SIMSPAR Version 1.3
A Spatially-explicit Individual-based Object-oriented
Simulation Model for the Cape Sable seaside sparrow
in the Everglades and Big Cypress Landscapes
Philip Nott, Jane Comiskey, John Curnutt and Louis Gross
The Institute for Environmental Modeling
Tennessee Knoxville, TN 37996-1610
Copyright 1998 - The University of Tennessee
Introduction
The Cape Sable seaside sparrow (Ammodramus maritimus mirabilis) is an
ecologically isolated subspecies of the seaside sparrow (Beecher 1955). Recent
surveys estimate a population size of approximately 3000 individuals (from 6000+
in 1992), and its range is restricted to the extreme southern portion of the
Florida peninsula, almost entirely within the boundaries of the Everglades
National Park and Big Cypress National Preserve (Werner 1975, Bass and Kushlan
1982). The sparrow currently breeds in marl prairies typified by dense stands
of graminoid species usually below 1m in height and naturally inundated by
freshwater during part of the year. As water levels recede during the dry season
in late winter and spring, the sparrows establish territories and start nesting in
dense grass tussocks. Several broods may result from a single male territory if
hydrologic conditions allow (Lockwood et al. 1997). If water levels do not
recede early enough in spring, nesting may be delayed, and if water levels
subsequently rise during the nesting season eggs or nestlings may be drowned.
Recent declines have occurred in the sparrow population across its entire range,
probably due to higher water levels in recent years (Pimm et al. 1995).
Because the current range of the sparrow is limited to a few hundred square
kilometers and because it is subject to flooding and fires, the population is
highly vulnerable. The Everglades restoration project (Hinrichson 1995) plans
to change to the hydrology of the southern Everglades. Changing water levels in
parts of the sparrow's range would affect the reproductive success of the
sparrow. This model attempts to quantify the relative impact of alternative
hydrologic regimes on the survival of sparrow populations.
SIMSPAR Version 1.3
Several important features are supported in Version 1.3. Significantly, the
ATLSS habitat layer (Florida GAP analysis) is spatially integrated with the
existing topology and with extensive sparrow survey data. Subsequent analysis
provides a set of rules with which to define potential breeding habitat. SIMSPAR
predicts population levels and spatial distributions of breeding activity that
correlate well with field data.
SIMSPAR landscape
The landscape of the sparrow's range is modeled explicitly as a set of spatial
cells of fine enough resolution (500m) to represent areas of similar vegetation,
topography, and hydrology. The spatial extent of the model is an area west of
Shark River Slough covering 18 x 28.5 km. In this implementation of SIMSPAR
only the western subpopulation is included, since it is the only region for
which field estimates of fine resolution topography are available. SIMSPAR is
capable of being applied to other subpopulations when these topographic data
become available.
1. Universal Transverse Mercator coordinates describe the spatial extent of the
area modeled, which encapsulates the western sub-population (UTM Zone 17:
Northing 2 851 070, Southing 2 821 570, Easting 517 665, Westing 499 665). This
area is expressed as a 500m-resolution landscape (59 rows x 36 columns) and
corresponds geographically to the top left-hand box shown in SPARLAND.PDF. The
symbols within this box mark the historical locations of breeding sparrows
determined from extensive surveys in 1981, and every year since 1992 (Curnutt et
al. 1998).
2. This portion of the 1994 FGAP habitat map overlays the topography previously
derived from water depth measurements made during the 1995 extensive sparrow
survey. Interpolating non-uniformly spaced estimates of elevation using an
inverse distance method creates this topography.
3. We determined from the fine-scale 30m-resolution FGAP habitat map, the
percentage cover of each habitat type within each 500m resolution cell. This is
represented by a false color image of the area shown in WESTFGPH.PDF, black
lines delineate the boundaries of each 500m cell considered. A key to the 25
commonest (by % cover) classes depicted in this image is given in WESTFGPK.PDF.
4. We analyzed the fine scale habitat data associated with the 79 known
locations of breeding sparrows observed in the 1992 extensive sparrow survey
(Nott in prep.). Sonny Bass of Everglades National Park conducted this survey.
The results of these analyses are summarized as a set of 41 habitat type
tolerances and 2 habitat preference rules, tabulated in FGPRULES.TXT and
described briefly below.
* The 41 individual habitat types do not cover more than a certain percentage
of any 500m-resolution cell in which breeding sparrows were observed in 1992.
* The majority (76) of the 79 known 1992 breeding locations contain greater than
70% coverage of habitat types 30,33, and 39.
* Only 4 of the 79 known 1992 breeding locations are associated with less than
30% coverage of 'wetter' marshland habitat types (0 23 24 25 29 31 32 34 38 40
41).
* The Cape Sable Seaside-Sparrow is known to avoid trees when establishing
breeding territories. Many trees exist across the western landscape that are
sub-30m resolution features and therefore do not appear in the FGAP habitat map.
We complemented the FGAP map by determining the locations of tree infested survey
sites from the 1992 extensive survey data. The trees Bass observed at these
sites include pine, hardwood, cypress, willow and mangrove.
5. The results of these analyses are summarized in WESTLAND.PDF. Cells which
do not meet the criteria set in the first three rules are marked as non-habitat
cells and are shown as diagonal crosses (x). The cells in which survey trees
occurred are marked as non-habitat; with vertical crosses (+). Cells in which
an asterix (*) appears are cells defined as non-habitat by both the analysis of
FGAP data and the known locations of trees from extensive survey data.
The resultant landscape contains fewer cells available for breeding than version
1.0 in which the available habitat was simply defined by an upper limit of
elevation. We might expect results to be more variable between simulations
given the patchiness of available breeding habitat in version 1.3 to those of
version 1.0 due to the effects of inherent stochastic processes on isolated
pockets of breeding individuals.
Individual based approach
SIMSPAR provides continuous, spatially-explicit estimates of population size and
structure by following the actions, growth, reproduction and mortality of each
individual sparrow in the region. It follows the state of each individual
sparrow including its sex, age, reproductive status and location. The model
increases the age of an individual each day and updates its status according to
movement and behavior rules. The backbone of the model is a simple flow of
decisions and actions that affect individuals in relation to abiotic factors and
other individuals. The model is stochastic in that there are probability
distributions for each state change of an individual, thus requiring a set of
Monte Carlo simulations for each scenario.
1. Each individual sparrow in the population is modeled during the breeding
period. This may occur sometime between mid-February and as late as early August
of each year, dependent upon hydrologic conditions. In particular, the model
tracks the sex, age, breeding status, of each model individual from egg to the
end of its life. For mature males, the model tracks establishment of breeding
territories, finding a mate, the start of nesting, and the status of eggs and
nestlings on a daily basis. The life cycle parameters used in the model are
taken from life history studies of the sparrow (Werner 1975, Lockwood et al.
1997).
One important such parameter is that of annual adult survival rate. Pimm
(pers. comm.) reports an analysis of banding data ( A. m. mirabilis) suggesting
a survival rate of 0.5. Post and Greenlaw's (1982) analysis of data for resident
Ammodramus maritimus maritimus in New York salt marshes suggests 0.57-0.6. Both
Werner (1975) and Post et al. (1983) suggest a value of 0.85 for territorial males,
but both these studies used low sample numbers over a single year (N=16, N=21
respectively). A review by Martin and Li gives an average value of 0.55 for both
sexes. This is based on a study by Nichols et al. (1981) of 112 permanent resident
seaside sparrows over a 10 year period using the Jolly-Seber technique. SIMSPAR
adopts an adult survival rate of 0.6.
2. The relationship of sparrow breeding activity to water depth is modeled.
Water depth in spatial cells is tracked daily through a hydrologic model. A
spatial cell is not available for breeding activity (specifically mating and
nest building) until the water level in that cell falls below a threshold level
of 5 cm. Any rise in the water level above 14 cm in a particular spatial cell
during the nesting season is assumed to cause nest abandonment for sparrows that
have nests in that cell. The elevation of a cell relative to the water stage
determines the length of the effective reproductive season for the pair of
sparrows and the vulnerability of the nest to flooding. The sparrows are not
modeled in detail during the non-breeding season. Age-specific mortality rates
are assigned during that period probabilistically, currently the age dependent
mortality is set at 1.0 - 0.6 = 0.4 for all age classes.
3. At the beginning of each breeding season males that previously owned
territories return to those nesting territories as do the females. Females may
not remain in that cell during the breeding season, as females are relatively
mobile and may range up to 4 kilometers to find a mate. Recruits return to a
habitable location in the vicinity of their natal site. Dispersal occurs before
the beginning of the breeding season, are dispersing individuals move stepwise
across the landscape searching for suitable breeding habitat.
An individual chooses a cell in its 8-cell neighbourhood with equal probability.
An individual cannot move to a non-habitat cell (as per WESTLAND.PDF), or a cell
it has previously visited but failed to occupy. This is defined as a 'self-
avoiding random flight' henceforth known as SARF. This assumes the individual
has gained a 'perfect knowledge' of the location and status of cells it
previously visited. At each step an individual tests the cell to determine the
availability of vacant territories. If a vacant territory exists, it will end
its search and is assigned to that cell. If no vacant territories exist it will
attempt to move again following the rules described above.
Individuals start their search in their natal cell and with each step there is a
0.1 probability of mortality. This simulates the danger of moving during which
time they risk predation and starvation. If an individual encounters a cul-de-sac
in which it cannot move to a cell it has not already visited it remains there as
a floater. This process results in cells being saturated with males, a roughly
equal number of females, and a number of 'floaters' able to occupy territories
that become available due to the death of the male 'owner' or its mate. It also
results in a number of individuals dispersing to adjacent potential breeding
habitat where they can attempt to breed.
These rules have the effect that after a productive year in which many individuals
fledged successfully the extent of the area occupied by sparrows increases. The
distribution of dispersal distances reveals that most individuals disperse locally
within 0-2 km, but a few may disperse as far as 5-6 km. After wet years recruits
have a higher probability of ending their search in or close to their natal site
thereby saturating previously productive cells. In this case dispersal is normally
restricted to 0-2km.
Applying these rules to the validation run (using real NP205 stage height data
1981-1997) results in simulations that resemble the temporal and spatial pattern
seen in the results of the extensive sparrow surveys of 1981,1992, and 1993-1997.
Model population initialization
The 'seed' population used in SIMSPAR Version 1.3 is both spatially and
numerically explicit, . It is based on extensive helicopter survey counts of
singing males at the vertices of a 1km grid. Bass and Kushlan (1983) multiply
the number of singing male sparrows by a factor of 16 to estimate the total
number of birds within 1km2.
The seed SIMSPAR population distribution consists of a composite distribution of
1993, 1995, 1996 and 1997 sparrow observations, not just the 1997 distribution.
We assume the extensive helicopter survey misses low levels of breeding activity
at these historical locations; the so-called 'veil' effect. This same effect is
observed in validation runs of SIMSPAR in which randomly chosen cells are
sampled for the presence of singing males. No males are detected in some cells
but breeding activity occurs at some time during the breeding season.
There are several possible reasons why breeding activity may be undetectable at
a site (i.e. no males were singing). Either water conditions are unsuitable,
the males are not singing while caring for walking fledglings or, singing
males are too distant to be heard. In summary, the initial population is
distributed among 5 main patches depicted by the white circles in SPARTOPO.PDF.
Determining the initial population size and distribution requires processing of
the raw data. Firstly, because several years of data are combined to produce
the initial distribution more singing males are present than are present in 1997
alone. Secondly, because the locations of breeding sparrows were surveyed at a
1km-resolution, these map out into every other 500m-resolution SIMSPAR cell. To
deal with this problem we extrapolate each observation to fill the remaining 3
cells of a 2 x 2 cell block equivalent to 1km2 in area. We multiply the total
number of birds per cell (given by the composite distribution above) by a factor
of 16/(No. of 500m cells in 1km2) = 4, in accordance with the correction factor
of Bass and Kushlan (1983). Finally, we halve this number to produce an initial
population size of 308 birds.
Life cycle and behavioral parameters
The life cycle and behavioral parameter values used in this model produce
population-level estimates close to the field estimates when the real NP205
hydrologic dataset (for the period 1977-1996) is used to drive the model. These
values are documented in table F5SPARAM.TXT. Many of these remain the same as
for Version 1.0 with a few notable exceptions:
The maximum allowable density of breeding territories per cell is set to 4,
equivalent to a maximum of 16 breeding pairs per 1km2. This equates to 4
singing males recorded in the extensive survey. SIMSPAR assumes a uniform
spatial distribution of the maximum density of breeding territories per cell.
Observed densities vary between 1 and 16 but the relationship between territory
density and habitat heterogeneity is not yet fully understood.
The onset of nesting activity is crucial to the number of broods a pair of
sparrows may produce. The earliest recorded nesting activity occurred mid-March
in both 1996 and 1997, despite the fact conditions were dry prior to that date
(Lockwood pers. comm.). Therefore, SIMSPAR assumes an environmental trigger,
probably photoperiod, is responsible for the onset of breeding. Accordingly,
the SIMSPAR breeding cycle begins at the end of February and continues until the
beginning of August. Given initially dry conditions, the peak of the first
breeding cycle predicted by SIMSPAR occurs between mid-March and mid-April,
similar to the peak of first brood activity observed in the field. During the
SIMSPAR breeding cycle males may mate a maximum of three times given suitable
short-hydroperiod conditions, and may successfully raise three broods.
Hydrology model
The model is driven by daily water level data from a single cell close to the
sub-populations under study. Specifically, we designed and tested the model
using historical data from a National Park Service hydrologic monitoring
station. The western area is influenced by the hydrologic patterns experienced
at the station NP205 located at (Northing 2825223 Easting 515235).
In light of the sensitivity of the model to variations in water depth,
we have reevaluated our choice of the particular SFWMM 2-mile x 2-mile
grid cell for representing water depth at Gage NP205. This gaging station
lies near the juncture of 4 SFWM grid cells. After reviewing graphs of
SFWMM calibration/validation depth values for these four cells, in
comparison with historical gage data, we have chosen the cell, numbering
from the upper left, in the 47th row and 16th column (which would be 46,15
if numbered from 0) to use for all comparisons in the upcoming evaluations.
It is our understanding that ENP hydrologists use the 46th row and 16th column,
which is our row 45 column 15 or, counting from lower left, starting from 1,
their row 20, column 16), or in some cases the mean of values for the
four cells.
Stochasticity
The model is initialized with a population of 308 sparrows reflecting the size
of the 1997 population estimate for this region (Curnutt et al. 1998). The
individuals are placed in higher elevation cells but not at ones for which trees
and shrubs dominate the vegetation, concordant with the findings of Nott et al.
(1998). The initial locations do not vary among replicates. A number of
replicate simulations (typically 20) are run for each hydrologic scenario. Each
replicate employs a different random number seed which causes variability in the
temporal and spatial patterns of successful breeding. The "variability"
incorporated in this model affects individuals' breeding locations, mate choice,
mating success, clutch size, gender, dispersal and mortality. Due to this
individual-level stochasticity, population- level estimates from the model vary
among replicates. This stochasticity allows for construction of a population
viability analysis, in which estimates are made of the probability of population
levels falling below or going above certain thresholds. These probabilities are
calculated by determining what fraction of the replicate simulations cross these
thresholds.
Population Viability Analysis (PVA)
All scenario comparisons are made using 3 distinct PVA techniques: time to
quasi-extinction; time to quasi- explosion; and interval explosion risk. For
each of these analyses we started calculations at 5 years after the start time
for the model.
a) Time to quasi-extinction (TQX) is the cumulative probability (based on
replicates - thus the term "quasi-") that the population will drop below a
specified number over time.
b) Time to quasi-explosion (TQE) is the converse of time to quasi-extinction.
It estimates the probability that the population will exceed a specified number
over time.
c) Interval explosion risk (IER) is the probability that the population will
reach a certain number over the period of the model.
Acknowledgements
Phil Nott wishes to thank all parties who attended his presentation of this
model on February 19th 1998 at the Center for Natural Resources in Everglades
National Park, and at the meeting on April 24th at SFWMD headquarters in Palm
Beach. Many invaluable comments and suggestions were made, some of which are
included in this version.
References
Bass, O. L., Jr., and J. A. Kushlan. 1982. Status of the Cape Sable
Sparrow. Report T-672, South Fla. Res. Ctr., Everglades National Park.
Homestead, Fla. 41 pp.
Beecher, W.J. 1955. Late-Pleistocene isolation in salt-marsh sparrows.
Ecology 36:23-28.
Curnutt, J. L., A. L. Mayer, M. P. Nott, O. L. Bass, D. M. Fleming, S.
Killeffer, N. Fraley and S. L. Pimm. 1998. Population dynamics of the
Endangered Cape Sable Seaside-Sparrow. Animal Conservation 1: in press .
Hinrichson, D., 1995. Waterworld. The Amicus Journal 17: 23-27.
Lockwood, J. L., K. H. Fenn, J. L. Curnutt, D. Rosenthal, K. L. Balent
and A. L. Mayer. 1997. Life history of the Endangered Cape Sable
Seaside-Sparrow. Wilson Bulletin in press.
Martin T.E and P. Li. 1992. Life history traits of open- vs. cavity-nesting birds.
Ecology 73: 579-592.
Nott, M. P., O. L. Bass, D. M. Fleming, S. E. Killeffer, N. Fraley, L.
Manne, J. L. Curnutt, T. M. Brooks, R. Powell and S. L. Pimm. 1998.
Water levels, rapid vegetational changes and the Endangered Cape Sable
Seaside-Sparrow. Animal Conservation 1: 23-31.
Nott, M. P. and J. L. Lockwood. Individual-based spatially explicit model
for the endangered Cape Sable seaside sparrow. in prep.
Pimm, S. L., K. Balent, T. Brooks, J. L. Curnutt, J. L. Lockwood, L.
Manne, A. Mayer, M. P. Nott, and G. Russell. 1995. Cape Sable Sparrow
Annual Report. NBS/NPS, Everglades National Park, Homestead, Fl.
Post, W., J. S. Greenlaw, T. L. Merriam and L. A. Woods. 1983. Comparative
ecology of the Northern and Southern popualtions of Seaside Sparrow. in The
Seaside Sparrow, its Biology and Management. Occ. Papers of the North Carolina
Biol. Surv. 123-136.
Werner, H. W. 1975. The biology of the Cape Sable Sparrow. Report to U. S.
Fish and Wildlife Service, Frank M. Chapman Memorial Fund, The International
Council for Bird Preservation and U.S. National Park Service, Homestead, Fl.
215 pp.
===============================================================
Output associated with the ATLSS Individual-based Cape Sable Seaside Sparrow
Model.
In accordance with ATLSS file naming conventions, each file name will consist of
the characters:
"X" or "_" => the Base, typically F for the F2050 base or E for the C1995
base
"X" or "_" => the alternative scenario or base
"SP" => the ATLSS Individual-based Cape Sable Seaside Sparrow Model
"XXXX" => 4 character mnemonic "." "PDF" or "TXT" or "DOC" => PDF, tabular text or documentation
===============================================================
ATLSS Individual-based Cape Sable Seaside Sparrow
1. Maps
Map outputs used to characterize results of the Individual-based Cape Sable
Seaside Sparrow component of ATLSS will consist of a number of image files in
PDF file format. A 30m-resolution habitat map of the western region is shown in
WESTFGPH.PDF with its associated key WESTFGPK.PDF. A 500m-resolution map of the
topology and associated non-habitat cells is shown in WESTLAND.PDF ,superimposed
on this map are the locations of the initial population distribution.
SIMSPAR also outputs a "Set" of model results, comparing a Base case
to a Scenario, following the conventions for ATLSS comparison of two model runs.
The extent of the map is the western region of the Everglades as shown on the
map "SPARLAND.PDF". Each map has 3 panels. The left panel is an alternative or
base scenario. The right panel is the base scenario, typically the Future
without Project Conditions Case. The middle panel is the difference between the
two scenarios.
Each map depicts the model area at a fine (500-meter x 500-meter) scale
of resolution, with each cell color coded to represent the annual productivity
value or the difference between the annual productivity value in the alternative
scenario and the base scenario. For more information on the productivity
values, see the section on Visualization above.
For each of a list of years, the images will provide a spatial display
of productivity values during that year. In addition, there is an image file
for the mean of all simulated years. The list of years are those that serve to
highlight the differences between the scenarios.
The mnemonic characters are composed according to the convention:
"XX" = Last two digits of the year
"XX" = PM - Productivity Map
An example of an ATLSS Individual-based Cape Sable Seaside Sparrow map file
name. F5SPMYPM.PDF - Comparison map displaying all years mean productivity of
F2050 Base and the Alternative 5 scenario.
========================================================
2. Time Series
Time series sets associated with the ATLSS Individual-based Cape Sable Seaside
Sparrow will display replicate population trends, the variability of populations
and population viability analyses. For information on population trends and
variability, see the section on Visualization above. For information on
population viability see the sections on Stochasticity and Visualization above.
File Name Description
--------- --------------------------------------------------------
XXSPTRAJ.PDF Shows a single line representing population trajectories for
each replicate simulation and both a base and a scenario run.
XXSPCVAR.PDF Summarizes the coefficient of variation in predicted population
levels for both a base and a scenario run.
XXSPPTQX.PDF Presents probabilities of time to quasi-extinction based upon
replicates for either a base or scenario run.
XXSPPTQE.PDF Presents probabilities of time to quasi-explosion based upon
replicates for either a base or scenario run.
XXSPPIER.PDF Presents interval explosion risk based upon replicates for both
a base and a scenrio run.
XXSPMPOP.PDF Presents the mean and standard deviation of the predicted
breeding population size for both a base and a scenario run.
========================================================
3. Histograms None.
========================================================
4. Tables None.