A parallel model of the hydrology, vegetation, and deer components of the current sequential SIMPDEL model has been presented. Results were very similar in both models and excellent speed improvements were obtained. In addition to promising results, the parallel model proved worthwhile in verifying the outputs produced by the sequential model. The original SIMPDEL model contained an extremely large number of bugs and was restructured and improved during parallel model development. Thus the parallel model provided a means for result comparison which aided in locating bugs in the sequential model that may not otherwise have been noticed. The results also provide strong evidence that grid-based parallelization schemes can be highly effective for individual-based ecological models with explicit spatial structure. Earlier concerns on parallelization for individual-based ecological models involving movement argued that distributing individuals over processors would be more efficient than dividing space over processors [Hae92]. A parallel implementation of SIMPDEL in which processors handle specified individuals was not developed since initial attempts indicated that it would be extremely difficult and inefficient to do so, in part due to the message passing required to update the underlying forage maps.
Future work on the parallel SIMPDEL model consists of porting the existing parallel code to a network of workstations using PVM [Gei+94]. Since the landscape is initially partitioned dynamically according to area, the number of processors used can be scaled up or down, with little or no change to the code structure beyond the replacement of CMMD function calls with the corresponding function calls from the PVM library. In addition, the panther component will be parallelized and incorporated into the model. Future plans for the sequential model include additional water level inputs, map data layers, and catastrophic events, which may also be included into the parallel SIMPDEL model.