Document: ALI-3-87-11

A Bayesian habitat model for the southern Green Frog (Rana clamitans).

WILLIAMS, A.K.* ¹, J.M.BERKSON ¹ and J.C.MITCHELL ²

Virginia Tech, Blacksburg, VA 24060 USA ¹
University of Richmond, Richmond, VA 23173 USA ²

Abstract:
Models that predict the probability of occurrence of wildlife species based on environmental measures are useful tools for resource managers. Several approaches, such as logistic regression, discriminant analysis, and artificial neural networks, have been used for probability of occurrence models with varying degrees of success. We selected a different approach and developed a pattern recognition model based on Bayes' theorem to predict the probability of occurrence for the Northern Green Frog (Rana clamitans melanota) across a landscape. The model was created using a subset of presence and absence locations from a three-year amphibian and reptile survey conducted at Fort A.P. Hill in Virginia. Environmental variables, including National Wetlands Inventory data, and information from locations of amphibians with similar habitat requirements were used to determine conditional and prior probabilities for model input. The model was applied to the landscape of Fort A.P. Hill using ArcView GIS software to create a probability of occurrence map. Posterior probabilities of occurrence across the landscape ranged from 0.217 - 0.997. The model predicted presence locations with greater accuracy and precision than absence locations. The mean predicted probability of occurrence for all presence locations was 0.84 (s = 0.18) and 0.56 (s = 0.24) for all absence locations. The model predicted probabilities of occurrence over 0.75 for 85% of the presence locations. For 57% of the absence locations, the predicted probability of occurrence was less than 0.50. Overall, the Bayesian modeling approach performed successfully and may offer advantages over standard modeling approaches.

Keywords: pattern recognition model, Bayes theorem, southern green frog, GIS

This abstract is being presented at: 10:30 AM in session:
Oral Session #38: Amphibian Ecology.