Environmental drivers of Antarctic biodiversity in ice-free areas at a continental scale

The many long-held hypotheses on the environmental variables driving distributions and abundance of terrestrial biodiversity in Antarctica are rarely underpinned by empirical tests, particularly at the continental scale. Here we use sophisticated species distribution models, with the most comprehensive compilation of biodiversity data, to identify the key environmental drivers of Antarctic biodiversity and predict the likelihood of biodiversity occurrence across the continent.We included 25 taxonomic groups, ranging from lichens to penguins, finding that elevation, slope, precipitation, and rugosity are significant drivers for all taxa. Other drivers, such as total degree days and distance to coast likewise had significant relationships, but responses varied between and within taxonomic groups. Our results also provide some of the first measures of species richness and empirically derived biodiversity hotspots in ice free areas across the continent.The model:We use area interaction processes (a member of the family of Gibbs processes), and an extension of the inhomogeneous Poisson point process (the most well-known and used point process) to model and predict species occurrences across the continent. One the main advantages of this point process is that it can account for inter-point interaction (spatial dependence). This interpoint interaction allows us to relax the assumption of independence between points, often a commonly used but unrealistic assumption of species distribution.Environmental Predictors:The environmental and climate data used as predictors included: distance to coast (coast), elevation (elev), an indicator of snow melt (melt), terrain roughness (rugosity), slope, solar radiation (solrad), total degree days (dd), total precipitation (precip), distance to nearest geothermal site (geothermal), mean cloud cover (cloud), mean wind speed (wind), and distance to nearest human infrastructure (base). Base, geothermal and coast covariates were derived from either point or polygon data sets which gave the locations of these features. In order to include these data in the area interaction process model, they were converted to raster data, with each cell containing the euclidean distance of that cell to the nearest point in the feature data set. The distance to coast map was generated from the SCAR Antarctic Digital Database medium resolution coastline dataset. All other covariates were interpolated or downscaled to 1 km raster spatial layers (covering the extent of ice-free Antarctica). Fixing the base covariate at a certain level during predictions was used to account for sampling bias. These spatial layers were chosen as predictors they cover most abiotic features that are believed to be relevant factors in determining species distribution (see Introduction) and were available at a continental scale for ice-free areas.A geotiff of each environmental predictor used in the models (at 1km resolution ) is provided in Ice_Free_Drivers_PREDICTORS folder.Model outputsWe were able to model 25 taxonomic groups, ranging from lichens to penguins, finding that elevation, slope, precipitation, and rugosity are significant drivers for all taxa. Other drivers, such as total degree days and distance to coast likewise had significant relationships, but responses varied between and within taxonomic groups. Our results also provide some of the first measures of species richness and empirically derived biodiversity hotspots in ice free areas across the continent.For each species two predictive outputs were generated (at 2km scale).1)    Conditional relative intensity all predictors – the predicted number of presences of a taxa per unit area (with all predictors from the best model) provided in the folder: Ice_Free_biodiversity_drivers_OUTPUTS_ALL2)    Conditional relative intensity (no base) - the predicted number of presences of a taxa per unit area (with all predictors from the best model but not “distance to base”) provided in the folder: Ice_Free_biodiversity_drivers_OUTPUTS_ENVThe habitat suitability maps were used as an input to the Antarctic Ecosystem Inventory (Toth et al. in review). The interpolated layers (for each taxa) that were used in this study are provided in the folder - Habitat_Suitability_Inputs_for_Antarctic_Ecosystem_Inventory.ReferencesTóth, A.B., Terauds. A., Chown S.L., Hughes K.A., Convey P., Hodgson D.A., Cowan D.A., Gibson, J., Leihy. R.I., Murray N.J., Robinson, S.A., Shaw, J.D., Stark J.S., Stevens M.I., van den Hoff, J., Wasley J. and Keith, D.A. (in review). The Antarctic Ecosystem Inventory: A classification, descriptions and map of Antarctica’s ice-free lands. Submitted to Scientific Data (June 2024).