Date of Award:

2013

Document Type:

Thesis

Degree Name:

Master of Science (MS)

Department:

Wildland Resources

Advisor/Chair:

Eric M. Gese

Abstract

Historically, kit foxes (Vulpes macrotis) once occupied the desert and semi-arid regions of southwestern North America, ranging from Idaho to central Mexico. Their range-wide decline has warranted the kit fox to be listed as endangered in Colorado, threatened in California and Oregon, and designated as a state sensitive species in Idaho and Utah. Once considered the most abundant carnivore in western Utah, the kit fox has been in steep decline over the past decade, creating a demand to determine kit fox presence. Currently there is little consensus on which survey methodology is best to detect kit fox presence. We tested 4 survey methods (scat deposition, scent station, spotlight, trapping) along 15 5-km transects within a minimum known population of radio collar kit fox. Home range sizes for kit foxes on the study site were extremely large, averaging 20.5 km2. Scat deposition surveys had both the highest detection probabilities (= 0.88) and were the most closely related to known fox abundance (r2 =0.50, P = 0.001). For detecting kit foxes in a low density population we suggest using scat deposition transects during the breeding season. This method had low costs, was resilient to weather, had low labor requirements, and entailed no risk to the study animals.
Next in determining kit fox presence is estimating kit fox distribution. We developed resource selection functions (RSF) using presence data from the noninvasive scat surveys to model kit fox distribution. We evaluated the predictive performance of RSFs built using three popular techniques (Maxent, fixed-effects and mixed-effects general linear models) combined with common environmental parameters (slope, aspect, elevation, soil type). Both the Maxent and fixed-effects models performed to an acceptable level with relatively high area under the curve (AUC) scores of 0.83 and 0.75, respectively. The mixed-effects model over valued higher elevations and had poor model fit. This study demonstrated that it was possible to create valid and informative predictive maps of a species distribution using a noninvasive survey method for detecting a carnivore existing at low density. By demonstrating the application of noninvasive surveying to model habitat quality for a small mesocarnivore, wildlife management agencies will be able to develop predictive maps for species of interest and provide more knowledge to help guide future management decisions.

Share

COinS