Multivariate Geographic Clustering Using a Beowulf-Style Parallel Computer

The authors present an application of multivariate non-hierarchical statistical clustering to geographic environmental data from the 48 conterminous United States in order to produce maps of regions of ecological similarity called ecoregions. Nine input variables thought to aflect the growth of vegetation are clustered at a resolution of one square kilometer. These data represent over 7.8 million map cells in a g-dimensional data space. For the analysis, the authors built a 126-node heterogeneous cluster--aptly named the Stone SouperComputer--out of surplus PCs. The authors developed a parallel iterative statistical clustering algorithm which uses the MPI message pawing routines, employs a classical master/slave single program multiple data (SPMD) organization, performs dynamic load balancing, and provides fault tolerance. In addition to being run on the Stone Souper-Computer, the parallel algorithm was tested on other parallel platforms without code modification. Finally, the results of the geographic clustering are presented