To assess predictive performance of our model, we collected actual U. thibetanus occurrence data. These additional observations were gathered fromnatural historymuseumdatabases, including data derived via the Global Biodiversity Information Facility (http://www.gbif.org/) and VertNet (http://www.vertnet.org) (Constable et al., 2010; Flemons
et al., 2007). Occurrences were placed on the binary map to establish omission error rates and develop a cumulative binominal test of the statistical significance of model predictions