Spatial autocorrelation and the analysis of invasion processes from distribution data: a study with the crayfish Procambarus clarkii

Complex spatial dynamics are frequent in invasive species; analyzing distribution patterns can help to understand the mechanisms driving invasions. We used different spatial regression techniques to evaluate processes determining the invasion of the red swamp crayfish Procambarus clarkii. We evaluated four a priori hypotheses on processes that may determine crayfish invasion: landscape alteration, connectivity, wetland suitability for abiotic and biotic features. We assessed the distribution of P. clarkii in 119 waterbodies in a recently invaded area. We used spatially explicit statistical techniques (spatial eigenvector mapping, generalized additive models, Bayesian intrinsic conditional autoregressive models) within an information-theoretic framework to assess the support of hypotheses; we also analyzed the pattern of spatial autocorrelation of data, model residuals, and eigenvectors. We found strong agreement between the results of spatial eigenvector mapping and Bayesian autoregressive models. Procambarus clarkii was significantly associated with the largest, permanent wetlands. Additive models suggested also association with human-dominated landscapes, but tended to overfit data. The results indicate that abiotic wetlands features and landscape alteration are major drivers of the species’ distribution. Species distribution data, residuals of ordinary least squares regression, and spatial eigenvectors all showed positive and significant spatial autocorrelation at distances up to 2,500 m; this may be caused by the dispersal ability of the species. Our analyses help to understand the processes determining the invasion and to identify the areas most at risk where screening and early management efforts can be focused. The comparison of multiple spatial techniques allows a robust assessment of factors determining complex distribution patterns.