Spatial Autocorrelation Analysis of Trend Residuals in Biological Data

Publisher Website
Google Scholar
Abstract
Geographic variation trends are often quite complex and consist of variation at different spatial scales. In such cases an analysis of spatial structure by spatial autocorrelation analysis is confounded by this intermixing of different scales. Trend surface analysis (TSA) or canonical trend surface analysis (CTSA) offer ways of overcoming this problem. We subjected residuals from two types of data to spatial autocorrelation analysis: TSA residuals of simulated population-genetic data (isolation by distance with or without migration), and CTSA residuals of morphometric data. This approach was able, in the first case, to separate the effects of isolation by distance from migration, and to give an estimate of the scale at which each process occurred; in the second case, it furnished an estimate of the mean size of the local variation pattern as distinct from its long range pattern. In populations where different variables show structure at different spatial scales, these methods can separate out such effects even if the point-values for the different variables are not located at the same coordinates.