURGO) database da ta use information. USDA, Lincoln, Nebraska. U.S. Environmental Protection Agency. 1993. Me thods for the determination of inorganic substances in environmental samples, EPA/600/R-93/100. U.S. EPA, Cincinnati, Ohio. U.S. Environmental Protection Agency. 1984. Th e determination of i norganic anions in water by ion chromatography, EPA/600/484-017. U.S. EPA, Cincinnati, Ohio. U.S. Fish and Wildlife Service. 2002. Na tional Wetlands Inventory. Version 2003. Florida Geographic Data Librar y. St. Petersburg, Florida. Villapando, R.R. and D.A. Graetz. 2001. Wa ter table effects on phosphorus reactivity and mobility in a dairy manure-impacted S podosol. Ecological Engineering 18:77-89. Walsh, S.J., D.R. Butler, and G.P. Malans on. 1998. An overview of scale, pattern, process relationships in geomorphology: a remote sensing and GIS perspective. Geomorphology 21:183-205. Wear, D., M. Turner and R. Naiman. 1998. Land cover along an urban-rural gradient: implications for water qualit y. Ecol. Applic. 8(3):619-630. Wetzel, R.G. 1983. Limnology. Saunders Co llege Publishing. Orlando, Florida. Wharton, S.W. 1989. Knowledge-based spectral classification of re motely sensed image data, p. 548-577. In G. Asrar (ed.) Theory and ap plications of optical remote sensing. John Wiley & Sons, New York. Whickham, J.D., T.G. Wade, K.H. Ritters, R. V. O’Neill, J.H. Smith, E.R. Smith, K.B. Jones, and A.C. Neale. 2003. Upstream-t o-downstream changes in nutrient export risk. Landscape Ecology 18:195-208. Wicks, T.E., G.M. Smith, and P.J. Curran. 2002. Polygon-based aggregation of remotely sensed data for regional ecological analyses Int. J. of Applied Earth Observation and Geoinformation 4:161-173. Wilson, J.P. W.P. Inskeep, J.M. Wraith, and R.D. Snyder. 1996. GIS-based solute transport modeling pplications: Scale effects of soil and climate data input. J. Environ. Qual. 25:433-439. Whigham, D.F. and T.E. Jordan. 2003. Isol ated wetlands and water quality. Wetlands 23(3):541-549.

PAGE 240

223 White, R.E. and A.W. Taylor. 1977. Effect of pH on phosphate adorption and isotopic exchange in acid soils at low and high a dditions of soluble phosphate. J. Soil Sci. 28:48-61. Wu, J. and R. Hobbs. 2002. Key issues and re search priorities in landscape ecology: An idiosyncratic synthesis. Landscape Ecology 17:355-365. Yin, Z-Y., S. Walcott, B. Kaplan, J. Cao, W. Lin, M. Chen, D. Liu, and Y. Ning. 2003. An analysis of the relationship between sp atial patterns of wate r quality and urban development in Shanghai, China. Computer s, Environ. Urban Syst. (Article in press). Yin, C. and Z. Lan. 1994. The nutrient retention by ecotone wetlands and their modification for Baiyangdian Lake restor ation. Water Sci. and Tech. 32(3):159167. Yuan, T.L., 1966. Characteristics of surface and spodic horizons of some Spodosols. Soil Crop Sci. Soc. Fla Proc. 26:163-174. Zhou, M., R.D. Rhue, and W.G. Harris. 1997. Phosphorus sorption characteristics of Bh and Bt horizons from sandy coastal plain soils. Soil Sci. Soc. Am. J. 61:1364-1369.

PAGE 241

224 BIOGRAPHICAL SKETCH Kathleen McKee was born in Anaco, Venezu ela, daughter of a petroleum engineer and an elementary schoolteacher. Her fam ily moved back to their hometown of St. Louis, Missouri when she was 10 years old. In 1989, she received her Bachelor of Arts degree from St. Louis University, with a double major in computer science and Spanish. The information technology career path led her through many fulfilling projects of database programming, software training, pr oject management, and pr ocess engineering. Her interest in environmenta l and physical sciences had her attending nigh t courses and volunteering for conservation organizations in Missouri, Massachusetts, and Ecuador. She spent 1 year studying soil science full-time at the University of MassachusettsAmherst, and she became interested in linking technology, soil science, and environmental problems. This thesis attempts to serve as a culmination of this pursuit.