A reproducible, data-driven approach to mapping species distributions using presence-only data and biogeographic templates

Expert-derived range maps are used extensively in macroecological and biogeographic analyses, yet they are subjective, taxonomically biased, and inconsistent in their treatment of species’ absences. We developed a reproducible, data-driven approach to estimate species’ extent of occurrence (EOO) using presence-only data and subregions of the Interim Biogeographic regionalization for Australia (IBRA). This approach employs a Gaussian kernel density estimator calibrated for spatial coherence and ecological realism, producing maps independent of arbitrary grid structures. We applied it to 610 Australian bird species and evaluated the concordance of our data-driven EOO maps against BirdLife International expert-derived range maps. The spatial association between these two map sources, measured on a 0-1 scale, ranged from near-zero to 0.93 across species, with higher concordance among terrestrial birds. Estimates of richness using both map sources varied most at the finest spatial scale (IBRA subregions), whit mean, and root mean square errors at the coarsest biogeographic scale (ecoregion) being 1/3 lower than those at the subregional scale. Likewise, we demonstrated the approach’s applicability across taxa by generating data-driven EOO maps for selected amphibians, mammals, reptiles, and vascular plants. Like for birds, discontinuities in the distribution of these non-avian species represented different subpopulations over Australia. Our approach minimizes the influence of sampling bias and internal discontinuities in the estimation of species’ EOO, while allowing for range edge refinement and subpopulation delineation. It provides an objective and scalable alternative to expert-derived range maps, well suited for large-scale ecological research requiring consistency in spatial precision. Given the plethora of biogeographic templates already in use, our approach is adaptable to many contexts; and thus, can readily support a better understanding and conservation of biodiversity at large spatial scales.