Cellular Ecolab

In the original speciication of the Ecolab model, spatial variation of the system's variables was included, with an additional migration operator introduced proportional to the divergence of the species density. Suitably discretized, the model then consists of a number of cells, each running the non-spatial Ecolab model, linked together by the migration operator. This paper reports on the implementation of such a cellular Ecolab. Originally, it was hoped that computational parallelism might be exploited by providing parallel versions of the underlying array operations used. However, it was found that the cost of the irregular data communication involved in the sparse matrix multiplication destroyed any gain from parallelism. Thus a Cellular Ecolab ooers a way of exploiting parallelism, that for certain situations as superlinear speedup is observed. As an example of the type of problem addressable by a spatial Ecolab, the species-area rule is examined, although the results are not good enough at this stage to be deenitive.