Modeling software used in the project

Generalized Dissimilarity Modelling (GDM) models beta-diversity across a landscape. Specifically, it is a matrix regression technique and predicts biotic turnover/ dissimilarity between sites based upon environmental dissimilarity and geographic distance. The advantages of GDM are that it makes few assumptions regarding the relationship between environment and genetic diversity and can explicitly take into account the potential influence of geographic distance (i.e. isolation-by-distance) on generating population divergence. See Ferrier et al. 2007, Diversity and Distributions 13:252-264. for more details on running GDM and underlying principles. Generalized dissimilarity models were run using an Avenue script in ArcView v 3.2 in conjunction with a SPlus v 4 script obtained from the authors of GDM. All final models for this project were created using 5000 randomly distributed classification training samples and 50 final GDM classes. We ran these models in four ways; using only environmental variables, using only Euclidean distance, using only anthropogenic highways, and using all of these variables combined (full model). These different runs were used to compare the relative influence of each category of predictor variable and to detect cross-correlations among variables.