Everkite has been developed as an individual-based, spatially explicit model to project population changes under diverse hydrologic scenarios. As basic information for any predictive model, good empirical studies are required. A large number of empirical studies have been completed on the Florida snail kite and provide the correlative relationships between specific aspects of its life-history and behavior with the hydrology of the system. These relationships form the building blocks of the hydrologically driven population-dynamics model.
For clarification we outline how Everkite works here. The model is designed to predict temporal and spatial patterns of snail kite numbers under various hydrolog-ical scenarios. It does so by following the lives of all individual kites in the model population on a weekly time step. The spatial structure of the model consists of a network of wetlands, each representing one of the major wetlands inhabited by snail kites in southern and central Florida. Some of the important characteristics of the Everkite model are as follows:
1. Each snail kite is individually modelled. Attached to each individual are state variables, representing sex, age, spatial location, and reproductive status.
2. In the model, snail kites nest across an array of 14 disjunct wetlands of southern and central Florida that are linked to GIS map layers of the hydrological model. The habitat quality within each of these areas is assumed to be relatively uniform. There is also one aggregated "peripheral" habitat (making fifteen total wetlands in the model), representing areas of inferior quality that snail kites may use for foraging, though not for nesting.
3. Each individual wetland is allowed to undergo changes in spatially averaged water level, which affect apple snail density. Foraging activities are not modeled explicitly, but instead water levels are directly translated into a habitat quality parameter, representing the foraging conditions. Decreasing conditions in a given wetland result in a tendency for the kites to move away and, for those that stay, a reduced reproductive rate and an increased mortality rate. Carrying capacities were not assigned to the wetlands, so the model does not attempt to examine the population dynamics close to an upper bound in the population size.
4. Although Everkite was built as an individual-based stochastic (Monte Carlo) model, a deterministic matrix model version is used here. Transition values for activities, such as the starting of breeding or the movements of the kites from any given site to another, are described by matrix elements representing the fractions of the population that start breeding or make the move to particular sites. There is a general tendency for kites to move to nearby wetlands rather than to very distant ones. Everkite produces the number of kites in each wetland in each week for the duration of the hydrological scenario. A good way to characterize a given scenario in a single number is to calculate the long-term yearly population growth rate (X) for the whole period covered by the scenario. The whole period used in the simulation was determined by the record of hydrological data, which was 31 years.
Two versions of Everkite have been developed, which differ in the level of detail in which the kites were described. A simple spatially-explicit individualbased model was developed to perform a sensitivity analysis of a system with the properties of the kite population in southern and central Florida. This model is described in details in Mooij et al. (2002). A more detailed version of the model was developed to incorporate most known details about kite population dynamics and behavior, based on extensive field studies. Both versions of the model handle space by distinguishing the fifteen major wetlands that constitute the fourteen main nesting habitats of the snail kite, plus a peripheral non-nesting habitat component, and both models were run with a time step of one week. They also both cover four biological processes: ageing, reproduction, movement, and mortality. They differ, however, in the way these processes are described. The detailed model can be run in a stochastic, individual-by-individual mode (Everkite version 3.01) and in a deterministic, fine scale matrix model mode (Everkite version 4.01). We used the deterministic version of the complex, parameter-rich model for this chapter (Everkite 4.01). It is totally data-driven. The hydrological scenarios were entered as forcing functions, based on detailed hydrological models for the various parts of southern and central Florida. Descriptions of the weekly dynamics of kites were also entered in tabular format.
The approach followed is best explained by making a comparison with the ANOVA approach in statistics. For each of the six major events (nest initiation, nest failure, nest desertion, nest success, movement and death) that a kite could perform in a given week we determined, using empirical data, which factors significantly modified the probability of a specific event happening. Here probability is interpreted as the percent of the population to which the event happens. For example, there are several modifiers that influence the probability of nest initiation, including a seasonal effect, a nesting attempt effect, a wetland effect, a habitat quality effect and a crowding effect. For each of these independent variables, an appropriate number of discrete states were defined (i.e., 12 months, five life stages, five environmental states). Then, for each state, a multiplicative parameter was entered in the model that represented the specific modifying effect of that state to the overall probability. The overall probability of a nest initiation was then calculated by multiplying the basic probability with all the relevant state specific multipliers. This approach produced a very versatile model. An inhibitory effect of a specific state on a specific process (e.g., drought on nest initiation) could easily be achieved by entering a value of zero for the modifier, which then automatically resulted in an overall probability of zero of that event happening.
The empirical information needed for this parameter-rich model comes from a detailed study of the demography and movements of kites (Bennetts and Kitchens, 1997). Fecundity and fledgling survival were estimated through nest studies. Radio telemetry of 282 birds and mark-resighting of banded snail kites were used to estimate survival, to evaluate the influences of environmental conditions (e.g., hydrology) on survival, to evaluate the movement patterns of snail kites in Florida, and to relate these movement patterns to environmental conditions.
Model input for the four scenarios included historical weekly water levels per wetland for the period January 1,1965, to December 31,1995. Water levels determine the environmental state of each of the 15 wetlands (Beissinger, 1995). The primary water level states are classified as high, low, drought, and lag (where lag refers to a year following a drought). After prolonged inundation the habitat status of vegetation cover, which affects foraging success will change from suitable habitat to one of two degraded states: moderately degraded or severely degraded. Degraded habitats provide less useful foraging habitat and are less suitable for reproduction. The time lags of degradation are much longer than those of the primary states because they represent a much slower process.
The within-year variation in water levels and environmental states interacts with the breeding cycle of the kite. The breeding cycle is implemented in Everkite by varying the nest initiation rate between months. Highest initiation rates are in the period February through May whereas nest initiation rate is assumed to be zero in September and October.
Was this article helpful?