Identifying aggregation and association in fully mapped spatial data

We describe a clump recognition process that may be used to analyze fully m apped spatial data. Any given spatial pattern can be made less aggregated b y replacing the closest-together pair of plants by a single individual at t heir centroid position. By repeatedly amalgamating pairs of individuals in this way, an initially aggregated pattern can be reduced to one indistingui shable from complete spatial randomness (i.e. a two-dimensional Poisson pat tern). The clump recognition process provides information on the size struc ture of aggregates within a population. Randomizing the position of clump c enters can be used to generate patterns that have similar aggregation chara cteristic to the original pattern. This property is used to develop Monte C arlo simulations for testing interspecific associations. We also discuss te sts of association that are based on measuring segregation between clump ce nters. We illustrate the methods with a series of patterns from (1) simple, stochastic processes, (2) a spatially explicit population model, and (3) a dune annual community.