Assessing suitable and critical habitat for wood bison ( Bison bison athabascae) using remote sensing and geographic information systems

Wood bison (Bison bison athabascae) are listed by the Committee on the Status of Endangered Wildlife in Canada (COSEWIC) as threatened. The Species at Risk Act (SARA) requires that listed species have their critical habitat identified in a recovery or action plan. Defining critical habitat requires that the species-habitat relationship for wood bison be clearly understood at several spatial scales that would be applicable to the management and conservation of this species. To create the most accurate picture of critical habitat a very accurate map of land cover was created. I explored a technique of image segmentation using ecological regions as a means of explaining the spectral variance in remote sensing imagery to increase classification. Results show a significant increase in classification accuracy (alpha = 0.05, one-tailed) over two-stage approaches (Z = 2.49, Z crit = 1.65 p=0.0063). Resource use was assessed by examining a series of models established a priori using logistic regression. The resultant models were compared by assessing the Aikake Information Criteria (AIC) scores and the final models was assessed by using k-fold cross validation and out-of sample validation with data for a separate study area. I found that a good model of resource use could be created using predictors of use that included measures of habitat type as well as several measures of landscape physiognomy, Contrast Weighted Edge Density (CWED), Patch Density (PD), and Contagion.