LAUSR

Estimating fossil species' geographic range is a major goal for paleobiologists. In the deep time, this is most commonly performed by using polygon‐based methods such as the minimum convex polygon (MCP) or the Alpha‐Hull. Unfortunately, such methods provide a poor representation of the fossil species' actual range, because they are unable to take control of the severe stochastic and taphonomic biases. Here, we introduce MInOSSE (massively interpolated occurrences for species spatial estimation), a model‐based method that combines a machine learning algorithm and geostatistical approaches to reconstruct a target fossil species' geographic ranges by relying on the distribution of other coeval species and without using environmental predictors. We tested MInOSSE by using many simulated fossil species' distributions, comparing its performance with MCP and Alpha‐Hull outcomes and applying it to real case studies. In all simulations, MInOSSE outperformed the competing methods. Interestingly, the superior performance of MInOSSE becomes more apparent when the fossil record of the target species is scarce, that is, when appropriate range reconstruction is most problematic with polygon‐based methods. MInOSSE is a powerful tool for researchers interested in studying geographic range evolution, effects of range size on extinction risk, as well as biodiversity dynamics and macroecological patterns in the deep time.