Finding a reliable and accurate solution
Given the difficulty of acquiring verified point data across all species, many researchers modify ERMs to justify mapping diversity at higher resolutions or “accuracy.” A common method is to trim species ranges by habitat and elevational range (Li et al., 2016, Ocampo-Peñuela et al. 2016, Brooks et al., 2019). Yet, to do so universally makes a number of critical errors and fails to address the root causes of issues in such analyses.
Firstly, these new approaches do not account for spatial biases associated with administrative boundaries, or missing key range areas (highlighted above). In addition, though refining by habitat is sensible if species are well-known, such data exist for relatively few species. As an example, clipping by altitude makes assumptions about both the level of knowledge on species ranges (defining species’ true elevational ranges is challenging for most, with high uncertainty) and the consistency of the data across species ranges (as there is likely to be a relationship between elevation range and latitude for most species with large ranges, and these may shift seasonally); basically, ranges will vary by latitude and may not be known for most species. Additionally, current minimum ranges do not necessarily represent climate-based ecophysiological thresholds for species, as lower-elevation range limits are most vulnerable to being converted to other land-use types and other disturbances, though our point data showed that many species listed as mid or high elevation by the IUCN had also been recorded at sea-level, thus, current IUCN assessments may overlook areas of the range without evaluating what data is available. Our analysis showed that species ranges regularly fell over a much broader area than in ERM assessments even when such a listing was given, and that estimates of range without sufficient data falsely represent true species ranges. Thus, basing future projections on thresholds generated by ERMs will over-inflate species’ perceived vulnerability by effectively removing higher-temperature areas that may actually be optimal for some species.
For these reasons, clipping an “expert-generated map” with “expert knowledge on species ranges” may amplify biases, and data-driven alternatives with sensible uncertainty measures should be developed, especially given that most range maps clearly disobey current guidance. A better approach is to sensibly use data of species localities to develop predictions of where species are known to occur, restrict extrapolation to the country or island where localities exist and use these predicted ranges as a replacement for “expert range maps.” Such initiatives could also be used to drive data aggregation and sharing, maximizing the availability of open data such as that made available through GBIF (though careful error checking is needed, Orr et al., in review). To further grow these resources, mechanisms must also be developed to better fund taxonomic data verifications and museum data digitization, including also mandated data sharing for projects receiving this funding (Orr et al 2020). Such data are more likely to capture less-accessible areas and rarer species than that generated by citizen scientists, making them invaluable for generating a representative view of the natural world (Hughes et al., in review).
Here, we show that ERMs are biased and inconsistently delineated across space and taxa, and that even simple approaches like minimum convex polygons produce more realistic diversity models. Trimming, which can be streamlined into a reproducible approach and applied in a standardized way across species (Figure 3), further improves the performance of MCPs, whilst many shortcomings in ERMs are not ameliorated. As a basis for regional conservation, the data now exist for many taxa to have such data driven approaches, whilst continued use of range maps without careful bias management could misdirect conservation attention.