PARALLEL-PROCESSING OF SPATIAL STATISTICS

The computational intensity of spatial statistics, including measures of spatial association, has hindered their application to large empirical data sets. Computing environments using parallel processing have the potential to eliminate this problem. In this paper, we develop a method for processing a computationally intensive measure of spatial association (G) in parallel and present the performance enhancements obtained. Timing results are presented for a single processor and for 2-14 parallel processors operating on data sets containing 256-1600 point observations. The results indicate that significant improvements in processing time can be achieved using parallel architectures.