ATLSS: ACROSS-TROPHIC-LEVEL SYSTEM SIMULATION: An Approach to Analysis of South Florida Ecosystems Biological Resources Division United States Geological Survey South Florida/Caribbean Ecosystem Research Group Miami, Florida ACROSS-TROPHIC-LEVEL SYSTEM SIMULATION (ATLSS): APPROACH FOR ANALYSIS OF SOUTH FLORIDA ECOSYSTEMS Progress Report January 1997 DRAFT South Florida/Caribbean Ecosystem Research Group Biological Resources Division United States Geological Survey Table of Contents Executive Summary 1 I. Introduction and Background 2 A. Overview 2 B. General description of ATLSS 2 C. How ATLSS developed 4 D. Objectives of ATLSS 4 E. Geographic scope of ATLSS 5 F. Empirical support for calibration and testing of ATLSS modules 6 G. Timelines for completion of ATLSS projects 6 II. Uses of ATLSS 8 A. ATLSS as a tool for managers 8 B. ATLSS as a tool for scientists in forming and testing hypotheses 9 III. Connection of ATLSS with Everglades Landscape Model (ELM) 10 IV. Limitations and Data Gaps 11 V. Individual Components within the ATLSS Framework 12 A. Landscape 12 B. Abiotic components modeling 16 C. Vegetation modeling 18 C1. Seasonal vegetation dynamics model 19 C2. Plant succession and diversity modeling 22 C3. Disturbance modeling 24 D. Lower trophic level modeling 26 E. Fish and aquatic macroinvertebrates modeling 28 F. Reptile and amphibian assemblage modeling 31 G. Crocodilian modeling 33 H. Wading bird assemblage modeling 35 I. Cape Sable seaside sparrow modeling 37 J. Snail kite population modeling 40 K. White-tailed deer and Florida panther trophic interaction modeling 43 VI. Integration of ATLSS Models 45 A. Need for integration 45 B. ATLSS integration design 45 C. ATLSS integration progress 45 D. Trophic network analysis 47 VII. Model Testing Procedures 49 A. Plans for model rationale justification, calibration, and validation of models 49 B. Considerations of error multiplication 50 VIII Long-Term Plans 51 A. Future additions to the biotic components of ATLSS 51 B. Future additions of environmental conditions 51 C. Broader roles for ATLSS 52 IX. Project Titles, Principal Investigators, and Institutions 53 X. Funding Agencies 55 XI. ATLSS Publications 56 XII. Other References 58 XIII. ATLSS Presentations 62 XIV. ATLSS Home Page Information 64 Executive Summary The Across Trophic Level System Simulation program, or ATLSS, is an integrated set of computer simulation models representing the biotic community of the Everglades/Big Cypress region and the abiotic factors that affect this community. The models are spatially explicit, using GIS map layers of topography, soil, vegetation type, etc. The spatial extent of the models is the entire Everglades/Big Cypress region and some surrounding areas, and the spatial resolution is generally 500 x 500 meter cells, though sometimes finer . Relevant abiotic processes, such as hydrology, fire, and major storms are modeled. The biotic community is represented by a hierarchy of models, beginning with the process models of the biota constituting the energy base, including vegetative biomass, lower trophic level invertebrates, and decomposers. Models that contain relevant detail on size and age structure simulate several important functional groups, fishes, macroinvertebrates, and small reptiles and amphibians, which utilize the production of the energy base and provide food for some of the top consumers. Higher trophic level species populations, such as the Florida panther (and an important prey species, the white-tailed deer), several species of wading birds, the American alligator, the Cape Sable seaside sparrow, and the snail kite are modeled using individual-based models. These populations do not include all the high trophic level consumers that might ultimately be of interest in ATLSS, but because of their ecological importance, or importance as indicator species, these species provide a good representation of the higher trophic components of the Everglades food web. The primary goal of the ATLSS Program is to produce an integrated set of models for assessment of the biological effects of water delivery scenarios. The ATLSS models will be both calibrated and validated, and ready for use by scientists and managers. Because ATLSS is intended for use by managers, it will have a user-friendly interface with a simple menu to guide the user both in selecting biotic components to simulate and analyzing the results. The ATLSS Program has been adopted by the Federal, State, and Tribal Taskforce for the Restoration of the South Florida Ecosystem as the primary ecological tool for assessing the ecological effects of alternative water management changes. The objectives of the ATLSS program over the longer term are to aid in understanding how the biotic communities of South Florida are affected by the hydrologic regime and by other abiotic factors, and to provide a predictive tool for both scientific research and ecosystem management. The distribution, volume, and timing of water flow influences the energy and material transfers among ecological components within and across the trophic levels of these systems. The ATLSS integrated model simulates mechanistically the causal relationships between hydrology and the biotic components of the Everglades/Big Cypress region. ATLSS is composed of several individual modeling projects that provide modules for ATLSS. These projects are matched where necessary with empirical studies to support the model with empirical information. The information available from ATLSS empirical studies and other data sources will be divided into parts that can be used, respectively, for calibration and testing of the component ATLSS modules. The ATLSS Program is scheduled to produce by the end of 1997 and early part of 1998 usable models of all of the biotic components that are part of the program. In subsequent years, the models of these components will be refined and models of additional components (e.g., small mammals, raptors such as ospreys) will be added to the program. In addition, other threats to the environment, such as mercury and global warming, may be modeled. I. Introduction and Background A. Overview The Everglades Forever Act calls for land acquisition, the rerouting of water flow, and other actions to be undertaken for the purpose of restoring Everglades National Park and Big Cypress Swamp National Preserve. Most specific decisions concerning possible management restoration actions have not yet been made and await evaluation their potential long-term effects. Unfortunately, restoration experiments are prohibited because of the large spatial extent and long time horizons required to restore these ecosystems. Quantitative modeling is one management tool that can be used to evaluate the long-term outcomes of specific restoration actions. Modeling is an indispensable tool for understanding complex ecological systems and for providing guidance in their protection. A model is a mathematical or computer code representation of the real world. The computer simulation models described in this report incorporate cause-and-effect relationships between the components of the biota and aspects of the environment that are influenced by humans. These models can make predictions about how these biotic components will respond to human manipulations. In particular we are interested in predicting the responses of selected biotic components in the Everglades/Big Cypress region of South Florida to such abiotic conditions as the seasonally varying water levels across the South Florida landscape. The modeling is designed to provide a basis for management decisions in this region. New modeling approaches for populations and communities have been developed, including individual-based models (e.g., DeAngelis and Gross 1994, also see definition in Table 1) and there is an increasing emphasis on a landscape-level viewpoint (e.g., Schafer 1990, Turner and Gardner 1991, Huston 1994, Forman 1996). Individual-based models allow the enormous store of knowledge of physiological and behavioral ecologists to be incorporated into models of populations and communities. The landscape perspective recognizes that the dynamics of a population depend on the landscape in which the population exists, so that population and community models must incorporate the features and heterogeneity of the landscape to an extent and degree of resolution that depend on the questions being asked. Enormous advances have also been made in GIS technology (e.g., Miller 1994). Mapping techniques using satellite and aerial photography, together with ground truthing, are providing detailed maps of vegetation and other characteristics of regions to a high degree of resolution. These GIS maps can be used as the underlying landscapes for regional scale models of populations and communities. Simulation modeling can describe how individuals, populations, and communities utilize the landscape and can help one determine the viability of these biotic entities over time scales of interest. This is the basic approach being used for developing predictive models for the ecosystems of South Florida. B. General Description of ATLSS The Across Trophic Level System Simulation program, or ATLSS, is an attempt to apply these new conceptual and technological advances in modeling and GIS to regional scale ecosystem analysis. ATLSS is an integrated set of computer simulation models. These models are set in the landscape of South Florida, represented by GIS map layers of topography, soil, vegetation type, etc. There are underlying models of the relevant abiotic quantities, such as hydrology, fire, and major storms. "On top" of these abiotic models are models of the energy base of the region, including vegetative biomass, lower trophic level invertebrates, and decomposers. Models that contain some detail concerning size and age structure simulate the main macrofaunal components, fish, small reptiles and amphibians that utilize the production of the food web base. At the very top are models of the higher trophic levels that utilize the production from the lower trophic levels. ATLSS is this integrated set of models, where the term "integrated" means that the different models and map layers, or "modules" (see definitions in Table 1) within ATLSS will be able to pass information back and forth to each other when necessary (e.g., when an herbivore module consumes a plant, or when a predator module consumes an herbivore). ATLSS uses a spatially-explicit landscape scale, modeling approach (see definitions in Table 1). This means that the distribution of the components of the biotic community are modeled across the landscape. This landscape structure is based on remote sensing information on elevation, vegetation, etc., across the region. There is an integrated system of landscape scale models: hydrologic models, plant community succession models, models of fire and other disturbances, and models of lower, intermediate, and selected higher trophic level functional groups or species populations. The hydrologic model used is interchangeable, and can be either the Natural System Model (Fennema et al. 1994), the Water Management Model (MacVicar et al. 1984), the Everglades Landscape Model (ELM, Fitz et al. 1993), which already exist in validated form, or any other model that is available, validated, and accepted by the research and management communities. ATLSS predicts how the ecosystem responds to different hydrologic regimes. ATLSS is a natural extension of the way in which spatial modeling is applied to predict hydrology across the South Florida landscape. It extends this spatial modeling to the biotic community. ATLSS is, therefore, a spatially explicit, landscape level modeling approach. ATLSS also embodies a food web approach, as it incorporates all trophic levels and simulates energy flow through the food web. ATLSS uses different modeling approaches at different trophic levels in the major food webs (aquatic and terrestrial) of the system. The different modeling approaches used in ATLSS include (see Figure I.1): 1) process-oriented models for key functional groups of lower trophic level organisms (periphyton and macrophytes, detritus, micro-, meso-, and macroinvertebrates); 2) size- and age-structured population models for key functional groups and species of intermediate trophic levels (five functional groups of macroinvertebrates and fish, four functional groups of amphibians and small reptiles); and 3) individual-based or -oriented models for key higher consumer species (American alligator, colonial wading birds, the Cape Sable Seaside Sparrow, the Snail Kite, White-tailed deer, Florida panthers). A method of linking together models that need information from, or provide information to, other models, is also needed. This integration of ATLSS model components in a common framework or integration shell is required to provide a system level approach to addressing the many scientific and management questions affecting South Florida wetlands and uplands. Because of this, we refer to ATLSS as an "integrated model". The computer code for the fully integrated model is in a standard, object-oriented language (C++) to allow easy modifications to incorporate future research and the need for possible changes in algorithms. The fully integrated ATLSS model will also have the capability of interacting through data passing with other landscape scale models, such as the Everglades Landscape Model of the South Florida Water Management District. It will provide user-friendly interfaces to allow regional managers to analyze the ecological effects of alternative hydrologic scenarios. A GIS object-oriented data storage structure will enable the saving of outputs of the integrated model for conducting multivariate analyses on the resulting database using statistical software packages. Such analyses also provide the scientific basis for conducting risk and cost-benefit analyses for each alternative hydrologic scenario evaluated. C. How ATLSS Developed ATLSS officially began as a fully funded program in 1995. However, the roots of ATLSS began a few years earlier with the need to formally conceptualize the Everglades/Big Cypress region as a dynamic, heterogeneous landscape. It is a landscape characterized by changes on several temporal scales, including within-season short-term changes, seasonal patterns, multi-year cycles, and long- term directional changes. This view of the Everglades/Big Cypress region was reflected in the proceedings volume, Everglades: The Ecosystem and Its Restoration (Davis and Ogden 1994). A second influence in the early development of ATLSS was the emergence of individual-based spatially explicit modeling as a way of modeling the dynamics of some populations (Huston et al. 1988, DeAngelis and Gross 1994), particularly higher trophic level species, in which there is rich physiological and behavioral information. In this approach, every member of a population is individually modeled. This allows population ecologists to model the spatial demographics of populations, in a realistic way, on temporally varying, heterogeneous landscapes. These models also relate directly to the research and monitoring data collected by field biologists, and make predictions that can be compared directly to field observations. The details of the time/energy budgets of the individuals, the heterogeneous distribution of food resources on the landscape, and stochastic factors are all simulated in determining the growth, reproduction, and mortality of individual members of the population. To obtain population level statistics, such as total population size through time, sums over all of the individuals being simulated are computed. The first components of the ATLSS integrated model were spatially explicit, individual-based models of wood storks (Wolff 1994, Fleming et al. 1994) and white-tailed deer/Florida panthers (Abbott 1995, Abbott et al. 1995, Comiskey et al. 1996). However, ATLSS was expanded as the need was seen to know the spatio-temporal pattern of food availability for these high-level consumers. First, a primary intermediate trophic level functional group, freshwater fishes, were modeled (DeAngelis et al. 1996). Then, to provide a mechanistic way of providing carrying capacities for the fish, the lower aquatic trophic levels (periphyton, meso- and macro-invertebrates) were modeled. Vegetation biomass production was modeled to serve as a forage base for the white-tailed deer. Progress in developing the lower trophic level and abiotic basis of the ATLSS integrated model has been aided by close collaboration with the ELM program (see section III. B.). Subsequently, other components have been added to ATLSS, both to fill out the trophic structure and to add higher-level components that are important for conservation reasons (e.g. Cape Sable seaside sparrow, snail kite). D. Objectives of ATLSS The immediate objective of the ATLSS program is to provide a quantitative, predictive modeling package for guidance of the South and Central Florida restoration effort. Given the complexity of this immense ecological system and the uncertainties involved in restoration, this program has been deemed absolutely critical to the restoration effort (Weaver et al. 1993). The ATLSS Program has been adopted by the Federal, State, and Tribal Taskforce for the Restoration of the South Florida Ecosystem as the primary ecological tool for assessing the ecological effects of alternative water management changes. The relationship of this and other modeling projects to the integrated set of tasks comprising the Central and South Florida Restudy Project is portrayed in Figure I.2. The objectives of the ATLSS program over the longer term are to aid in understanding how the biotic communities of South Florida are linked to the hydrologic regime and to other abiotic factors, and to provide a predictive tool for both scientific research and ecosystem management. The distribution, volume, and timing of water flow influences the energy and material transfers among ecological components within and across the trophic levels of these systems. The ATLSS integrated model simulates mechanistically the causal relationships between the hydrology and the biotic components of the Everglades/Big Cypress region. Stated in scientific terms, the ATLSS Program has the goal of predicting the spatial and temporal patterns of biota in response to changes in the hydrology and other physical aspects of the environment by simulating mechanistically the causal relationships between hydrology and the biotic components of the Everglades/Big Cypress and surrounding ecosystems. To help with inferring the causes for declines in key species over the past few decades, ATLSS will be used to compare trophic responses to the natural (pre-drainage) patterns of water flow and to the current (post-drainage) patterns, simulated with the same time series of rainfall data. Analyses of such comparisons will allow: 1) identification of the effects of altered landscape characteristics; 2) testing of related hypotheses concerning landscape alterations as possible causes of species declines; and 3) qualitative evaluations of the minimum hydrologic threshold requirements of the biota as a guide to restoration efforts. ATLSS will then be used to predict the responses of biotic communities to several proposed alterations of the hydrologic regime in South Florida and, from these predictions, to provide advice to the South Florida Ecosystem Restoration effort. Beyond the immediate aims of providing information for the restoration, ATLSS has as its long-term goals the study of other impacts on the South Florida ecosystem, such as pollution (e.g., mercury, phosphorus), land-use change, global warming, and the invasion of non-native species. ATLSS will also form a framework for the design of empirical studies and testing of hypotheses. E. Geographic Scope of ATLSS The geographic scope of ATLSS is currently confined to the area of South Florida that is included in hydrologic models. This geographic area is pictured in Figure I.3. All ecosystem types within this area except the urban and agricultural areas to the north and east are included in the ATLSS integrated model. Although Figure I.3 excludes the mangrove estuaries to the west, ATLSS modules are being developed for these areas in anticipation of hydrologic models being available at some time in the future. ATLSS must also be expanded to the northwest as soon as possible, to incorporate some of the habitat of the Florida panther that is currently excluded. It will eventually be extended farther north, and, in fact, the snail kite model does include patchy sites from Lake Okeechobee north to the Kissimmee lakes and Upper St. John's River. Expansion southward into Florida Bay, or linkage to proposed models of the Florida Bay ecosystem, is also planned. F. Empirical Support for Calibration and Testing of ATLSS Modules The individual modeling projects that provide modules for ATLSS are matched where necessary with empirical studies to support the model with empirical information. The information available from ATLSS empirical studies and other data sources will be divided into parts that can be used, respectively, for calibration and testing of the component ATLSS modules. The ATLSS package will provide the best possible forecast of how the ecosystems of South Florida should respond after the implementation of the restoration has begun. As the ecosystem response to restoration is monitored, this will provide further data to test and refine the ATLSS integrated model. In the sense of adaptive management, the combination of monitoring data and improved model predictions will be used to recommend modifications of the restoration scenario where necessary, a procedure shown schematically in Figure I.4. G. Timelines for Completion of ATLSS Projects Originally, ATLSS was scheduled to produce an integrated set of models for use in the analysis of hydrologic restoration scenarios by the middle of the year 2000. However, the recent Water Resources Development Act sets a schedule for a final feasibility report by the Restoration Task force for July, 1999. Therefore, to produce useful input for this report, ATLSS models must be completed much faster than the original schedule for completion of the overall package. In fact, most of the component models that make up the integrated ATLSS will be completed long before the integrated ATLSS is finished. We expect most of the component models to become available in usable form during 1997 and 1998. These can feasibly be used in aspects of the restoration evaluation prior to final completion of the overall integration. There are some caveats, however: 1) It is unlikely that all of the component models will be linked together in a central shell and accessible through an easily used menu before the middle of 1998. Each application will have to be "hard-wired", rather than being selected merely by used of a menu, in order to meet the revised and shortened time frame. 2) Some of the models will be missing sound information on key parameters, as these data are still being collected. In particular, some models will depend on having the hydrologic models extended to the coastal estuaries. 3) Most models will not have undergone full testing ("validation"), so there will be continuing modifications of these models to improve their predictive capabilities. (It should be stressed that model testing must always be an ongoing process in any case.) Table 2 shows the current schedule for completion of components of ATLSS. In some cases the table distinguishes two or three versions of the model. These cases are elaborated in somewhat more detail below, and is discussed in Section V. It should be understood that at least minor modifications will continue to be made in all models after their scheduled completion data. Vegetation biomass production - The first version of the model is complete now (denoted as Cv1). But this does not include certain geographic regions. The next version will extend the model to the coastal mangrove (tidal) areas (denoted by Cv2). The final version will extend the model to the western uplands, outside the coverage of the currently available hydrologic models. Vegetation succession - The first version with be available by the end of 1997. The final version, which will add disturbance sensitivity, will be available in the middle of 1998. Fish functional groups - The first version of the model is available now. The final version will add models of two important aquatic macroinvertebrates; apple snails and crayfish. This version will also extend the model from the freshwater marshes to the coastal estuaries. Reptile and amphibian community - The first version, or "descriptive phase," will be completed in May 1997. This will describe the standing stocks of key functional groups and the energy fluxes between these groups. The second, or "predictive phase," of the model will be completed by June 1998. This will allow predictions to be made concerning how these standing stocks and fluxes will respond to abiotic scenarios. Wading birds - The first version of this model, simulating a single nesting colony of one species (wood storks), has been completed. The final version will model simultaneously many breeding colonies of up to five species, including mixed colonies. Snail kite - An initial version of the snail kite model is available. This operates on yearly time steps. The final version will be available in December 1997. This will include more behavioral details and will operate on daily or 5-day time steps. Florida panther/deer - The first version of this model is available. The final version will include more behavioral details, such as panther marking behavior. This will be available in April, 1997. II. Uses of ATLSS A. ATLSS as a Tool for Managers The primary goal of the ATLSS Program is to produce an integrated set of models for assessment of water delivery scenarios, where the models are calibrated and validated, and ready for use by scientists and managers. Because ATLSS is intended for use by managers, it will have a user-friendly interface with a menu to guide the user both in selecting biotic components to simulate and analyzing the results. For example, suppose the user wishes to assess the effect of a given water delivery scenario on the population of Florida panthers. In making this choice, a suite of modules will automatically be selected to be run interactively; 1) topographic map 2) rainfall data from historical record (20-year period) 3) hydrologic module with given water delivery scenario 4) vegetation change module a. disturbance module b. vegetation succession module 5) vegetative biomass production module 6) white-tailed deer module 7) Florida panther module The user will be able to choose certain parts of the ATLSS landscape, including the entire landscape, on which to simulate the panther. The user will also be able to specify the duration of the simulation. This will normally be a 20-year period. Because the ATLSS biotic modules for higher trophic levels are stochastic, Monte Carlo models, each simulation is only one possible realization. To calculate means and variances for population behavior, a large number of simulations will be performed for a given water delivery scenario. For each of these simulations, a different pseudo-random number generator initiator will be used. The model will perform a number of simulations, where this number of scenarios will be chosen to give a chosen level of confidence. The user will be able to choose to observe a wide array of the output produced both by individual simulations and the summary data averaged over many simulations. In particular, the output data on spatial locations for each model panther through time can be observed on a GIS map and related to various environmental conditions across the landscape. It is the further objective of ATLSS to present these results in a form that allows scientists and managers to easily compare how each of the scenarios affects each of our model outputs, and to compare also each of these scenarios in terms of costs and of effects on other functions, such as flood control. This will be done through an interface that will take the output from our models and allow the scientists and managers (users) to view it in a convenient format that makes comparisons easy. For example, the user may want to compare the effects of each of the water delivery scenarios on Cape Sable seaside sparrows. The user should be able to select "Cape Sable seaside sparrow" and be presented with GIS maps and summary statistics showing the yearly variations in Cape Sable seaside sparrow densities over 20 years for each of the scenarios. As another example, the user may want to look at all of the effects of a single scenario. In this case, the user selects "Scenario 1" and is presented with GIS maps showing how the densities of each of the above outputs change through time over 20 years for Scenario 1. The user may also want to determine which scenario maintains all key species populations above certain critical levels, for the least amount of costs and impairment of other functions, such as flood control. This should also be possible through the interface. B. ATLSS as a Tool for Scientists in Forming and Testing Hypotheses An important long-term benefit of ATLSS will be in its use in formulating and exploring, through simulation modeling, key scientific hypotheses regarding South Florida ecosystems (some of which hypotheses may have important practical implications as well). A brief list of such questions or hypotheses, ranging from the ecosystem to the individual species level, is as follows: 1) the issue of spatial extent and population viability. What are the threshold hydrologic conditions for viability of the species? To what extent can populations be maintained on ranges that are smaller than their historical ranges, through manipulation of the functional attributes, such as landscape heterogeneity or productivity, of the shrunken range? 2) what are the effects of habitat fragmentation on key species in South Florida ecosystems? 3) what are the relative importances in the overall energy budget of South Florida ecosystems of the main biotic communities? For example, it has been hypothesized that the herpetological community may play a larger role in the Everglades than in most other ecosystems. 4) both the loss of short-hydroperiod wetlands and decreases in coastal estuary productivity have been hypothesized to be key causes in the decline of wading birds in the southern Everglades. ATLSS will attempt to compare the two factors through simulation. 5) there are numerous questions regarding the adaptations of particular species. For example, alligators are highly sexually dimorphic. Various hypotheses are possible and some may involve energetic and dispersal constraints that could be studied in a spatially explicit individual-based model such as ATLSS. Migratory patterns of snail kites and site-selection for colonial nesting wading birds are also phenomena that may be elucidated by the modeling of large numbers of individuals on complex landscapes to compare the advantages of various strategies. III. Connections of ATLSS with Everglades Landscape Model A key step in the development of a complete modeling approach for ecosystem analysis in South Florida is collaboration and possible eventual integration with the Everglades Landscape Model (ELM) of the South Florida Water Management District (e.g., Fitz et. al. 1993). The primary objectives of the ELM are to: 1) provide a spatial modeling tool to estimate water demands of the Everglades. 2) predict changes in the landscape pattern of vegetation that is associated with hydrology, water quality, and fire frequency. ELM includes the following processes: 1) water movement vertically in a cell 2) nutrient fluxes through compartments 3) primary production 4) decomposition 5) organic/inorganic sediment suspension and deposition 6) vegetation succession (function of hydrology and fire frequency) ELM thus provides information that is complementary and of use to ATLSS (items 1,2,3,4) and provides the opportunity for useful collaboration in areas that are important to both models (items 5,6) that are not well developed in the current ATLSS integrated model. The ATLSS and ELM modeling groups plan initial collaboration in three phases: Phase 1. ELM results will be used to drive the fish "resource" component of the ATLSS fish model (development of file sharing system; no model modifications). Phase 2. ELM hydrology, nutrients, and vegetation will be used to drive the ATLSS fish model (some modification of ELM and the fish model will be necessary). Phase 3. ELM and ATLSS will be coupled, so that ELM nutrients and vegetation include feedbacks from the ATLSS fish model. Interactions on the fire disturbance model will be undertaken. Status of the three phases. Phase 1 should be complete by January 1997, Phase 2 by May 1997, and Phase 3 by August 1997. Long-range plans The ELM and ATLSS groups will work together towards plans formulating general improved methodologies of ecosystem modeling and object-oriented design. IV. Limitations and Data Gaps It is important to point out gaps that exist in data needed to quantify the models within ATLSS. These gaps impose limitations on the accuracy of the models. Some of the most important gaps are described below. 1) Perhaps the most pressing general need is for higher resolution of ground surface topography and hydrology, both at a macro- and microscale. Current hydrologic resolution is 2 mile x 2 mile in the South Florida Water Management Model and 1 km x 1 km (ELM) model. Improvement is needed as follows: a) Macroscale topography and hydrology to at least 500m x 500m resolution for responses of some of the higher level consumers where water level can directly impact nest sites (e.g., Cape Sable seaside sparrows, alligators) and food availability (e.g., wading birds). Also, this or finer resolution topography will allow more detailed predictions of the water conditions that various vegetation types experience across the landscape. b) Microscale topography within 500m x 500m cells. Microtopography is needed, at least in a statistical way, to predict the occurrence of local refugia for fish, amphibians, and alligators during drydowns, or for terrestrial organisms during floods. 2) There is a need for development of further GIS layers, such as soil nutrient content and hydrologic properties, is needed. 3) More information is needed on the properties and dynamics of South Florida vegetation. Specifically: a) Information on growth rates, maximum biomass, response to water level, annual seasonality, soil type, etc. b) Responses of vegetation to stresses and disturbances such as salinity and extreme temperature. 4) Basic physiological knowledge is needed on some organisms (e.g., alligators). 5) Behavioral/physiological information is needed; e.g., conditions that trigger nesting behavior in wading birds or dispersal in snail kites. 6) Insect biomass is of direct importance to some of the modeled biotic components (e.g., Cape Sable seaside sparrows) and is indirectly important to nearly all. However, little is known about insect standing stocks and dynamics, so this functional component is omitted from the present ATLSS integrated model. V. Individual Components within the ATLSS Framework Below, the individual component modules within ATLSS are briefly described. It will be noticed that especially with regard to the higher trophic levels, the model coverage of the biotic community is not complete. This is due ultimately to the huge amount of effort that would be required to model all higher level species. Choices had to be made and these were based in part on the ecological importance of the species, its usefulness as an indicator species, and data availability. In some cases species satisfying these criteria coincided with species that are Federally listed as threatened or endangered. However, each of the higher- level trophic level species in ATLSS has some significance besides its own intrinsic value. The Florida panther is representative of species that require large home ranges. The wading birds, Cape Sable seaside sparrow, and white-tailed deer are species that utilize patchily distributed resources. The American alligator requires a relatively stable hydrologic environment, and the snail kite is an example of an extreme ecological specialist. A. Landscape Structure 1. Purpose of this component Restoring and preserving the aesthetic and functional properties of the South Florida landscape is the central goal of the South Florida Restoration. Consequently, developing a predictive understanding of the biological and physical processes that define this landscape is a central goal of the research and modeling that supports the restoration effort. The vast extent of this landscape, along with its high spatial and temporal variability, provide a major modeling challenge. The high spatial variability of the landscape requires that the basic spatial unit be relatively small, while the large extent requires a very large number of the basic spatial units. There must always be some compromise between the time required to run a model and the size and/or spatial resolution of the model. The goal of the ATLSS landscape structure is to represent the basic physical properties and processes of the landscape as simply as possible while still providing the information needed for the much more complex biological models. The general objectives of the ATLSS landscape structure are: 1) To provide information on the spatial and temporal distribution of the key physical properties of the South Florida landscape in the form that is needed by the various biological and physical components of the ATLSS modeling program. 2) To provide the capability to calculate and record temporary and permanent changes in the physical properties (e.g., water depth, topography, soil depth, vegetation biomass) of the South Florida landscape that result from physical or biological processes. 3) To provide an appropriate interface between specific biological models and the physical properties of the landscape (e.g., water depth) that are relevant to that model. 4) To provide an appropriate interface within sets of biological models that interact with one another through their distribution across the landscape (e.g., interaction between deer and panthers, or between fish and wading birds). The ATLSS landscape structure is not a single "model" but is rather a group of components, some of which function as models, others which are digital maps and datasets, and others which serve as interfaces between individual-based models and the specific landscape properties that are important to them. 2. Modeling approach The basic modeling approach has been to develop an object-oriented (C++) structure that is sufficiently general to store spatial information of any type (along with standard FGDC metadata) and to allow appropriate calculations to be preformed between different types of information. The "landscape structure" includes: 1) "landscape classes" that store spatial (GIS) information that can be accessed (and changed, if appropriate) through interactions with other ATLSS models; 2) "data structure operators," which are system level processes that handle the mechanics of data input-output, maintain structure and associated metadata; and 3) "interface operators" that modify the landscape classes, either by updating a landscape class through internal processes (e.g., changing water levels, vegetation biomass, or fish density) or by exchanging information between the landscape classes and the various ATLSS plant and animal models, some of which may simply respond to the landscape, while others may actually alter the landscape. As a simple example, the high-resolution water depth map updates itself internally through the subtraction of ground surface elevations from water stage heights, which allows calculation of surface water depth (or depth of water table). Note that in addition to storing such standard GIS information as surface elevations, soil types, and road locations, the landscape structure also stores model outputs that are distributed across the landscape, such as the output of the vegetation model or the fish model. The landscape model structure is not limited to a specific set of spatial and temporal scales, and could operate, if necessary, at very fine spatial scales over limited portions of the entire region. However, for typical applications, the landscape model runs at a spatial resolution of 100 x 100 meter cells, at temporal resolutions of 1 to 5 days. For computational efficiency, the landscape model is designed to operate over a set of scales that can be simply aggregated through nesting (e.g., sizes of 1, 4, 16, 64, 256 units). Any environmental variable can potentially be included as information in the landscape structure, if adequate data on its spatial (and possibly temporal) distribution are available. However, the most important environmental variable in the landscape structure is the one with the greatest spatial and temporal variability across South Florida: water depth. Consequently, one of the major functions of the landscape structure is to predict the spatial and temporal variation in water depth across the entire region at scales of resolution that are relevant to the biological processes being modeled. The ATLSS project is not attempting to develop an independent hydrologic model for the region. Rather, the landscape model is designed to use standard stage height output from scientifically-reviewed and accepted hydrologic models as input for model development, testing, and scenario evaluation. Currently, the landscape model is being run using stage height output from the South Florida Water Management District's Water Management Model (SFWMM). As a consequence, the spatial domain of the ATLSS landscape model, and thus all of the component biological models of ATLSS, is currently limited to the area covered by the SFWMM. The spatial scale at which hydrology is modeled in the SFWMM (with 2 x 2 mile cells as the basic unit) is much larger than the scales at which the landscape must be modeled to predict ecological responses. It is well known that small scale variation in topography interacts with larger-scale varation in water level to determine the timing and patterns of drought, wetness, and water depth that are critical to understanding the population dynamics of different organisms. Because currently available topographic data for the region are far too coarse to be relevant to the fine-scale spatial biological patterns that characterize the Everglades, Big Cypress, and other critical areas in South Florida, the ATLSS landscape model has taken the approach of inferring the land surface elevation from current vegetation patterns. This approach generates "pseudotopography," which is ground surface elevations that are predicted on the basis of vegetation maps, rather than being measured directly. As direct measurements of topography become available at the scales of resolution needed by biological models, direct topographic data will replace "pseudotopography." 3. Current progress The ATLSS landscape structure has been completed and tested for internal consistency. It is now being used as the underlying structure of the vegetation model, the deer model, and the panther model, all of which now operate over the entire region of the SFWMM. In addition, the landscape structure is being used as the basis of the fish model, which is currently operating over portions of the entire region. A key component of the landscape structure, the "pseudotopography," has been completed and is now being tested and refined (Fig. V.A.1). It is expected that refinement of the "pseudotopography" will continue as more information becomes available, until eventually it will be completely replaced by direct, high resolution topographic measurements. Until that time, however, some form of pseudotopography will be necessary to provide sufficient spatial resolution of water depth variability for biological models. We have developed and tested a program for calculating pseudotopography, and have prepared a manuscript based on the approach described (briefly) below. The method we have used for generating pseudotopography is based on the relationship between hydroperiod and vegetation type that has been documented by research in South Florida over the past 30 years. The basic assumption is that the average hydroperiod of an area (the amount of time that the area is under water each year) can be predicted from the vegetation of that area. While this relationship apparently holds over a range of spatial scales, our application is based on a vegetation classification derived from standard LandSat imagery. We are using a new vegetation map recently prepared by the Gainesville unit of the USGS/BRD (Pearlstine et al.) for the Florida GAP Analysis Project, with a spatial resolution of approximately 28 x 28 m cells. We will continue to update our pseudotopography as the vegetation map is tested and revised and new vegetation maps become available. In particular, the new vegetation maps being prepared from aerial photography by the University of Georgia and the South Florida Water Management District will provide both higher spatial resolution and more accurate vegetation classifications that can be used to generate improved pseudotopography. The two primary steps in the generation of pseudotopography are: 1) calculation of the relationship between water level (stage height) and hydroperiod using the annual hydrograph for the area, and 2) for each vegetation type within the area, calculation of the land surface elevation that is required to produce the appropriate hydroperiod for that vegetation type. The first step is relatively straightforward, and requires the creation of a cumulative frequency curve for hydroperiod generated by the SFWMM. This is accomplished by summing the number of days that the water is at or above a specific level, from the highest water level that occurs during the year (corresponding to the elevation with the shortest hydroperiod) to the lowest water level (corresponding to the elevation with the longest hydroperiod). This calculation produces a curve of the potential hydroperiod for all elevations within the range of the annual hydrograph, and is performed independently of the average ground surface elevations used by the Water Management Model (i.e.,potential hydroperiod can be calculated whether the water level is above or below the ground surface elevation assumed by the model). This calculation is repeated for each of the 2 x 2 mile cells within the area covered by the Water Management Model (~1877 cells for the current version of the model). The calculations are independent of the spatial scale of the data, and could be repeated for water level elevations at different scales produced by other hydrologic models, or other versions of the SFWMM. The second step is somewhat more complex, because of the necessity of preserving the water volumes produced by the hydrologic model (SFWMM). For a given volume of water, the actual elevation of its surface will depend on the shape of volume it is allowed to occupy, i.e., it will be deep (have a high elevation) if confined to a volume with a small surface area, or shallow if confined to a volume with a large surface area. As the topography of the ground surface underlying the water volume in a given area is altered to produce the appropriate hydroperiod for each vegetation type, the actual elevation of the water surface will change as the volume that it can occupy is altered. The water volume output of the SFWMM is partitioned between surface water (which occupies 100% of the available volume) and subsurface water (which occupies some fraction of the available volume). In the SFWMM, the subsurface waterstorage capacity of bedrock is approximately 20%. However, because most of the small-scale topographic variability in the Everglades (except for pine rocklands) results from variation in marl and peat accumulation, we use storage capacities ranging from 20 to 85% for generating pseudotopography. Water volume is maintained by adjusting the water surface elevation as the land surface elevation is altered to produce the hydroperiod required for each vegetation type, beginning with the longest hydroperiod vegetation (occupying the lowest positions on the landscape) and progressing through vegetation types with decreasing hydroperiods (higher elevations). We are continuing to test the pseudotopography against the directly measured elevations that are becoming available for selected parts of the region. 4. Empirical data Ideally, the basic information in the ATLSS landscape structure would be completely based on field data. Unfortunately, actual measurements at the needed spatial and temporal resolutions are simply not available for most of the region. Consequently, much of the physical information in the landscape structure is based on the output of the best models currently available (e.g., the SFWMM for hydrology, the GAP Analysis vegetation map for vegetation, and the Everglades Landscape Model for nutrients). As these models are further developed and improved, the ATLSS models will become more accurate in their predictions of spatial and temporal patterns. The ATLSS landscape structure is designed to be as flexible as possible, and will be able to exchange information with models being created by other groups as part of the South Florida Restoration Project, as well as exchange modules or subroutines with other models as appropriate. Any additional data that are collected on high resolution topography, soil depth and nutrient content, and other landscape properties can be directly incorporated into the landscape GIS layers. Better topographic data are particularly important for improving model accuracy. 5. Timeline for completion of work The ATLSS landscape structure is now operational. Minor modifications will be made as new or better data become available, as well as to meet the specific information needs of other ATLSS submodels. B. Abiotic Components Modeling 1. Hydrologic models The seasonal hydrologic pattern in space and time is the key driving force in the Everglades. Changes in the hydropattern will affect other abiotic factors, such as fire severity and nutrient fluxes, as well as all biotic components of the system. The ATLSS integrated model uses the output from existing hydrologic models of the Everglades/Big Cypress region. There are three available models, all covering most of the area south of Lake Okeechobee, excepting the urban, agricultural, and coastal mangrove estuary areas. South Florida Water Management Model (SFWMM) (MacVicar et al. 1984). The SFWMM was developed by the South Florida Water Management District to describe coupled surface and ground water movement and stage across the region shown in Figure I.3. The model includes the complete network of canals, levees, pumps, and well fields. The area modeled is divided into 2-mi x 2-mi grid cells. The main input driving variables are rainfall and potential evaporation. The model includes as its main processes overland flow across the marsh, infiltration, evapotranspiration, ground water flow, and channel flow through canals. The Natural System Model (NSM) (Fennema et al. 1994). This was developed by the South Florida Water Management District using the calibrated algorithms and parameters from the South Florida Water Management Model. It resembles the SFWMM in most ways (e.g., a cell size of 2-mi x 2-mi), but in the NSM all of the water regulation structures built by humans are removed. The NSM does not simulate the natural Everglades, however, as it does not reliably estimate the overflow from Lake Okeechobee. Everglades Landscape Model (ELM) (Fitz et al. 1993). The hydrology model component of the ELM was developed primarily to study interactions between water and vegetation and the transport of nutrients. This model uses the same algorithms as the SFWMM, but omits some of the canal structure of the system. The spatial cell is 1-km x 1-km. 2. Nutrient models The two nutrients thought to be most limiting in the Everglades/Big Cypress region are phosphorus and nitrogen. The ELM simulates the kinetics of both of these within and between individual spatial cells. Both of these nutrients are divided into dissolved components in the surface water and sediment pore water. The surface water component can be carried by horizontal flows between cells. These movements are modeled as mass flows, but concentrations of the nutrients are calculated at each time step. Mineralization and biotic uptake by plants and microbes are simulated, but most of the detailed kinetics of nitrogen chemistry are omitted. 3. Salinity The ELM model also can simulate the movement of NaCl, though no effects on biota are currently simulated. 4. Dissolved oxygen and temperature A module, PhysDyn, has been developed (Fitz et al. 1996) that simulates the changing oxygen concentration and temperature of representative water parcels. The temperature is simulated by a function involving the water column and the atmospheric temperature. The oxygen in the water column is modeled using the processes of exchange with the atmophere, input from plant production, and depletion due to detrital decomposition. C. Vegetation Modeling General overview Of all the component models of the ATLSS program, the vegetation models are most closely tied to the landscape. Virtually the entire landscape is covered at a detectable density with vegetation of some type. Most of the vegetation is relatively permanent, or at least does not move from one location to another very frequently. Virtually all animals, on the other hand, move from one area to another, and many occur at extremely low densities across most the landscape. Consequently, the ATLSS vegetation models are closely integrated with the ATLSS landscape structure (see Section V.A). The basic landscape properties such as elevation and sediment depth directly affect the vegetation, and the amount and type of vegetation on the landscape are extremely important to the distribution, growth, and survival of virtually all animal species. Information on vegetation properties is one of the major types of information that is supplied to the ATLSS animal models by the ATLSS landscape structure. The following subsections describe three distinct types of models that address different aspects of the vegetated landscape of South Florida. All three are essential to meet the goal of understanding and managing the South Florida landscape, but each addresses a separate set of processes. C1. Seasonal Vegetation Dynamics Model - addresses the spatial and temporal variation in the amount and quality of forage available for herbivores, the fuel available for fire, the structure available for animals, and the light available at the water surface. C2. Plant Succession and Diversity Models - address the changes in the abundance of particular plant species in response to changing environmental conditions or as a result of successional change under relatively constant conditions. This approach is essential for predicting how species distributions, including rare species and exotics, are likely to change in response to natural or human-caused environmental change, and how the overall patterns of biodiversity will respond to water manipulation and other management activities. C3. Disturbance Models - South Florida is subject to several types of disturbance that can suddenly decrease the amount of living plant material on the landscape, and thus indirectly (as well as directly) affect many types of animals. Existing disturbance models for fire and hurricanes will be integrated into the ATLSS landscape structure, and development of a freeze model is planned. C1. Seasonal Vegetation Dynamics Model 1. Purpose of the component The energy provided by plants (net primary production or NPP) is the foundation for nearly all of the animal life on Earth. In South Florida, the high temporal variability of the environment (both within a year and between wet versus dry years) and its high spatial variability (from deepwater sloughs to pinelands) produce tremendous variation in the amount of food available to animals, both the herbivorous animals that feed directly on plants and the carnivores that feed on the herbivores. This variation in food availability determines the movement patterns, the population sizes and growth rates, and ultimately the survival of animal species ranging from mosquitofish to panthers. The purpose of the seasonal vegetation model dynamics is to predict the spatial and temporal availabililty of the biomass produced by vascular plants. Vascular plants (herbaceous macrophytes, shrubs, and trees) produce most of the net primary production available to terrestrial herbivores, and directly influence the productivity of aquatic algae (an important source of energy for aquatic food chains) through their effect on light and nutrient availability. Vascular plants respond strongly to variation in water and nutrient availability that result from both natural processes and human activities (e.g., water management). Plant material that is not directly eaten by animals is important as a source of energy for detritus-based food chains (e.g., bacteria, crustaceans, some fish), as a source of organic matter for peat formation, as fuel for fires, and as a source of structure in the environment (e.g., mangrove systems, hardwood hammocks). Specific issues being addressed by the ATLSS vegetation and related lower trophic level models include the amount and quality of food available to deer and other herbivores, and the distribution and structure of vegetation that affects the aquatic food chain through its impacts on temperature, light availability, oxygenation, nutrients, and foraging habitat. Because some of the birds and mammals in South Florida move great distances in their search for food, the landscape vegetation model is designed to predict the seasonal and interannual changes in biomass of all vegetation types across the entire region. The landscape vegetation model is not designed to predict how plant species composition and biodiversity may change in response to variation in water, nutrients, climate, and disturbances such as hurricanes, fire, and frost. These issues will be addressed by the vegetation succession and diversity models described in the next section. 2. Modeling approach The requirements that the landscape vegetation model predict vegetation properties over a large region dictates that the model be relatively simple. Simple predictive models of this type are often described as "empirical models", which use mathematical descriptions of observed responses, (e.g., plant growth in response to water level), in contrast to "mechanistic models," which use mathematical descriptions of the processes that produce the observed responses (e.g., photosynthesis, transpiration, individual processes). In reality, there is no clear dividing line between empirical and mechanistic models, since most "mechanistic" models include empirical descriptions of at least some of the mechanisms. The ATLSS seasonal vegetation dynamics model is designed to produce high spatial and temporal resolution across the entire South Florida region for a few vegetation properties of importance to the higher trophic levels in both aquatic and terrestrial food chains. These vegetation properties relate to the amount and quality of different types of plant tissue, with the types classified on the basis of their quality as forage. The definition of forage quality used in the model is based on the energy content that is available to a ruminant herbivore, specifically white-tailed deer. Although the three vegetation classes are defined based on their digestible energy content (e.g., Kcal/kg), this classification relates directly to a number of important vegetation properties and ecosystem processes in addition to energy content. In general, vegetation with high digestible energy content also has relatively high availability of mineral nutrients and protein. Digestibility by herbivores is usually highly correlated with digestibility by microbes and fungi, and hence correlated with rates of decomposition and nutrient cycling. Within some vegetation types, digestibility may be correlated with fuel properties, such as flammability (i.e., leaves ignite more readily than wood), but not necessarily with gross energy content. The three classes of vegetation are defined separately for each of the vegetation types, based on differences in the digestible energy content of the vegetation (the energy content per unit mass is 2000, 1500, 800 kcal/kg for classes 1,2, and 3). Although the properties of individual species are not specifically represented by the vegetation classes, in most vegetation types, the most palatable class is composed of only a few species that represent a very small proportion of the total plant biomass (e.g., Crinum, Hymenocallis, Salix). Consequently, most of the biomass in all vegetation types is in the lowest quality class, and represents primarily mature and dead herbaceous tissue plus wood. This vegetation class can thus be used to track total dead and woody biomass as it relates to vegetation structure, successional stage, and fuel availability for fire. The spatial variability of vegetation properties in the model is determined by the resolution of available information on the vegetation types of South Florida. Currently, the highest resolution vegetation maps are based on satellite images with a maximum spatial resolution of 28.5 meters. These approximately 30 x 30 meter cells are the basic spatial unit of the vegetation map (produced by the NBS Gainesville Laboratory, L. Pearlstein et al.) being used in ATLSS, although for interaction with many of the ATLSS submodels (e.g., deer, panther, wading birds) this scale is aggregated to coarser resolutions of 100 m or 500m. Each vegetation type is assigned a local elevation consistent with its hydroperiod in relation to the other vegetation types in the area (see discussion of landscape models and pseudotopography, Section A). Surface elevation for any cell on the landscape is subtracted from water stage height information (from the SFWMD or ELM) to provide water depth (above or below the surface) which determines how rapidly any specific vegetation class grows or senesces. The temporal dynamics of the vegetation are driven primarily by water level, as it influences plant growth, senescence, and drying. Water level influences the rate of increase in plant biomass as a multiplier (from 0 to 1) of the maximum growth rate for a particular vegetation class (e.g., class 1 vegetation in the sawgrass vegetation type). Seasonal variation in the maximum plant growth (independent of water level effects) is driven by seasonal variation in total solar radiation (a multiplier from 0.8 to 1.0). The temporal resolution of vegetation change can be varied within the model structure, but for most purposes is changed in weekly increments (Figure V.C.1). Because the dynamics of the vegetation are driven by water, the spatial scale of the water data (actually, the output of the SFWMD Water Management Model) allows a substantial increase in computational efficiency. Since the water surface is essentially level over areas much larger than the 28.5 x 28.5 meter units of the vegetation map, the water levels and vegetation responses of many cells can be calculated simultaneously. Specifically, all of the 30x30m cells of a particular vegetation type within the spatial area defined by the Water Management Model (with output as 2 mile x 2 mile cells), can be assumed to experience the same water conditions and thus be calculated in a single step. Thus, rather than performing separate calculations for each of the 12,752 of the 28.5 x 28.5m cells within a 2 x 2 mile area, the calculations are performed only once for each of vegetation types within the 2 x 2 mile area (Figure V.C.2). Typically, there are less than 6 vegetation types, with a maximum of 25 (depending on the total number of vegetation or land use types in the current map). Individual 28.5m cells need be considered separately only when they become unique as a result of herbivory by deer or other disturbances such as fire. Unpredictable events that occur on real landscape, such as fires, hurricanes, freezes and lightning strikes will always cause detailed model predictions to diverge from observations, unless the models are rerun with the actual events included. Predictions of future conditions can include the effect of stochastic disturbance processes in order to quantify the potential variation in future conditions, but the exact details of future conditions can never be predicted. 3. Current progress The basic model has been completed and checked for internal consistency. Future work will involve: 1) testing the model against empirical data on spatial and temporal variation in plant biomass that will be provided by the new monitoring and research program; 2) expansion of the model to areas beyond the current extent of the SFWMM; 3) refinement of model parameters, as better data become available from field research and monitoring, plus possible modification or addition of processes that influence seasonal variation in vegetation growth and biomass (e.g., anoxia in flooded systems, seasonal nutrient dynamics). 4. Empirical data The primary limitation for vegetation modeling is the lack of adequate information on the properties and dynamics of the South Florida vegetation. Specifically, the information needed to parameterize and test the vegetation model with regard to growth rates, maximum biomass, responses to water level, annual seasonality, soil type, nutrient availability, etc., across the system over a multi-year period, is not available at a sufficient level of spatial, temporal, and functional resolution to meet the needs of restoring and managing South Florida's ecosystems. Continued model development, testing, and application must proceed in concert with a coordinated field research and monitoring program that can provide long-term records of the conditions and process rates of all the major vegetation types and environments of the region. A network of long-term permanent sites for monitoring seasonal and/or annual changes in biomass, along with process-level research on plant growth and nutrient cycling is essential both for further model development and empirical input into adaptive management decisions. Addition of new vegetation types, such as mangroves and salt marsh vegetation, will require appropriate parameter sets (including groundsurface topography), plus the development of functions for responses to environmental conditions such as salinity. Higher quality vegetation maps that may be developed in the future, as well better data on growth response parameters can easily be incorporated into the existing model. Improved information on physical drivers, such as salinity and water quality, will be essential for extension of the model into the intertidal and marine areas. Because of the critical importance of water-borne nutrients in parts of the South Florida system, ATLSS will require quality-assured input data for nutrient concentrations in order to model vegetation in areas where nutrient concentrations are high or highly variable. Currently, the only potential source of such inputs is the ELM model being developed by the SFWMD. The ATLSS vegetation model could potentially use nutrient concentration output from ELM as input, and incorporate a high resolution nutrient uptake and flux component that could be tested against the coarser resolution of the ELM model. C2. Plant succession and diversity modeling 1. Purpose of component In addition to the food they provide to herbivores, other aspects of vegetation are important to the landscape of South Florida, including both functional and aesthetic properties. Many of these vegetation properties are associated with individual species that may vary in abundance across a number of different vegetation "types." Some individual species are important because they cause a problem, such as Melaleuca, Schinus, and Eichornia. Other species are important because they are attractive, unusual, or rare, such as many of the tropical hardwoods, orchids, and threatened or endangered species. A few species are important because they are so abundant that they dominate the ecosystem function and structure of the landscape over large area, such as sawgrass. Plant species differ in many ways, such as size, growth rates, physiological tolerances, chemical composition, and importance to particular species of birds, mammals, and insects. These differences must be taken into account in resource management and assessments, since they influence how a species responds to either management actions or natural variation in conditons. In addition, since individual plants may live many years and grow from small seeds to large sizes, it is often necessary to know not simply how the species as a whole is responding, but the relative number and condition of individuals of different sizes and/or ages. For example, a population composed of all mature individuals and no young individuals may indicate a population headed toward local extinction. The need for predictions about the status and distribution of particular plant species (including endemic species and exotic species), as well as population structure and overall species composition as it relates to biodiversity, requires a very different modeling approach than that used in the ATLSS seasonal vegetation dynamics model. 2. Modeling approach An approach that is well-suited to address questions about individual species status, population structure and aggregate biodiversity is individual-based modeling, which is being used in many of the higher trophic level components of ATLSS. Individual-based models of plant communities are particularly good for predicting changes in plant size and species abundance across spatial gradients in environmental conditions (e.g., water depth, nutrient, or salinity gradients) as well as for environmental conditions that change through time (e.g., secondary succession, or response to physical changes in the environment). Numerous individual-based plant models have been developed and published over the past 25 years, so it will not be necessary to develop an entirely new model concept and structure, as was the case for many of the ATLSS animal models. Adaptation of the best features of existing models, and rewriting them into the object-oriented structure of C++ compatible with the ATLSS landscape structure, will be relatively simple. Although many of the individual-based plant models that have been developed in the past were models of tree and forest dynamics, the basic conceptual approach and model structure is the same for any form of vascular plant, from herbs and grasses, through shrubs and trees. The primary differences between models applied to different vegetation life forms are in time step (shorter for herbs than for trees, to capture more rapid competitive interactions), in growth and size parameters, and in light extinction patterns (different for grasses versus broad leafed plants). Individual-based plant models are not appropriate for completely modeling areas as large as South Florida (the number of individual plants would overwhelm even the largest computers). Rather, these models are useful for predicting the detailed changes that would be expected to occur in relatively small areas under a specific set of conditions, which can then be extrapolated to other areas with the same conditions. In addition, a well-validated individual-based plant model could be used to refine the parameters and predictions of a less detailed vegetation model, such as for seasonal dynamics. The plant succession models will differ significantly from the seasonal dynamics model. Although the spatial area that any particular implementation of the succession model will directly address is much smaller that the total area covered by the seasonal dynamics model, it will be much more detailed in its representation of plant structural and process complexity, and thus can be considered to be much more "mechanistic." The plant succession models will not be based on specific vegetation types or classification, but rather will create their own vegetation types on the basis of which species survive the physical conditions that are input to the model and the modifications to those conditions that result from interactions with other plants. This continuum of species composition can be classified into vegetation types on the basis of the same criteria used by The Nature Conservance and other public and private conservation organizations to classify actual vegetation (and which were used to define the classes of the satellite-based vegetation map used by the ATLSS landscape structure). Addressing the species-specific effects of the major natural disturbances will be an important effort, which may contribute to the restoration of areas damaged by human activities as well. This model may be useful for specific management applications where interspecific interactions are important for either endangered species or exotic species. Issues of plant dispersal and plant genetics could potentially be addressed this model. Detailed plant species composition and vegetation structure predictions may be important for some ATLSS submodels that require higher spatial resolution than the Phase I model. These include small vertebrates such as reptiles, amphibians, and small birds, as well as insects. 3. Current progress No funds have yet been allocated for this modeling effort. However, the well-developed state of individual-based plant competition models, along with the experience and working models of several of the plant ecologists already involved in South Florida (e.g., Doyle 1981, 1994, Chen and Twilley, in review), suggest that these models could be developed relatively quickly. As with the seasonal dynamics model, the primary limitations to the development and application of these models are in the data currently available for parameterization and testing, and not in the conceptual or code developments aspects of the model. 4. Empirical data Fortunately, the data requirements of both types of vegetation models can be met with the same research and monitoring program. Data on vegetation biomass and growth should always be collected on a species-specific basis, so the same sampling design will meet most of the needs of both models. The individual-based vegetation models will always be subject to improvement with better parameters for individual species, and better mathematical representation of critical plant processes, such as growth responses to environmental resources, allocation responses, responses to disturbance etc. The primary environmental drivers of water level, solar radiation, and nutrients that are used for the seasonal dynamics model will also be used for the succession models. The accuracy of both types of model can be increased through more accurate input for the environmental drivers. The succession models will also be linked directly to output from the SFWMM and ELM models. In addition, vegetation changes predicted by the individual-based models could be used to adjust the aggregate vegetation parameters using in the seasonal dynamics model, or in ELM. C3. Disturbance Modeling 1. Purpose of the component Disturbances such as fires, freezes, and hurricanes are integral features of the environment of South Florida and shape the vegetative structure of the landscape. Extensive disturbances that can cause massive mortality of plants and animals, as well as substantial damage to the infrastructure of civilization, are also a characteristic feature of South Florida. The biological and economic impacts of disturbances such as hurricanes, floods, fires, and freezes are obvious to anyone at all familiar with the region. Predictive models of any component of the future condition of the region must include the potential effects of major disturbances as a factor, even though the precise details of any disturbance can never be predicted in advance. Consideration of both the positive and negative impacts of the major disturbance types is essential for any type of resource management or restoration planning. 2. Modeling approach The two components of any disturbance model are 1) the differential responses of organisms of different types and sizes, and 2) the physical disturbance process itself. The responses of organisms to disturbance is extremely context-specific, with small differences in the location, disturbance properties, and species-specific and even individual-specific properties having a major effect on the impact of the disturbance. This level of detail can be very important for predicting disturbance impacts for management purposes, such as the effect of a particular fire regime or hurricane intensity on the size distributions of different species. Individual-based models are virtually the only feasible approach for predicting disturbance impacts and recovery in a way relevant to natural resource or wildlife management. Thus, the plant succession and diversity models described above are a prerequisite for developing disturbance models for the South Florida landscape. The physical process models for each of the major disturbance types present very different challenges. While hurricanes and freezes occur independently of the condition or properties of the vegetation, fires are highly dependent on the amount and flammability of the vegetation, as well as the immediate weather conditions. Fortunately, these physical disturbance processes operate in basically the same way wherever they occur (in contrast to the varying sensitivity of different species of plants and animals). Consequently, disturbance models that have been developed for application in other regions or to other issues are likely to be applicable to South Florida vegetation with relatively little modification. Hurricanes: Hurricanes affect large areas, with fairly predictable consequences based on distance and direction from the eye, as well as local topography and surface conditions. The importance of including hurricane disturbance in models of tropical forest succession has been recognized for some time, and such individual-based models have been effective at explaining forest structure and predicting future patterns of forest succession (Doyle 1981, 1994). The recent increase in detailed studies of hurricane impacts on forests (Loope et al., 1994; Smith et al., 1994, Doyle et al., 1995a and b) as well as progress in calculating hurricane dynamics using historical meteorological data, suggests that hurricane disturbances can be readily incorporated into the ATLSS landscape structure through collaboration with ongoing research and modeling projects. Fires: Fire has long been a feared natural phenomenon that was thought to be so destructive that massive investments were made in preventing and fighting fires across the United States. Recently, there has been increasing recogition that fire is an important, and in many cases essential, process for maintaining desirable forest and landscape properties. A major revision of resource management fire policies is underway across the entire country, with the goal of using the positive effects of fires to enhance desireable ecosystem properties. Because of the ubiquity and potential destructive power of fires, a major effort has been made to understand and model the dynamics of natural wildland fires under a wide variety of conditions, particularly by the US Forest Service (e.g., Rothermel, 1972). This work has been the basis for the development of fire models for many of the vegetation types of North America, including the Florida Everglades (Wu et al., 1996). Existing fires models will be modified in collaboration with their developers and adapted for use with the ATLSS landscape and vegetation models. Freezes: The effects of occassional cold air masses that penetrate to South Florida can be quite dramatic. Depending on the temperature and duration of the freezing conditions, aboveground plant parts of increasing diameter are killed, particularly of cold-sensitive tropical species such as mangroves. Because belowground parts are rarely killed, regrowth following a freeze can be rapid, although the initial effects appear devastating. Freeze impact models have been developed for agricultural purposes, and we will modify one of the models for use with the woody and herbaceous plant communities of South Florida. As with the above two disturbance types, modeling the potential impacts of freezes is only feasible in the context of individual-based plant models that include the relevant size and sensitivity properties of different species and individuals of different sizes and degrees of exposure. 3. Current Progress No funding has yet been provided for this component of the ATLSS project. However, as discussed above, the work that has already been accomplished by other research and modeling groups working on these issues should allow rapid development of this component of ATLSS through collaboration with these programs. 4. Empirical data The effects of major disturbances on the vegetation types of South Florida are quite well documented in the publications of Everglades National Park and Big Cypress National Preserve, including those based on the recent impacts of Hurricane Andrew. These data represent an invaluable resource for parameterizing and testing the disturbance models described above. In addition, ongoing research related to fire management will continue to provide important data for refining species-level predictions. The major efforts that have already gone into developing disturbance models for South Florida, both from the perspectives of model development and model testing, will allow this component of the ATLSS project to be developed rapidly. D. Lower Trophic Level Modeling 1. Purpose of the component The primary goal of ATLSS is to assess the consequences that changes in hydrology exert upon the higher animals of the South Florida wetlands, such as, deer, panther, alligators, and wading birds. Variations in water level do affect such "charismatic megafauna" directly; however, many effects of hydroperiod upon the larger animals are more indirect. The amount of standing water, and especially the lack thereof, can have strong influence upon faunal species, such as fish, prawns, and apple snails, that serve as forage for the large predators. Furthermore, these food items depend in turn upon vegetational and microscopic resources -- collectively referred to as "lower trophic level (LTL) components." These include submerged aquatic vegetation (macrophytes and periphyton), detritus, and very small heterotrophs, e.g., insect larvae, small snails and copepods. It is of utmost importance to the success of ATLSS that how the forage and LTL elements mediate the effects of changes in water level upon higher species be included in the modeling package. Accordingly, the modeling of forage resources will be described in the next section (V.E). The focus here is upon models of how LTL compomnents respond to changing water levels. To begin with, it should be noted these LTL elements also depend upon physical factors other than water level, so that account must also be taken of how they respond to changing temperature (or season), nutrients, and (where appropriate) salinity. At the heart of the model lie the trophic interactions among the several LTL entities and the predation by higher level species modelled in other objects of ATLSS. 2. Modeling approach In comparison with ATLSS modules for the higher trophic elements, the models of the LTL populations appear far simpler. Whereas the species of megafauna are modelled on the level of the individual organism, and the size and/or age structures of the forage resources are followed closely, the populations of LTL elements are reckoned simply as lumped variables in a set of coupled, ordinary differential equations. That is, the biomass level of any LTL entity is considered not to vary over the spatial domain (ca. 500m square) of each ATLSS cell in which it appears. Trophic interactions among LTL variables are simulated on an continuous and instantaneous basis, whereas physical factors and predation from above are updated only at 5-day intervals. The LTL model consists of five "objects" or components: (1) Macrophytes, such as Utricularia, or sawgrass, (2) Periphyton, the mostly diatomaceous film of algal "slime" that coats almost all submerged vegetation, (3) Detritus, or dead plant and animal matter,(4) Mesoheterotrophs, such as water fleas, copepods and nematodes, and (5) Macroinvertebrates, which consist mostly of insect larvae. The principal trophic interactions among these LTL components consist of grazing by meso- and macroheterotrophs upon periphyton. The remaining transfers are of dead material from each LTL component to the detrital compartment (See Ulanowicz 1994 and Figure V.D.1.) Losses of LTL elements to predation by species higher in the food chain will be effected by the predator object modules. For example, a fish or prawn object module will call the LTL module once every five days. From the amounts of LTL resources presented to it, the predator routine will subtract its 5-day consumption of items and pass the remainders back to the LTL routine. The remainders then serve as initial conditions for the next 5-day interval of LTL growth, which will be simulated within the LTL module. The parameters that govern growth, predation and death all vary on a seasonal basis. The average magnitude and amplitude of seasonal change for each parameter was determined from curve- fitting techniques that were applied to seasonal data on variations of these components, as collected by Dr. Joan Browder of NOAA and Mr. William Loftus of USGS/ENP. The response of all components to changes in water level resembles a threshold "ramp- function." That is, stocks of all components are assumed to be independent of water level until the depth falls to 7cm. Beyond that depth, biomass is transferred from all living stocks to the detrital pool in proportion to how far the level has dropped below the 7cm threshold. For example, by the time the water level falls to 1.75cm, 75% of the biomass in each stock will have died off and become detritus. Such "die-off" continues until only 5% of the original stock remains, and that residual is maintained for the duration of drought as "seed- stock" for the subsequent episode of reflooding. Models for the LTL communities in four different habitat types have been constructed. Separate models exist for gramminoid wetlands with either short or long- hydroperiods. The models for both gramminoid biotypes are identical and differ only in the values of their dynamical parameters. Models also have been created to represent the ecosystems of the forested (cypress) wetlands and the mangrove estuaries (Ulanowicz 1995a.) These latter models have essentially the same components as in the gramminoid models, except that autotrophic production (of macrophytes and periphyton) is held very low due to the dense tree canopies that shade these wetlands. They are driven instead via a seasonally-varying input of detrital "litterfall", which is decomposed and processed by the heterotrophic compartments. The responses of ecosystem processes to dissolved nutrients (phosphorus), salinity (in the mangroves), and temperature have been built into the models. These physical variables are to be generated by other object modules of ATLSS. (See Section B above.) 3. Current progress Models of the LTL communities in all four habitats are now operational. The modules were coded originally in FORTRAN and subsequently translated into C++ by the University of Tennessee ATLSS contingent. The gramminoid models already have been coupled with the fish and other forage resource models, and the combination yields plausible outputs. The models for the cypress wetland and mangrove biotopes have been written in FORTRAN, "calibrated", and currently are being translated into C++. They soon will be linked to the forage resource models now being created for those habitats. Initial results indicate that all LTL models behave in stable fashion (as designed.) The recovery of the gramminoid ecosystems from drydown appears to mimic reality well, when compared with the little independent data that are available (Figure V.D.2.) 4. Empirical data All LTL models have been calibrated using data that antedate the ATLSS project. In some cases, data available for calibration remain sparse to nonexistent. At some date prior to incorporating these particular modules into the ATLSS shell, these components should be recalibrated using data to be acquired in the interim. Special attention is drawn to the lack of almost any data pertaining to the faunal communities of the mangrove estuaries. 5. Time line for completion of work The LTL modeling task is essentially complete. Recalibration will be done as new data appear. E. Fish and Aquatic Macroinvertebrate Modeling 1. Purpose of the component Fish biomass constitutes a major energy resource for the wading bird communities and other top-level predators of the Everglades and Big Cypress ecosystems of southern Florida. Fish communities are exposed to annual fluctuations in water level (Loftus and Kushlan 1987). Fish populations expand during the period of flooding, while the annual drydown concentrates many of the fishes in shallow waterbodies, where they are easily available to predators. The purpose of the fish computer simulation model in ATLSS is to predict the fish population responses to this seasonal pattern of water levels in all the spatial cells across the landscape, and thereby the pattern of prey availability for wading birds. Long-term hydrology can have important effects on this pattern, so modeling is important to predict changes that various restoration water management scenarios would produce. For example, droughts can produce massive losses of fish numbers, and threaten the residual "seed" populations of fishes needed to repopulate the marshes in the next wet season. One of the factors helping to mitigate the severity of the effects of drought are the quasi- permanent waterbodies that exist in many areas, such as creek channels, alligator holes, solution holes, and other depressions, that are refugia for fish during these "drydowns". However, the small fish are exposed to high predation, mostly from larger fishes in the larger refugia. Prediction of speed of recovery of the fish population in a cell following a drought is one of the main goals of the model. 2. Modeling approach The model describes the seasonal dynamics of the community of small fishes (e.g., mosquitofish and killifishes) as water levels change through the year for each 500 x 500 meter cell in the Everglades/Big Cypress region. Each cell is modeled as having a statistical distribution of depressions that can serve as refugia for fish if the cell dries down during a year. There are two fish functional groups in the model; small fishes, which are a primary prey of wading birds, and large fishes, which are predators on the small fishes. The fish in each of these two functional types are modeled as 1-month age-classes. Growth in age, growth in size, and mortality occurs on these 5-day time steps, but an increase to the next age-class occurs only on the first time step within a given month. For a given age the size of the fish is calculated from a von Bertalanffy equation and weight in grams dry weight is given by a weight-length relationship. The changes in water level are modeled, as are the interactions of the fishes with their resource base of periphyton, macrophytes, mesoinvertebrates, macroinvertebrates, and detritus (see Figure V.E.1 and see section V.D for a description of the resource base of the fish). The simulation also includes the interaction of large and small fishes to address the question of whether the larger fishes may be a major regulating factor for the small fishes. Movement of fish into and out of ponds and depressions is modeled as a function of changing water levels. Intercell movement can also occur. Reproduction and probability of mortality of fish are modeled as dependent on the age of the fish, the season of the year, and water depth. 3. Current progress The model has been used to demonstrate changes in fish population and biomass numbers over multi-year periods under a variety of hydrologic conditions. The response of the model to the different scenarios of water depth (or hydroperiods) such as shown in Figure V.E.2, is especially important. Figure V.E.3 shows the model predictions of number densities (numbers per square meter, averaged over the entire 25-ha spatial cell) of the small type fish greater than 30 days of age throughout the year, for each of the scenarios, corresponding to hydroperiod scenarios a, b, c, d, e, and f of Figure V.E.2. The population density for hydroperiods a and b reach a maximum of only slightly more than 1 fish/m2 and slightly more than 2 fish/m2, respectively. Apparently, a flooded period of only seven or eight months is not sufficient for large population growth. A strong periodicity is shown in the two intermediate scenarios, c and d, with maximum population densities toward the end of the flooded period of more than 10 fish/m2. When the cell is flooded continuously, the populations of the small type fish fluctuate between about 13 and 20 fish/m2. The reason for the seasonal changes in this case is that the lower trophic levels are undergoing seasonal variations. The model makes three further predictions that appear to us to be fairly robust in the model system. First, there appears to be a threshold of about nine or so months in the length of the hydroperiod. If the hydroperiod is less than this, the small- fish-type population stays small, no more than a few fish per square meter. For longer hydroperiods, the fish population can reach levels of 10 to 20 fish per square meter (averaged over the whole spatial cell) by the end of the hydroperiod, roughly what the population would be under continuously flooding (but during the drydown these fish will be concentrated by the receding waters to much higher local densities). The second prediction is that the large, piscivorous fish do not have a significant impact on small fish populations in the marsh, even though they do in the pond. The third prediction is that the repopulation of the marsh by small fishes following a drydown, even a prolonged drydown, occurs very rapidly, within a little less than a year (though this year could be a critical one for wading birds dependent on fish prey). If the above predictions are, indeed, robust and general ones, then the prediction of the impact of fluctuating water levels on the numbers of small fishes, aggregated over species, in a marsh will be made much more simple. The ATLSS single-cell lower trophic level (LTL) and fish models have been combined with the general landscape structure code to produce an integrated model that may be run across any suitable portion of the South Florida landscape. This allows for the spatially-explicit simulation of LTL and Fish functional group responses to hydrology by coupling together many spatial cells, within each of which the LTL and Fish cell models are run. For the LTL model, cells are treated as independent, in that there is no flux of LTL components across cells. This does not imply that there are not strong spatial correlations in LTL components however, since hydrology is a strong driving variable in the LTL model and there are spatial correlations in hydrology. In the Fish components, while the underlying dynamic model of the functional groups is the same for each cell of the landscape, movement of fish does occur between cells as well as between locations within cells (e.g. the proportion within a cell in ponds, marsh and solution holes changes as a function of hydrology). The model has been coded in object-oriented form in C++. There are significant memory requirements in order to track the size (or age) structure of functional groups in marsh, pond and solution holes across a landscape. Thus, although the model may be run across any size grid, to date simulations have been limited to regions of about 50x50 grid cells of 500m each (thus covering a spatial extent of about 625 square kilometers). The model currently requires about 1 hour of cpu time per year of simulation on an Ultra Sparc 1 with 256 MB RAM and a 167MHz processor for this size grid. A variety of issues arise in the landscape model that are required to extend the single cell models. These include: 1) The solution holes distribution must be randomized across the spatial extent of the landscape. The model assumes 10 depth classes of solution holes. For each cell the model randomly assigns (according to a distribution chosen a priori that could be estimated from appropriate field data) the total fraction of the area of the cell that is occupied by each solution hole depth class. 2) The ponds distribution must be randomized across the spatial extent of the landscape. The model randomly assigns (according again to an a priori distribution that could be estimated from data) pond area for each cell of the landscape, and assumes a constant water level for all the ponds in the landscape for all the time steps. 3) The initial density of LTL components and fish functional group size structure must be assigned across the landscape. These may be assigned to be equal across cells, or may be read in from a previously saved output file from a simulation. 4) The landscape model allows movement of fish functional group density between cells. Fish are assumed to only move between marsh areas of adjacent cells, so there is no movement from the pond of one cell to any area within an adjacent cell. Fish movement is based on a combination of water level differences and functional group density differences between neighboring cells, with higher fluxes at higher values of these differences. Several alternative assumptions may be made about movement at the edge of the simulated area - presently it is assumed that there is no movement across this boundary. 4. Empirical data This modeling work is being done in collaboration with empirical scientists studying fish in the Everglades/Big Cypress pregoin (William Loftus, Joel Trexler, Jerry Lorenz), who helped in estimating parameter values for the model. Data on the seasonal changes of fish number and biomass densities are available for some areas of Everglades National Park (Loftus and Eklund 1994), and these data can be used for calibrating and testing the model. Conveniently, these authors have divided the fishes they surveyed into the two types of large and small fishes that are used in the model. Loftus and Eklund (1994) sampled fish populations in the Everglades over a period of several years, and found mean annual densities of the small-fish type to range from 15.5 to 17.1 fish/m2 in the early part of the study and 30.2 to 34.5 in the later part of the study. This compares with values of about 5 to 15 small- type fish/m2 in the model for the longer hydroperiod simulations. Loftus and Eklund estimated the mean annual density of large-type fishes at about 0.012 fish/m2. The model densities ranged from about 0.005 to 0.013 for the longer hydroperiod simulations. The empirical data of Loftus and Eklund (1994) showed strong seasonal oscillations, with the highest small-fish densities within a year approaching four or five times the lowest densities. This contrasts with the oscillations over a year in the model, where dry season densities dropped to a very small fraction of the peak densities. It may be that the model drydown periods were more severe, or at least more continuous, than those occurring in the Everglades areas sampled by Loftus and Eklund. The model simulations are in general comparable with data from the field in the Everglades (Loftus and Eklund 1994), even though the only part of the model that has been calibrated to data was that for the lower trophic levels. This indicates that the model is probably correctly transferring the energy into the two fish functional groups. 5. Timeline for completion of work Currently the fish model has been parameterized only for the freshwater areas of the Everglades/Big Cypress. Because fish production in the coastal mangrove areas is probably critically important for wading birds at certain times, the model must be extended to these areas. One of the main hindrances to doing this is the absence of a hydrologic model for the coastal estuaries. As soon as a model for major parts of the coastal estuary zone is available, the current model can be extended. F. Reptile and Amphibian Assemblage Modeling 1. Purpose of the component The herpetological assemblage of the Everglades and Big Cypress is a understudied but vital component of these systems. Populations of reptiles and amphibians can be large (G. H. Dalrymple, unpublished data), and a keystone species in the region, the American alligator (Alligator mississippiensis), is a member of this assemblage. The herpetological assemblage, by itself, may constitute a major pathway for energy flow and its interaction with other groups (e.g., insects and fishes) may have large impacts on the overall structure of energy flow in these aquatic ecosystems. In particular, amphibians and reptiles are a primary prey item of alligators (B. Barr, unpublished data) and may play a vital role in the population dynamics of this species. Hydroperiod and periodic drydowns alter the structure of the herpetological assemblage in the freshwater regions of the Everglades and Big Cypress (Duellman and Schwartz 1958, Dalrymple 1988, Dalrymple et al 1991a, b). Some individuals find solution holes, depressions, subterranean habitats or other microsites during drydowns and remain active, while other individuals leave dry areas (Dalrymple 1988, Dodd 1993). During severe droughts, specific members of the assemblage aestivate (Hansen 1958, Gibbons et. al 1983, Etheridge 1990). Furthermore, egg and larval stages of the amphibians require varying durations of standing water to successfully survive and mature into adults. Given the potential impacts reptiles and amphibians can have in the freshwater aquatic systems of southern Florida, and their observed responses to hydrology, our goal is to develop a computer simulation that will allow us to predict population level responses in the herpetological assemblage to fluctuations in water level. 2. The modeling approach The modeling of the herpetological assemblage has a both a descriptive and a predictive phase. Descriptive phase. Because so little is known about the herpetological assemblage, we are developing a food web and description of energy flow using extensive field data provided by Dr. George Dalrymple and linear programming techniques. Linear programming is an optimization routine and is excellent for estimating energy fluxes in food webs. In our linear programs, we assume a steady state in each functional group (i.e., energy in equals energy out). Furthermore, we constrain the solution set using empirically derived ranges of standing crop biomass, diet, and respiration for each herpetological functional group, thus assuring the results fall within biologically reasonable bounds. Constraints on non-herpetological components, as well as alligators have currently been left broad, because data are lacking. Thus, the results indicate the possible energy flows and biomasses, given the internal constraints found just within the herpetological assemblage. To develop the food webs presented here, we minimized the differences between the empirically derived estimates of energy fluxes, and those possible given the constraints on the system. We have three general goals for the descriptive phase. First, we will describe the pattern of energy flow within the herpetological food web. Second, we will determine how the structure of the web and the energy fluxes change with hydroperiod and habitat type. Third, we will determine the energetic interaction between the herpetological assemblage and other assemblages in the Everglades as well as estimate the potential flow of energy to alligators. Predictive phase. Using the results from phase 1 to generate the general ranges in biomasses as well as the potentially key members of the assemblage, a simulation will be created to predict the impacts of hydrology on the herpetological assemblage. The structure of the model will be based on additional empirically derived estimates of demographic parameters of the functional groups supplied by Dr. George Dalrymple. The model will function with 500 m x 500 m spatial cells, and time steps of 5 days and include the impacts of spatial heterogeneity within and across cells, responses to changes in water level (including dry-downs), and biological interactions within the herpetological assemblage (Figure V.F.1). The model will interact with the fish, crocodilian, and lower trophic level models. 3. Current progress We are currently analyzing the data necessary to complete phase 1; the description of energy flow through the food web. We have constructed food webs and run linear programs for three generalized habitat types; marsh, prairie and upland. The structure of the food web changes, as interactions between members of the herpetological assemblage are more common in the marsh than in the other habitat types, which are similar in structure (Figure V.F.2). Biomass of the more aquatic functional groups generally decreases in the upland habitats relative to the prairie and marsh. Lizards make up the largest biomass of any functional group in the prairie and upland habitats, whereas snakes and medium frogs have the highest biomass in the marsh. Despite these differences across habitats, the total amount energy flowing through the assemblage is similar across all habitat types though there is a slight trend for more energy flow in wetter habitats (marsh = 27,566 g/ha; prairie = 25,089 g/ha, upland = 22,105 g/ha). A common trend in all three habitat types is that estimated energy fluxes within the herpetological assemblage can be strongly influenced by changes in the constraints or estimated biomasses of non-herpetological functional groups. Thus, there is indeed a need to link the disparate trophic groups (i.e., birds, alligators, fishes, reptiles and amphibians) under one general simulation. Work on phase 2, the predictive simulation, has not formally begun. However, we have developed a preliminary conceptual basis for the model (described above, Figure V.F.1). 4. Empirical data Both phases of the modeling are based on a large amount of empirical data. Dr. George Dalrymple is supplying the data based on long-term censuses, analyses of large numbers of preserved specimens, and literature reviews. He has estimated upper and lower bounds, as well as average and median estimates of biomass, consumption, respiration and diets for each functional group in marsh, prairie and upland habitat types. Dr. Dalrymple will also supply data on reproduction, turnover, survival, and other demographic information for each functional group for the purposes of parameterizing the predictive simulation in phase 2. 5. Time line to completion Phase 1 will be completed by May of 1997 and will be presented in at least two deliverables in the form of journal articles. Once these results are complete, we will have nearly 18 months to develop and test the simulation, thus completing phase 2 near the end of 1998. G. Crocodilian Modeling 1. Purpose of the component The American alligator (Alligator mississippiensis) is a keystone species in the Everglades and Big Cypress Swamp, as defined by its role as a top-level carnivore and architect within these systems and its influence on the structure, distribution and abundance of native plant and animal communities. Although the alligator is a large, mobile carnivore that represents a versatile and selectively opportunistic predator, it depends upon stable resources within its local environment, e.g. presence of surface water and related prey resources. During a long life span of nearly continuous growth, an individual feeds on a variety of prey species and sizes that vary according to the alligator's size. As alligator populations consist of overlapping size classes, they are consumers at all trophic levels. The alligator also modifies its environment through construction and maintenance of "alligator holes" to regulate its body temperature. These holes also serve as critical dry-season refugia for a variety of other aquatic animals upon which wading birds and other predators feed. Because loss of such a keystone species can lead to drastic re-ordering of some parts of the floral and faunal community, the recent severe declines in the abundance of the American alligator in South Florida are of concern. These declines are attributed to alterations in the natural hydrologic regime of the region. 2. The modeling approach The alligator model simulates alligator responses to varying hydrologic regimes in a variety of freshwater, local environments. These local environments include short- and long-hydroperiod wetlands. The model consists of two parts. One part simulates the life stages of individual adult and subadult alligators and, in particular, nesting female alligators and their reproductive performance. The second part simulates the life stages of hatchlings and juveniles in typical nest areas. Hatchlings and juveniles can be modeled as cohorts because small alligators suffer from a high mortality. The module tracks their daily growth and survival based upon a set of environmental factors and stochastic survival probabilities. Adults and sub-adults, however, are modeled as individuals. The basic time step to model individual alligators is a day. Food intake during each day is monitored, as well as the amount of energy spent for activities during the day (and the night). An average assimilation coefficient from feeding experiments is used to determine the daily energy budget. These budgets are accumulated over a longer time period (e.g. a month) and are then converted into growth or regression, depending on whether the accumulated energy budget is positive or negative. The model simulated individual alligators, in particular females at their nesting areas (a schematic is shown in Figure V.G.1). Nest areas are categorized according to some basic characteristics, such as location, size and depth of pond, food availability, and suitability for building a nest. An alligator pond, or the presence of surface water provides a means for the alligator to regulate its body temperature. A simple limnological model is used to calculate approximate water temperatures in relation to weather conditions. The availability of food is simulated using output from the models for fish and aquatic macroinvertebrates. Food intake also depend on an alligator's body temperature and its current nutritional status. The model assumes that alligators do not eat every day, but capture some larger prey items once every several days, and a few smaller prey items in between. The smaller prey items are treated on an average basis, whereas the larger prey captured are treated as separate and distinct food items. The model assumes that these larger prey are 'offered' to the individual alligator, which accepts or rejects them, depending on its state variables, e.g. satisfaction/hunger. Given the food intake and the composition of the diet, the model then calculates the energy budget of the animal using food conversion rates and maintenance energy requirements determined from feeding experiments. 3. Empirical data The relationships between the amount and composition of the alligator's diet and its growth and reproductive performance have been determined for captive animals in experimental studies as well as for animals in the wild. Because the model for adults and sub- adults is based on individuals, it provides a variety of outputs that may be directly compared to the field data on individuals, in addition to comparisons of data on nesting success under differing environmental conditions. The model does not include detailed individual-based components for the hatchling and juvenile stages. 4. Timeline to completion of work The alligator model is currently under construction. The model has been coded, but not yet parametrized and calibrated. There are few data available on characteristic properties of alligator ponds, e.g. size and depth distribution, temperature profiles, faunal characteristics. The majority of the data available were obtained for ponds in the Big Cypress swamp (e.g. Kushlan and Hurt, 1979). Although it is probably not appropriate to generalize these data over the entire Everglades/Big Cypress ecosystem, they are nevertheless used to parametrize parts of the model. However, studies to obtain data on limnological characteristics of alligator ponds are currently underway and will be used in a reparametrization of the model. In addition, telemetry studies to determine the movement of alligators, in particular during the dry season, were begun in the fall of 1996. These studies will provide the data necessary for a full parametrization of the alligator model. H. Wading Bird Assemblage Modeling 1. Purpose of the component As top-level carnivores, wading birds are important components in the aquatic food web. Wading birds are highly mobile animals that influence the structure and dynamics of their freshwater and estuarine prey communities, and transport nutrients from their feeding sites to their colonies. Wading birds are also highly dependent on patchy resources with a high prey density. Since water management regulation ponding of overland flows in northern reaches of the Everglades catchment area and severe overdrainage of the southern reaches downstream of these impoundments has occurred. Taken in combination, these have altered the locations or spatial arrangement, as well as reduced the areal extent of seasonal foraging habitats for wading birds. Declines in wading bird populations have occurred in all feeding guilds concurrent with these landscape changes. Decreasing numbers attempt to nest each year, particularly at traditional colony sites within the downstream reaches of the southern Everglades. The focus of present modeling efforts for wading birds in ATLSS is to develop simulation models that allow one to investigate the dynamics of colonies and the nesting success of wading birds in relation to different hydrologic scenarios and the resulting spatial and temporal distribution of their prey. The species selected for modeling in ATLSS are tactile (wood stork, white ibis) and visually (great egret, great blue heron) feeding species and represent the majority of the wading bird population in the Everglades and Big Cypress ecosystem of southern Florida. 2. The modeling approach Because wading birds depend on patchy resources and can travel long distances during their foraging flights, aggregated model are inappropriate to describe the dynamics of colonies in a heterogeneous and rapidly changing environment as the Everglades/Big Cypress ecosystem. Although most activities of wading birds, such as whether to forage solitarily or in flocks, occur on small time scales (minutes/hours), they nevertheless have a large influence on the foraging success of individual birds as well as the depletion of local prey resources. This necessitates the use of an individual-oriented modeling approach in which individual birds are described by a set of species-specific rules that govern their behavioral activities. Contrary to other ATLSS models, the wading bird models do not operate on fixed time scales, because the various activities of individual birds are of different duration. Instead, the models use an event-driven simulation approach where each bird sets its own individual time-scales that depend on the duration of its current activities. The model consists of species-specific sets of rules for the behavior and the energetics of nesting adults, as well as rules for the energetics and growth of their nestlings. The model makes extensive use of the output of the landscape, hydrology and fish models within ATLSS. These data are then used by the individual birds; for example, to determine where they can forage and how successful they are at these sites. The wading bird model thus operates on the same spatial grid but not on the same time scale as other ATLSS models. 3. Current progress A first version simulating a single species nesting colony (wood storks) has been completed (Wolff, 1994) and was already applied to test various hypotheses about the influence of hydrological patterns on the timing of nesting and the subsequent nesting success (Fleming et al. 1994). We modeled a single colony at a traditional colony site and ran different scenarios in which we decreased the areal extent of peripheral wetlands that represent early shallow water foraging habitat for wading birds (Figure V.H.1). Colonies are formed if the birds find sufficient food in high density patches which then triggers a nesting response and provides female birds with the additional energy required for egg production. In our simulation, the larger the reduction in peripheral wetlands area, the later was colony formation (Figure V.H.2). In addition, the number of nesting attempts decreased for larger delays in colony formation (Figure V.H.3). Both of these patterns can be attributed to a shortage of high density food patches during the early dry season. There was enough food available for the birds to meet their basic energy requirements, but not enough to trigger nesting and energy required for egg production. The number of fledglings in a colony can serve as an indicator for the possible number of new recruits into the breeding population. This number decreased sharply with the reduction in peripheral wetlands area and was always well below the possible production if every nesting attempt had produced three fledglings (Figure V.H.4). If the parents could not provide their chicks with sufficient food, brood reduction occurred. If the adult birds could not meet their own energy requirements they subsequently deserted the colony as well. These general patterns, delay in colony formation, reduction in the number of nesting attempts, and lower numbers of fledglings, were quite robust for all scenarios in which the areal extent of peripheral wetlands was decreased. The current version simulates wading birds in several, mixed- species colonies and has been largely completed for the freshwater areas of the Everglades/Big Cypress ecosystem. The rules describing the behavioral activities and the energetics of individual birds for wading birds other than wood storks were parametrized as far as existing data permitted. Fine tuning of the model as well as interfacing with the landscape, hydrology and fish models will proceed as soon as high-level code for these models becomes available. These models were specifically constructed to evaluate the relation between hydrology and nesting colony success. Due to this focus, the time span in the model is one breeding season, and therefore a variety of assumptions must be made to evaluate colony success over several years. In particular, to date there is no explicit model component dealing with survival outside of the nesting season because data about interannual survival of wading birds are virtually nonexistent. In addition, the model does not yet include any adult or nestling mortality that originates from disease or predation. 4. Empirical data There exists an extensive data base of published and unpublished data on wading birds. Nevertheless, some critical data (for all species!) are still missing from the literature. The most prominent missing information is foraging success in relation to prey availability. However, field studies are currently under way to fill most of these data gaps so that individual rules can be more accurately formulated and parametrized. 5. Timeline to completion of work The mangrove areas are assumed to be critically important during the early part of the breeding season and the wading bird model must therefore be extended to these areas as well. An extension to these areas will be possible as soon as the corresponding models for the hydrology and the prey resources are available for these areas. I. Cape Sable Seaside Sparrow Population Modeling 1. Purpose of the component The Cape Sable seaside sparrow (Ammodramus maritima mirabilis) is an ecologically isolated subspecies of the seaside sparrow (Beecher 1955, Funderburg and Quay 1983, Post and Greenlaw 1994). 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 breeds in marl prairies either side of Shark River Slough (Figure V.I.1). Marl prairies are typified by dense mixed stands of graminoid species usually below 1m in height, naturally inundated by freshwater for two to four months annually. The potential of such habitat, for sparrow breeding, is dependent upon regimes of fire, hydrology and catastrophic events (hurricane and frost). Recent declines in the sparrow population across its entire range, especially the western portion, highlight the need for an effective ecological management strategy. The remaining core of the population occupies approximately 60-70 sq.km. in the area adjacent to the south east of Mahogany hammock (Area B in Figure V.I.1). This sub-population currently represents 73% of the total population (1996 estimate), and because of the spatial restriction it is seriously at risk to the effects of hurricane or wildfire. Changes to the hydrology of the southern Everglades may also increase the threat of extinction. Increased hydroperiods affect the sparrow in two ways : a) directly shorten the potential breeding season, and b) indirectly by causing changes in the vegetation. Recent studies (Nott et al. in press) show that wetter conditions cause typically short-hydroperiod vegetation (Muhlenbergia) to become dominated by sawgrass (Cladium jamaicense) and spikerush (Eleocharis spp.). This kind of habitat is less suitable for breeding purposes but remains available for foraging. The main objective of the model is to investigate the effects of fire and hydrology regimes upon various measurements of the sparrow population. These include lifetime reproductive success of individuals, movement patterns and spatial distributions of the population, fluctuations in the size and structure of the population and local densities. The model adopts an individual- based spatially explicit approach. Such an approach is preferable for modeling small populations that are dependent upon limited resources (Uchmanski and Grimm 1996). In this model, individual sparrows in the population explore a variable landscape consisting of 100m x 100m cells. This resolution is ecologically appropriate, considering the minimum territory size, the scale of many landscape features, and the length of typical 'neighborhood' flights. 2. The modeling approach A set of state variables describes each individual in the population. They differ from one another and respond to both the landscape, and to other individuals in the population.The minimum set required to model the observed complexity of the sparrow's behavior include spatial location, age, sex, weight, reproductive status, fitness and associations with others. The model updates each individual's status daily according to movement and behavior rules. The spine of the model is a simple flow of decisions and actions that affect individuals. A flow diagram for the breeding season portion of the model is shown in Figure V.I.2. At each step the model updates the breeding status and tracks associations between individuals. Each individual (in random order) moves around the landscape according to a simple set of movement rules. These are dependent upon the time of year, water levels, the status of the individual, the attributes of the cells it encounters and the attributes of neighboring cells. Important landscape attributes include elevation, the vegetation classification, and fire history. Some types of cells represent 'reflective' barriers to movement (pine forest, hammock and open water), other 'transparent' cell types allow movement but do not represent breeding habitat (sawgrass/spikerush marsh). Temporal and spatial patterns in water levels represent the main environmental driving force behind the model. A set of behavioral rules mimics observed interactions between individuals. The outcome probability of