Monitoring-based assessment of gap-analysis models

Gap-analysis models of vertebrate species richness are primarily created based on literature and expert review to predict individual species' occurrences and overall richness of vertebrates. Such models need validation based on empirical data to assess their accuracy. We describe and apply a new technique for assessing the accuracy of spatially explicit models. We evaluated the accuracy of South Carolina gap-analysis vertebrate models of predicted occurrence for reptile, amphibian, and mammal species on the Savannah River Site, SC, by comparing the agreement between gap-analysis models with models derived from multi-year monitoring data. We determined the species model agreement, commission and omission errors, and spatial correspondence in both single-species and richness models, and spatial correspondence of nodes of high richness. Average species agreement (accuracy) between models was 63%, with similar commission and omission error rates. Where there was spatial correspondence in single-taxon analyses, up to 15% of species identities differed in richness maps. Further refinement of vertebrate models will improve their accuracy, critical for the application of gap analyses to conservation decision-making.