Perils of Using Range Maps in Macroecological Models of Species Richness

Aim: As range maps do not imply the continuous presence of species across their extents—and are commonly developed with species conservation in mind—their underlying assumptions unsurprisingly lead to richness overpredictions for a given area. Despite this expectation, species richness extracted from overlapping species range maps continues to be used in macroecological models (MEMs) of species richness. In this paper, we demonstrate the various implications of using range map richness in MEMs. Location: Puerto Rico. Taxon: Anolis lizards. Methods: We used random forests to build two MEMs of Puerto Rico Anolis species richness: (a) using richness extracted from superimposed species range maps and (b) using biodiversity surveys conducted by the Puerto Rico non-profit, Para la Naturaleza. We then compare richness predictions, variable importance and evaluation metrics between the two models. Results and Main Conclusions: We demonstrate how (1) using range map richness in MEMs results in richness overprediction, (2) variable importance taken from MEMs built with range map richness may be unreliable or indicative of variable importance at a coarser spatial scale and (3) correlation metrics used for evaluating MEMs are potentially misleading. Through this work, we shed light on the implications of using range map species richness in MEMs.