Adequacy of wildlife habitat relation models for estimating spatial distributions of terrestrial vertebrates

In the U.S. National Biological Service's gap analysis, potential distributions of terrestrial vertebrate species are based on the synthesis of wildlife habitat relation data and then modeled using a vegetation cover map derived from Landsat Thematic Mapper imagery. Using long-term species lists from eight National Parks in Utah we evaluated the adequacy of the wildlife habitat relations data generated by gap analysis in predicting species distributions at landscape scales. Omission and commission error rates were estimated for major taxonomic groups and for each national park. Depending on the taxonomic group, omission error ranged from 0 to 25%, whereas commission error ranged from 4 to 33%. Error rates were highest in amphibians and reptiles and lowest for birds and mammals. In general the error rate declined as the size of the park increased. The Utah wildlife habitat relation models performed well when used to predict the presence or absence of terrestrial vertebrates in eight national parks in Utah and should provide valuable information for making conservation decisions. They also provide a measure of support for the use of these models within the gap analysis framework. Although it is likely that accuracy of wildlife habitat relation models will vary from state to state, and even considerably within a state, the modeling process seems robust enough to provide a reasonably high level of accuracy for use in conservation planning at the ecoregion level.