Distribution of Wading Birds Relative to Vegetation

WATERBIRDS
JOURNAL OF THE WATERBIRD SOCIETY
VOL. 25, NO. 3
2002
PAGES 265-391
Distribution of Wading Birds Relative to Vegetation and Water Depths
in the Northern Everglades of Florida, USA
G. THOMAS BANCROFT,1,3 DALE E. GAWLIK2 AND KEN RUTCHEY2
1
2
3
National Audubon Society, 115 Indian Mound Trail, Tavernier, FL 33070, USA
Everglades Division, South Florida Water Management District, 3301 Gun Club Road
West Palm Beach, FL 33406, USA
Present address: The Wilderness Society, 1615 M Street NW, Washington, DC 20036, USA
E-mail: [email protected]
Abstract.—The response of Great Blue Herons (Ardea herodias), Great Egrets (Casmerodius albus), Wood Storks
(Mycteria americana), and White Ibises (Eudocimus albus) to water level (index of depth) and vegetation in the northern Everglades of Florida was studied in two years, each with dissimilar water levels. A regression model was constructed for each species in an average year (1988) and a dry year (1989) to examine the relationship of bird
abundance to water depth and area of eight vegetation classes. The analyses showed that bird abundance is related
to both water level and the vegetation community, but water level generally had the greatest effect. Models showed
that in the average year (1988), there was a water level threshold, above which bird abundance was predicted to
decline. The level threshold varied among species and may have reflected species-specific foraging constraints.
However, in the dry year (1989), the relationship between bird abundance and water level was positive and linear,
indicating that few places had water deeper than the thresholds observed in the average year. Overall, the area of
slough vegetation appeared to have the second greatest effect on bird abundance. Generally, all models had low
coefficients of determination (R2 range 0.06-0.42) suggesting that factors other than water level and vegetation were
important, or birds were responding to variables in the model, but at different spatial scales than that which the
data were collected. Models for Great Blue Herons and Great Egrets had higher coefficients of determination than
models for Wood Storks and White Ibises. The more solitary feeding behavior of the herons and egrets resulted in
a more even distribution across the marsh than for storks and ibises, which were usually found in flocks. Our study
suggests that if restoration of the Everglades results in more natural hydrologic cycles, an increase in the amount of
slough habitat, and a decrease in the proportion of cattails, foraging conditions for wading birds may improve.
Received 8 March 2001, accepted 14 February 2002.
Key words.—Everglades, Florida, foraging distribution, Great Blue Heron, Great Egret, hydrologic cycles, White
Ibis, Wood Stork, vegetation
Waterbirds 25(3): 265-277, 2002
Hydrologic patterns affect many aspects
of wading bird ecology in a variety of ecosystems (Kahl 1964; Dimalexis and Pyrovetsi
1997; Gonzalez 1997; Arengo and Baldassarre 1999). In the Everglades, the dependency of wading birds on hydrologic patterns
is so strong that changes in the historic pattern of water level fluctuation are thought to
be a major factor causing the decline in numbers of breeding birds of several wading bird
species (Bancroft 1989; Ogden 1994).
Water depth is one hydrologic variable
that has been repeatedly shown to play an
important role in determining where and
when wading birds forage (Kushlan 1976;
Willard 1977; Recher and Recher 1980; Powell 1987; Gawlik, in press) and in determining reproductive success (Frederick and
Spalding 1994; Cézilly et al. 1995; Smith and
Collopy 1995). Water depths in wetlands can
change within days and thereby rapidly
change the suitability of sites as foraging habitat for wading birds or abruptly lead to nest
abandonment (Frederick and Collopy 1989).
Hydroperiod (the number of days per
year an area is inundated with water) is a
266
WATERBIRDS
hydrologic variable that could potentially
affect wading birds, but it acts more slowly
than does water depth. Hydroperiod determines vegetation community composition and
density (Gunderson 1994). In the Everglades,
shortened hydroperiods over large areas have
coincided in an increase in dense Sawgrass
(Cladium jamaicense) stands and a decrease in
more open slough habitats (Davis et al.
1994). Lengthened hydroperiods in other
areas have coincided with an increase in
open water habitat (Davis et al. 1994). Previous analyses have shown that wading bird distribution in the Everglades is correlated with
dominant vegetation type (Hoffman et al.
1994; Strong et al. 1997), but the resolution
of the analyses were too coarse to evaluate
the response to suites of vegetation types. It
is possible that changes in the vegetation mosaic may have contributed to shifts in the distribution of wading birds in the Everglades,
but this relationship has yet to be quantified.
Substantial effort is now being invested in
the development of restoration plans for the
Everglades ecosystem in hope that improved
hydrologic conditions will help restore
breeding numbers of wading birds within
the system. Considering that changes in hydrologic patterns cause gradual changes in
vegetation, it is important to understand
how vegetation, in addition to water depth,
affects wading bird distributions.
In this study, we test two qualitative hypotheses: First, does the vegetation mosaic,
in addition to water depth, influence the distribution of wading birds? Second, is this relationship different in a year when most of
the Everglades is covered with shallow water
as compared to a dry year when less water is
present and many areas are dry? We test
these hypotheses by quantifying the abundance of the Great Blue Heron (Ardea
herodias), Great Egret (Casmerodius albus),
White Ibis (Eudocimus albus), and Wood
Stork (Mycteria americana) across Water Conservation Areas (WCA) 1 and 2A from January through May of 1988 and 1989. These
two years differed in rainfall and, therefore,
the general hydrologic conditions in the
marshes; 1988 had average rainfall and 1989
was a drought year (Bancroft et al. 1994).
STUDY AREA
Water Conservation Areas 1 (also known as the
A.R.M. Loxahatchee National Wildlife Refuge) and 2A
comprise 57,240 and 42,206 ha of wetlands, respectively
in the northern Everglades (Fig. 1). These shallow-water
marshes were impounded by a levee system in 1961/
1962 to provide flood protection to the cities and farms
to the east and north and to provide water for agricultural and municipal use during the dry season (Light et al.
1989, Light and Dineen 1994). Under management of
the South Florida Water Management District (SFWMD), these impoundments are currently viewed as an
important ecological component of on-going Everglades
restoration efforts. Regulation schedules, that vary due
to seasonal weather conditions and rainfall, attempt to
maintain a natural hydrologic cycle, but are constrained
by water supply and flood protection demands.
The WCAs are not reservoirs but marshes in which
the levee and canal network influences the rate of flow
through the system. Local rainfall is the major factor influencing the depth and flow of water (Fennena et al.
1994). Water flows southward through the WCAs as a
shallow sheet, and pools in the southern ends of both
WCA 1 and 2A. The size and depth of the pools vary with
general water levels.
METHODS
Bird data. Distribution and abundance data for wading birds were taken from the Systematic Reconnaissance Flight (SRF) program (Porter and Smith 1984;
Bancroft et al. 1992; Bancroft et al. 1994; Hoffman et al.
1994) conducted in WCA 2A and WCA 1. The Systematic Reconnaissance Flight program was developed during the 1980s to document distribution and abundance
of wading birds within the Everglades system. East-west
aerial transects spaced two km apart were flown at an altitude of 60 m and a ground speed of 129 km h-1. All
wading birds detected in strips 150 m wide along both
sides of the transect were recorded. Positions were fixed
by recording the longitude. Transect lines were then divided longitudinally into 4-km2 cells. After excluding
boundary cells that included either levees or canals, we
had 208 interior 2 × 2 km grid cells (83,200 ha). We excluded boundary cells because we were concerned that
birds would be attracted to canals during low water periods. For the analyses presented here, one survey in
each of five months (Jan-May) conducted in both 1988
and 1989 for WCA 2A and WCA 1 were used. Arc/Info
software was utilized to create a 2 km by 2 km grid layer
of the SRF bird abundance data for these areas (Fig. 1).
The number of birds counted in each cell represented
the dependent variable in the analyses. We analyze data
for 1988 and 1989 separately because water depths were
substantially lower in 1989 than 1988, as illustrated by a
hydrograph for one location in north-central WCA 1
(Fig. 2). We only used two years of this multi-year dataset
because we were most interested in investigating the
role that vegetation played in addition to water depth in
the distribution of wading birds. Including more years
would create a statistical problem in that vegetation did
not vary between years. Water depth variation would
therefore overwhelm the analyses.
We combined water level and vegetation data (see
below) with the SRF bird abundance data to form the
resource databases for the analyses. We created separate
WADING BIRDS, WATER DEPTH, AND VEGETATION
267
Figure 1. Wading bird Systematic Reconnaissance Flight grid and vegetation map of Water Conservation Areas 1 and
2A of the northern Everglades, Florida.
268
WATERBIRDS
Figure 2. Mean monthly water depth in 1988 and 1989 at
a gage in north central WCA 1.
databases for 1988 (average) and 1989 (dry). Each database included bird numbers, water level, and vegetation
composition for each cell in five months. Thus, each
analysis had 1,040 observations (5 months × 208 cells).
Water level for each 2 × 2 km SRF grid was calculated
by intersecting the SRF and water level data (see below
for description of source of water level data) and calculating the weighted average water level for each SRF
grid cell. Hectares of vegetation for each SRF grid cell
were calculated by summing the area that each vegetation category contributed to each SRF grid cell (see below for source of vegetation data).
Water depth data. The South Florida Water Management Model (SFWMM) is a spatially explicit computer
model that simulates the hydrology of South Florida
(Fennema et al. 1994, South Florida Water Management
District 1997). The regional model uses a 3.218 km ×
3.218 km grid system. The SFWMM assumes homogeneity in hydrologic characteristics within each grid cell and
runs at a fixed time step of one day. The SFWMM has
gone through an extensive testing process and water
patterns predicted by the model match observed patterns (South Florida Water Management District 1997).
Uncertainty analyses were conducted throughout the
extent of the model to determine confidence limits
around water depth predictions (South Florida Water
Management District 1997). The model is used to make
management decisions in south Florida by the Water
Management District and is being used in evaluating
restoration alternatives.
We use this model to provide an index of water
depths, hereafter called water level, available for foraging wading birds across each grid cell. The water level
index is a continuous variable measured in cm. Individual cells are relatively flat, with small topographic
change from the northern to southern ends. Although
only one water level value is available for each cell, it is
important to note that each cell does possess microtopographic variation that provides variable feeding
opportunities for wading birds. As water levels rise in a
cell, fewer places in the cell would have depths appropriate for wading bird feeding. Similarly, as the “levels”
decrease, more places would be available for foraging.
Even with a negative water level value (indicating water
had receded below ground) foraging sites may still be
available because of the micro-topographic variation
across a cell. The SFWMM was run to calculate surface
water level for each grid cell on the same dates that SRF
bird abundance data were gathered.
Vegetation data. Rutchey and Vilchek (1994) and
Richardson et al. (1990) utilized digital image processing techniques to classify multispectral SPOT satellite
images of Water Conservation Area 2A (from 1991) and
Water Conservation Area 1 (from 1987), respectively.
Detailed vegetation data were not available for other
areas in the Everglades where SRF surveys were conducted. Image processing software (ERDAS, Atlanta, GA)
was utilized to recode and combine the classified images
produced by Rutchey and Vilchek (1994) and Richardson et al. (1990) into one image with eight wetland vegetation categories. Vegetation categories were 1)
Sawgrass, 2) Cattail (Typha domingensis), 3) Slough/
Open Water/Wet Prairie mix (called slough from here
on), 4) Brush, 5) Cattail mix, 6) Sawgrass/Cattail mix,
7) Brush mix and 8) Tree Islands. Brush species typically
included Southern Willow (Salix caroliniana), Wax Myrtle (Myrica cerifera), and/or Button-Bush (Cephalanthus
occidentalis). Tree island vegetation consisted primarily
of dense vegetation with canopies of Dahoon Holly (Ilex
cassine) and Red Bay (Persea borbonia), with Wax Myrtle
occupying lower tiers and the outside edges of the
islands. The “mix” categories were dominated by the
named mixture, but included other plant species.
Slough/open water/wet prairie habitats are found in
the wettest sites in the Everglades. Wet Prairie includes
associations of spikerush (Eleocharis spp.), beakrush
(Rhynchospora spp.), maidencane (Panicum spp.), Arrowhead (Sagittaria lancifolia), and Pickerel Weed (Pontederia
lanceolata). Slough/open water habitats include White
Water Lily (Nymphaea odorata), Floating Hearts (N. acquatica), and Spatterdock (Nuphar advena). Periphyton and
a number of submerged aquatics are found in this habitat. The final vegetation coverage was converted to an
Arc/Info Grid file with 9.144 m × 9.144 m grid cells (Fig.
1). Although the SPOT classified vegetation maps used
in this study represent the best available data for all vegetation classes, maps of dominant cattail coverage based
on aerial photography have shown that SPOT images
overestimate area of that vegetation class (Rutchey and
Vilchek 1999) by approximately thirty-five percent.
Statistical analyses. The first step of the model selection process was to consider the potential shape of the
response of bird abundance to each independent variable. For water depth, the a priori expectation, based on
previous studies (e.g., Powell 1987; Hoffman et al.
1994), was that bird numbers would be highest at some
optimum water depth and lowest at either end of the
depth gradient. This quadratic relationship could also
occur for any particular vegetation type. This would be
the case if birds preferred a vegetation mosaic rather
than more monotypic stands. For example, if birds preferred an equal mix of two vegetation types, predicted
bird abundance would increase as a function of area for
each vegetation type until the amount reached 50%.
Thereafter, predicted bird abundance would decrease
with increasing amounts of either vegetation type. Thus,
for each variable, a linear and quadratic relationship
with bird abundance was considered. If the model included a variable raised to the power two, it was also required to include the variable in its lower-order form
(Freund and Littel 1991; Freund and Wilson 1993). The
dependent variable, number of birds per cell, was transformed using ln + 0.1 to stabilize the variances.
For each species and year, analyses were performed
to identify the most parsimonious regression model
WADING BIRDS, WATER DEPTH, AND VEGETATION
consisting of a subset of the nine variables (i.e., water
level, and eight vegetation categories). The most parsimonious model is one with an optimum balance between low precision (i.e., too many variables in the
model) and high bias (i.e., too few variables in the model; Burnham and Anderson 1992). There are two interrelated aspects to identifying the most parsimonious
model. First, for each species we calculated 18 “optimum” models for 1 to 18 terms. For each number of
terms, the “optimum” model was one that contains the
set of variables that provides the minimum residual sum
of squares (Freund and Wilson 1993). This was achieved
using PROC REG and the RSQUARE selection procedure in SAS (SAS Institute Inc. 1988). The RSQUARE
selection procedure was preferred over other screening
procedures, such as stepwise, because it calculated all
possible combinations of variable and guarantees that
the “optimum” model can be identified (Freund and
Littel 1991; Freund and Wilson 1993). Second, the most
parsimonious model was selected from the set of 18 optimum models for each species. This was done using the
Cp (Mallows 1973) criterion, which partitions out the
bias error component from the random error component of the total error mean square (Neter et al. 1996).
Cp was calculated as:
2
C p = ( SSE ⁄ σ ) + 2 p – N
where SSE is the error sum of squares, σ2 is an estimate
of the variance, p is the number of terms in the model,
including the intercept, and N is number of observations. The final model selection was based on criteria
that (1) the value of Cp is small and (2) Cp ≈ p (Neter
et al. 1996). Values of Cp > p indicate a model is underspecified (i.e., biased) and values of Cp < p indicate a
model is overspecified (Freund and Littel 1991; Neter
et al. 1996).
After identifying the most parsimonious models, a
variety of diagnostic tools were used to evaluate the
choices. To determine if interaction terms would substantially improve the final models, we plotted the residuals against the cross-products of each pairwise
combination of independent variables (Neter et al.
1996). If the resulting pattern indicated that the interaction term influenced the residuals, it was included in
the final model. Plots of residuals versus predicted values to evaluate overall model adequacy were examined
to test the assumption of equal variances. A “fanshaped” pattern of residuals suggests that the variance is
proportional to the mean, thus requiring further remedial steps (Freund and Wilson 1993; Neter et al. 1996).
A formal diagnostic statistic, the variance inflation factor, was used to test for the presence of a correlation
among independent variables. The variance inflation
factor provides a measure of how much larger the variance of a parameter estimate was than if it were uncorrelated with other independent variables (Freund and
Wilson 1993). Values greater than ten are commonly
thought to indicate serious multicollinearity (Freund
and Wilson 1993; Neter et al. 1996). In polynomial models such as these, quadratic terms are likely to be correlated, which could technically lead to multicollinearity.
However, the implications are not the same as with linear terms (Freund and Wilson 1993), and therefore the
variance inflation factor for quadratic terms was not
considered. Although water depth and vegetation are
likely to be correlated at some temporal scale, there was
269
very little multicollinearity among these variables in our
data set. It is likely that the influence of water depth on
the vegetation community occurs over longer time
scales than our monthly surveys. Water depth changed
daily, whereas vegetation was static for the period of
study. None of the final models were changed after evaluating their results with the suite of diagnostic tools.
After the diagnostic phase of the analysis was completed, regression coefficients for each independent
variable in the final models were used to calculate predicted bird abundance as a function of each independent variable, holding the effects of all other
independent variables in a model constant. If a quadratic term was included in the final model, coefficients for
both the linear and quadratic form were included in the
equation. To get the predicted change in untransformed
bird abundance as a function of the independent variables, the antilog of predicted bird abundance - 0.1 was
calculated. These values were plotted against observed
values for each independent variable in the model to illustrate only the shape of the relationship between the
dependent and independent variables and the relative
magnitude of the response. We caution against using the
plots to predict actual bird abundance because the predictive power of the models was usually low.
To identify the variables that were most explanatory
of wading bird distribution, we examined the “optimum” models containing 1 to 4 terms. These models
were not the most parsimonious but they identified the
variables that appeared to be most important in explaining wading bird distributions.
RESULTS
The mean number of Great Blue Herons
per survey did not vary significantly between
1988 ( x = 101) and 1989 ( x = 112; Table 1;
F1,9 = 0.10, n.s.). The number counted varied
from 27 in May 1989 to 185 in January 1989.
On average, significantly more Great Egrets
were counted in 1988 ( x = 714) than 1989 ( x
= 349; Table 1; F1,9 = 8.21, P = 0.02). The number of Great Egrets varied from 72 in April
1989 to 845 in January 1988. On average, significantly more White Ibises were counted in
1988 ( x = 4,428) than in 1989 ( x = 870; Table
1; F1,9 = 8.79, P = 0.02). The number of White
Ibis varied from 130 in May 1989 to 8,559 in
January 1988. In 1989, most White Ibis in the
Everglades were south of our study area (Bancroft et al. 1994). The number of Wood Stork
did not vary significantly between 1988 ( x =
109) and 1989 ( x = 160; Table 1; F1,9 = 0.27,
n.s.). The number counted was highly variable among months, varying from three in
May 1989 to 485 in January 1989.
All final models included terms for water
level and various vegetation categories (Appendix 1). Coefficients of determination for
270
WATERBIRDS
Table 1. Numbers of wading birds counted in the interior cells of Water Conservation Areas 1 and 2A of the Florida
Everglades. The survey transects represent 15% of the study area.
Month
January
February
March
April
May
January
February
March
April
May
Mean
Year
Great Blue Heron
Great Egret
White Ibis
Wood Stork
88
88
88
88
88
89
89
89
89
89
106
140
86
108
66
185
165
118
63
27
845
794
743
670
517
746
405
317
72
205
8559
3951
4481
2754
2393
2712
819
531
160
130
238
38
115
122
34
485
208
98
5
3
106
531
2649
135
all models were low, ranging from 0.06 to
0.42, but model fit always was enhanced with
the addition of vegetation terms. Models
provided a better fit to the data in 1989 than
1988 for all species (Appendix 1), although
the increases in coefficients of determination in 1989 were greatest for White Ibis (increase of 0.22) and Wood Stork (increase of
0.18). The most parsimonious model for
Great Blue Heron and Great Egret contained fewer vegetation terms in 1989 than
in 1988, whereas the opposite trend was evident for White Ibis and Wood Stork.
The optimal models, with between one
and four terms, suggest that the inclusion of
water depth in models was robust and further, that area of slough was frequently correlated with bird abundance. Models with a
single term contained water depth for six of
the eight species-year combinations (Appendix 2). Area of slough was the single term in
the Great Egret and Wood Stork models in
1988. Overall, area of water, slough, and
brush had the greatest effect on bird abundance. Area of tree island, cattail, cattail mix,
and sawgrass-cattail mix also were included
in models containing up to four terms, but
they had less of an effect on bird abundance.
During the average hydrologic year of
1988, the models showed that as water level
increased, numbers of all species increased
initially, then reached a plateau, and finally
decreased. The plateau, or threshold of maximum use, varied among species. The plateau was lowest for the White Ibis and only
slightly deeper for the Wood Stork. The
abundance plateau occurred at intermediate
water levels for the Great Egret and at the
highest water levels for the Great Blue Heron (Fig. 3). During the drought year of 1989,
water levels rarely were greater than 30 cm
above ground and the relationship between
bird numbers and water levels never reached
a plateau for any species. For all species,
abundance increased with water depth either linearly or as a quadratic function.
DISCUSSION
These analyses suggest that water depth
and the vegetation community influences
the abundance of wading birds within a cell
and therefore the distribution of wading
birds across the Everglades. Generally, water
depth had the greatest effect. During 1988,
an average year, there was a water depth
threshold beyond which predicted bird
abundance declined. This threshold of maximum use reflected species-specific foraging
constraints, not simply leg length. Wood
Storks have the longest legs of these four species (Palmer 1962), but their numbers began
to decline at intermediate water levels whereas the numbers of the Great Blue Heron did
not decline until water levels were much
deeper. The shallower threshold for storks
could exist because successful foraging requires high concentrations of fish and/or
larger sizes of fish, which are more easily
caught in shallower water (Kahl 1964; Kush-
WADING BIRDS, WATER DEPTH, AND VEGETATION
271
Figure 3. Relationship between predicted numbers of birds/cell and independent variables in regression models
for White Ibis, Wood Stork, Great Egret, and Great Blue Heron during 1988 and 1989.
lan et al. 1975; Ogden et al. 1976). High concentrations of fish occur in the Everglades,
mostly when water levels are shallow and especially when receding (Kushlan et al. 1975).
The predicted numbers of White Ibis began
to decline at shallower depths than those
documented for any other species. Ibis have
short legs and frequently probe their bill into
the sediment to find invertebrates. Thus, water depths cannot be much greater than the
272
WATERBIRDS
length of the bill and head of an ibis. Of the
visual feeders, the Great Blue Heron has
longer legs and exploited deeper water than
did the Great Egret. In 1989, when water levels across the northern Everglades were lower than in 1988, all species increased in
abundance as a function of water level. In
1989, few locations had water levels deeper
than the thresholds levels observed in 1988.
Area of slough appeared to have the second greatest effect on bird abundance overall. For Great Egrets and Wood Storks in
1988, area of slough had greater effect than
water depths on bird abundance. Although
the importance of this open habitat, especially for flocking species, seems intuitive and has
been reported (Bancroft et al. 1992; Hoffman
et al. 1994), there has been little quantitative
evidence to support the idea. Hoffman et al.
(1994) demonstrated that on a coarser spatial
scale (i.e., one vegetation class per 4 km2),
wading birds tended to avoid sawgrass but
there was no indication that birds preferred
sloughs. The importance of slough in this
study provides support for the notion that the
loss of sloughs, because of either shortened
hydroperiod (Hoffman et al. 1994; Gunderson 1994) or cattail invasion (Davis et al. 1994;
K. Rutchey pers. obs.), may have negative
consequences for foraging wading birds.
Tree islands are an important habitat for
the Great Blue Heron and White Ibis (Hoffman et al. 1994). Results here support this assertion, and also suggest that tree islands are
important for the Wood Stork. Tree islands
provide both roosting and nesting sites for
wading birds. However, the landscape configuration where tree islands occur may be at
least as important as the islands proper. Tree
islands are abundant across the central portion of WCA 1. This area contains a complex
mosaic of several vegetation communities
and contains more micro-topographic relief
than other places within the study area. This
topographic relief concentrates prey over a
wider range of hydrologic conditions and
may provide good feeding conditions during
much of the dry season. Wading birds forage
in a succession of isolated pools formed during the dry season in the marshes within this
tree island mosaic.
Cattails were included in five of the final
models, but did not appear to contribute
substantially to changes in bird abundance
based on the magnitude of their coefficients.
In three of the models, where cattails were
most abundant, predicted bird abundance
was lowest. At sites where cattails cover large
areas, the plant often grows very dense and
may preclude birds from foraging effectively.
The models for the Great Blue Heron in
1988 and Great Egret in 1989, predicted bird
abundance increased at high Cattail densities. Cattails currently cover large areas of the
northeastern and western portions of WCA
2A. Although dense Cattails reduce available
foraging habitat, they occurred in relatively
deep water and therefore may have supported the larger fish frequently taken by the
Great Blue Heron. Cattails also are indicative
of high nutrient areas that support a greater
biomass of fish than more oligotrophic sites
(Turner et al. 1999). In addition, a network of
airboat trails, which in the average year of
1988 may have provided open foraging sites
for the long-legged Great Blue Heron, occurred in Cattail areas. In 1989, when conditions were dryer, the airboat trails provided
depressions which may have concentrated
food for the Wood Stork. Wood Stork numbers were highest during the two months in
1989 when these areas had shallow water.
Personal observations indicated that willows and Pond Apples, which were classified
as brush, are commonly used by wading birds
for roosting and nesting. These areas frequently contain deeper depressions than the
surrounding marsh and therefore, they concentrate fish after other marsh habitats have
dried. Several large concentrations of feeding wading birds were found in these willow
and pond apple communities during 1989.
Differences between years in the vegetation terms included in final models were evident for all species. This pattern suggests
that the way birds respond to vegetation is affected by the hydrology (Hoffman et al.
1994). Strong et al. (1997) examined habitat
use of the Tricolored Heron (Egretta tricolor)
and Snowy Egret (E. thula) in Everglades National Park during three years (1987, 1988,
1989) that varied dramatically in hydrologic
WADING BIRDS, WATER DEPTH, AND VEGETATION
conditions. Differences in habitat use between years for each species was greater than
between species differences in habitat use
within a year. During 1987, a average year,
the inland freshwater marshes had water levels too deep to allow foraging and most birds
fed in coastal areas. During 1988 when water
levels were intermediate, most birds flew to
freshwater areas to feed. In 1989 when the
freshwater areas were dry, birds fed in coastal
areas near their colony. Strong et al. (1997)
showed that, within years, reversals in the
seasonal water recession resulted in shifts in
habitat use by the Snowy Egret.
All eight models had low to moderate
coefficients of determination. The variation
among cells in the degree of micro-topographic variation would contribute to the
low R2. Central WCA 1 has greater microtopographic relief than the rest of the study
area. This greater relief has previously been
noted as one of the reasons why this area is
used extensively by wading birds over a
broad range of hydrologic conditions (Hoffman et al. 1994). Within-cell topographic
variation in WCA 2A appeared less and birds
used it through a narrower range of water
levels (Hoffman et al. 1994).
Another possible reason for the low coefficients of determination is that birds were
responding to variables not included in the
models. For example, prey composition and
availability is an important factor affecting
wading bird abundance and distribution
(Kushlan et al. 1975; Ogden et al. 1976; Gawlik,
in press). Vegetation and water depth, which
determine the vulnerability of prey to capture (one component of prey availability),
were included in our models. However we
had no measure of prey abundance, a component of prey availability. Prey abundance
changes quickly because wading birds can
consume much of the prey in an area over a
relative short time and then move to new areas (Strong et al. 1997; Gawlik, in press). In
addition, if water levels rose between one
month and the next, the prey base in reflooded areas may not have had sufficient
time to recover (Hoffman et al.1994).
A third reason for the low coefficients of
determination may have been that birds
273
were responding to water levels and vegetation, but were doing so at spatial scales different from those at which the data were
measured. In a hierarchical model of habitat
selection, birds respond to environmental
cues at different spatial scales (Wiens 1989a).
For example, wading birds may leave the
morning roost and fly in a particular direction to forage based on general water levels
in a region (i.e., spatial resolution of several
km). But, the subsequent decision to land at
a particular point in the marsh may be a
function of vegetation type or micro-elevation in the immediate vicinity (spatial resolution less than one meter). Because of the
large spatial extent of our study area (83,200
ha), water depth was modeled at a spatial resolution of 4 km2 even though micro-topographic variations likely produced depth
variations at smaller scales. Even if landscape
data were available at such fine scales, there
may be differences among species in the
scale at which they respond to a particular
habitat variable (Wiens 1989b; Pearson
1993). Understanding the spatial scale at
which animals respond to various environmental variables has been the focus of recent
work (Keller 1990; Gascon and Travis 1992;
Collins and Glenn 1997) and should lead to
more robust habitat models.
The generally higher coefficients of determination for Great Egret and Great Blue Heron in 1988 than for White Ibis and Wood
Stork in 1988 may reflect their fundamentally
different foraging behavior. Great Egret and
Great Blue Heron are much more habitat
generalists while Wood Stork and White Ibis
have more restricted requirements for foraging habitat (Gawlik, in press). Although all
four species sometimes feed in flocks, White
Ibis and Wood Stork do so more often and
rarely are found singly (Smith 1995). Great
Egret and Great Blue Heron, however, are
regularly found as single individuals or feeding spread out across an area (pers. obs.,
Smith 1995). The greater flocking behavior of
White Ibis and Wood Stork resulted in a more
clumped distribution of birds in the landscape, thus reducing the fit of the models.
The results of this study provide evidence
that wading bird distributions are related to
274
WATERBIRDS
the vegetation community and that this relationship is altered by water depth. It is also
clear that other factors affect distribution
patterns. Additional work is needed to quantify how these other factors in combination
with water depth and the vegetation mosaic
affect prey availability and ultimately, wading
bird foraging opportunities. Second, a better
understanding of the importance of microtopographic variation in providing good foraging conditions would enhance the understanding of wading bird distribution. Is microtopographic variation critical for increased
prey populations or do they mainly form
structural places for the concentration of
prey? Third, we need greater understanding
of why slough habitats in the Everglades ecosystem appear to be extremely important to
wading birds. Are slough habitats the best for
increasing prey population size or are these
habitats structurally ideal for concentrating
prey and increasing their vulnerability to
capture by wading birds?
Restoration of more natural hydrological
patterns in the Everglades is a national and
international issue (Davis and Ogden 1994;
Bancroft 1996). Altered hydrologic patterns
have been blamed for the dramatic decrease
in breeding populations of wading birds
over the last 50 years (Ogden 1978; Bancroft
1989; Ogden 1994). As efforts to restore the
Everglades move forward, careful evaluation
of how this restoration is affecting vegetation
communities, prey availability and wading
birds will need to be done. The working hypothesis is that restoration of more natural
hydrologic cycles should allow recovery of
extent and functionality of slough habitat,
which should improve foraging conditions
for wading birds and contribute to the recovery of wading bird numbers.
ACKNOWLEDGMENTS
The bird survey work was supported by a grant to
GTB from the South Florida Water Management District. We thank Wayne Hoffman and Richard J. Sawicki
for their help with collecting and processing the field
data. Discussions with Peter Frederick, Wayne Hoffman,
John C. Ogden, and Richard Sawicki have improved various aspects of this work. We thank Tom Fontaine, Fred
Sklar, John Ogden, Quan Dong, Lou Toth, Peg Gronemeyer, and Stefani Melvin for comments on an earlier
draft of this manuscript.
LITERATURE CITED
Arengo, F. and G. A. Baldassarre. 1999. Resource variability and conservation of American Flamingos in
coastal wetlands of Yucatan, Mexico. Journal of Wildlife Management 63: 1201-1212.
Bancroft, G. T. 1989. Status and conservation of wading
birds in the Everglades. American Birds 43: 12581265.
Bancroft, G. T. 1996. Case Studies: United States. Pages
199-216 in Human population, biodiversity and protected areas: Science and Policy Issues (V. Dompka,
Ed.). American Association for the Advancement of
Science, Washington, D.C.
Bancroft, G. T., W. Hoffman, R. J. Sawicki and J. C.
Ogden. 1992. The importance of the Water Conservation Areas in the Everglades to the Endangered
Wood Stork (Mycteria americana). Conservation Biology 6: 392-398.
Bancroft, G. T., A. M. Strong, R. J. Sawicki, W. Hoffman
and S. D. Jewell. 1994. Relationship among wading
bird foraging patterns, colony locations, and hydrology in the Everglades. Pages 615-657 in Everglades:
the ecosystem and its restoration (S. M. Davis and
J. C. Ogden, Eds.). St. Lucie Press, Delray Beach, FL.
Burnham, K. P. and D. R. Anderson. 1992. Wildlife 2001:
populations. Elsevier Applied Science, New York.
Cézilly, F., V. Boy, R. E. Green, G. J. M. Hirons and A. R.
Johnson. 1995. Interannual variation in Greater Flamingo breeding success in relation to water levels.
Ecology 76: 20-26.
Collins, S. L. and S. M. Glenn. 1997. Effects of organismal and distance scaling on analysis of species distribution and abundance. Ecological Applications 7:
543-551.
Davis, S. M. 1994. Phosphorus inputs and vegetation sensitivity in the Everglades. Pages 357-378 in Everglades:
the ecosystem and its restoration (S. M. Davis and J. C.
Ogden, Eds.). St. Lucie Press, Delray Beach, Florida.
Davis, S. M., L. H. Gunderson, W. A. Park, J. R. Richardson and J. E. Mattson. 1994. Landscape dimension,
composition, and function in a changing Everglades
ecosystem. Pages 419-444 in Everglades: the ecosystem and its restoration (S. M. Davis and J. C. Ogden,
Eds.). St. Lucie Press, Delray Beach, Florida.
Davis, S. M. and J. C. Ogden. 1994. Everglades the ecosystem and its restoration. St. Lucie Press, Delray
Beach, FL.
Dimalexis, A. and M. Pyrovetsi. 1997. Effect of water level fluctuations on wading bird foraging habitat use
at an irrigation reservoir, Lake Kerkini, Greece. Colonial Waterbirds 20: 244-252.
Fennema, R. J., C. J. Neidrauer, R. A. Johnson, T. K.
MacVicar and W. A. Perkins. 1994. A computer model to simulate natural Everglades Hydrology. Pages
249-289 in Everglades: the ecosystem and its restoration (S. M. Davis and J. C. Ogden, Eds.). St. Lucie
Press, Delray Beach, Florida.
Frederick, P. C. and M. W. Collopy. 1989. Nesting success
of five Ciconiiform species in relation to water conditions in the Florida Everglades. Auk 106: 625-634.
Freund, R. J. and R. C. Littel. 1991. SAS system for regression. Second ed., SAS Institute Inc., Cary, NC.
Freund, R. J. and W. J. Wilson. 1993. Statistical Methods.
Academic Press, Inc., San Diego.
Gascon, C. and J. Travis. 1992. Does the spatial scale of
experimentation matter? A test with tadpoles and
dragonflies. Ecology 73: 2237-2243.
WADING BIRDS, WATER DEPTH, AND VEGETATION
Gawlik, D. E. In press. The effects of prey availability on
the numerical response of wading birds. Ecology.
Gonzalez, J. 1997. Seasonal variation in the foraging
ecology of the wood stork in the Southern Llanos of
Venezuela. Condor 99: 671-680.
Gunderson, L. H. 1994. Vegetation of the Everglades:
determinates of community composition. Pages 323340 in Everglades: the ecosystem and its restoration
(S. M. Davis and J. C. Ogden, Eds.). St. Lucie Press,
Delray Beach, Florida.
Hoffman, W., G. T. Bancroft and R. J. Sawicki. 1994. Foraging habitat of wading birds in the Water Conservation
Areas of the Everglades. Pages 585-614 in Everglades:
the ecosystem and its restoration (S. M. Davis and J. C.
Ogden, Eds.). St. Lucie Press, Delray Beach, Florida.
Kahl, M. P. 1964. Food ecology of the Wood Stork (Mycteria americana) in Florida. Ecological Monographs
34: 97-117.
Keller, J. K. 1990. Using aerial photography to model
species-habitat relationships: the importance of habitat size and shape. Pages 34-46 in Ecosystem management: rare species and significant habitats. New
York State Museum Bulletin 471.
Kushlan, J. A., J. C. Ogden and A. L. Higer. 1975. Relation of water level and fish availability to Wood Stork
reproduction in the southern Everglades, Florida.
Open File Report 75-434. U.S. Geological Survey,
Tallahassee.
Light, S. S., J. R. Wodraska and S. Sabina. 1989. The
southern Everglades: The evolution of water management. National Forum 69: 11-14.
Light, S. S. and J. W. Dineen. 1994. Water control in the
Everglades: A historical perspective. Pages 47-84 in
Everglades: the ecosystem and its restoration (S. M.
Davis and J. C. Ogden, Eds.). St. Lucie Press, Delray
Beach, FL.
Mallows, C. L. 1973. Some comments on Cp. Technometrics 15: 661-675.
Neter, J., M. H. Kutner, C. J. Nachtsheim and W. Wasserman. 1996. Applied linear statistical models. Forth
ed. Richard D. Irwin, Inc., Chicago.
Ogden, J. C. 1978. Recent population trends of colonial
wading birds on Atlantic and Gulf coast plains. Pages
135-153 in Wading Birds (A. Sprunt, IV, J. C. Ogden
and S. Winckler, Eds.). National Audubon Society,
New York.
Ogden, J. C. 1994. A comparison between wading bird
nesting colony dynamics (1931-1946 and 1974-1989)
as an indication of ecosystem conditions in the
southern Everglades, Pages 533-570 in Everglades:
the ecosystem and its restoration (S. M. Davis and
J. C. Ogden, Eds.). St. Lucie Press, Delray Beach.
Ogden, J. C., J. A. Kushlan and J. T. Tilmant. 1976. Prey
selectivity by the Wood Stork. Condor 78: 324-330.
Palmer, R. S. 1962. Handbook of North American birds.
Yale University Press, New Haven.
Pearson, S. M. 1993. The spatial extent and relative influence of landscape-level factors on wintering bird
populations. Landscape Ecology 8: 3-18.
275
Porter, K. M. and A. R. C. Smith. 1984. Evaluation of
sampling methodology—systematic flight/pilot
wading bird survey, technical report. Everglades
National Park, Homestead, Florida.
Powell, G. V. N. 1987. Habitat use by wading birds in a
subtropical estuary: implications of hydrography.
Auk 104: 740-749.
Recher, H. F. and J. A. Recher. 1980. Why are there different kinds of herons? Transactions of the Linnaean Society of New York 10: 135-158.
Richardson, J. R., W. L. Bryant, W. M. Kitchens, J. E.
Mattson and K. R. Pope, 1990. An evaluation of refuge habitats and relationships to water quality, quantity, and hydroperiod (A synthesis report prepared
for the Arthur R. Marshall Loxahatchee National
Wildlife Refuge, Boynton Beach, Florida), Florida
Cooperative Fish and Wildlife Research Unit,
Gainesville.
Rutchey, K. and L. Vilchek. 1994. Development of an Everglades vegetation map using a SPOT image and
the global positioning system. Photogrammetric Engineering & Remote Sensing 60: 767-775.
Rutchey, K. and L. Vilchek. 1999. Air photointerpretation and satellite imagery analysis techniques for
mapping cattail coverage in a northern Everglades
impoundment. Photogrammetric Engineering and
Remote Sensing 65: 185-191.
Smith, J. P. 1995. Foraging Sociability of Nesting Wading
Birds (Ciconiiformes) at Lake Okeechobee, Florida.
Wilson Bulletin 107: 437-451.
Smith, J. P. and M. W. Collopy. 1995. Colony turnover,
nest success and productivity, and causes of nest failure among wading birds (Ciconiiformes) at Lake
Okeechobee, Florida (1989-1992). Archiv Für Hydrobiollogie Special Issues Advances in Limnology
45: 287-316.
SAS Institute, Inc. 1988. SAS/STAT User’s Guide, Release 6.03 Edition. SAS Institute, Inc., Cary.
South Florida Water Management District. 1997. South
Florida Water Management Model. Hydrologic Systems Modeling Division, Planning Department,
West Palm Beach, Florida.
Strong, A. M., G. T. Bancroft and S. D. Jewell. 1997. Hydrological constraints on Tricolored Heron and
Snowy Egret resource use. Condor 99:894-905.
Turner, A. M., J. C. Trexler, C. F. Jordan, S. J. Slack,
P. Geddes, J. H. Chick and W. Loftus. 1999. Targeting ecosystem features for conservation: standing
crops in the Florida Everglades. Conservation Biology 13:898-911.
Willard, D. E. 1977. The feeding ecology and behavior
of five species of herons in southeastern New Jersey.
Condor 79: 462-470.
Wiens, J. A. 1989a. The ecology of bird communities,
Vol. 2: processes and variations. Cambridge University Press, Cambridge.
Wiens, J. A. 1989b. Spatial scaling in ecology. Functional
Ecology 3: 385-397.
276
WATERBIRDS
Appendix 1. The most parsimonious multiple regression models of the relationship among water depth, vegetation
area, and bird abundance for four species of wading birds in the northern Everglades (all models, N = 1,040, P <
0.0001).
Species
Year
Model a
R2
MSEb
Great Blue Heron
1988
0.18
1.31
Great Blue Heron
1989
0.23
1.27
Great Egret
1988
0.41
2.10
Great Egret
1989
0.42
1.69
White Ibis
1988
0.14
6.14
White Ibis
1989
0.36
2.24
Wood Stork
1988
0.06
1.03
Wood Stork
1989
Abundance(ln + 0.1) = -2.17 + 0.79(depth) – 0.14(depth2)
– 0.01(cattail) + 0.00008(cattail2) – 0.01(brush) +
0.00009(brush2) – 0.04(brushmix) + 0.002(brushmix2) +
0.003(slough)
Abundance(ln + 0.1) = -1.55 + 0.69(depth) – 0.05(treeisl)
+ 0.001(treeisl2) + 0.006(slough) – 0.00002(slough2) +
0.008(brush)
Abundance(ln + 0.1) = 15.55 + 1.56(depth) – 0.81(depth2)
– 0.05(catmix) + 0.00003(catmix2) – 0.05(cattail) +
0.0001(cattail2) – 0.05(brush) + 0.00007(brush2) –
0.05(sawcatmix) + 0.00007(sawcatmix2) – 0.04(sawgrass) –
0.00001(sawgrass2) – 0.03(treeisl) – 0.03(slough) –
0.00002(slough2)
Abundance(ln + 0.1) = -0.91 + 1.35(depth) – 0.06(treeisl)
+ 0.001(treeisl2) + 0.01(slough) – 0.00004(slough2) –
0.009(brush) + 0.0001(brush2) – 0.007(sawcatmix) +
0.00006(sawcatmix2) + 0.003(cattail)
Abundance(ln + 0.1) = 0.16 + 0.26(depth) – 0.62(depth2)
– 0.07(treeisl) + 0.001(treeisl2) + 0.04(brushmix) +
0.02(brush) – 0.0001(brush2) + 0.01(slough) –
0.00004(slough2) – 0.008(sawcatmix)
Abundance(ln + 0.1) = 14.19 + 1.50(depth) + 0.23(depth2)
– 0.07(treeisl) + 0.0006(treeisl2) – 0.05(sawcatmix) +
0.00001(sawcatmix2) – 0.04(catmix) + 0.00001(catmix2) –
0.04(brush) + 0.0001(brush2) – 0.03(sawgrass) –
0.00001(sawgrass2) – 0.03(slough) – 0.00003(slough2) –
0.02(cattail) – 0.0001(cattail2)
Abundance(ln + 0.1) = -1.98 + 0.15(depth) – 0.14(depth2)
+ 0.005(treeisl) + 0.005(slough) – 0.000001(slough2) –
0.005(sawgrass) + 0.00001(sawgrass2) + 0.004(brush) –
0.00004(brush2)
Abundance(ln + 0.1) = -1.23 + 1.10(depth) +
0.032(depth2) + 0.01(cattail) – 0.00006(cattail2) –
0.007(brush) + 0.00007(brush2) – 0.005(sawcatmix) +
0.00002(sawcatmix2) – 0.005(brushmix) – 0.001(catmix)
0.24
1.08
b
MSE = Mean Square Error.
Abbreviations: depth = water depth, catmix = area of cattail mix, sawcatmix = area of sawgrass/cattail mix, treeisl
= area of tree islands, brush = area of brush, brushmix = area of brush mixed with other vegetation, sawgrass = area
of sawgrass, cattail = area of cattails, slough = area of slough.
a
WADING BIRDS, WATER DEPTH, AND VEGETATION
277
Appendix 2. Optimuma multiple regression modelsb containing between one and four terms. The dependent variable is the number of birds per cell for each of four species of wading birds in the northern Everglades and the
possible set of independent variables are water depth and area of eight vegetation types (all models, N = 1040).
Species
Year
One term
Two terms
Great Blue Heron
1988
depth
(0.09, 1.45)
depth, slough
(0.14, 1.37)
depth, slough, brush
(0.15, 1.35)
Great Blue Heron
1989
depth
(0.14, 1.41)
depth, brush
(0.19, 1.32)
depth, brush, treeisl
(0.20, 1.32)
Great Egret
1988
slough
(0.24, 2.70)
slough, slough2 slough, depth, depth2
(0.24, 2.67)
(0.37, 2.24)
Great Egret
1989
depth
(0.33, 1.95)
depth, brush
(0.35, 1.91)
depth, slough, slough2
(0.39, 1.78)
White Ibis
1988
depth
(0.05, 6.72)
depth, depth2
(0.09, 6.46)
depth, depth2, slough
(0.12, 6.25)
White Ibis
1989
depth
(0.25, 2.58)
depth, catmix
(0.27, 2.53)
depth, slough, slough2
(0.32, 2.36)
Wood Stork
1988
slough
(0.03, 1.06)
slough, depth
(0.03, 1.06)
slough, depth, depth2
(0.05, 1.04)
Wood Stork
1989
depth
(0.19, 1.14)
depth, depth2
(0.21, 1.12)
depth, depth2,
sawcatmix
(0.21, 1.12)
a
Three terms
Four terms
depth, slough,
brush, brush2
(0.16, 1.34)
depth, brush,
slough, slough2
(0.21, 1.29)
slough, slough2,
depth, cattail
(0.38, 2.19)
depth, slough,
slough2, cattail
(0.40, 1.77)
depth, depth2,
slough, slough2
(0.12, 6.20)
depth, slough,
slough2, catmix
(0.32, 2.34)
slough, depth,
depth2, treeisl
(0.05, 1.04)
depth, depth2,
brush, brush2
(0.23, 1.10)
The “optimum” model for each number of terms was selected based on the lowest error mean square.
The numbers in parentheses are R2 and Mean Square Error.
c
Abbreviations: depth = water depth, catmix = area of cattail mix, sawcatmix = area of sawgrass/cattail mix, treeisl
= area of tree islands, brush = area of brush, brushmix = area of brush mixed with other vegetation, sawgrass = area
of sawgrass, cattail = area of cattails, slough = area of slough.
b