Missing the point – errors of omission need solving
The pervasive biases in these datasets are not a new issue. Range
overestimations have been explored (Brooks et al. 2019), but errors of
omission have received comparatively little focus. Researchers seemingly
assume that, because models overestimate habitat suitability within
their borders, the total area of inhabitance should also be excessive.
However, a recent analysis (Li et al., 2019) found that though BirdLife
range maps were typically around ten times the area of predictive
models, ERMs still missed areas with recorded, verified records,
potentially highlighting areas of limited value whilst missing key areas
species may require to survive. We found the same issue in our analyses;
even in the best-studied group (birds), 25% of records fall outside
individual ERMs on average, and up to 46% of records for other groups.
This varies significantly between regions, with up to 90% of records
for reptiles in oceanic zones potentially falling outside their ERMs.
When we average each region, rather than all data (which emphasizes
better-sampled regions in the West), the average accuracy of the results
drops for most groups such that many mapped ranges are spatially
incorrect and may, thus, misdirect conservation efforts. Altogether,
these issues can lead to entirely incorrect estimates of biodiversity
hotspots, as seen when comparing the validated models (Hughes, 2017)
which showed ERM hotspots were much larger (upto 40% of the region for
some groups relative to upto 5% for models: Li et al., 2019). Despite
this, ERMs still failed to capture 15% of the most diverse hotspots
according to models, and may actively hinder effective conservation
efforts by both over-estimating some and missing other key biodiversity
regions. Thus, at least in developing and tropical regions, these errors
in ERMs mean that aggregating distribution data and trimming MCPs based
on environmental factors may more accurately map species ranges and have
a lower probability of omission than most groups.
Critical hotspots in transboundary areas are particularly likely to be
overlooked and missed due to these issues, as some borders show
particularly high purported species turnover based on ERM analyses
(Figure 2, Figure S1). These issues only further complicate pre-existent
challenges with working in close vicinity to many political boundaries.
Already, transboundary conservation assessments and recommendations are
being made based on these data (Mason et al. 2020). These concerns are
even more important for rare and range-limited species, which cannot
necessarily be protected merely by conserving more charismatic
megafauna. One consequence is that the largest groups are often removed
from analyses, for example, some analyses have used only 8% of mammal
species (Visconti et al. 2016), the rough equivalent of assessing plants
without angiosperms. Further efforts are needed to increase the breadth
and empiricism underlying ERMs and to ensure that data-driven analyses
are applied to remove such biases.