Species Distribution Models (SDMs) are an important tool for ecologists and environmental scientists. SDMs allow researchers to use a species’ observed spatial distribution, together with other important factors like temperature, rainfall and elevation, to predict the location of species over a wider range. SMDs are becoming especially important as climate change alters the spatial distribution of rainfall and temperature, which impacts where species may be threatened in the future.

SDMs usually rely on one of two fundamentally different types of data as the background information about the species in question: site-occupancy (SO) data or presence-background (PB) data. SO data are high quality, but also high cost and cover small areas. By contrast, PB datasets are large and cover large geographic extents, but may be affected by biased sampling. RMIT researchers Vira Koshkina, Yan Wang, and Lewi Stone from the School of Science, and Ascelin Gordon from School of Global, Urban and Social Studies, together with their colleagues Robert Dorazio (DELWP), and Matt White (Tel Aviv University), developed a new SDM which for the first time incorporates both SO and PB data types. Their integrated SO-PB model incorporates the strengths of both data types while simultaneously minimising their respective shortcomings.

Using simulated data, the researchers evaluated the performance of their integrated model’s performance against SDMs that use PB or SO data alone. They found that their integrated model improved the predictions around the distribution of species compared to those models, even when the SO data was sparse and collected over small geographic extents. Additionally, correlated environmental variables (such as temperature and rainfall) have been shown to impact the ability of SDMs to reliably estimate species distributions. The authors showed that their correlations among environmental variables did not affect the outcomes of their integrated model.

The authors also tested their new approach using real world data about the distribution of the yellow-bellied glider (Petaurus australis), a gliding marsupial found in eastern Australia, from Victoria to northern Queensland. PB data were obtained from the Victorian Department for Environment, Land Water and Planning (DELWP) biodiversity atlas, which records individual sightings of the glider across the state. SO data were collected as part of planned DELWP biodiversity surveys, targeting specific priority areas in Victoria. Each of these datasets was used to fit to their respective models, and then used together in an integrated model, in combination with a suite of environmental variables, including elevation, maximum temperature in January, minimum temperature in July, rainfall in both months, evaporation in both months, distance to streams, wetness, and visible sky. The integrated model produced noticeably different estimates of species density across Victoria, compared to the PB model: the integrated model estimated higher densities in the south-western and eastern regions of the study area, and lower densities in the southern and northern most regions. The authors also showed that the integrated model had higher predictive performance, compared with both the PB and SO models. This RMIT-led spatial research shows that these novel integrated models, combining data from the two main data sources, are better at predicting species abundance than using them alone. It can be used to predict spatial abundance at any spatial scale because it uses a point-process model.