Does GARP really fail miserably? A response to Stockman et al. (2006)

Abstract

Stockman et al. (2006) found that ecological niche models built using DesktopGARP 'failed miserably' to predict trapdoor spider (genus Promyrmekiaphila) distributions in California. This apparent failure of GARP (Genetic Algorithm for Rule-Set Production) was actually a failure of the authors' methods, that is, attempting to build ecological niche models using single data points. In this paper, we present a re-analysis of their original data using standard methods with the data appropriately partitioned into training/testing subsets. This re-evaluation generated accurate distributional predictions that we contrast with theirs. We address the consequences of model-building using single data points and the need for a foundational understanding of the principles of ecological niche modelling. © 2006 The Authors.