A NEW METHOD FOR DETECTING SPECIES ASSOCIATIONS WITH SPATIALLY AUTOCORRELATED DATA

Many organisms display patchiness in their distribution patterns over a wide range of spatial scales. Patchy distribution patterns can be ca used by processes such as growth, migration, reproduction, and mortali ty, which result in neighboring areas being more likely to contain a s pecies than distant areas, a phenomenon known as positive spatial auto correlation. When species are patchily distributed, the within-species spatial randomness assumptions of the standard statistical tests for detecting species associations are seriously violated. Using these tes ts under such circumstances can lead to incorrect rejection of the nul l hypothesis. To address this problem we introduce a new test for dete cting species associations-the random patterns test. This test takes i nto account spatial autocorrelation by including the characteristics o f the spatial pattern of each species into the null model. A randomiza tion procedure was used to generate the null distribution of the test statistic. The random patterns test is illustrated with data collected from an herbaceous understory community of a Eucalyptus forest near C anberra, Australia.