|
SPATIAL AND TEMPORAL DYNAMICS IN THE EVERGLADES ECOSYSTEM
WITH IMPLICATIONS FOR WATER DELIVERIES TO EVERGLADES
NATIONAL PARK
By
LANCE H. GUNDERSON
A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
1992
Copyright 1992
by
Lance H. Gunderson
Dedicated to Gene, Dorothy and Sara Gunderson
My immediate links with the past and the future
ACKNOWLEDGMENTS
The author is indebted to numerous individuals and agencies, without
whom this work would not have been completed. First and foremost, my
deepest thanks go to Buzz Holling, for his support and guidance in the midst of
shifting paradigms, but most of all for teaching me that it is important to have
fun in the endeavor of science. I am very grateful to J.J. Delfino, who has been
extremely supportive as chairman of the committee. Committee members
Warren Viessman, Ronnie Best and Dan Spangler all provided expert comments
and assistance. Clay Montague and H.T. Odum are acknowledged for their
input in the early stages of the work.
Friends, family and colleagues all assisted with various tasks. John
Stenberg is gratefully acknowledged for his help on tasks too numerous to list.
Dave Sikkema, George Schardt, and Bob Johnson provided data from Everglades
National Park. John Stenberg and Alan Herndon were integral parts of the
evapotranspiration studies. Steve Davis, John Richardson, and Jennifer Silviera
all provided maps used in the cross scale analyses of the vegetation. Steve Light
helped enlighten (groan) me on the theories and practice of public policy and
water resources. No thanks would be too much for Carl Walters. Candy Lane,
Toni Carter, Keiley and Kenny Pilotto all helped with tasks along the way.
Finally, I am grateful to Bev, for taking care of all the details that enabled me to
tackle this project, and everything else.
This work was supported by grants from the South Florida Water
Management District, the Division of Sponsored Research at the University of
Florida and funds from the Arthur R. Marshall, Jr. Chair in Ecological Sciences.
TABLE OF CONTENTS
pae
A CKN O W LED G M EN TS ............................................................................................ iv
LIST OF TABLES .......................................................................................................... vii
LIST O F FIG U R ES ...................................................................................................... viii
A BST R A C T ............................................................................................................. xvi
CHAPTER I. INTRODUCTION TO THE EVERGLADES ECOSYSTEM ......... 1
The Everglades Ecosystem........................................ ........................... 3
H ydrology ......................................... ................. ........................... 7
History of Human Use ........................................ .................................... 10
W after M anagem ent.................................................................................... 12
Water Deliveries to Everglades National Park .......................................... 18
Managing Ecological Systems ............................................................. 23
Sum m ary ..................................................................................................... 25
CHAPTER 2. POSING THE QUESTIONS................................................... 26
Views of Ecosystem Structure and Function............................................ 27
H ypotheses .................................................................................................. 31
Water Deliveries to Everglades National Park The First
H ypothesis Set ................................................ ........................ 32
Cross-Scale Patterns In The Everglades Ecosystem The
Second Hypothesis Set ............................................ ........... ... 33
O objectives ................................................................................... . ............... 34
CHAPTER 3. MODELING THE "RIVER OF GRASS"......................................... 36
Background ................................................................................................. 38
Evapotranspiration in Wetlands ................................... ............ 38
Measurements of Evapotranspiration in southern Florida............ 40
Transpiration from Three Everglades Plant Communities............ 43
Measurements of community evapotranspiration...................... 46
Flow in W wetlands ...................................... .................................. 51
Ecological Models and Scale............................................................... 54
Model Description and Development................................................... 56
Hydrologic Components ................................................................. 56
Vegetation Com ponents ...................................................................... 59
Changes in Landscape Vegetation Types...................... ............ 62
Development of Flow Coefficients for Landscape Units ............ 65
Development of Evapotranspiration Coefficients for Landscape
U n its ...................................................................................... . ............... 67
R esu lts .......................................................................... . ............ .................. 70
Sensitivity Analysis- Flow and Evapotranspiration ..................... 71
Agreement with Historic Data .................................... ........... .. 73
Linkages between hydrology and vegetation................................ 77
Flow and upstream area.................................................................... 84
Summary ..................................................................................................... 89
CHAPTER 4. A CROSS SCALE EXPLORATION OF THE EVERGLADES
LANDSCAPE. .................................................................................................... 93
Methodology........................................................................ .. ........................ 94
Methods to Detect Discontinuities........ ...................................... 100
Methods to Analyze Patterns of Self-Similarity............................. 103
Techniques of estimating fractal dimensions................................. 107
Fourier analysis .................................................................................... 112
Summary of Methodology................................................................ 113
Data Sets Used In The Cross-Scale Analyses ............................................... 116
Fire data ........................................................... .............................. 122
Hydrologic data sets .......................................................................... 127
Evaporation Data ................................................................................. 127
Sea Level Data................................................ .................................... 128
Results Of Cross-Scale Analyses .................................................................. 129
Topographic data ................................................................................. 132
Vegetation Patterns ............................................................................ 138
F ire s ........................................................................................................ 144
R ain fall ................................................................................................... 153
W ater Levels ......................................................................................... 173
W ater Flow ............................................................................................ 184
F ires ........................................................................................................ 18 7
Sea Level .......................................................................................... 187
Temperature and Pan Evaporation ................................................. 192
Discussion of Results ...................................................................................... 203
Su m m ary ........................................................................................................... 214
CHAPTER 5. AN END AND A BEGINNING...................................................... 217
A Su m m ary ....................................................................................................... 217
Understanding Ecosystem Dynamics through Alternative
P arad igm s.......................................................................... ............................. 221
Prognosis for System Restoration ................................................................ 223
LITERATURE CITED .................................................................................................. 226
BIOGRAPHICAL SKETCH .................................................................................. 239
LIST OF TABLES
Table 1. History of major water management structures in the
Everglades ecosystem that influenced water deliveries to
Everglades N national Park. ..................................... ............. 14
Table 2. History of major water management policies that influenced
water deliveries to southern Everglades and Everglades
N national Park ............................................... ......................... 15
Table 3. Daily transpiration rates for three vegetation types in the
Everglades ....................................... .................... .................. 45
Table 4. Description of vegetation categories (landscape units) used
in Everglades m odel. ............................................ ........... .... 61
Table 5. Vegetation density and related flow coefficients as a function
of depth for sawgrass, tree island, wet and marl prairie
vegetation types..................................................................... ..... 64
Table 6. Spatially weighted flow coefficients for each landscape unit
used in the Everglades M odel. .................................. ............ 67
Table 7. Average daily and annual evapotranspiration for plant
communities used to develop evapotranspiration
coefficients for Everglades model.............................................. 68
Table 8. Summary of data sets used in cross-scale analyses of
Everglades ecosystem .................................................................. 130
Table 9. Summary of cross-scale analyses of spatial data sets,
indicating break points detected by fractal and gap
an alyses. .......................................................................................... 131
Table 10. Summary of cross-scale analyses of temporal data sets,
indicating dominant frequencies found in each set................ 131
LIST OF FIGURES
Figure 1. Location of the Kissimmee River, Lake Okeechobee and
Everglades Drainage Basin in Southern Florida.................... 4
Figure 2. Current broad scale land-use designations in the historic
freshwater Everglades drainage basin...................................... 6
Figure 3. Mean daily transpiration rates from sawgrass, tree island
and marl prairie plant communities.......................................... 47
Figure 4. Time course of mean daily evapotranspiration (cm/day) at
study sites P33 and P37 from February 1985 through
Septem ber 1986. ............................................. ....................... 49
Figure 5. Model grid used to depict hydrologic and vegetation
dynamics in Everglades ecosystem. .......................................... 57
Figure 6. Location of sample rain and stage gauges, flow sections and
pan evaporation sites within the Everglades region................. 60
Figure 7. Percent cover of sawgrass, wet prairie, and tree island in
various w indow sizes. .......................................... ............ ... 63
Figure 8. Effects of doubling and halving flow coefficients on
simulated flow through Tamiami flow section. ...................... 72
Figure 9. Effects of varying base evapotranspiration rates on
simulated flow through Tamiami flow section. ......................74
Figure 10. Time series of simulated (solid lines) and actual (dashed
lines) stages at gauge P33 from 1960 to 1988........................ 75
Figure 11. Time series of simulated (solid lines) and actual (dashed
lines) stages at gauge P35 from 1960 to 1988............................ 76
Figure 12. Map codes for initial landscape vegetation types in the
Everglades model. Dominant code in the Everglades
proper is type 1. Cover types developed from Davis
(1943). Explanation of codes is found in Table 4 .................... 79
Figure 13. Map codes of landscape vegetation types in grid cells of
Everglades model at end of 28 year simulation run. Note
presence of type 3 codes in right -central portion of array.
See Table 4 for explanation of codes....................................... 81
Figure 14.
Figure 15.
Figure 16.
Figure 17.
Figure 18.
Figure 19.
Results of changing vegetation patterns on simulated flow
through three flow sections in the southern Everglades.......... 83
Difference in predicted stage between models with and
without vegetation linkages of succession,
evapotranspiration and flow. The model without
linkages tended to predict higher stages (less than 0.5 ft),
under both the natural and managed scenario........................ 85
Three dimension plot showing increase in annual flow (z) as
upstream area is increased (y) over a 28 simulated year. ........ 87
Log-log plot of simulated flow versus upstream area. Scatter
due to differences in annual rainfall. Gray area
represents bounds of actual flow and equivalent area............. 90
Rank order plot, difference indices 1 and 2 for mock data set
designed to simulate continuous normal distribution.
Distribution has mean of 50. For smooth curves such as
this, no value of DI1 is significant. No sequential values
of D I2 are significant ............................................. ........... .... 99
Rank order plot, difference indices 1 and 2 for mock data set
derived from uniform random distribution. Values of
both DI1 and DI2 are significant, yet values between gaps
exhibit no pattern ......................................................................... 100
Figure 20. Rank Order Plot, Difference Indices 1 and 2 for mock data
set designed to show gaps. Gaps are values of difference
indices (1 and 2) outside a 95% confidence interval.
Clumps are not evident, but indicated by a pattern of
uniformity small groups of difference indices. ....................... 101
Figure 21.
Figure 22.
Sample Koch curves to show patterns of increasing fractal
dimension (1.5 1.7), from Mandelbrot (1983)......................... 108
Example of first four levels or divisions for rendering data
into a quadtree storage form at. .................................................. 111
Figure 23. Log-log plots of tile size versus number of tiles for three
Koch Curves of known fractal dimension. Fractal
dimension is estimated by the negative slope of curves.
Actual fractal dimensions (D) of curves are A) D=1.5, B)
D=1.63, C) D=1.71, as shown in Figure 21. .............................. 113
Figure 24. Fourier analysis of random mock time-series data, showing
time-series plot (top) and spectral plot (bottom). This
data set was created by taking 64 samples from a uniform
random number distribution. No peaks are significant in
the spectral plot. ........................................................................... 116
Figure 25. Fourier analysis of structured mock time series data,
showing time-series plot (top) and spectral plot (bottom).
This data set was created by combination of two sine
w av es ................................................................... ......................... 117
Figure 26. Top transect m aps................................................................................ 120
Figure 28. Sample vegetation map, showing patterns of sawgrass
w within a 160 m w indow ............................................................... 125
Figure 29. Sample vegetation map, showing pattern of sawgrass within
a 1600 m sam ple w window ............................................................ 126
Figure 30. Sample vegetation map, showing pattern of sawgrass within
a 16 km sample window. ............................................... 127
Figure 31. Location and size of sample window within which recent
fire histories were analyzed for spatial and temporal
patterns. ....................................................................................... 128
Figure 32. Topographic surveys from transects 1 through 4 ( see Figure
4 for locations) showing elevational variation with east
w est distance. .................................................................................. 135
Figure 33. Topographic surveys from transects 5 through 8 ( see Figure
4 for locations) showing elevational variation with east
w est distance. .................................................................................. 136
Figure 34 Topographic surveys from transects 9 through 11 ( see
Figure 4 for locations) showing elevational variation with
east w est distance. .......................................................................... 137
Figure 35.
Figure 36.
Figure 37.
Figure 38.
Log-log plot of transect length versus step length (top plot)
used to estimate breaks in fractal dimension. Bottom plot
shows results of rolling regressions. Break in top plot is
indicated at point of maximum regression coefficient
(bottom plot). ................................................................................. 138
Log-log plots of box size versus box count used to estimate
fractal dimensions of sawgrass vegetation within three
sample windows. Estimates of fractal dimension are
given by slope of regression. Data from three windows
are combined in lower right plot, and suggest a break in
the fractal dim ension. ................................................................. 141
Log-log plots of box size versus box count used to estimate
fractal dimensions of wet prairie vegetation within three
sample windows. Estimates of fractal dimension are
given by slope of regression. Data from three windows
are combined in lower right plot, and suggest a break in
the fractal dim ension. ................................................................. 142
Log-log plots of box size versus box count used to estimate
fractal dimensions of sawgrass vegetation within three
sample windows. Estimates of fractal dimension are
given by slope of regression. Data from three windows
are combined in lower right plot, and suggest a break in
the fractal dim ension. ................................................................. 143
Figure 39. Rank order plot (top plot) and difference indices (bottom
plot) for sawgrass patches sampled from 1600 m
w window s. ........................................................................................ 145
Figure 40.
Figure 41.
Figure 42.
Rank order plot (top plot) and difference indices (bottom
plot) for wet prairie patches sampled from 1609 m
w window s. ............................................................................ 147
Rank order plot (top plot) and difference indices (bottom
plot) for tree island patches sampled from 1609 m
w window s. .................... ........................... ....... .... 148
Rank order plot (top plot) and difference indices (bottom
plot) for sawgrass patches sampled from 16 km
w indow s. ............................. ................. .... .... 149
Figure 43. Rank order plot (top plot) and difference indices (bottom
plot) for wet prairie patches sampled from 16 km
windows...... ............................... ..........150
Figure 44.
Figure 45.
Figure 46.
Figure 47.
Figure 48.
Figure 49.
Figure 50.
Figure 51.
Figure 52.
Figure 53.
Figure 54.
Figure 55.
Figure 56.
Rank order plot (top plot) and difference indices (bottom
plot) for tree island patches sampled from 16 km
w indow s. ......................................................................................... 151
Rank order plot of log fire sizes from Shark River Slough,
1958-1979. ........................................................................................ 153
Difference Indices 1 and 2 used to determine gaps in fire
sizes, Shark River Slough, 1958-1979 ......................................... 154
Time series plot of daily rainfall data from Tamiami Ranger
Station, 1949 through 1977. ......................................................... 156
Time series plot of daily rainfall data from Royal Palm
Station, 1949 through 1977. ......................................................... 157
Spectral plots from Fourier analysis of daily rainfall from
Tamiami Ranger Station, indicating dominant annual and
m monthly cycles. ............................................................................. 158
Spectral plots from Fourier analysis of daily rainfall from
Royal Palm Station, indicating dominant annual and
m monthly cycles. ............................................................................. 159
Rank order plots of daily rainfall data, Tamiami and Royal
Palm Stations. ................................................................................ 160
Difference index 2 versus rank order for daily rainfall data.
Bars represent value of index, gray areas represent mean
+ 95% C .I........................................................................................ 161
Time series plot of monthly rainfall data from Royal Palm
and Tamiami Ranger Stations, 1949 through 1977.................. 163
Spectral plots from Fourier analysis of monthly rainfall from
Tamiami Station, indicating dominant annual and
m monthly cycles. ............................................................................. 164
Spectral plots from Fourier analysis of monthly rainfall from
Royal Palm Station, indicating dominant annual and
m monthly cycles. ............................................................................. 165
Rank order plot and difference index 2 for monthly rainfall
data from Tamiami Ranger Station. Arrows locate
significant gaps. ...................................... 166
Figure 57.
Figure 58.
Figure 59.
Figure 60.
Figure 61.
Figure 62.
Figure 63.
Figure 64.
Figure 65.
Figure 66.
Figure 67.
Figure 68.
Figure 69.
Figure 70.
Figure 71.
Rank order plot and difference index 2 for monthly rainfall
data from Royal Palm Station. Arrows locate significant
g ap s. ........................................................................................ . 167
Time series of total annual rainfall at Tamiami and Royal
Palm Stations. ............................................................................... 168
Spectral plots from Fourier analysis of annual rainfall data
from Royal Palm and Tamiami Stations................................. 170
Rank order plot and difference index 2 for monthly rainfall
data from Royal Palm Station. Arrows locate significant
g ap s. ........................................................................................ . 171
Hierarchical Cluster Tree for annual rainfall at Royal Palm
Station ............................................................................................. 172
Rank order plot and difference index 2 for monthly rainfall
data from Tamiami Ranger Station. Arrows locate
significant gaps. ............................................................................. 173
Hierarchical Cluster Tree for annual rainfall at Tamiami
R anger Station................................................... ........................... 174
Time series plots of monthly water levels at gauging stations
P33 and P38, from 1958-1979. ..................................................... 176
Time series plots of monthly water levels at gauging stations
P35 and P37, from 1958-1979. ..................................................... 177
Spectral plots from Fourier analysis of monthly water level
data from four P stations............................................................. 178
Spectral plots from Fourier analysis of daily water level data
from four P stations. .................................................................... 179
Rank order plots of daily water level data from stations P33
and P 35. .......................................................................................... 180
Rank order plots of daily water level data from stations P37
and P 38. .......................................................................................... 181
Rank order plot and difference index 1 for monthly water
level data at station P33. .............................................................. 182
Rank order plot and difference index 1 for monthly water
Figure 72.
Figure 73.
Figure 74.
Figure 75.
Figure 76.
Figure 77.
Figure 78.
Figure 79.
Figure 80.
Figure 81.
Figure 82.
Figure 83.
Figure 84.
Figure 85.
level data at station P35. .............................................................. 183
Rank order plot and difference index 1 for monthly water
level data at station P37. .............................................................. 184
Rank order plot and difference index 1 for monthly water
level data at station P38............................................................... 185
Time series plot of monthly water flow across Tamiami Trail
flow section 1940-1982. ................................................................ 187
Spectral plots from Fourier analysis of monthly flow across
Tamiami Trail flow section. ........................................................ 188
Rank order plot of monthly water flow across Tamiami flow
section ............................................................................................. 190
Difference indices 1 (top plot) and 2 (bottom plot) for
monthly flow data, Tamiami flow section................................ 191
Time series plot of log fire sizes (top plot) from 1958-1979,
and spectral analysis (bottom plot) indicating dominant
cycles in fire data. ........................................................................... 192
Time series plots of mean monthly sea level elevation at Key
West, Miami, and Naples for the time period 1910
through 1990. ................................................................................. 193
Spectral plots from Fourier analysis of detrended sea level
data from Key West. Dominant cycle is the annual
period .............................................................................................. 195
Spectral plots from Fourier analysis of detrended sea level
data from Miami. Dominant cycle is the annual period.......... 196
Spectral plots from Fourier analysis of detrended sea level
data from Naples. Dominant cycle is the annual period......... 197
Time series plot of mean monthly minimum and maximum
air temperatures from Belle Glade and Tamiami Ranger
S tatio n s............................................................................................. 198
Spectral plots from Fourier analyses of maximum and
minimum monthly temperature data from Belle Glade.......... 199
Spectral plots from Fourier analyses of maximum and
Figure 8(
Figure 8
Figure 8
Figure 8
Figure
Figure
Figure 9
Figure
Figure 9
Figure 95
minimum monthly temperature data from Tamiami
R anger Station............................................................................... 200
6. Time series plots of monthly pan evaporation from Belle
Glade and Tamiami Ranger Stations, 1965 through 1991........ 201
.7. Spectral plots from Fourier analyses of monthly pan
evaporation from Belle Glade and Tamiami Ranger
Stations, 1965 through 1991. ....................................................... 202
8. Time series plots of daily evapotranspiration and pan
evaporation at site P33, January 1985 through October
1986 ......................................................................................... . 203
9. Time series plots of daily evapotranspiration and pan
evaporation at site P37, January 1985 through October
1986 ................................................................................................. 204
90. Spectral plots from Fourier analyses of daily
evapotranspiration at sites P33 and P37. Bars represent
magnitude of cycle, gray 95% C.I................................................ 206
)1. Spectral plots from Fourier analyses of daily pan
evaporation at sites P33 and P37. Bars represent
m agnitude of cycle, gray 95% C.I. ............................................... 207
2. Results of mini-model frequency of rain input (A), ET
outflow (B), and stage fluctuations (C). Note difference
in time range, stage plot covers 40 simulated years................ 211
)3. Hydrologic hierarchies in the Everglades ecosystem,
showing scales of dominant frequencies in surface water
and atm ospheric variation. ......................................................... 213
,4. Topographic hierarchies in the Everglades ecosystem,
indicating breaks between microtopographic and
macrotopographic features and processes. .............................. 215
. Vegetation hierarchies in the Everglades ecosystem, showing
scales of plant species, communities, landscape units and
the Everglades ecosystem as defined by breaks in the
fractal dim ension............................................................................ 217
Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
SPATIAL AND TEMPORAL DYNAMICS IN THE EVERGLADES ECOSYSTEM
WITH IMPLICATIONS FOR WATER DELIVERIES TO EVERGLADES
NATIONAL PARK
Lance H. Gunderson
December 1992
Chairman: Joseph J. Delfino
Cochairman: C.S. Holling
Major Department: Environmental Engineering Sciences
The Everglades is a unique wetland ecosystem. During this century, the
ecosystem has been partitioned for disparate uses of human habitation,
agriculture, water conservation and ecosystem conservation in a national park.
The sustainability of Everglades National Park is dependent upon upstream
water sources. Water management in the Everglades and water deliveries to the
Park are linked to human perceptions of ecosystem dynamics.
One line of inquiry used expansion of a state-of-the-art computer model to
examine the upstream area that once contributed water to the Park. Linkages
between vegetation and hydrology were added as vegetation mediation of
evapotranspiration and flow and hydrologically induced vegetation changes, but
neither addition appreciably improve understanding of hydrodynamics in the
Everglades system at the scale of the model. Prior to management, the entire
system, south of Lake Okeechobee, contributed flow to Everglades Park except
during dry years. Since the onset of intensive water management, an equivalent
area of only about one-third of the historic drainage basin has supplied water
into the Park. But these conclusions are dependent upon the assumptions made
to represent the system at a specific spatial-temporal scale in a model. At other
scales the conclusions could well be different. That led to the second major topic
of this thesis; that of cross-scale structure and dynamics.
A cross-scale mode of inquiry suggests that ecosystems exhibit
discontinuities in spatial structures and temporal patterns across time and space
due to the interaction of key processes operating over different scale ranges.
Spatial patterns in the topography, vegetation and fire data sets exhibited scale
regions of self-similarity separated by distinct breaks. Temporal patterns of
rainfall, stage, flow, evaporation and sea-level exhibited multiple cycles. These
analyses support the theory that ecosystems are structured around a few
keystone variables of mixed spatial and temporal dimensions. Dramatic
discontinuities appear in patterns as a result of the interactions of processes
operating at different space and time domains. This emerging viewpoint of
ecosystem structure and dynamics will hopefully provide a basis for new
understanding and hence improved management of this unique ecosystem.
xvii
CHAPTER 1.
INTRODUCTION TO THE EVERGLADES ECOSYSTEM
There are no other Everglades in the world.
-Marjory Stoneman Douglas
The Everglades is a wetland ecosystem unlike any other on Earth.
Situated in the subtropics of southern Florida, the unique combination of
physiography and biota blend into a landscape whose name is internationally
recognized. Undoubtedly some of the values and distinctions that the area now
holds are due to attributes of the natural system. During the twentieth century,
the human population in and around the Everglades ecosystem has increased
dramatically, resulting in a myriad of demands on and uses of a unique
ecosystem. Many of the current management problems are associated with the
historical spatial partitioning of resources within a once contiguous ecosystem.
The pattern unfolding throughout the past century is one of a transition of land
uses, from a pristine wetland with negligible human use to one dominated by a
variety of human uses each with characteristic spatial and temporal domain.
These varied land uses range from intensive agriculture in the northern
Everglades to Everglades National Park in the south.
The primary purpose of this work is to improve understanding of the
critical processes and factors in the Everglades ecosystem that influence water
deliveries to Everglades National Park. The issues surrounding water deliveries
to the park cannot be described from a single moment in history nor from a
spatial perspective of the current border of the park. Indeed, the water problems
of the park are woven throughout a rich tapestry extending back thousands of
years and covering the southern half of peninsular Florida. In order to make the
problem tractable, the dissertation is divided into five chapters. The introductory
chapter contains descriptions of the natural and human histories in the system
that lead to a conclusion that water management is fundamentally linked to
concepts of ecosystem dynamics. The second chapter in this work compares and
contrasts alternative concepts of ecosystem dynamics from which the hypotheses
and objectives of the dissertation are derived. The third chapter presents the
results of an attempt to invalidate the hypothesis that the entire Everglades
drainage basin contributed water to the park. The "upstream area" hypothesis
was tested using a state-of-the-art ecosystem model that couples vegetation and
hydrologic dynamics. The fourth chapter contains the results of analyses of a
series of data sets that are used to test the second hypothesis, based on an
alternative concept, and seeks to understand system dynamics based upon a
recognition of the role of discontinuities in both structure and processes The
final chapter contains a summary that compares and contrasts the understanding
of system dynamics and water deliveries that were developed in chapters three
and four and presents implications of these results on water management and
policy.
This introductory chapter is devoted to describing both the natural and
human components in the Everglades ecosystem. First, the components and
processes of the pristine or natural Everglades ecosystem are described. The next
three sections are historical accounts, documenting the increasing human
involvement with the system. These historical accounts include a brief history of
relevant human activities, a review of how water management developed in
southern Florida, and finally, how the understanding and policies of water
deliveries to Everglades Park have changed. This chapter concludes with a
description of the linkages between management, policy and understanding of
system dynamics, and how these relationships have evolved in this complex
wetland system.
The Everglades Ecosystem
The Everglades is a distinct physiographic region located in southern
Florida, and its natural features have been described for over 100 years (Heilpren
1887, Willoughby 1898, Harshberger 1914, Harper 1927, Davis 1943, Craighead
1971, Gunderson and Loftus In Press). Prior to intervention by man, the
Everglades encompassed approximately 10,500 km2 of freshwater marshes,
sloughs and hardwood tree islands (Davis 1943). The system was approximately
210 km along the north-south axis, bordered by Lake Okeechobee on the north
and Florida Bay on the south. The widest east-west dimension was 77 km, from
the higher Atlantic coastal ridge on the east to the Big Cypress Swamp on the
west (Figure 1).
The wetland complex is a result of a large arcuate trough in the
underlying limestone bedrock. Three surficial formations are recognized, and all
were formed by shallow marine accumulations during the Pleistocene Era,
primarily the Sangamon interglacial stage (Parker and Cooke 1944). The Fort
Thompson formation underlies the northern Everglades, and is comprised of
marine and freshwater marls beds interleaved with limestone and sandstone.
The Anastasia formation is found in the northeastern Everglades, and is
characterized by sandy limestone, calcareous sandstone. The surficial feature of
the southern Everglades is Miami limestone, comprised of oolitic and bryozoan
facies (Hoffmeister 1974).
ATLANTIC
OCEAN
BIG
CYPRESS
SWAMP
GULF OF
MEXICO
Florida Bay
Figure 1. Location of the Kissimmee River, Lake Okeechobee and Everglades
Drainage Basin in Southern Florida.
Historic Freshwater
Everglades
.-- Kissimee Lake
S Okeechobee-Everglades
Drainage Basin
I 0 16 32 km
S 0 10 20 mi
67 ev4
5
The wetland soils of the Everglades are Holocene sediments, categorized
as peats, mucks and marls, and are biogenic. The oldest soils in the Everglades
are approximately 5500 years old (Gleason et al. 1984), dating back to an
approximately 3 m transgression of sea level (Robbin 1984). Peats and mucks are
histosols, named by the dominant recognizable plant remains from which the
soils are derived, and accumulate under extended periods of inundation. The
marls are a calcitic mud, produced by reprecipitation of calcium carbonate from
saturated water during photosynthesis by blue-green algae (Gleason 1972).
The topography of both the bedrock and the soil surface is flat,
characterized by almost no relief with extremely low gradients. The maximum
elevations in the northern Everglades are approximately 5.3 m above the national
geodetic vertical datum (NGVD), and now occur in the Arthur R. Marshall
National Wildlife Refuge (Figure 2). The elevational gradient is mostly north to
south, with an average slope of 2.8 cm/km (Parker et al. 1955). The variation in
elevation is attributed to the underlying bedrock structure and accumulations of
organic sediments. The microtopographic variation is caused by and contributes
to differences in vegetation cover and type.
The vegetation of the Everglades region is a complex of gramineous and
woody wetland associations. The spatially dominant communities are sawgrass
marshes, wet prairies, and hardwood swamp forests (Davis 1943, Craighead
1971, Olmsted et al. 1980, Gunderson and Loftus In Press). Sawgrass marshes,
monotypic stands of sawgrass, Cladium jamaicense found over peat and marl, are
the ubiquitous, characteristic association of the Everglades. Wet prairies over
peat are sparsely vegetated, generally dominated by either spikerush Eleocharis
cellulosa, or maidencane, Panicum hemitomon. Wet prairies on marl are diverse
association, dominated by sawgrass and muhly grass, Muh-lenbergia filipes, and
contain over a hundred other species (Olmsted et al. 1980).
20 m
20 nru
FLORIDA
BAY
Figure 2. Current broad scale land-use designations in the historic freshwater
Everglades drainage basin.
The hardwood swamp forests of the Everglades are called tree islands,
descriptive of the isolated clumps of trees surrounded by the lower stature
wetland grass communities. Dominant species in the tree islands are mostly bay
trees; swamp bay, Magnolia virginiana, red bay, Persea palustris, dahoon holly, Ilex
cassine, wax myrtle, Myrica cerifera.
The Everglades is a unique wetland due in part to the spatial and
temporal patterns of the components of the hydrologic regime. Year to year
variation in the hydrologic cycle results in oscillating periods of flood and
drought. The intra-annual variation is also great, characterized by wet summers
and dry winters. Rainfall and overland flow are the principal inputs, yet the
relative contribution of each to the hydrologic budget is debated. The magnitude
of direct rainfall contribution to the hydrologic regime of the Everglades
distinguishes it from other large freshwater wetland systems such as the Llanos
in Venezuela, the Okavango in Botswana or the Pantenal in Brazil where most of
the marsh water originates from rivers.
The climate of southern Florida is subtropical, classified by Hela (1952) as
a tropical savanna, with insufficient rainfall during the summer months to
compensate for a winter dry season. The area has also been classified as
subtropical moist forest type (Dohrenwend 1977, Greller 1980) due to the high
annual rainfall and moderate annual biotemperatures.
Rainfall over the Everglades exhibits both spatial and temporal variability.
Annual rainfall over the system averages 130 cm (Thomas 1970, Bradley 1972,
MacVicar and Lin 1984). Annual rainfall extremes for the period of record 1940
through 1980 range from low of 95 cm in 1961 to a high of 270 cm in 1947
(MacVicar and Lin 1984). Thomas (1970), using spectral analysis, found about a
seven year pattern within the record, indicative of a cyclical pattern with this
return interval. Rainfall patterns exhibit distinct seasonality; approximately 85%
of the annual average rain falls between May and October. Thomas (1970) found
that rain totals during the wet season were bimodally distributed, with peaks in
June and September. Spatial analysis of rainfall records indicate that the coastal
region receives on the average 30 to 35 cm more than the interior marshes. The
northern Everglades and southern Everglades receive more rainfall than the
central regions (MacVicar and Lin 1984).
The rainfall patterns can be related to different processes which influence
the timing and amount of precipitation. The summer rainy season is attributed
to convective thunderstorms, which are linked to mesoscale land-sea breeze
patterns. During the summer, insolation results in differential heating of the air
over the land mass compared to air over the water. The heated air over the land
rises, creating low pressure and establishing a pressure gradient along which
maritime air flows toward the center of the Florida peninsula. The moisture-
laden air rises, cools adiabatically, condenses, and forms convective
thundershowers. This process has been described as the 'rain machine' (Pardue
1982, Yates 1982). Some authors suspect that rainfall totals have decreased
because drainage and development have altered the net radiation budget
(increased reflectance due to a higher albedo of developed areas) which in turn
decrease the rate of convection (Gannon 1978, Pardue 1982). Statistically higher
rain amounts measured during September have been explained by the greater
incidence of tropical cyclones during this month. Rain during the winter dry
season is associated with the passage of cold fronts that pass on the average
every seven days. Annual variations in frontal passage have been linked to jet
stream location.
Historically, surficial flow left the Everglades system through a number of
pathways. In the northeast, water moved through the cypress-dominated
Hungryland and Loxahatchee Sloughs. Major rivers that carried Everglades
waters through the coastal ridge include the New River, Little River and Miami
River. Surface water moved through the higher coastal ridge in a series of
transverse glades. Water flow in the southern Everglades occurred in the broad
shallow depressions of Taylor and Shark Sloughs, named for the rivers that
received the bulk of the flow.
Although the Everglades are recognized as a distinct physiographic
region, it is part of a much larger drainage system, containing a number of river
systems and Lake Okeechobee to the north (Figure 1). Prior to development,
hydrologic connections were traceable to central Florida where the Kissimmee
River originates. The Kissimmee is the largest of the rivers and other smaller
creeks and sloughs which empty water into Lake Okeechobee. Surface water
entered the Everglades from the southern boundary of the lake at two points
when stages exceeded 4.5 m NGVD, and a 52 km long spillway when stages
exceeded 5.6 m NGVD (Parker 1984). The entire system has been referred to as
the Kissimmee-Lake Okeechobee-Everglades (KLOE) system.
The hydrology of the historic Everglades ecosystem can be summarized as
follows. The system is a wide, shallow flat basin, with an overall small
topographic gradient. The primary hydrologic input is rainfall, and although
averages 1.3 m/yr, is characterized by wide spatial and temporal variability.
Other inputs occurred as surface and subsurface flow from Lake Okeechobee and
from wide sloughs in the Big Cypress Swamp. Evapotranspiration is the primary
avenue of water loss, estimated to be 80% of rainfall. Remaining water in the
system flows slowly to the south either to the east recharging the surficial aquifer
of the coastal ridge or the southwest entering the estuarine mangrove zone prior
to reaching Florida Bay. The seasonal patterns of rainfall and evapotranspiration
interact to yield distinct annual wet and dry periods as well as variations in
overland flow.
History of Human Use
Evidence of human inhabitance in southern Florida dates to well over
10,000 years. B.P. (Carr and Beriault 1984), prior to the existence of the vast
wetland ecosystem. Written accounts, which date back almost 500 years,
describe native humans using the resources of the wetland ecosystem. Nunez de
Cabeza (1514) relays descriptions of the fierce native Indian tribes that inhabited
the coastal portions of southern Florida and the peaceful Mayami tribes which
colonized the edge of Lake Okeechobee. These early Americans probably burned
the Everglades (Robertson 1954), and used the area for hunting and fishing
purposes.
The name Everglades first appeared on British maps in the early 1800s
(Vignoles 1823) probably a contraction of "Never a glade," descriptive of the large
treeless expanses. With the expansion of European derived settlers throughout
the southeastern coastal plain, native Americans translocated from the Carolinas
to southern Florida. The term "Seminole," which is the name for the major tribes
of current native populations that persist today, means "runaway." The Seminole
term for the area is "Pa-hay-okee", which loosely translates into "grassy lake",
again, descriptive of a non-forested wetland system. These native Americans
used the elevated tree islands for homesites and cultivation of crops, as well as
hunted and fished throughout the system. The United States fought a series of
wars with the Seminoles during the mid 1800s, restricting their territory to a few
reservations through south Florida. One remains within the Everglades proper,
where the Miccosukee Indians still retain land use rights.
The latter part of the nineteenth century marked the first influx of white
settlement and attempts at "reclamation" of the wasteland known as the
Everglades. Soon after Florida became a state in 1845, early settlers and their
governments embarked on programs to drain the Everglades for habitation and
agriculture. Buckingham Smith was commissioned by the U.S. Senate to
reconnoiter the Everglades for development potential (Smith 1848). In 1850,
under the Swamp and Overflowed Lands Act, the federal government deeded
7500 mi2 to the state, including the Everglades. The Florida legislature
established the Internal Improvement Fund, whose board was to sell and
improve these swamp lands through drainage. Attempts at manipulation of the
water were ineffective in the 1800s, as the magnitudes of the variations in
hydrology were far greater than the minor control structures could handle.
By 1900, initial colonization of the coastal regions east of the Everglades
was underway. The population of Palm Beach, Broward and Dade counties in
southern Florida in 1900 was 28,000 (U. S. Dept. of Commerce 1990). By 1920, the
major land uses now found in southern Florida had started. Urban development
was occurring along the railroad line down the east coast. Agriculture was
developing in the peat lands south of Lake Okeechobee. Conservation of the
natural resources had begun with the formation of Royal Palm State Park in 1917
in the southern Everglades.
During the period 1920 through 1990, the spatial extent of these land uses
grew, in large around these three general loci. In the 1940s, 283,000 ha of the
northern Everglades was designated as the Everglades Agricultural Area (EAA).
By the mid 1940s Water Conservation areas were designated in the central
regions of the glades to manage water resources for multiple purposes.
Conservationists work started during the 1920s came to fruition in 1935 with the
establishment of Everglades National Park in the southern Everglades, although
the park was not formally dedicated until 1947. The park area was increased in
1989 from 1.4 million to 1.6 million acres by the addition of Northeast Shark
Slough. Urban development along the east coast has followed the exponential
increase in population, and resulted in the drainage and colonization of former
wetland areas. As of 1990, 5.1 million people live within the confines of the
historic drainage basin (U. S. Dept of Commerce 1990). The current
configuration of the Everglades ecosystem depicting agricultural areas in the
north, water conservation areas in the central areas and Everglades National Park
in the south is shown in Figure 2.
Through the past century, the spatial extent of the historic Everglades
ecosystem has been slowly whittled away, to the degree that perhaps one-half of
the original system has been irrevocably converted to specific land uses. As of
1985 the historic Everglades ecosystem as defined by Davis (1943), was
partitioned into at least five major use types. Gunderson and Loftus (In Press),
estimate that 32% of the historic Everglades is in areas designated for water
management, 27% in agriculture, 17% for preservation of natural resources, 12%
has been developed for urban purposes, and 12% remains as drained,
undeveloped lands. Davis et al. (In Press), estimate that only half of the original
land area of the Everglades is still in native vegetation types, and that certain
landscape types, including a large pondapple forest in the north as well as
marshes and cypress forests in the east, are gone. The remaining natural areas
have probably been hydrologically altered, and their future viability is largely
dependent upon water management actions.
Water Management
Water management within the Everglades is accomplished by physical
structures and operational criteria. The physical structures consist of levees or
dikes, canals, water control gates (mainly weirs), and pumps. The operational
criteria are constructed around the multiple objectives of the system. The two
primary objectives are flood protection and water supply, having evolved with
the changes in land use within the system and the nature of the historical
ecosystem. The history of water management appears to be one where natural
events or crises precipitated plans and activities that resulted in more
infrastructure and attempts to control the variation in the natural system.
Reactions to natural crisis have resulted in changes and development of two
components of water management; the physical structures of water manipulation
(Table 1) and the policies and programs by which water is managed (Table 2).
Canal construction typified the earliest period of water management in the
Everglades. The first large canal in the system was completed in 1882, when
dredges excavated a channel between Lake Okeechobee and the Caloosahatchee
River. Water levels in the Lake were reported to have declined approximately 50
cm (Johnson 1958). During the next 45 years, canal construction proceeded
sporadically as a result of intermittent funding. By 1917, four major canals, the
Miami, North New River, Hillsboro and West Palm Beach had penetrated the
interior of the Everglades, probably resulting in some drainage of the wetland
system.
Hurricanes during the 1920s devastated human developments along the
east coast and south of Lake Okeechobee. Earthen dams which had been
constructed to exclude waters of the Lake were breached during the hurricane of
1928, resulting in extensive flooding and a loss of about 2400 lives (Blake 1980).
In response, the federal government funded the construction of the Hoover Dike
around the Lake, which was completed by 1938, in order to contain floodwaters.
During the 1940s, federal and state laws established the system of water
management as it now exists. Rainfall during this decade varied wildly, creating
History of major water management structures in the Everglades
ecosystem that influenced water deliveries to Everglades National
Park.
STRUCTURE
RESULT
Construction of Miami,
North New River, Hillsboro
and Palm Beach Canals
Construction of
Caloosahatchee Canal
Lake Okeechobee levees
Muck levee constructed
Hoover dike constructed
Construction of Tamiami Trail
Everglades Agricultural
Area levees completed
Water Conservation Areas
1, 2, and 3 enclosed by levees
S-12 structures complete
L-67 canal and levee
Drainage of coastal areas
and interior Everglades
wetlands
Lowering of water level in
Lake Okeechobee
Impound water in Lake
Okeechobee, control water
movement to south
Alteration of flow patterns
channel through culverts
Control of water movement
in northern Everglades
Control of water movement
in middle Everglades
Flow spatially constricted to
four flowways
Canal to deliver water into
center of Shark Slough
Table 1.
YEAR
1917
1924
1926
1938
1928
1959
1962
1962
1967
History of major water management policies that influenced water
deliveries to southern Everglades and ENP. (Blake 1980, Wagner
and Rosendahl 1985).
POLICY
PURPOSE
Everglades Drainage District
Flood Control Act PL 80-858
Deliver water from WCA 3A,
based upon stages
Deliver water based upon
stage in Lake Okeechobee
- Minimum delivery schedule
(PL 91-282 guarantee park a
amount of water)
- Flow through plan
(PL 98-181 allowed for
experimental deliveries)
1985 -
present
Rainfall plan
To drain Everglades for
agricultural and development
Ameliorate flood effects by
construction of conservation
areas, levees
Store water in WCA 3A,
park to receive after
storage requirements met
Increase flow to park during
hurricane season, restrict
flow during drought
To assure ENP 260,000 acre-
feet/year, and share in certain
drought adversity
Allow S-12 structures to
remain open, no regulation
schedule
Deliver water based upon
upstream climatic conditions
Table 2.
YEAR
1907
1948
1962
1966
1966
1970
1970
1982
1982
1985
conditions which prompted action. The early 1940s were extremely dry,resulting
in saline intrusion into the freshwater aquifers of the coast and subsequent salt
dam construction. Extensive flooding occurred during 1947, following an
extremely wet summer and the passage of two cyclonic storms. Over 105 inches
(270 cm) of rain was reported to have fallen (MacVicar and Lin 1984) during
1947. This flood resulted in the passage of the federal Flood Control Act in June
1948 (PL 80-853). The act authorized the U.S. Army Corps of Engineers to
develop a plan known as the Central and Southern Florida Project for Flood
Control and Other Purposes, which would address the water management needs
of the area. The plan contained three basic elements: 1) designation of the EAA,
2) construction of water conservation areas in the central Everglades and 3)
construction of an eastern levee. The purposes of the water conservation areas
were to protect the east coast and agricultural areas from flooding, recharge
regional aquifers and prevent salt water intrusion. In 1949, the state legislature
created the Central and Southern Florida Flood Control District (FCD) to act as
local sponsors for the federal project. The FCD was renamed in 1977 as the South
Florida Water Management District, at which time an additional objective,
enhancing environmental resources, was added to the above mentioned
purposes.
Construction of the physical structures of the project began in the early
1950s and continues to be modified to date. Three water conservation areas
(WCA) were surrounded by levees (Figure 2). Water conservation area 1, was
also given designation as the Loxahatchee National Wildlife Refuge in 1951. (In
1984, the area was renamed the Arthur R. Marshall National Wildlife Refuge in
honor of an eminent ecologist). Water Conservation Areas 2 and 3 were divided
into subunits A and B, primarily to decrease infiltration losses in the southeastern
portions of these areas. By 1962, the conservation areas were closed in and
functionally intact. Canal construction to date has resulted in approximately
1400 miles of canals.
Operational criteria for water management in the southern Everglades
revolves around the stated regulation schedules for the water conservation areas.
The schedules are target stages which vary over the year, which tend to revolve
around two objectives: 1) minimizing flood risk during the hurricane season
(June-October) and 2) maximizing storage during the dry season (November-
May). When levels are below regulation schedule, water outflow is minimized to
allow stages to increase to the regulation level. When the schedule is exceeded,
water is released to lower levels. Modifications to these schedules have been
made during recent years. The schedule for WCA 3A has been modified to allow
zones around a certain stage value, within these zones water input and outflow
are moderated so that rapid movement of water is negated. The regulation
schedule of WCA 2 has been modified to allow periodic drydown (Worth 1987).
Currently, the schedule for the Marshall NWR (WCA1), is being evaluated for
changes that would improve wildlife habitat.
The water conservation areas are not only spatially central, but
functionally central to water management in the Everglades. These areas are
designed to be used for many purposes, primarily flood control and water
supply. These areas act as surge tanks in receiving water during flood periods.
Runoff from agricultural areas to the north is placed in these areas. Water in the
WCA's is also kept from flowing into areas to the east, in order to lessen flood
impacts. During dry periods, water is also stored in order to meet demands
along the coast and to the south, especially Everglades Park.
Water Deliveries to Everglades National Park
Estimates of pre-drainage water flow into the area now in Everglades park
are tenuous due to at least two reasons. No measurements of flow were made
prior to 1940 and by 1940 many upstream canals were in place and may have
siphoned upstream waters to the coast. The Miami Canal was cut through the
ridge as of 1917 (Blake 1980) thereby removing water from the area immediately
north and east of the park. Historic (pre-drainage) average annual flows to the
area of the park were calculated to be 2 to 2.5 million acre feet. (Parker 1984). The
U.S. Army Corps of Engineers (1968) calculated a smaller mean value,
approximately 1.25 million acre feet. These flows were estimated to be the
amount of overland flow into the southern Everglades. Smith et al. (1989) using
a correlation between freshwater flow and the annual band width of a coral in
Florida Bay, estimated that during the period 1881-1939 annual flow averaged
1.15 million acre feet (1.4 billion cubic meters), whereas flow during 1940-1986
was estimated to be 0.47 million acre feet (0.5 billion cubic meters). Dynamic
flow models (Walters et al. 1992) driven by actual rainfall during the period 1960-
1987, predict flow to have varied between 0.5 and 2.5 million acre feet (0.62 and
3.1 billion cubic meters ), depending upon rainfall.
Overland flow has entered the area now in northern Everglades Park
(primarily Shark Slough) through man-made structures since about 1928 when a
series of round and square culverts were placed beneath the roadbed of Tamiami
Trail (US Highway 41). Most of these culverts are still in place and deliver water
to northeast Shark River slough. As part of the plan to enclose southern WCA
3A, a levee (named L-29) was constructed on the border between WCA 3A and
ENP. This effectively altered the distribution of flow through the western half of
the historic Shark Slough. Four sets of gates (designated S12A through S12D)
were placed in Levee-29 to allow water movement between the conservation area
and the park. Each of the four gates is comprised of six 25 foot wide vertical lift
gates. Each set of gates is designed for a maximum flow of 8000 cfs (226 m3/sec),
with a maximum headwater stage of 12.4 ft. and maximum tailwater stage of 11.9
ft. (U.S. Army COE 1968). The L-29 borrow canal provides the headwater to the
gates. Other structures that were constructed for various reasons to direct flow
in the Shark Slough, but no longer used, include the L-67 extension canal, S-12-E,
S-12-F and S-14 (Wagner and Rosendahl 1982). The alignment of park boundary
also bisects the other main drainage basin (Taylor Slough) from its headwater.
The physical structures that deliver water at the boundary into Taylor Slough
include a pump station (S-175) that delivers water out of canal L-31 W.
There have been at least eight different time periods each with varying
hydrologic regimes under which water has flowed into the Shark Slough area of
the park. Prior to initiation of construction of L-29 and the S-12 structures, water
flow into the park was unregulated in the sense that water across the boundary
was dependent upon hydraulic gradients within the upstream marshes and only
restricted by the capacity of the culverts. Starting in 1961, overland flow was
entirely cut off to Shark Slough while construction was underway, marking the
second flow regime. From December of 1963 through March of 1965, water was
moved from WCA 3A only after regulation schedule was met, that is, the park
only received excess water after upstream storage were met (Wagner and
Rosendahl 1985). During 1965 and 1966, three zones within WCA 3A were used
to deliver a monthly amount of water. From the period of March 1966 through
September 1970, the stage in Lake Okeechobee was used to determine water
deliveries to the park, with totals scaled from no delivery if the stage was below
12.5 ft, 150 cfs if the stage was above 12.5 and below 13.5, and 1000 cfs if the stage
was greater than 13.5 ft. (Wagner and Rosendahl 1985).
During the 1960s the park experienced low rainfall years and was
concerned about the quantity of water it received in context of increasing urban
demands. Two studies defined the water needs of the park using existing flow
and stage data. Dunn (1960) analyzed data for the period 1947-1952 and found
that the median annual flow into the Shark Slough area was 273,000 acre feet
(3.36 x 108 m3), a value that he recommended be adopted as the minimum flow
requirement. Hartwell et al. (1964) developed stage-duration curves for station
P-33 in the park and stage-discharge correlation between P-33 and flow into the
park, and used these relationships to determine an annual discharge requirement
of 243,000 ac. ft. (2.97 x 108 m3). A crude average of these two figures was
incorporated into a congressional act in 1970 (PL 91-282) which guaranteed the
park an annual minimum delivery of 315,000 ac. ft. (3.85 x 108 m3) or 16% of the
water in the system. These annual deliveries were to be partitioned into the three
flow sections into the park. Shark Slough was to receive a minimum of 260,000
ac. ft (3.18 x 108 m3) annually, 37,000 ac. ft.( 0.45 x 108 m3) were to be delivered
into Taylor Slough, and 18,000 ac. ft. (0.22 x 108 m3) into the eastern panhandle
area of the park (Wagner and Rosendahl 1985). This law established the legal
right of the park to a minimum amount of water and to share adversity
associated with periods of drought. During the 1970s the minimum delivery
concept was altered from a minimum threshold to one of a static portion of water
allocated to the park each year. The annual flows through the S-12s were
regulated tightly, and during the years 1970 through 1982, met minimum
delivery requirements, but tended to release water over the schedule during
summer months.
In 1983, following a wet year and changes in the operating criteria for
backpumping into Lake Okeechobee, the park service requested alterations to the
"minimum" delivery schedule. Fearing too much water would come into the
park, a number of alterations to the structures of the system were requested,
along with changes to methods of delivery. In order to remove water from
eastern and southern WCA 3A through pathways other than into the park,
culverts were placed in Levee 28 in order to allow water to flow into the Big
Cypress. Other outlets for WCA 3A were requested but not implemented. Water
was to be diverted into WCA 3B. In response to these requests, Congress passed
a law (PL 98-181) that allowed for experimental water deliveries to the park. For
the next two years (1983-1985), the S-12 structures were left entirely open, so that
water would enter the park as a function of hydraulic gradients between WCA
3A and the park. Although the flow-through plan may have achieved objectives
of restoring the natural timing of flow, the situation of leaving the gates open did
not bode well with water managers faced with the necessity of storing as much
water as possible in WCA 3A for meeting other needs on the coast.
The latest act in the unfolding play of water deliveries to the park was the
Rainfall plan developed by Tom MacVicar of the SFWMD and staff of the COE.
They analyzed rainfall-runoff data from the period 1940 through 1952, and
developed a statistical model which predicted weekly flow based upon net
rainfall (rainfall minus evapotranspiration) over WCA 3A from the previous ten
week period and the previous weeks' discharge. The model achieved two
objectives; the timing and quantities of deliveries were linked to upstream
weather conditions, and the flow would be re-distributed spatially as it was prior
to the construction of the water conservation area. The regulation schedule was
also modified to allow for variation in water level conditions within WCA 3A in
order to avoid the rapid releases of water into the park (MacVicar and VanLent
1984, MacVicar 1985, Neidrauer and Cooper 1988). In essence, the rainfall plan
limits the source basin of the park to WCA 3A by directly timing delivery to
rainfall over the area. Buried in this delivery plan is the unknown contribution
of other areas in the Everglades (and Lake Okeechobee) to water in WCA 3A and
eventually to the park. This is manifest in the supplemental deliveries, by which
more water than the rainfall formula predicts is delivered. The supplemental
deliveries are linked to a modified regulation schedule. The key elements of the
rainfall plan are 1) to link timing and quantity of baseline deliveries to upstream
rainfall, 2) to increase quantity of flow during periods of high water, 3) to
decrease quantity of insured deliveries during dry periods, and 4) to supplement
the baseline quantity of water depending upon a wider range of water depths
(regulation schedule) within WCA 3A.
Major determinants to the constantly changing methods and policies of
water delivery to the park have been the observed degradation of biological
resources in the southern Everglades and Everglades National Park. Dry years
and accompanying fires during 1962 and 1971 prompted the appeal for more
guaranteed water. Increased mortality of alligator young (Kushlan and Kushlan
1980) was attributed to rapid water level rises associated with regulatory releases
during the early dry season. Browder (1985) developed relationships between
flow into the estuary of Florida Bay and shrimp production. The most attention
has been drawn to a dramatic decline in the number of wading birds; nesting
success of wading birds has decreased by 95% of levels in 1930s (Robertson and
Kushlan 1984). Reasons for the declines have been intimately linked with
decreases in flow through the park (Ogden 1978, Ogden 1987, Powell et al. 1989,
Walters et al. 1992). Other authors believe that too much water in the Everglades
has contributed to the population decreases (Kushlan 1987) and that the park
should receive less water.
The preceding review of the Everglades ecosystem has followed two
separate paths: one recounts the natural history and the other the human
41
history. These histories intertwine, and are linked by the ways in which humans
perceive, understand and react to nature, the subject of the next section.
Managing Ecological Systems
One interpretation of the history of water management in the Everglades
is that it appears to follow a pattern of crisis and reconfiguration (Light et al, In
press). The crises arise from dramatically unexpected system behavior, such as
floods, droughts and fires. Crises in the past have appeared suddenly, as
surprises, and the subsequent responses have dramatically changed the way the
system has been managed (Table 1). The central reconfiguration occurred
following the flood of 1947, when the Central and Southern Florida Project was
spawned. Since then, other crises have occurred with subsequent changes in the
policy and practice of delivering water to the park. The reaction by humans to
these surprises takes the form of policies and management actions. The
responses are shaped by perceptions and interpretations of how nature operates.
At least two concepts are involved in the interpretation of nature that
create the basis of policy and management formulations. The first concept relates
to various views of system stability. The second concept deals with the assumed
or perceived uncertainties associated with either system understanding or
impacts of management actions. At least three views of system stability have
been abstracted: equilibrium, dynamic and evolutionary (Holling 1987). An
equilibrium view is defined as one dominated by the assumption that key
response variables always return to a point or set of points. The dynamic
perspective recognizes that system variation occurs within and between a range
of stability regions so that system behavior appears at times constant, other times
continuously changing, and at times jumping abruptly into another stability
regime. In the evolutionary view, the stability landscape can change, implying
fundamental structural and organizational changes in the system. Dealing with
the inherent and fundamental uncertainty associated with shifts within and
among these stability domains is at the heart of adaptive management (Holling
1978, Walters 1986). During the last decade both the policy and management
philosophy in the Everglades crossed thresholds involving both changes in views
of stability domains, and in strategies of management.
Policy in the Everglades is still largely rooted in the equilibrium-centered
perspective, although the dynamic view has been recognized and partially
incorporated into management schemes in the mid 1980s. Water movement is
largely determined by regulation schedules in the different components of the
system (Lake Okeechobee and the water conservation areas). These regulation
stages reflect an equilibrium view of water management, that is, it always returns
to an ideal stage within a retention pool. The modifications to WCA 3A schedule
associated with the rainfall delivery plan, however, indicate a shift to a dynamic
viewpoint, allowing variability in the managed system.
Water management in the southern Everglades during the last decade has
developed more attributes of adaptive environmental management (Holling
1978, Walters 1986). Within the last decade, programs such as the iterative
testing plan (Light et al. 1989) have been applied using the concept that water
management necessarily has some experimental attributes. This has even been
codified, by the adoption of PL 98-181 which allows for experimental deliveries
to the park. The rainfall plan can be classified as a passive adaptive technique
(Walters 1986), whereby historical data are used to construct a model that guides
management plans. Two problems with this technique are that 1) environmental
and management effects are confounded, and 2) little opportunity exists for
improving the model or testing new models (Walters 1986, Walters and Holling
1990).
Summary
In this chapter, the key pieces of the interplay between the natural and
human dimensions of the Everglades are described. The undisturbed Everglades
ecosystem can be characterized as an oligotrophic, sub-tropical wetland system
with high temporal variability in rainfall input. The landscape is flat, yet
supports a complex spatial mosaic of marsh and woody vegetation plant
associations. Humans have interacted with the system for as long as it has been a
wetland. Dramatic changes have occurred during this century, within which
time about half of the land area has been converted to agriculture and urban
development. Over the past 50 years, Everglades National Park has been
established, as has one of the largest water management infrastructures in the
world. Water management and deliveries to Everglades Park have undergone
dramatic, non-linear changes resulting from recurrent crises and surprises. The
foundations for policy and management development during these periods of
reconfiguration are intimately linked to and dependent upon our understanding
of ecosystem dynamics. Understanding ecosystem dynamics and the different
paradigms regarding system organization is the point at which the first chapter
ends and the second chapter begins.
CHAPTER 2.
POSING THE QUESTIONS
Using all the weapons of our logical, mathematical and technical armoury we try to prove
that our anticipations were false--in order to put forward, in their stead, new unjustified
and unjustifiable anticipations, new 'rash and premature prejudices' as Bacon derisively
called them
-K.R. Popper
As indicated in the preceding sections, the problem of water deliveries to
Everglades Park has many dimensions, including how ecosystems vary over
different time spans, and the subsequent reactions and adaptations of people to
these fluctuations in the system. Resource policy and management is
fundamentally related to how humans perceive and attempt to comprehend the
vagaries of nature. Even though institutional and human dynamics of the system
are important, they are fundamentally rooted in basic paradigms about how
ecosystems function. The study of ecosystems can be particularly difficult
because of the variety of components, processes and variables. Attempting to
incorporate all variables makes the problem overwhelmingly complex (Gallopin
1991). Simplifying assumptions allow for these studies to become tractable. The
remainder of this chapter will outline and contrast three existing approaches to
simplify understanding of ecosystem dynamics and a description of a new
emerging paradigm. Hopefully, this theoretical background will lay the
framework from which hypotheses and objectives of this work are derived in the
concluding sections of this chapter.
Views of Ecosystem Structure and Function
Understanding and interpreting ecosystem structure and function are
based upon underlying methodological assumptions and paradigms held by the
observer. At least three such concepts are recognized, while a new one is
emerging to account for the paradoxes that emerge from applying the first three.
These viewpoints can be characterized and contrasted by two components of the
paradigms: 1) the factors or variables that are important in the system and, more
fundamentally, 2) the manner in which these variables interact. The first
assumption is that variables interact in such a way that the strength or
significance of the interaction can be tested against a null model that is random.
The second paradigm is based upon a view that the world is structured in a
hierarchical manner, with distinct levels of causation defined by the observer.
The third view of ecosystem science is rooted in mathematical modeling, and
concentrates on system dynamics across a limited range of scales. The newest
belief, emerging because of limitations in the other views, is cross-scalar in scope
and implies a world of lumps and discontinuities in which a small number of key
processes determine function, each over its own range of scales. None of these
views are wrong; indeed, all represent partial truths and continue to thrive
because they are useful. Following a brief characterization of each existing
assumption and their limitations, the emergence of a new view will be presented.
Ecosystem science as practiced by the 'Stochastics' is characterized by
multi-variate statistical approaches. The implicit assumption is that the variables
are operating within similar domains in space and time and therefore have
correspondingly similar ranges of variation. Explicit assumptions include that
the variables are essentially derived from continuous distributions and therefore
have certain properties that can be estimated from sampling. In the extreme,
variation in response variables is partitioned to either the variation of other
variables or to a random error term. In all cases, the null hypothesis is a random
one; relationships can only be inferred by rejection of the null. Examples of work
in the Everglades ecosystem of this type include analyses by Smith et al. (1989)
and Browder (1985) in correlating freshwater flow in the system with biotic
responses in Florida Bay. Indeed, the current rainfall formula developed by
MacVicar (1985) is a statistical approach, whereby flow through the southern
Everglades is regressed against rainfall and stage. This approach is powerful,
because the tools are readily available, and only a statistically significant number
of samples are necessary for application.
The hierarchical view of ecosystems, on the other hand, while a powerful
concept, is still struggling for widespread application after being introduced at
least 50 years ago. The basic framework offered by hierarchists is one of
partitioning variables and interactions into distinct levels or "holons" (Allen and
Starr 1982). Variables that operate at similar scale ranges occupy the same level
within a hierarchy and interact more than with variables between levels. One
feature of hierarchies is dubbed as asymmetry, where the variables at slow levels
constrain the variables at fast levels. The underlying assumptions of variability
are similar to the random approach, in that the variables are assumed to have
continuous distributions and that these distributions are predictable. Another
common belief of hierarchists is that hierarchies are relatively stable, static and
operate near equilibrium. The most current view of hierarchists is that the world
should be partitioned into the appropriate hierarchical level such as landscape,
ecosystem, population, organism for study, analysis and understanding (Allen
and Hoekstra 1992). To my knowledge, no applications of this theory has been
applied to studies of the Everglades.
A tremendously rich understanding of ecosystem structure and function
has been achieved by modellers who apply the third approach to explanation.
Although inductive, the approach can improve understanding by testing
dynamic interactions among variables. Assumptions regarding variables and
interactions can be clearly stated by mathematical formulae translated into
computer code. The utility or power of modeling also carries related costs.
Empirical rules such as "parsimony in the selection of variables" (Clark et al.
1979), the "power of two" (Walters 1986) or the optimum trade-off between
articulation and predictability (Costanza and Sklar 1985) all attest to constraints
on modeling.
The limitations imposed on each of these views has to do with issues of
scaling. All of these approaches to explanation treat both variables and
interactions as scale invariant. Scale invariant means that the behavior of
variables and the rules or properties of interaction do not change within the scale
limits imposed by the observer. The power of these approaches comes from the
knowledge of a rule set, and how far (over what range of scales) the rules apply.
Limitations related to scale are a result of underlying assumptions (stochastics),
theoretical frameworks (hierarchists) and of practical experience modellerss). For
variables to be analyzed, compared or contrasted using the available statistical
approaches, they must change within a similar manner. If there are dramatic
differences in the space or time dynamics of variables, then statistical methods
either give erroneous conclusions or, flat out, don't work. An example of this is
the rainfall formulation (MacVicar 1985) mentioned above, where the regression
analysis indicated that no statistical relationship existed between rainfall and
flow!
Similar problems of the mismatch between variable spatial or temporal
domain arise in hierarchical theories and in the application of modeling
techniques. The hierarchists (Allen and Starr 1982, O'Neill et al. 1986) recognize
that "slower" variables constrain "intermediate" variables while "faster" variables
are essentially meaningless, or noise. Little progress has been made with linking
these variables together other than in a conceptual or qualitative sense.
Modellers have come to essentially similar conclusions, as expressed in the
practical "Rule of Two". The empirical rule states that the best models never
extend more than two orders of magnitude in either space or time. The basic
approach of scaling in modeling, is to treat "slower" variables as constants, and to
treat "faster" variables as random or stochastic events. Existing space-time
models of hydrodynamics in the Everglades fall within this guideline and will be
described later in the modeling section. The result of limitations imposed by
these various approaches is the breakdown of understanding, as evidenced by
inherently unpredictable system behavior (Holling 1986). These limitations and
inevitable surprises helped prompt the development of a theory that attempts to
embrace the cross-scale dynamics of ecosystems.
This emerging cross-scale theory has roots in both the hierarchical and
modellers perspective and can be traced to a review and synthesis of the
dynamics of a number of ecosystems. Holling (1986) compared the dynamics of
23 ecosystems, and concluded that the essential behavior of the system could be
traced to three or four sets of variables, each of which operated at distinctly
different rates. The sample ecosystems were categorized into one of four classes
of systems: forest insects, forest fires, grazing in savannas and aquatic harvesting.
Models of the reviewed ecosystems all generated complex behavior in space and
time that qualitatively correlated with observations of the actual systems. The
essential dynamics of the systems could be attributed to a small number of
keystone variables. The speeds of each keystone variable differed from each
other by as much as an order of magnitude, so that the time constants were
discontinuous in distribution (the hierarchists would consider each keystone
variable as a part of different levels or holons) or as a small number of nested
cycles. The results of the review led to the hypothesis that ecosystem dynamics
are organized around the operation of a few key variables, each operating at
distinct speeds.
The next critical step in development of theory was the proposed
hypothesis that the system should be structured in such a way that the keystone
variables entrain other variables. The entrainment should occur in both spatial
and temporal dimensions creating structural features that exhibit distinct gaps
and lumps and temporal processes that exhibit distinct periodicities. An overt
manifestation of a lumpy, discontinuous world should be expressed by attributes
of the animals that live in these systems. This hypothesis was challenged by a
series of tests using data from three biomes (Holling 1992). The tests using adult
body mass of birds and mammals from the boreal forests, prairies and pelagic
ecosystems, indicated the presence of discrete gaps that defined groups (Holling
op cit.). Alternative hypotheses using developmental, historic or trophic
explanations for the groupings were all invalidated, leaving only the strong
inference that ecosystems (abiotic and biotic components) were similarly
organized (Holling op cit.) into discreet lumps.
Hypotheses
Two hypotheses are posed in this work and arise from two of the
approaches mentioned above. Both are aimed at improving understanding of the
structure and function of the Everglades ecosystem that specifically relates to
system dynamics and flow to Everglades Park. The first hypothesis derives from
the approaches that understanding complex system behavior can be induced
from modeling non-linear interactions among continuous keystone variables
within a constrained range of scales. The second hypothesis is developed from
the cross-scale arguments, and attempts to invalidate the lumpy, discontinuous
view of the world.
Water Deliveries to Everglades National Park The First Hypothesis Set
A dynamic water budget approach is a powerful conceptual tool for
evaluation of the factors influencing deliveries to Everglades Park. The amount
of water in the southern Everglades at any time is a net result of changes between
inputs (rainfall and inflow), and outputs (outflow, evapotranspiration, and
groundwater infiltration). Theoretically, the rates of flow and evapotranspiration
are related to vegetation type and structure. Since Everglades National Park is
situated at the downstream end of the historic ecosystem, it is dependent upon
water from upstream sources. The water that enters the park comes from two
sources: rainfall over the park and overland flow from the north. Assuming that
local rainfall contributions to the park water budget are relatively unchanged, the
first hypothesis deals with the contribution of upstream sources to the water
requirements of the park.
Hypothesis: The effective drainage basin that supplied water to
Everglades Park was the entire Kissimmee, Lake Okeechobee and
historical Everglades ecosystem. Implicit in this hypothesis is that
overland flow is a dominant pathway of water movement, and that
hydrologic continuity throughout the historic system is crucial to
maintaining water supply to the park.
Corollary: In an area as flat as the Everglades, the vegetation and
hydrology are intimately coupled. The structure and type of vegetation
affect both evapotranspiration rates and resistance to overland flow.
Vegetation type and structure are in turn, affected by water depths and
hydroperiod. The coupling between hydrology and vegetation
determines the relationship between the amount of water that flows
through the system and the amount that evapotranspires to the
atmosphere.
Null: The effective drainage basin was a much smaller geographic area.
Overland flow into the park system originates from an effectively smaller
drainage basin. This is because evapotranspirative losses are high relative
to rainfall, therefore water would evaporate before moving very far
downstream. Other users in the system can remove water without major
disruption of the flow that entered the park historically.
Cross-Scale Patterns In The Everglades Ecosystem The Second Hypothesis Set
The processes that influence flow to the southern Everglades cover a wide
range of space and time scales. Rainfall results from atmospheric processes,
ranging from meso-scale (Florida peninsula) to global dynamics. Vegetation
structure can be identified at scale ranges from parts of individual plants (stems,
leaves) to the organization of plant associations in the landscape. The combined
processes of evaporation and transpiration occur from the level of leaf stomata to
entire ecosystems. Other processes have similarly wide ranges of variation in
space and time.
In order to attempt to invalidate the first hypothesis, the methodology
requires that the world of the Everglades be "squeezed" into a framework of fixed
spatial and temporal domains. The second hypothesis is based upon the
emerging theory which suggests that across scale ranges, ecosystems are
organized in such a way so that dumps and gaps appear in structural features
while a small number of cycles and harmonics occur in temporal features.
Hypothesis: The Everglades ecosystem is structured by a small number of
processes, of which hydrology and vegetation are one set of keystone
variables. Over a range of scales, patterns produced by vegetative and
hydrologic processes should have a few characteristic domains in space or
time that are separated by discontinuities. Within the time domain, a few
dominant frequencies or cycles should emerge in the hydrologic processes
of rainfall, water level, flow and evapotranspiration. Within the spatial
domain, a few groupings of object size (such as vegetation patches), or
texture will emerge that correspond to levels in a spatial hierarchical
structure. Other ecosystem level processes, such as topography and fire,
will exhibit similar patterns.
Null: Over a range of scales, the temporal patterns of hydrology and
spatial patterns of vegetation in the Everglades will exhibit structures that
correspond to underlying continuous distributions No dominant or
nested cycles will appear in the time patterns of hydrologic processes such
as rainfall, water level, flow or evapotranspiration. No breaks or
discontinuities in the spatial patterns will be found, and patterns will be
self-similar over a wide range of scales.
Objectives
The aim of this work is to develop new understanding of ecosystem
dynamics by testing hypotheses regarding water and vegetation dynamics that
are rooted in two different viewpoints. There are two main objectives: 1) use
"scale-bound" modeling techniques to help understand factors influencing water
deliveries to Everglades Park, and 2) apply cross-scale analyses to Everglades
data sets to test for breaks or discontinuities in patterns.
The first objective will involve the construction of a spatially and
temporally explicit model to capture the dynamics of the system in order to test
the first hypothesis. The model will couple dynamics of hydrology and
vegetation within a spatial domain of two orders of magnitude and a temporal
domain of almost three orders of magnitude. The model will be used to attempt
invalidation of the proposal that the entire basin contributed water to the
southern Everglades. This objective is the focus of Chapter 3.
The second objective will be to develop, test and apply a variety of cross-
scale methods to identify patterns in keystone variables in the Everglades
ecosystem. Since the theory of cross-scale interactions is just emerging, a great
deal of this work has been devoted to the development of new methods and
methodology. Fortunately, this work was able to reap the benefit of data sets of
many variables that have been collected over the years on different portions of
the Everglades system. The methodology, and results of the cross-scale analyses
are the subject of Chapter 4.
CHAPTER 3.
MODELING THE "RIVER OF GRASS"
When your only tool is a hammer, the answer to every problem is a nail.
-R. Yorque
Almost 50 years ago, Marjory Stoneman Douglas created a dramatic image
when she described the Everglades as a "River of Grass" (Douglas 1947).
Technically, neither of these terms are appropriate. The system is hardly a river
because there is no defined water course and water flows very slowly, only about
60 kilometers a year. The "grass" in the title refers to sawgrass, which is properly
classified as a sedge. However, the metaphor is still appropriate because it can
be interpreted to depict the coupling of the vegetation and hydrology in this
ecosystem that at one time, was a united ecosystem.
In this chapter, the test of the first hypothesis that the entire Everglades
system provided water to Everglades National Park and test of the corollary
hypothesis regarding the coupling of vegetation and hydrology, are presented.
These hypotheses were tested with a model that incorporates coupled vegetation
and hydrologic dynamics over time within an explicit spatial array. Hydrologic
models of the Everglades system have been used to evaluate management
options within the current system configuration (MacVicar 1985) or to create
views of the system prior to human intervention (Walters et al. 1992, Perkins and
MacVicar In prep.). Such models provide a robust methodology from which the
contributions of upstream areas to flow into the southern Everglades can be
evaluated. Key uncertainties in these models include information about
overland flow resistance and evapotranspiration and infiltration to groundwater.
In a system as flat as the Everglades, vegetation influences surficial
hydropatterns by mediating resistance to flow and controlling
evapotranspiration. Hydrologic regimes also influence the vegetation pattern
(Davis 1943, Craighead 1971, Gunderson 1989). The corollary hypothesis posits
that the interactions between hydropatterns and vegetation patterns are coupled
and create feedback loops. None of the previously developed hydrologic models
of the Everglades incorporate complete feedbacks between hydrology and
vegetation. The SFWMD models (MacVicar 1985, Perkins and MacVicar, In
prep.) vary flow and evapotranspiration by land cover types, but the land cover
types do not change as a function of hydrology. The model of Walters et al.
(1992) changes vegetation types as a function of hydrology and other factors, but
has spatially fixed flow and evapotranspiration rates. The approach in this work,
therefore, is to couple vegetation and hydrologic feedback dynamics in the
framework of existing models to test hypothesis about upstream area
contributions to the park.
This chapter is divided into four sections: background, model
development, results and summary. A fair portion of this chapter is devoted to
improving understanding of the interactions among evapotranspiration, flow
and vegetation since they are key uncertainties in the model. The background
section will develop a theoretical base for understanding these processes and the
results of studies that compares evapotranspiration rates among vegetation
communities of the Everglades in order provide a foundation for the linkages in
the model. Following the background section, the model is described including
components and their interactions. The section following the model description
presents the results of testing the upstream area hypothesis and the corollary
hypothesis regarding the coupling of vegetation and hydrologic processes. This
chapter is concluded with a summary of modeling the "River of Grass" and
implications of the results to policy and management.
Background
Evapotranspiration in Wetlands
Evapotranspiration is the combined processes of water flux into the
atmosphere by evaporation from water or soil surface and transpiration from
vegetation. In an area such as the Everglades, both processes are in effect, as
there are areas of relatively sparse vegetation (open water marshes), grassy
wetlands and forested wetland communities. In addition to the interaction
between evapotranspiration and vegetation, other physical variables influence
the rate of water flux. All of these variables, and measures of evapotranspiration,
appear to vary across scales of interest. Previous studies on evapotranspiration
in the Everglades have been made at different scales, and will be reviewed below
in relation to spatial and temporal groupings, but first a review of theoretical
background.
Evapotranspiration is a component of the energy budget. A general
formulation for steady state system at a specific location is shown in equation (1),
modified from Brutsaert (1984), and Viessman et al. (1989). The amount of net
solar radiation (ambient minus reflected) determines the amount of energy
available for other processes. The energy can be used to increase the
temperatures (sensible heat) of both the atmosphere and the soil substrate. Some
of the energy is used in photosynthesis, and some may be moved by advection
(wind) to other areas. The other energy is used for the phase transition of water
from liquid to vapor. Since the evaporative process requires energy for the phase
transition, the energy is not measurable or latent. The latent heat of evaporation
times the rate of evaporation describes the evaporation term in the energy budget
equation.
(1) Rn = LeE+H+G+P+A
Rn = specific flux of incoming net radiation
Le = latent heat of evaporation
E = rate of evaporation
H = specific flux of sensible heat into the atmosphere
G = specific flux of sensible heat into the ground
P = rate of energy used in photosynthesis
A = rate of lateral advection of energy
Due to difficulties in measurement of the latent heat of evaporation and
the confounding effects of vegetation influences on the processes, a number of
techniques have been developed to measure evapotranspiration. The techniques
fall into two categories, those that derive from energy budget, with certain
simplifying assumptions and those that are empirical. Penman (1948) derived a
formula that includes a wind advectionn) term with assumptions of constant
Bowen ratio (ratio between sensible and latent heat) that allows for measurement
at one level. Other techniques derive from the Dalton formulation that estimates
a vapor gradient between the surface and the atmosphere. Empirical methods
include formulations by Blaney-Criddle (1950) who related evapotranspiration
with average temperature; Holdridge (1967) who related vegetation form to
temperature and potential evaporation and simple techniques such as pan
evaporation or lysimetry.
Evapotranspiration is influenced by a mixture of processes that occur at
different scales in space and time. Solar insolation at a spot on the earth
fluctuates on daily, annual and multiple year cycles. The solar radiation is also
influenced by fast (on the scale of minutes) fluctuations in processes, such as
cloud cover. Vegetation both directly and indirectly influences the
evapotranspirative processes. The fastest controls vegetation occur at the level
of the stomata, where water flux is linked to photosynthesis (Jarvis and
McNaughton 1986). Individual plant species' genetic composition and
adaptations influence the size and density of stoma, leaf orientation, responses
to various changes in insolation, wind, humidity and temperature (Jarvis and
McNaughton 1986). The composite architecture of the canopy in wetlands can
influence the reflectance of both short and long wave radiation, with implications
to net energy and water use (Odum 1984, McClanahan and Odum 1991). At the
landscape or regional level, the vegetation cover type influences the reflectance
or albedo.
The preceding paragraphs gave a brief review of evapotranspiration
theory to provide a basis for understanding the various approaches and
techniques for measurement of this complex process. The following section will
present the results of previous investigations and present published
measurements of rates of evapotranspiration at different spatial and temporal
scales.
Measurements of Evapotranspiration in southern Florida
Measures of evapotranspiration in southern Florida have also been made
at different scales ranging from local up to the entire peninsula. For convenience,
these can be grouped for discussion into broad-scale measures that include the
region and basins, medium scale measures (less than 10 m on a side) typified by
lysimeters, and evaporation pans and small scale measures that examine losses
from leaf surfaces.
For the region, Dohrenwend (1977) used an empirical formula developed
by Holdridge (1967) that related evaporation and mean annual biotemperature to
calculate an annual evapotranspiration of about 1000 mm. Input-output analyses
of basins in and around the Everglades calculate a range of values of
evapotranspiration that are similar. Allen et al. (1982) estimated annual
evapotranspirative losses from 890 to 1040 mm from Taylor Creek basin north of
Lake Okeechobee. Leach et al. (1971) estimated evapotranspiration from the
Water Conservation Areas at 965 mm/yr. Shih et al. (1983) estimated water
losses from the Everglades Agricultural area at 1018 mm/yr using a water
budget approach and also compared a number of other techniques and found
annual means ranging from 1018 to 1035 mm.
Most of the smaller scale investigations involve the use of lysimeters
(tanks with planted vegetation), evaporation pans or shallow wells to derive
monthly and annual estimates of evapotranspiration. Clayton et al. (1949)
planted sawgrass in lysimeters and found monthly ranges of 78 to 208 mm and
mean annual losses of 1735 mm. Parker et al. (1955) estimated
evapotranspiration from pan evaporation and reported annual values from 1016
to 1143 mm. The Army Corps of Engineers (1968) reported monthly values from
63 to 135 mm. Shih (1981) compared average monthly and annual
evapotranspiration among sugarcane, alfalfa, and bahiagrass plants planted in
lysimeters with data from Clayton (1949). Shih (1981) found a range of monthly
values from 35 mm to 212 mm for sugarcane. Other crops were within these
monthly averages, and had lower annual totals. Carter et al. (1973) derived
annual estimates of 1100 mm for the Big Cypress Swamp area, immediately west
of the Everglades.
A number of studies attempted to develop relationships between
evapotranspiration (from lysimeters) and climatological data. Most assume
constant Bowen ratio; that is, a constant proportion between the sensible heat
flux and the evapotranspiration. Stephens and Stewart (1963) found that best
approximation to lysimeter values of potential evapotranspiration were based
upon a fraction of ambient radiation. Shih (1981) found good correlation
between lysimeter losses and monthly temperature corrected for cloudy days, a
modification of the Blaney-Criddle technique. Stewart and Mills (1967) and Shih
(1983) both measured decreasing evapotranspiration rates with declining,
subsurface water levels.
Perhaps the fewest studies have been done at the level of the individual or
on a daily basis. Brown (1981) studied pondcypress in southern Florida, and
measured average daily losses at 1.3 mm/day. Dolan et al. (1984) working to the
north measured daily marsh evapotranspiration from 0.5 to 10 mm/day. One
small scale study (Alexander et al. 1976) compared evapotranspiration rates of
potted seedlings of sawgrass and Melaleuca..
Different variables operating at different space and time scales have been
shown to influence evapotranspiration. Some variables, such as radiation are
spatially global; that is, they remain the important input to the process across all
scales. Other variables, such as leaf stomata, dictate fine scale (leaf level) control,
but cease to be important at the regional scale (Jarvis and McNaughton 1986).
More work has been done on scaling measures of evapotranspiration over the
time domain by increasing the extent or window (day to month to year).
Authors working in southern Florida (Stephens and Stewart 1963, Shih 1981, and
Carter et al. 1971) all recognized a decrease in variation as the time unit is
increased.
In reviewing the available literature, few measurements of evaporation or
transpiration from the native plant communities in the Everglades have been
published. In order to incorporate community level measures into the model, a
series of studies were done to develop measures of water loss by community
type. The studies were important for two reasons, 1) to establish that
evapotranspiration rates varied among the major vegetation communities, and 2)
to estimate the magnitude of any differences. If there was no difference among
the vegetation types, then evapotranspiration could be modeled by a variable
that only changed in time and not over space. If there was a difference in water
loss rates among the vegetation communities, then maps of vegetation
communities could be used to develop spatial patterns of evapotranspiration.
The next section of this chapter reports on the summary of two studies
using two different methods to calculate rates of evapotranspiration among the
dominant native plant communities. The first study uses transpiration rates
reported in Herndon and Gunderson (1989), and, with a few assumptions,
attempts to aggregate from the leaf level to the community. The second study
uses data reported by Gunderson and Stenberg (1990) on evapotranspiration
from two wet prairie sites. The measures reported in these two studies will be
summarized for use in the model.
Transpiration from Three Everglades Plant Communities
The rates of water flux from the leaves of dominant species in three
community types; sawgrass, tree island and marl prairie. (wet prairies had no
macrophytes) were measured. The measurements of leaf transpiration were
multiplied by the leaf area per vegetation type to yield community estimates.
Water flux rates (millimoles/cm2/sec) from the leaf surfaces were measured
using a steady state porometer (Li-Cor Model 1600). During the period from
December 1984 through February 1986, measures were made at 10 am, 12 pm
and 2 pm local time during one day every other month. At each sample period a
total of thirty measurements were made at random locations within the
community. The thirty measurements were combined to give an average flux for
each period. The rates of water loss were integrated over the day in order to
yield a water loss per day. To translate or scale these measures to a community
level, estimates of leaf surface area were made for each community type. All
leaves within a one meter square area were counted and leaf areas measured to
yield a leaf area/ m2.
Transpiration rates did not vary much among the vegetation types
sampled. Mean and range of water fluxes (transpiration) from the leaf surface
were determined from the field measures. Sawgrass transpiration ranged from a
low of 0.9 mmol-m2.sec-1 during December 1985 to a high of 3.2 mmol-m2.sec-1 in
July of 1985 (Table 3). Muhly grass, Muhlenbergia filipes, the co-dominant species
in the marl prairie, had similar rates, ranging from 1.6 to 2.4 mmolm2.sec-1 (Table
3). The swamp forest species showed similar transpiration, with values ranging
from 1.8 to 3.0 mmol.m2.sec-1 (Table 3). A one-way analysis of variance was
performed, and indicated no significant difference in daily transpiration among
vegetation type. A second analysis of covariance that removed the seasonal
trend in the data, also indicated no significant difference in transpiration rates
among the species monitored.
Daily water losses per community type were estimated by first calculating
daily transpiration and multiplying by leaf area per community type. Daily
transpiration was calculated by multiplying a six hour day length times the leaf
transpiration rates. A six hour day was thought to reflect an average period of
daily metabolic activity through out the year and probably underestimates daily
transpiration in the summer and overestimates wintertime values. Estimates of
oO O M 10 C) an in .o .- .N C' C M M C.o M
oocoo6ooo6ooooooocoooo
0 0 0 0 0 0 0 0 0 0 0 0 0
ddddddddddddddddddddddd
M O M 0 M ) 00 M CN 00 a0 M ma
cc CCcCcdddddddddddddddddddddddc
~< ~rNNN cNN oococo
c- . o 6 6 6 6 6 6 6 OO
(N- C\ N 00 00 (O ON ir) M 00 Nt N (O (O 4 4 N clO C) C4
--o
"t \D M M CD0 (O M m (N 0 Om
o 2
Z s
r-
z u
0)
Up.. a:
0 41----------
" Z
)- L]m ^ b G ^ h ^' ^ ii ^ ' ^ Lr U i ^ tn Lr t ^
-" M o e ^ ^ o o o o cpg o o o o g oo o c oog d
>- S-( O i t C- ^i C - C 1 ; -
o E
u
0
0a
[-\
leaf area per m2 of ground area were multiplied times the daily transpiration to
yield a community water loss.
Seasonal trends in community transpiration among the three types were
most evident in the swamp forest (Figure 3), and less evident in the marl prairie.
Transpiration was measured from all three vegetation types throughout the year,
indicating year round metabolic activity and no dormant period.
Analyses indicated a significant difference in water loss among the
vegetation types, due to differences in leaf area. A one-way analysis of variance
and one-way analysis of covariance (removing seasonal trend) both indicated a
significant difference in daily water loss among the three vegetation types.
Water loss rates from the marl prairie vegetation was the lowest, with an annual
mean of 0.016 cm/day. Average loss from the sawgrass marsh was 0.16 cm/day
and from the swamp forest 0.45 cm/day. Posteriori contrasts indicated that these
three types were significantly different.
In summary, estimates of water loss using the technique of scaling from
small scale transpiration to community values are dependent upon the
vegetation structure more than the flux rates. The water flux from the leaf
surfaces tend to vary seasonally, and do not exhibit differences among vegetation
type. Significant differences in water loss do appear to exist among community
types and appear to be related to the amount of leaf area present.
Measurements of community evapotranspiration
Other estimates of community evapotranspiration were made using
recorded tracings of water levels in shallow wells at two sites in Everglades
National Park. One well is designated P33 and is surrounded by wet prairie on
peat vegetation type. The other well is designated P37 and is situated in a wet
prairie on marl.
1
0.9
0.8
- 0.7 -
0.6 -
v^-
S0.5 -
S0.4
S0.3
0.2
0.1
0 I I I
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
MONTH
Figure 3. Mean daily transpiration rates from sawgrass, tree island and marl
prairie plant communities.
B Sawgrass
] Tree Island
SMarl Prairie
Evapotranspiration was estimated by comparing nighttime water losses
with daytime water losses, similar to the technique reported by Dolan et al,
(1984). The method is based upon the assumption that the only difference
between daytime and nighttime recession rates is due to evapotranspiration. The
technique is not useful on days with rain.
Rates of community evapotranspiration were different between the wet
prairie on peat and marl prairie sites. Mean daily water loss rates from the marl
prairie, calculated for each month of available data, ranged from a low of 0.10
cm/day during December to a high of 0.28 cm/day during June (Figure 4). This
translates to a mean annual total water loss of about 77 cm at the marl prairie
site. Daily rates were higher at the peat site. Anomalously high rates were
observed during June 1985, when the mean daily rate was 1.15 cm/day. This
occurred during a period of high temperatures, little rainfall and low ambient
water levels. Annual water loss at the peat wet prairie site was about 114 cm.
The mean difference between sites was 0.10 cm/day, (significant at P = 0.001),
indicating dramatically higher water loss rates at the peat site than at the marl
prairie site.
Mean daily evapotranspiration rates at the marl prairie (P37) followed a
smooth sinuous pattern over the time course of a year (Figure 4), whereas the
peat prairie site had dramatic anomalies during the early part of the summer.
The variation over time was summarized as percentages of annual water loss for
each month for use in the modeling section.
Summarization of Evapotranspiration Studies
The transpiration studies indicate that a difference in community
transpiration exists among the types studied, that comprise a hierarchy of water
00
o,6
00
'-U)
m m 00
00
00,
C13 0"
V)V
CZ0
00.0
00 00. \0
0\
ZO
o cz
00,
Pm CZl
(~~p/tn3)~kr U~lJdUf~~~~ppUA
005
>E
00X
.6,
IT C4 00IT C4 C
-4 6 C5 C
(X~/Lo)uopiisuzoeAQXpp uo
use in the landscape. The variation in community rates is due to dramatic
differences in leaf area among the vegetation types, more than losses per leaf
area. The marl prairie has low rates of transpiration, has a small total leaf area
per unit ground area, and is the most oligotrophic of the sites. The sawgrass sites
on peat have higher leaf area and community transpiration rates. The highest
transpirative water loss is from the tree island/swamp forest community, due to
higher rates and highest leaf areas. The swamp forest appears to be the most
eutrophic of the sites.
The studies of community evapotranspiration, indicate that differences
among types exist, the components of evaporation and transpiration vary among
the types. The transpiration and community evapotranspiration data both
indicate the fluctuation in rates over an annual cycle. Even though the rates
fluctuate seasonally, a mean daily rate will be used as the basis of comparison
among types and pathways of water loss. The P37 site is the only one of the sites
to have both transpiration and community estimates. The transpiration estimates
(0.02 cm/da) were about an order of magnitude lower than the community
estimates (0.2 cm/day), indicating that transpiration is not a large pathway of
water loss. The wet prairie on peat site (P33) had a higher daily rate (0.3
cm/day) than the marl prairie. The wet prairie rates were higher than the daily
transpiration rates in the sawgrass and lower than the transpiration rates at the
tree island site.
Evapotranspiration is one major pathway of water flow out of wetland
ecosystems. The other major outflow is via overland flow. The theoretical and
practical work studying this process will be reviewed in the next section.
Flow in Wetlands
Surface water flow in wetland ecosystems has been studied using
principles of applied hydraulics. The theoretical foundations for flow arise from
hydraulic principles, although there appears to be disagreement in the literature
as to the fundamental nature of flow regimes. A unique functional feature of
wetland systems is the periodic flooding and flow followed by periods of no
flow. This periodicity involves transitions among flow regimes from no flow to
laminar flow to turbulent flow. Most studies of wetland flow (Ree 1949, Petryk
and Bosmajian 1975, Lin and Shih 1979, Rosendahl and Probst 1980, Rosendahl
1981, Shih and Rahi 1982) assume a turbulent flow regime and utilize Manning's
formula (Manning 1890). Kadlec (1990) argues that laminar flow is dominant in
wetlands due to the low energy gradients and suggests a formulation such as the
one given in Equation (2) be used. Wetland flow is of the magnitude that
momentum terms in flow equations are usually ignored (Kadlec 1990, Hammer
and Kadlec 1992). The disagreement involving flow regimes can be partially
resolved as one of parameter values, as shown in Equation (2). Equation (2) is a
generalization of the Manning equation relating flow velocity as a function of
hydraulic radius and hydraulic slope. The critical parameters are a and 0 and
are assigned values of a = 0.5 and 3 = 0.67 in the Manning formula. Kadlec
(1990) reports values of a range from 0.4 to 1.0 and 3 from 2.5 to 3.75 for laminar
flow in wetlands.
(2) V = K S Rhe
V = velocity
K = general resistance coefficient
S = Hydraulic slope (difference in elevation potential)
Rh = Hydraulic radius (Cross-sectional area/wetted perimeter: Note
for very wide channels, Rh is approximated by water depth)
aX = slope exponent
3 = hydraulic radius exponent
In Manning's formula, K is represented by 1/n in SI units (1.49/n in
English units), and n is referred to as a roughness coefficient or as Manning's n.
Assuming turbulent regimes, most prior work on wetland flow has
attempted to develop better estimates of Manning's n, and particularly how this
coefficient varies with factors of water depth and vegetation density. The earliest
work relates n with water depth (Ree 1949) and the product of velocity and water
depth (Palmer 1945). Ree worked with flow in short grasses and found an
increase in n with depths up to the height of the grass, then a decrease, similar to
Palmer. Petryk and Bosmajian (1975) laid much conceptual framework, and
related n as a function of vegetation characteristics (primarily vegetation
density), boundary roughness and hydraulic radius. Petryk and Bosmajian
(1975) thought that vegetation resistance was much greater than boundary
roughness, and hence related n to vegetation density and inversely to hydraulic
radius; for n to remain constant vegetation density had to decrease if depth (Rh)
increased. Shih and Rahi (1982) using principles of Petryk and Bosmajian (1975),
developed estimates of n from 0.16 to 0.55 for marshes in the Kissimmee basin,
where n varied with seasonal changes in vegetation density.
Studies of flow in the marshes of the Everglades date back to the 1940s,
with the earliest work (Parker et al. 1944) measuring decreased flow in canals
resulting from infestations of water hyacinths. The U.S. Army Corps of
Engineers developed estimates of Manning's n in design memoranda for the
C&SF project, that averaged 0.035 and ranged inversely with water depth (U.S.
Army 1954). Leach et al. (1971) investigated data from a series of years and
found maximum flow rates of 1600 ft/day (0.6 cm/sec), which translate to a
cumulative annual distance of 50 miles (81 km). Rosendahl and Probst (1980)
and Rosendahl and Rose (1981) measured flow rates and resistance coefficients in
sawgrass and open marshes in Everglades National Park and reported greater
flow rates; from 0 to 0.022 ft/sec ( 0.67 cm/s) in dense sawgrass strands and
from 0 to 0.034 ft/sec (1.0 cm/s) in open marshes. Rosendahl (1981) calculated a
range of values of Manning's n between 0.4 and 2.4, with a mean of 0.99 and
found little correlation with depth.
Most of the models of water flow in the Everglades marshes use
Manning's equation and with varying reports as to the sensitivity of model
output to variations in the roughness coefficient. Lin and Shih (1979) used values
of n between 0.4 and 1.2, with an inverse relationship between n and depth. Lin
and Shih (1979) found that seasonal variation in n was necessary to achieve
model calibration, with lower values in the dry season and higher values in the
wet season. MacVicar et al. (1983) relate flow coefficients as a function of land
cover type, with values ranging from 0.1 to 2. Perkins and MacVicar (In press)
did a sensitivity analysis using very low value of n (0.05) and very high (2.0) and
found more effect on flow volume than stage, recommending development of
better coefficients with vegetation type. Walters et al. (1992) developed a
coefficient of flow equivalent to K in equation 2, of 2, which translates to a
Manning's n of 0.75.
A review of previous studies involving flow in wetland systems can be
summarized as follows. Although there is still uncertainty regarding the nature
of the flow regime, a generalized form of Manning's equation can be used. The
equation equates flow velocity as a function of water depth, hydraulic slope, and
a resistance coefficient, such as Manning's n. The flow coefficient can be
estimated from vegetation density, defined as the total cross-sectional area of
vegetation per unit length of flow (Petryk and Bosmajian 1975)
In the preceding sections, the processes influencing the measures of
evapotranspiration in south Florida and flow relationships in wetlands were
discussed in context of scale. The way in which these processes are incorporated
into a landscape model is primarily a scaling issue. The details of how this
scaling process was done is described below in the section on modeling
methodology. A brief review of some general concepts and approaches to
scaling in ecological models is included as a final piece in this background
section.
Ecological Models and Scale
Most descriptive landscape or ecosystem models have explicit domains in
space and time. Temporal domains are defined at the small end by the time step
and at the large end by the time horizon. Similarly, spatial grain is defined by
the size of grid cell and extent by the number of grid cells. Bounding the model
along spatial and temporal dimensions, defines what is inside and outside the
model domain in terms of scale. The empirical rule of thumb is that models
cover no more than two to three orders of magnitude in either space or time. The
rule is probably not related to technical constraints such as computer processing
power (Costanza and Maxwell 1992). The scale restriction may be related to
practical factors, such as debugging problems, validation criteria (Clark et al.
1979 ), or understanding the model (Costanza and Sklar 1985).
By using an ecological model with a fixed domain, decisions must be
made about what to do about processes that occur at different scale ranges. The
common approach in model construction is to treat processes that occur at slower
speeds and over broader ranges as constants. For example, if a model is
constructed to examine seasonal dynamics in sea level, then the global processes
that created a dramatic sea level rise between 5 and 10 thousand years ago, are
assumed to not change much over the course of a few years and therefore are
treated as constant. Faster processes are generally treated as noise or random
fluctuations within the system and can be averaged. Using the same example,
tidal influences on sea level occur over a short term and therefore can be treated
as noise over time spans of a year. The short term (daily) influences are averaged
to study seasonal or annual dynamics.
Components both inside and outside model structures are dealt with by
processes of aggregation and disaggregation. The simplest form of aggregation
is linear scaling. Scaling is defined as the translation of units based upon a fixed
relationship or ratio among metrics used in measurement. For example,
temporal metrics of minutes, hours, and days have a fixed relationship, therefore
one can trivially determine that one day is equal to 1440 minutes. Broadly, the
issue of aggregation has been dealt with either in terms of applying standard
statistical methods to derive "best" estimates and minimize error (O'Neill et al.
1986, Gardener et al. 1982), or by assuming linear aggregations among complex
variable sets (Iwasa et al. 1986, 1989). Basically, aggregation works if the
assumptions and rules used remain valid over the scales of translation.
Incorporation of evapotranspiration, flow and vegetation dynamics into a
spatially and temporally explicit hydrologic model of the Everglades involves
"fitting" these into a model with explicit bounds in time and space. The next
section describes the structure and development of the model used to investigate
hydrodynamics in the system.
Model Description and Development
The framework for the model was developed during a series of
workshops held between 1989 and 1990. The initial objective of the model was to
improve communication among scientists, engineers and practitioners in order to
discuss issues related to Everglades restoration. The model resisted a series of
attempts at invalidation (Walters et al. 1992, Richardson et al. 1990) and hence
has become a credible tool for examination of movement of water across the
landscape. The model framework depicts the time dynamics of the hydrology
within approximately 800 4 x 4 km grid cells (Figure 5) that cover the historic
Everglades ecosystem and surrounding areas. The model is bounded by Lake
Okeechobee to the north, the Atlantic coastal ridge to the east, Florida Bay to the
south and the Big Cypress National Preserve to the West (Figure 5). The basic
framework reported by Walters et al. (1992) was modified to include coupling of
vegetation and hydrology. The hydrology and vegetation components of the
model are described in the next two sections, followed by the development of
subroutines of evapotranspiration and flow that link the hydrologic and
vegetation components.
Hydrologic Components
Water depths within a cell change over time due to inflow associated with
rainfall, losses via evapotranspiration, net flux associated with overland runoff
from adjacent cells and net flux of water into or out of canals. The model is
driven by historic rainfall data, covering the period from January 1960 through
December 1988. Input data are of total monthly rainfall, averaged over the entire
basin area. Even though spatial gradients exist in the system (MacVicar and Lin
1984) equal amounts are added to each cell at the beginning of each simulated
month. Annual rainfall during the model time period ranges from 100 to 150
Figure 5. Model grid used to depict hydrologic and vegetation dynamics in
Everglades ecosystem.
cm/yr. The input to a cell is actually as net rainfall; that is, the recorded rain
total minus evapotranspiration.
Water is moved to adjacent cells based upon the Manning formula, where
velocity is a function of hydraulic slope, water depth and a resistance coefficient
(Equation 2). The alpha and beta values from equation 2 are 0.5 and 0.67.
Hydraulic slope is determined by difference between adjacent cells in the sum of
ground elevation and water depth. Levees in the model stop flow movement.
Water management is incorporated in the model by a series of water
management schedules within each conservation area and rules of water
movement around the schedules. If water levels at index cells are above the
monthly schedule, then water is removed from certain output cells. If water
levels are below scheduled levels, then water is retained. Water is diverted to
coastal areas by removal from key index cells to simulate removal via canals.
Target diversion rates are a function of maximum diversion (flow allowed in a
canal) and water depth.
Only surface water movement is calculated; no losses to groundwater are
included. In the peat areas of the system, with high infiltration resistance, this is
not considered to be a major source of error. In the transitional, sandy and marl
areas, movement from surface to groundwater is considerable, resulting in
overestimates of water depths.
The modeled area exchanges water with the surrounding areas. Water is
input from Lake Okeechobee, based upon decision rules and schedules within
the Lake. The Big Cypress regions to the west receives the same rainfall inputs as
areas over the Everglades proper. Water exchange with this area is only
constrained by structures in the model. No exchange (other than management
diversion) is made with the east, even though under historic conditions water
moved through the coastal ridge through a number of rivers, sloughs and finger
glades. Boundary conditions at Florida Bay vary seasonally from a low value in
winter to a high value in late fall, to reflect the annual dynamics of sea level.
Information on water depth and flow over time can be output for each
grid cell. Target cells that correspond to the locations of sites P33, P35, P37 and
P38 (Figure 6) were used as key indicators of model results. Cumulative annual
flow amounts were determined for three flow sections. One is for the set of cells
that coincide with the Tamiami Trail Flow section (Figure 6). The other sections
are at the boundary cells at the mouth of Shark River Slough, and at the mouth of
Taylor Slough.
The preceding paragraphs summarized the hydrologic variables and
interactions within the model framework. The next section deals with the
structure of the vegetation component of the model.
Vegetation Components
A total of 26 cover types were created to capture the variety of vegetation
patterns in the landscape of the model area (Table 4). The types were based upon
combinations of plant associations, as a grid cell of four kilometers generally
covers a non-homogeneous combination of plant communities. Some of the
vegetation types, such as sawgrass, can be the only type within some grid cells.
Other types, such as sloughs and tree islands are always smaller than the grid
cells. To determine the percent cover of vegetation communities within the
dominant landscape types, and how robust the measures of percent cover were
with a change in scale, the following exercise was done.
Vegetation cover was measured in a series of subsamples from the two
classified sixteen kilometer SPOT satellite scenes used in the vegetation map
section above. A total of thirty two samples were made (sixteen from each
scene), for window sizes of 500, 1000, 2000 and 4000 m. Percent cover was
Figure 6. Location of sample rain and stage gauges, flow sections and pan
evaporation sites within the Everglades region.
* RAIN GAUGE
* WATER LEVEL GAUGE
< FLOW SECTION
O PAN EVAPORATION
TEMPERATURE
61
Table 4. Description of vegetation categories (landscape units) used in
Everglades model.
Map LANDSCAPE UNIT
No. Description of Components
1 sawgrass, slough, tree island
2 sawgrass/slough
3 slough, with periphyton
4 slough, no periphyton
5 sawgrass marsh
6 sawgrass with woody invasion
7 sawgrass,cattail
8 wet prairie, on peat, tree islands
9 wet prairie, on peat, native woody plants
10 wet prairie, on peat, exotics
11 wet prairie, on marl and tree islands
12 wet prairie, on marl, muhly grass
13 wet prairie, on marl,native woody plants
14 wet prairie, on marl, exotics
15 Pine, hammock
16 Pine, hardwood
17 Pine, prairie
18 Upland hardwood scrub
19 tall mangrove forest
20 short mangrove forest
21 mangrove prairie
22 dwarf cypress
23 cypress dome
24 cypress hardwood
25 agriculture
26 no vegetation
calculated from each sample for each of the three major vegetation types;
sawgrass, tree island and wet prairie. The percent cover of sawgrass, wet prairie
and tree island communities in the southern Everglades did not vary over the
sampled window sizes. For windows sizes from 500 through 4000 m, the cover
of each of the three types was relatively constant (Figure 7). The mean cover was
about 66% for sawgrass, 21% for wet prairie and 13% for tree island. A two-way
analysis of variance indicated no significant difference in cover among the
window sizes. Although variances tended to decrease with window size, the
differences were not enough to violate assumptions of homoscedasticity. These
results indicate that using the percent cover of plant communities to describe the
landscape units is very robust to changes in the cell size of the landscape units.
A map editor is established in the model to allow for creating the initial or
starting vegetation array within the active cells. The map editor creates an array
that consists of a vegetation type code for each cell that can be addressed and
updated during the simulation. The rules in the model for changing the
vegetation types are the subject of the next section.
Changes in Landscape Vegetation Types
The vegetation change module of the model is set up to switch among
vegetation types, based upon rules relating to hydroperiod, nutrient
concentration and fire. At the end of each year of simulation, the hydroperiod
(number of months per year that a site is wet) and soil nutrient concentration are
calculated for each grid cell. The annual hydroperiod value is used to update a
running average hydroperiod for each cell. Fire is a stochastic event, with
probability related inversely to the annual hydroperiod value. If a random
number is less than the assigned probability for the hydroperiod value, then a
fire event is said to have burned the cell. The average hydroperiod, soil nutrient
500 1000 2000 4000
Window Size (m)
WET PRAIRIE
500
1000 2000
Window Size (m)
4000
500 1000 2000 4000
Window Size (m)
Percent cover of sawgrass, wet prairie, and tree island in various
window sizes.
80
60
40
20
Figure 7.
concentration and a fire event are all used as critical values to make changes
among vegetation types. If average hydroperiods are wetter or drier than a
certain threshold value, or if the soil nutrient concentration exceeds a critical
value, or if a fire occurs in the cell, then the vegetation cover type for that cell
changes. The threshold values of hydroperiods and target transition types are
user defined.
The transitions among vegetation types were determined using
information derived from a combination of sources, namely, literature and field
experience. The transitions involve the loss or gain of certain community types
within a cover type. The primary cover in the peat portions of the Everglades
system is a mosaic of sawgrass, slough and tree islands, designated as type one.
The hydroperiod ranges from 9 to 11 months in this type. If hydroperiods exceed
11 months, then the tree island and sawgrass types disappear, leaving only
slough (Craighead 1971, Worth 1987). If hydroperiods are less than 9 months,
then woody plants invade the sawgrass and slough (Craighead 1971, Gunderson
and Loftus In Press) resulting in a change to type 9. The other dominant peat
landscape type was a monospecific stand of sawgrass in the area now known as
the EAA (Davis 1943). If the sawgrass burns or dries out, woody plants invade,
(Wade 1980) resulting in type 6. The vegetation dynamics associated with a
change in nutrient status include a loss of periphyton in a slough system (Swift
1984), and a transition from sawgrass to cattail (Davis 1989).
The preceding paragraphs describe the "landscape" vegetation units used
in the model. These units are comprised of combinations of identifiable plant
communities. The transitions among landscape types involve the addition or
replacement of plant communities within each landscape unit. Since the plant
communities provide the "building blocks" for each landscape unit, they will be
used as the basis of relating flow and evapotranspiration.
Development of Flow Coefficients for Landscape Units
Two steps were involved in the determination of flow coefficients by
landscape unit. The first step related estimates of vegetation density for each of
the dominant plant communities to flow coefficients. The second step used the
percent cover relationships within a landscape type to develop a spatially
weighted flow coefficient.
To develop relationships between vegetation type and flow regimes,
estimates of vegetation density were derived. Vegetation densities were then
translated to Manning's flow coefficients using relationships developed by
Petryk and Bosmajian (1975) and Shih and Rahi (1982). Vegetation density was
determined for four vegetation types: sawgrass marsh, wet prairie over peat,
marl prairie and tree island. The literature was surveyed for measured values of
stem density for the gramineous vegetation types (marsh and prairie) and for
values of basal area for the forest type (tree island). Average stem densities in
the graminoid vegetation types were multiplied by the average stem size to yield
an average cross sectional area per length of flow. Total basal area was divided
by the stem density to yield an average tree size, then average cross sectional
area was determined per unit of flow length. The values of vegetation density
were then correlated to a Manning's n value based upon data compiled by Petryk
and Bosmajian (1975).
Stem density varied from a low value in the wet prairie (0.05 /m2)
to the highest (42/m2) in the marl prairie (Table 5). Even though the marl prairie
had the highest stem density, the sawgrass had the highest vegetation density.
The cross-sectional area of the plants comprising the marl prairie were much
smaller than sawgrass. The vegetation density reflects the total cross-sectional
area (m2) per ground area in the direction of flow (m3), and is in units of m-1.
The vegetation densities were highest in the sawgrass areas (0.15 m-1),
Table 5.
Vegetation density and related flow coefficients as a function of
depth for sawgrass, tree island, wet and marl prairie vegetation
types.
Vegetation Type Sawgrass Wet Prairie Tree Island Marl Prairie
Stem Density
(#/m2) 28 *(1) 3 *(2) 0.6 *(3) 42 *(4)
(#/ft2) 8.5 0.9 0.2 12.8
Stem Size (ft) 0.08 0.03 0.30 0.02
Stem Area
(ft2/ft2) 0.71 0.03 0.05 0.27
Depth
Vegetation (ft)
Density 0.5 1.42 0.06 0.11 0.53
(ft2/ft3) 1 0.71 0.03 0.05 0.27
1.5 0.47 0.02 0.04 0.18
2 0.36 0.01 0.03 0.13
2.5 0.28 0.01 0.02 0.11
3 0.24 0.01 0.02 0.09
3.5 0.20 0.01 0.02 0.08
Mannings *(5) 0.5 1.12 0.22 0.31 0.68
n* 1 1.26 0.25 0.35 0.77
1.5 1.35 0.27 0.37 0.82
2 1.41 0.28 0.39 0.87
2.5 1.47 0.29 0.41 0.90
3 1.51 0.30 0.42 0.93
3.5 1.55 0.31 0.43 0.95
Model Flow *(6) 0.5 1.3 6.7 4.8 2.2
Coefficient 1 1.2 5.9 4.3 1.9
K 1.5 1.1 5.5 4.0 1.8
2 1.1 5.3 3.8 1.7
2.5 1.0 5.1 3.7 1.7
3 1.0 4.9 3.5 1.6
___3.5 1.0 4.8 3.5 1.6
Herndon et al. 1991
Goodrick 1984
Gunderson 1982
Olmsted et al. 1980
Petryk and Bosmajian, 1975
Walters et al., 1992
n = (depth)^0.67*(veg density)^0.5
K=1.49/n
REFERENCES
intermediate in the marl prairie (0.06 m-1) and tree islands (0.05 m-1), and lowest
in the wet prairie (0.01 m-1). The estimated flow coefficients for sawgrass was
about twice that of tree island and wet prairie, and substantially greater than in
the wet prairie.
A spatially weighted average flow coefficient was determined for each
landscape unit. First, a mean flow coefficient was calculated over a range of
depths for each plant community type. The percent cover of each plant
community in a landscape unit was used as the weighting factor. For example,
in landscape unit 1, sawgrass covers on the average 66% of a cell, wet prairie
covers 22%, and tree islands cover 12%. The depth averaged flow coefficient for
these three types are 1.09, 5.44 and 3.93, respectively. The flow coefficient for this
landscape unit is (0.66*1.09+.22*5.44+.12*3.93) = 2.4. The spatially weighted
coefficients are given in Table 6. Even though the values vary among plant
community types, the spatially weighted averages are similar among the
landscape units
Development of Evapotranspiration Coefficients for Landscape Units
A spatially weighted average evapotranspiration rate was also calculated
for each landscape unit. The plant community rates were derived from average
daily rates calculated from the evapotranspiration data. The values ranged from
a low of 0.22 cm/day in the marl prairie to 0.4 cm/day in the swamp forest
(Table 7). The values for cattail and melaleuca were estimated from other
sources (Koch, unpub. data; Woodall 1980). The calculation of annual totals is
extremely sensitive to the daily rates, a difference of 0.1 cm in the daily rates
results in an annual difference of 36 cm. The spatially weighted mean annual
evapotranspiration values ranged from 95 cm (marl prairie, type 11) to 159 cm for
the unit of exotic trees on peat (Table 7). Annual totals for each landscape type
Table 6. Spatially weighted flow coefficients for each landscape unit used in
the Everglades Model.
Spatially
Map LANDSCAPE UNIT Spatial % Weighted
No. Description of Components Components Average K
1 sawgrass, slough, tree island 66-22-12 2.4
2 sawgrass/slough 78-22 2.0
3 slough, with periphyton 100 5.4
4 slough, no periphyton 100 5.4
5 sawgrass marsh 100 1.1
6 sawgrass with woody invasion 88-12 1.4
7 sawgrass,cattail 50-50 1.3
8 wet prairie, on peat, tree islands 80-20 5.1
9 wet prairie, on peat, native woody plants 60-40 4.8
10 wet prairie, on peat, exotics 20-80 4.2
11 wet prairie, on marl and tree islands 80-20 2.2
12 wet prairie, on marl, muhly grass 80-20 2.2
13 wet prairie, on marl,native woody plants 60-40 2.6
14 wet prairie, on marl, exotics 40-60 2.6
15 Pine, hammock 100 1.8
16 Pine, hardwood 100 1.8
17 Pine, prairie 100 1.8
18 Upland hardwood scrub 100 1.8
19 tall mangrove forest 100 10.0
20 short mangrove forest 100 10.0
21 mangrove prairie 100 10.0
22 dwarf cypress 100 2.0
23 cypress dome 100 2.0
24 cypress hardwood 100 2.0
25 agriculture 100 2.0
26 no vegetation 100 2.0
Table 7. Average daily and annual evapotranspiration for plant
communities used to develop evapotranspiration coefficients for
Everglades model.
Plant Community Daily ET Annual ET Annual ET
(cm) (cm) (in)
Sawgrass 0.33 120 47
Wet Prairie/Slough 0.26 95 37
Tree island 0.42 153 60
Marl Prairie 0.22 80 32
Exotic-Melaleuca 0.48 175 69
% Spatially Relative
Map LANDSCAPE UNIT Spatial Weighted Annual Rate
No. Description of Components Components Annual ET X/114 cm
(cm)
1 sawgrass, slough, tree island 66-22-12 119 1.04
2 sawgrass/slough 78-22 115 1.01
3 slough, with periphyton 100 95 0.83
4 slough, no periphyton 100 95 0.83
5 sawgrass marsh 100 120 1.06
6 sawgrass with woody invasion 88-12 123 1.08
7 sawgrass,cattail 50-50 145 1.27
8 wet prairie, on peat, tree islands 80-20 107 0.93
9 wet prairie, on peat, native woody plants 60-40 118 1.04
10 wet prairie, on peat, exotics 20-80 159 1.40
11 wet prairie, on marl and tree islands 80-20 95 0.83
12 wet prairie, on marl, muhly grass 80-20 95 0.83
13 wet prairie, on marl,native woody plants 60-40 110 0.96
14 wet prairie, on marl, exotics 40-60 118 1.04
15 Pine, hammock 100 114 1.00
16 Pine, hardwood 100 114 1.00
17 Pine, prairie 100 114 1.00
18 Upland hardwood scrub 100 114 1.00
19 tall mangrove forest 100 114 1.00
20 short mangrove forest 100 114 1.00
21 mangrove prairie 100 114 1.00
22 dwarf cypress 100 114 1.00
23 cypress dome 100 114 1.00
24 cypress hardwood 100 114 1.00
25 agriculture 100 114 1.00
26 no vegetation 100 114 1.00
were expressed as a ratio of 114 cm/yr, the fixed coefficient for the model. As
with the flow values, there appears to be some spatial convergence of averages.
That is, a relatively constant percentage of plant community types within a
landscape unit combined with a difference in rates among plant communities,
appears to result in a global average for a grid cell.
The preceding section of this chapter described the hydrologic
components, vegetation components and linkages of flow and
evapotranspiration in the model. Water depths within each grid cell change
monthly as a function of historic rainfall, net flow, and evapotranspiration. At
the scale of the model, 26 landscape types comprised of plant communities, are
used to describe vegetation patterns. Flow and evapotranspiration rates are
linked to the "landscape" type. This linkage was done through two steps: first to
determine coefficients for the dominant vegetation communities, then to develop
a spatially weighted average coefficient for each landscape unit based upon the
percent cover of vegetation types within the landscape unit. The transitions
among the landscape units are a function of cumulative water depths
(hydroperiod), fire and nutrient concentration. The next section of the chapter
presents the results of sensitivity analyses and testing the hypotheses.
Results
The results section has three parts. The first part assesses the sensitivity of
key parameters to flow calculations and compares model output with historic
data. The second part presents attempts to invalidate the linkages between
vegetation and hydrology. The third portion of this section reviews the tests of
the upstream area hypothesis.
Sensitivity Analysis- Flow and Evapotranspiration
Tests were done to explore the sensitivity of the model output (primarily
flow) to uncertainties in parameters associated with flow and evapotranspiration.
The sensitivity analyses of flow coefficients were done by doubling and halving
all coefficients calculated in the above paragraphs, running the model for a full
28 year scenario under natural conditions (no water control structures in the
system). The sensitivity to evapotranspiration was tested by running a full 28
year natural scenario with the annual evapotranspiration rate set at 89, 102, 107,
and 114 cm (35, 40, 42 and 45 inches).
Doubling and halving the flow coefficients did not appear to have an
appreciable affect on flow through the Tamiami flow section (Figure 8). The
largest deviations among the three sets of coefficients occurred during wet years
(model simulation years 1967, 1969, 1970; Figure 8). During these periods, the
flow differed by about 500 x 106 m3/yr between either of the runs with adjusted
coefficients and the unadjusted coefficients. This difference during wet years,
between the adjusted coefficient flows and the unadjusted flow, was about 12%
of the unadjusted flow. Differences during dry years were much less. There was
no evidence that changes in vegetation landscape types during any of these runs
altered or confounded the flow relationship among the adjusted coefficients. The
flow results from the unadjusted run were always intermediate between the
higher flows calculated by doubling the flow coefficients, and the lower flows
associated with halving the flow coefficients.
The results of varying base evapotranspiration indicate counter-intuitive
effects on rates of annual flow through the Tamiami flowsection. The results are
unexpected because there is not a constant relationship between the
evapotranspiration rate and amount of flow. The lower evapotranspiration rates
should generate higher stages and higher flow. The simulated flow data start out
4000-
3500
t3000
2500
,2000
0
S1500
:1000
500
1960
1960
1965 1970 1975 1980 1985
YEAR
Figure 8. Effects of doubling and halving flow coefficients on simulated flow
through Tamiami flow section.
--- FLOW COEFFICIENT* 1/2
-0- FLOW COEFFICIENT 2
-A- NATURAL
with this relationship, but the relationship changes through the time course of
the model run (Figure 9). These discrepancies are the result of the linkages
between hydrologic conditions and the vegetation landscape conditions. If the
system gets too dry, such as the scenario of higher base annual
evapotranspiration (114 cm/yr), then woody species invade the landscape,
increasing the evapotranspirative loss and lowering flow rates. The system
appears to entrain flows at lower levels of evapotranspiration. The lower
evapotranspiration maintains "wetter" landscape types that result in higher flow
rates.
Even with these interesting and counter-intuitive effects of vegetation-
hydrology-evapotranspiration linkages on flow, the model is apparently more
sensitive to changes in evapotranspiration than to variations in flow coefficients.
Changing the base transpiration rate only 25 cm/yr, results in a variation of
annual flow on the order of 500 x 106 m3/yr (Figure 9). A similar effect was
achieved by doubling and halving the flow coefficients. After establishing the
sensitivity of changing the base settings on flow regimes, model output using the
base settings will be compared to historic data.
Agreement with Historic Data
The model output, both stage and flow, indicates periods of agreement
and divergence with measured data. The actual stage at P33 and P35 and flow
through Tamiami tend to be lower than the model output during the period from
1961 through 1965 (Figures 10 and 11). This is probably due to the management
policy in effect during this period, when little or no flow was delivered to the
park (Wagner and Rosendahl 1985, Gunderson 1989). During other years, the
actual and modeled stage data tend to qualitatively agree. Since there was no
groundwater component to the model, agreement was only possible with surface
Figure 9. Effects of varying base evapotranspiration rates on simulated flow
through Tamiami flow section.
4000
3500-
S3000-
0 2500
I 2000
S1500
1000
500
1960 1965 1970 1975
YEAR
- 89 ---- 102 107 -- 114
Annual Base Evapotranspiration (cm)
1980
1985
MANAGED ------- ACTUAL
Time series of simulated (solid lines) and actual (dashed lines)
stages at gauge P33 from 1960 to 1988.
100
80
60
40
L 20
0
H-20
-40
-60
-80 E
1960 62 64 66 68 70 72 74 76 78 80 82 84 86
Year
Figure 10.
- MANAGED --..... ACTUAL
Time series of simulated (solid lines) and actual (dashed lines)
stages at gauge P35 from 1960 to 1988.
80
60
S40
u
S20
0
H
5-20
-40
-60 II
1960 62 64 66 68 70 72 74 76 78 80 82 84 86
Year
Figure 11.
water conditions. The modeled flow through Tamiami trail also agreed fairly
well with measured flow in periods other than the early 1960s and early 1970s.
The vegetation patterns at the end of the simulation period of the natural
scenario appear to agree with early descriptions of vegetation in the Everglades.
The initial array of landscape units in the model grid consisted of a sawgrass
plain south of Lake Okeechobee, the tree island/sawgrass mosaic in the central
core of the system, and marl prairies units in the south. (Figure 12). At the end of
the run, the sawgrass plain and marl prairie units had persisted (Figure 13). The
tree island/sawgrass mosaic had changed to a sawgrass/slough type in the area
of the persistent pool described in the paragraph above. Davis (1943) and Jones
(1948) observed and mapped similar patterns; the tree island/sawgrass mosaic
was only mapped in the northeast and southwest portions of the central
Everglades and a tree-less marsh was in the topographically lower southeast
region. The tree-less sawgrass/slough vegetation type is also captured in the
native Americans description of the system as "Pa-hay-okee", which loosely
translates to a grassy lake (Douglas 1947).
The output from the model agrees fairly well with historic hydrologic and
vegetation information. Even with uncertainties and sensitivities to
understanding flow and evapotranspiration processes, the model captures key
aspects of hydrologic and vegetation dynamics of the system. The next section
puts the model at risk, and attempts to determine the bounds of the relationships
between the hydrology and vegetation.
Linkages between hydrology and vegetation
Since the model output agreed fairly well with historic data sets, a test was
developed to determine the limits of the influence of vegetation dynamics on the
relationship between rainfall and runoff. The test consisted of a series of model
5 5 5 5 5 5 5 5 5 5 5 5 5 1
23 923
9 24 24
24 24 23
24 24 23
23 24 23
24 24 24
24 23 23
232322
11 17 17 23 23 23 23
11 11 22 17 11 17 22
11 11 11 17 17 17 17
23 11 11 17 17 17 17
11 11 9 22 17 17 17
11 11 24 24 23 11 17
11 11 24 24 23 17 11
. 21 11 23 23 11 23
. 20 11 11 11 23 23
20 21 20
S21 20
S. 21
S. 21
. ,
232323
232323
23 24 23
23 24 23
17 17 22 22 22
2311222222
2323222222
21 11 11 23 23 23 22 22 22
20 21 11 23 23 17 22 22 11
21 20 21 11 11 17 22 11 11
19 20 20 21 17 17 11 11 11
21 21 19 21 21 11 11 11 11
19 21 21 19 21 21 11 11 11
19 19 21 19 19 21 11 11 13
. 21 21 19 21 19 20 11 1
. 19 19 19 21 21 20 1
S. 19 19 19 21 19 20 1
.19 21 21 19 20 1
222222
222222
222222
222222
11 11 11
11 11 11
11 11 13
13 13 1
13 13 1
1 1 1
1 1 1
1 1 1
1 122
1 1 17
. . . 19 19 19 19 19 20 1 17 22
. . . .. 19 19 19 19 19 19 19 21 11 11
. . . 19 21 19 19 19 19 19 19 21 21
..... 555
. 555
. . 5 5 5
. . . 5 5 5
. . . 5 5
. . . 5 5
. . . 5 5
5
5
5
23 22 23 23 22 22 22 5
. . . 5
. . . 5
23 232322 11 22 5 5
22232322 11 222223
2323232222222222
2323232222222222
2222222222222222
17 17 22 22 22 22 22 22
17 17 17 22 22 22 22 22
17 17 22 22 22 22 22 22
17 17 17 22 22 22 22 22
. . . 19 21 19 19 19 19 19 19 19 21 21 11
. . . 19 21 21 19 19 19 21 19 19 19 19 20 2
. . . .. 21 21 21 21 19 19 19 19 19 19 19
. . . . 21 21 21 19 19 . .191
. . . . . 19 19 . . .
5 5
5 5
5 5
5 5
5 5
5 5
5 5
5 5
5 5
5 5
5 5
1 1
1 1
1 1
1 1
1 1
1 1
1 1
1 1
1 1
1 1
1 1
1 1
2 1
11 1
1 1
1 1
1 1
1 1
1 1
1 11 1
11 17 1
17 15 1
17 18 1
13 13 1
22222
1322
Figure 12. Map codes for initial landscape vegetation types in the Everglades
model. Dominant code in the Everglades proper is type 1. Cover
types developed from Davis (1943). Explanation of codes is found
in Table 4.
5 5 5 5 5 5 5
5555555
5555555
5555555
5 5 5 5 5 5 5
5 5 5 5 5 5 5
5 5 5 5 5 5 1
5 5 5 5 5 5 1
5 5 5 5 5 5 1
5555555
5555555
5555555
5555551
5 5 5 5 5 5 1
5555551
5555551
5555551
1 1 15555551 1
1 1 1 1 1 1 1
1 1 1 1 1 1 1
1 1 1 1 1 1 1
1 1 1 1 1 1 1
1 1 1 1 11 1
1 1 1 11 1 1
1 1 1 11 11
1 1 1 1 1 11
1 1 1 1 1 1 1
1 1 1 1 1 1
1 1 1 1 1 1 1
1 1 1 1 1 11
1 1 1 1 1 1 1
1 1 1 1 1 1 1
1 13 11 13 8 11 1
1 11 11 13 13 11 11
1 11 13 13 13 11 11
1 12 1313 1311 11
8 11 13 11 8 11 11
9 20 20 20 20 21 21
20 19 19 19 20 20 20
.9 19 19 19 .
9 . . .
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
11
111
111
11
11
23 10
23 24
23 24
24 24
24 24 9
24 9 9
24 24 24
24 24 24
13 15 17
11 1313
111313
231313
11 13 9
11 13 9
11 13 24
.2113
.2013
202120
.2120
.21
..21
24 24 24 24 24
24 24 9 24 24
9 24 24 24 22
24 24 24 22 24
24 24 23 24 24
1513151313
15 15 15 15 15
17 15 15 15 15
13 15 15 15 15
24 24 14 1517
24 24 15 13 15
24 24 13 24 24
14 13 24 24 24
21 13 13 24 24
20 21 13 24 24
2120211313
19 20 20 21 15
21 21 19 21 21
19 21 21 19 21
19 19 21 1919
21 21 19 20
.191919
.19 19 19
S. 1920
S. 1919
.191919
S.19 21 19
S. 19 21 19
S. 19 21 21
S. 21 21
.. 21
Map codes of landscape vegetation types in grid cells of Everglades
model at end of 28 year simulation run. Note presence of type 3
codes in right -central portion of array. See Table 4 for explanation
of codes.
13 24 24
24 24 13
24 24 13
24 24 13
9 24 13
1313 13
15 14 13
15 15 13
15 13 13
15 15 13
151313
131313
24 13 13
24 13 13
151313
15 13 13
17 13 11
1313 8
21 8 8
21 8 8
19 20 3
202020
20 19 20
20 19 20
19 19 19
19 19 19
19 19 19
19 19 19
19 19 19
21 21 19
21 21 19
. 5 6 6 6
. 5 6 6 6
. 5 6 9 6
5 5 5 6
5 6 6
5 6 6
S. 5 55
. . 6 6
. 6 6
. 5 5
. . 5 5
13 13 13 6 5
1.5666
13 13 13 24 1
13 13 13 22 1
13 13 13 22 1
22 22 22 22 1
13 13 13 22 1
13 13 22 22 1
13 13 22 22 1
13 13 22 22 1
14 13 22 22 1
14 13 22 23 1
13 13 3 23 1
13 13 23 23 1
13 13 9 1 1
13 14 13 1 3
13 13 13 1 4
13 13 8 1 3
13 9 3 3 3
14 1 3 3 3
3 1 3 3 1
3 3 3 1 8
3 1 10 22 17
1 9 9 17 17
20 9 17 13 13
19 21 13 13 13
19 19 21 21 13
19 19 19 21 21
21 19 19 19 19
19 19 19 19 19
1.5666
.5696
.5556
..566
..566
..555
...66
...66
...55
...55
13 13 13 6 5
1313 6 6 1
13 13 13 24 1
13131322 1
131313 22 1
222222221
13131322 1
1313 22 221
13132222 1
13132222 1
14 13 22 22 1
14 13 22 23 1
1313 323 1
13132323 1
13 13 9 1 1
13 14 13 1 3
131313 1 4
1313 8 1 3
14 1 3 3 3
3 1 10 22 17
1 9 9 17 17
20 9 17 13 13
19 21 13 13 13
19 19 21 21 13
19 19 19 21 21
21 19 19 19 19
19 19 19 19 19
19 .
6 6 9
6 6 6
669
666
666
666
665
665
655
665
555
555
655
565
111313
1 1 1
1 1 19
1 1 17
1 1 1
1 1 1
13 913
13 920
120 20 1
1 1 19
1 1 1
3 3 3
3 3 3
3 3 3
3 3 3
3 3 3
3 3 3
1 1 1
11 13 13
17 13 13
17 17 9
18 18 17
13 14 13
13 22 9
13 9 13
13 9 20
20 20 19
19 19 19
19 19 .
66
66
96
96
66
65
69
55
56
52
52
5 5
1 1
1 1
1 1
1 1
1 3
3 3
3 3
3 3
3 3
3 3
3 3
33
3 3
3 3
3 3
3 3
113
113
1313
1313
8 13
1113
1313
13 14
1313
1313
2020
19 19
19 19
Figure 13.
696
666
966
666
666
6 6 6
666
565
551
6 6 1
661
691
561
5 6 1
561
1 1 9
1 3 1
3 3 3
3 3 3
3 3 4
3 3 3
3 3 3
3 3 3
3 3 3
3 3 3
3 3 3
3 3 3
3 3 3
3 3 3
1 1 3
13 13 1
3 11
13 8 3
13 14 3
13 13 13
13 9 14
10 9 13
9 13 13
8 13 13
9 11 11
20 21 21
20 20 20
333
333
333
333
333
333
333
13131
3811
1383
13 14 3
131313
13 914
10 913
91313
81313
9 11 11
20 21 21
202020
69
69
10 9
10 10
99
99
99
19
11
11
10 1
1 1
11
11
11
3 3
3 3
3 3
3 3
3 3
3 3
3 3
3 3
33
3 3
3 3
3 3
3 3
2 2
39
312
1311
1313
13.
8 .
runs. For each run, a set amount of rain was delivered each year of a 20 year
simulation. The rain varied seasonally, as modeled by composite sine waves to
emulate the natural annual pattern, but remove any interannual variation. Four
runs were made, ranging from a very dry year to a very wet year. The model
inputs were equivalent to annual totals of 105, 118, 142 and 176 cm (36, 46, 56 and
68 inches).
The relationships between rain and runoff appears constant over a wide
range of rainfall inputs, then dramatically shifts if the system becomes very wet.
The flow and stages at key stations all reached a seasonally oscillating
equilibrium with rainfall inputs less than 142 cm. At steady rainfall up to 142
cm, the landscape vegetation types remained constant, and hence, the
relationship between rainfall and runoff was linear. At an annual input of 176
cm, a dramatic shift occurred around year 6, when the landscape units shifted to
a treeless wet prairie. Without tree islands or sawgrass, the overland flow rates
were greatly increased, resulting in a new equilibrium of flow coming through
the Tamiami flow section (Figure 14). No such vegetation shifts occurred in the
mangrove areas and hence, no dramatic change in the flow relationships in either
the Shark Slough or Taylor Slough flow sections.
These results indicate that the vegetation-hydrology linkages can be
invalidated at an extreme. The key point is that the vegetation array is fairly
stable with constant levels of average rainfall input and the relationship between
rainfall-runoff is constant. However, if the system has a prolonged (at least five
year period) of surplus rainfall, then the vegetation structure is destroyed and a
different equilibrium in the rainfall-runoff relationship occurs. Since the
vegetation units are fairly stable with constant levels of average rainfall, the
relationship between rainfall and runoff may not be dramatically influenced by
the coupling of landscape vegetation dynamics with the hydrology.
5000
4500
4000
3500
3000
2500
2000
1500
1000
500
0
1 6 11
Simulated Year
Results of changing vegetation patterns on simulated flow through
three flowsections in the southern Everglades.
--- TAMIAMI
-0- SHARK SLOUGH
TAYLOR
SLOUGH
Figure 14.
The output of two models, one with and one without coupled vegetation-
hydrologic dynamics, was compared to see if addition of the linkages changed
the predicted hydropatterns. The output of the model with linkages qualitatively
agrees with the models without linkages (Walters et al. 1992, Perkins and
MacVicar In press) All models predict a persistent pool of water north and east
of Tamiami Trail in the area now known as the Pennsuco wetlands. Flow and
stage results are similar between the model with and without linkages (Figure
15). Key uncertainties to these results are in assumptions about the amount of
water that moved surficially and as groundwater into the coastal ridge, and
contributions from Lake Okeechobee.
The results in the preceding paragraphs lead to the conclusion that the
addition of the vegetation and hydrologic dynamics does not improve the
models' ability to predict stage and flow. Dramatic changes in rain and runoff
relationship occurs only after persistent wet conditions. Model output of flow
and stage is not different between models with and without vegetation linkages.
One of the reasons may be that the dominant influence on surficial hydropatterns
is the rainfall. Certainly this is partially true. Other reasons are related to
problems of scaling. The landscape units are composites of vegetation
communities. There is good evidence for dramatic differences in evapo-
transpiration rates and flow resistance coefficients among the vegetation
communities. However, when composite values are calculated at the landscape
levels, the differences are spatially homogenized and converge towards singular
coefficients. The net result of modeling at the landscape scale is that the addition
of more complexity at "smaller" scales does not dramatically improve accuracy of
the model to predict surficial hydropatterns. The failure to improve model
accuracy by the addition of complexity agrees with previous workers (Walters
1986, Clark et al. 1979, Costanza and Sklar 1985). In spite of these limitations, the
|
RSS
HIDE
