The Florida Everglades is considered to be one of the most threatened ecosystems in the nation, due in part to decades of intense, adverse water management impacts [DO94]. Restoration alternatives are currently being developed to restore natural water flows to the area. Hence, there is a need to predict and compare the effects of alternative hydrologic restoration scenarios on the Everglades wildlife. Future projections of these effects can be realistically accomplished only by computer modeling [FDG+94].
The SIMPDEL ( Spatially-Explicit Individual-Based Simulation Model of Florida Panther and White-Tailed Deer in the Everglades and Big Cypress Landscapes) model was jointly developed by a group of modelers and biologists at the Institute for Environmental Modeling, University of Tennessee [CGF+95]. SIMPDEL is a member of the ATLSS family of models ( Across- Trophic- Level System Simulation for the Everglades and Big Cypress Swamp) under development at a variety of institutions with support from the National Biological Service. Upon completion, the SIMPDEL model will be used to analyze the effects of alternative water management scenarios on the long-term populations of white-tailed deer and Florida panther in the Everglades. The model tracks the growth, movement, foraging and reproduction of individual deer and panther across South Florida as affected by the underlying landscape, the spatial dynamics of hydrology, and the interactions between individuals. Typical model simulations include 20,000 deer and 50 panthers, with the population sizes varying throughout the simulation due to stochastic mortality and fecundity.
Execution time of the SIMPDEL model increases as the number of deer increases and as the abundance of forage decreases. For example, execution times ranging from to 39 hours were recorded for increasing population sizes. Hence, a primary goal of this work is to reduce substantially execution time of the sequential SIMPDEL model via parallel programming while producing population trends similar to those produced by the current sequential model. An additional design goal is to construct the parallel model in such a way as to ensure that the same results will be achieved regardless of the number of processors used, in order to simplify code conversion for a future implementation of the parallel model designed for a distributed network of workstations using Parallel Virtual Machine (PVM) [Gei+94]. As the panther component of the sequential SIMPDEL requires minimal CPU time, the focus of this study is the parallelization of the deer portion of the model.
The parallel computing environment used in this research is described in Section 1.1. Section 2 introduces the sequential SIMPDEL model and describes the Everglades study area. Section 3 discusses the parallel SIMPDEL model, and Section 4 presents results and speed improvements of the parallel model over the sequential model. Finally, a summary and future work considerations are provided in Section 5.