Quantifying the fundamental unit of biogeography: Assessing different methods to measure geographic range size and why it matters

Background/Question/Methods

Geographic range size is a fundamental property of a species and is a key criterion in determining the conservation status and prioritization of species. Yet, many different methods of measuring geographic range size are used in the literature, at times confounding range occupancy and range extent. As various methods can lead to drastic differences in species richness and extinction risk estimates, how we measure range size is profoundly important. We used a very large dataset of New World plant species to investigate several popular approaches of estimating the geographic range sizes and distributions of species. Our approaches varied from methods estimating the size of a species’ geographic extent, through methods approximating the size of a species’ occupied range, to methods estimating both the species’ occupied range and distribution. We explored how different Maxent model settings influenced the resulting range size estimates using combinations of bioclim layers, spatial filters, and different thresholds. For each approach, we computed the geographic range size and accuracy metrics and compared the resulting range sizes with those of expert-drawn maps. Finally, we created species richness maps from the ranges of various approaches to assess potential geographic variations in diversity estimates resulting from each method.

Results/Conclusions

For range size measures derived solely from occurrence data, latitudinal extent substantially overpredicted range size, while the area of occupied cells considerably underpredicted range size. In contrast, the area defined by a convex hull was a fairly good predictor of range size. Range size estimates derived from Maxent models were relatively unaffected by sample size, yet significantly improved when including both spatial filters and bioclim layers. Range size estimates were also highly influenced by the choice of threshold used to create presence/absence maps. In addition, species richness maps derived from the various range size estimation methods were markedly different than maps derived using expert range maps. Finally, a number of problems in the species occurrence data challenged our ability to produce accurate range size estimates at every level of analysis. Estimates were greatly influenced by taxonomic errors in species names, erroneous geographic coordinates, the presence of cultivars in species occurrence datasets, and incomplete coverage and representation of some species in expert maps.