Mapping habitat of threatened reptiles in Western Downs, Queensland: a spatial modelling approach using 'presence-only' occurrence data

Abstract

Biodiversity conservation planning requires spatial data. Information on the spatial distribution of wildlife species, as well as the location and extent of their habitat, is essential in developing conservation strategies. However, traditional ground-based survey and mapping methods cannot always deliver the necessary information in a timely and cost-effective fashion. For habitat mapping, the development of probabilistic modelling approaches offer some merits for the threatened reptiles in the Western Downs, Queensland, Australia. This study was conducted to identify the predictor variables significant in reptile habitat modelling, and to develop a spatially explicit predictive model for a group of reptiles. The predictor variables and 'presence-only' reptile occurrence data (representing seven threatened species of lizards and snakes) were examined using the weights-of-evidence (WofE) approach in a GIS platform. Of the 18 initial variables, seven of these were excluded from the modelling process due to their weak spatial association with reptile occurrences. These refer to topography-related variables (slope, aspect, topographic wetness index, and elevation) and ‘vegetation amount’ variables (foliage projective cover and NDVI). Conversely, land use, regional ecosystems type, evapotranspiration, land cover and major vegetation group variables exhibited strong spatial association with the observed reptile data. The results also show that the 4-map combination of 'regional ecosystems type' (RE), 'distance from water', 'soils', and 'distance from stream' produced the highest prediction accuracy (up to 87%). This study identified the regional ecosystems layer as the most significant variable, and it highlighted the importance of selected vegetation communities in the region. Since only 9% of the total area has high to moderate habitat preference for the reptiles, further vegetation clearing should be avoided to prevent new habitat loss. The weights-of-evidence approach was found highly suitable for predictive habitat mapping of threatened reptiles with 'presence-only' occurrence data.