Habitat suitability modeling using the software package MaxEnt (Phillips, Anderson, & Schapire, 2006) is a popular method for describing the habitat of rare species. MaxEnt uses “presence only” data to develop models; however presence data are highly skewed towards areas of high detection probability and these areas may not represent the full range of habitat use. Thusly, predictions from models developed using only data from areas with high detection probability may not represent all suitable habitat. This study tested the ability of MaxEnt models developed using three different data sets to accurately describe Western Massasauga (Sistrurus t. tergeminus) habitat at a local scale. Models were evaluated by their ability to predict high suitability values at locations of known snake occurrence. The first model was developed using only presence data from areas with the of highest detection probability (i.e. roads). This model was only able to identify half of the locations where snakes actually occurred as highly suitable. A second model was developed using presence data from one season of radio telemetry and road surveys. This model performed well, and when interpreted alongside telemetry observations, it indicated that the most suitable habitat for Western Massasaugas in the western rolling plains of Texas are areas with level uplands, well-drained loamy, sandy soils, with mixed grasses, Sand Sage prairies and mesquite savannahs. A model developed using the locations of the snakes’ brumation sites showed that the snake’s selected distinct wintering habitat based on the burrowing suitability of the soil.

Date of publication

Winter 12-16-2016

Document Type




Persistent identifier


Committee members

Neil B. Ford, John S. Placyk, Joshua A. Banta


Master of Science in Biology