Sampling units derived from geopolitical boundaries bias biodiversity analyses

Abstract Aim Sampling decisions are known to influence the estimation and interpretation of biodiversity. The use of geopolitical and administrative boundaries (GABs) for mapping biodiversity variables and identifying hotspots has a long history in biogeography and conservation biology. However, potential biases associated with these units and their effects on conservation decisions has not been adequately explored. Location Global. Time period 21,000 years ago to present day. Major taxa studied Amphibians, birds, mammals and reptiles. Methods I calculated biodiversity metrics for four major taxa at eight different spatial resolutions to understand scale‐dependencies. I then compared these estimates to those using GABs (e.g., countries and other large units) as sampling units. I also used both grid‐ and GAB‐based sampling designs to model the relationship between biodiversity patterns and relevant climate variables. Results Larger sampling units led to higher estimates of species richness and phylogenetic diversity. Taxonomic and phylogenetic endemism remained relatively stable across scales due to competing biases related to grain size and estimation of range size. Using GABs led to both under‐ and overestimation of endemism when compared to fine‐resolution grid‐based estimates, with differences varying across space and taxonomic groups. Importantly, using GABs resulted in different locations being identified as biodiversity hotspots when compared to grid‐based analyses. Likewise, different climate drivers of species richness and endemism were identified when using a GAB‐ versus grid‐based sampling design. Main conclusions Using GABs will remain an important method for mapping variation in biodiversity metrics across the globe. However, appropriate understanding of the biases associated with these and other ad‐hoc sampling designs should be better understood and accounted for in macroecology analyses, including for understanding environmental drivers of biodiversity patterns and identifying local and regional diversity hotspots. This is especially true for measures of diversity that incorporate species range size, such as taxonomic and phylogenetic endemism.