Spatial models of giant burrowing frog distributions

Abstract

Spatial models of species distributions are becoming a common tool in wildlife manage- ment. Most distributional models are developed from point locality data which may limit the model- ling process. Habitat variability within a species environment could be included by modelling areas of occurrence rather than point records. This may be particularly important for rare and cryptic spe- cies which often have only a small number of known localities. The giant burrowing frog Heleioporus australiacus is a threatened and cryptic species in south-eastern Australia. Previous attempts at mod- elling its distribution have been largely unsuccessful due to the extremely small number of known localities. We aimed to improve knowledge of the distribution of this species in south-eastern New South Wales (NSW) by comparing point- and area-based models. Our area-based model used water- sheds as the area unit based on population data for this species. Generalized linear models were used to compare the environmental variables at 37 known localities with a set of 1000 random sites. Model performance was compared using the area under the curve from the receiver operating characteris- tic curve. Both modelling techniques suggest that the species may be more widely distributed than current records indicate. The species is most commonly associated with dry forest environments with high habitat complexity but avoiding large river valleys, high peaks and steep areas. These trends are consistent with an earlier BIOCLIM model for the species distribution adding support to the influ- ence of these features as important to the species.