Skip to main content
U.S. flag

An official website of the United States government

Accuracy of gap analysis habitat models in predicting physical features for wildlife-habitat associations in the southwest U.S.

January 1, 2010

Despite widespread and long-standing efforts to model wildlife-habitat associations using remotely sensed and other spatially explicit data, there are relatively few evaluations of the performance of variables included in predictive models relative to actual features on the landscape. As part of the National Gap Analysis Program, we specifically examined physical site features at randomly selected sample locations in the Southwestern U.S. to assess degree of concordance with predicted features used in modeling vertebrate habitat distribution. Our analysis considered hypotheses about relative accuracy with respect to 30 vertebrate species selected to represent the spectrum of habitat generalist to specialist and categorization of site by relative degree of conservation emphasis accorded to the site. Overall comparison of 19 variables observed at 382 sample sites indicated ???60% concordance for 12 variables. Directly measured or observed variables (slope, soil composition, rock outcrop) generally displayed high concordance, while variables that required judgments regarding descriptive categories (aspect, ecological system, landform) were less concordant. There were no differences detected in concordance among taxa groups, degree of specialization or generalization of selected taxa, or land conservation categorization of sample sites with respect to all sites. We found no support for the hypothesis that accuracy of habitat models is inversely related to degree of taxa specialization when model features for a habitat specialist could be more difficult to represent spatially. Likewise, we did not find support for the hypothesis that physical features will be predicted with higher accuracy on lands with greater dedication to biodiversity conservation than on other lands because of relative differences regarding available information. Accuracy generally was similar (>60%) to that observed for land cover mapping at the ecological system level. These patterns demonstrate resilience of gap analysis deductive model processes to the type of remotely sensed or interpreted data used in habitat feature predictions. ?? 2010 Elsevier B.V.

Publication Year 2010
Title Accuracy of gap analysis habitat models in predicting physical features for wildlife-habitat associations in the southwest U.S.
DOI 10.1016/j.ecolmodel.2010.08.034
Authors K.G. Boykin, B.C. Thompson, S. Propeck-Gray
Publication Type Article
Publication Subtype Journal Article
Series Title Ecological Modelling
Index ID 70036159
Record Source USGS Publications Warehouse