A Parallel Structured Ecological Model for High End Shared Memory Computers

This paper presents a new approach to parallelize spatially-explicit structured ecological models. Previous investigations have mainly focused on the use of spatial decomposition for parallelization of these models. Here, we exploit the partitioning of species age structures (or layers) as part of an integrated ecosystem simulation on a high-end shared memory computer using OpenMP. As an example, we use a parallel spatially-explicit structured fish model (ALFISH) for regional ecosystem restoration to demonstrate the parallelization procedure and associated model performance evaluation. Identical simulation results, validated by a comparison with a sequential implementation, and impressive parallel model performance demonstrate that layer-wised partitioning offers advantages in parallelizing structured ecological models on high-end shared memory computers. The average execution time of the parallel ALFISH model, using 1 computational thread, is about 11 hours, while the execution of the parallel ALFISH model using 25 computational threads is about 39 minutes (the speedup factor being about 16).