Site icon Conservation news

When it comes to the IUCN Red List, accuracy is the order of the day (commentary)

  • It is clear that the IUCN tries to ensure that its criteria for determining threat status can be uniformly applied across all species. It is also clear that IUCN has applied this valuable service for a long time, for which we should all be grateful.
  • The approach we have used takes advantage of the enormous amounts of freely available geo-referenced data to build much more accurate range maps based on expert filtered citizen science sightings, a plethora of high resolution ecological and geophysical data from the Western Ghats, and well-tested statistical tools.
  • Since these data can now be found for many places in the world and for many taxonomic groups, we call on IUCN to embrace this approach.
  • This post is a commentary. The views expressed are those of the authors, not necessarily Mongabay.

When Mongabay first covered the publication of our paper, Ramesh et al. (2017), we anticipated a vigorous response from the IUCN and, indeed, the IUCN did not disappoint.

A follow-up commentary by Craig Hilton-Taylor, head of the IUCN Red List, added additional information on IUCN’s position. It is our hope that through this back and forth between IUCN and scientists working on more efficient and rigorous ways to assess species threat levels, most of which we assume will occur out of public view, the conservation community will arrive at the most robust way forward. Here are a few issues to consider.

First, for researchers and practitioners to use IUCN’s information with confidence, we must be clear about the terms to which we are referring and the comparisons we are making. The Extent of Occurrence (EOO) that Hilton-Taylor refers to is, by IUCN’s definition (on page 46 of the linked document), “the area contained within the shortest continuous imaginary boundary which can be drawn to encompass all the known, inferred or projected sites of present occurrence of a taxon, excluding cases of vagrancy.” The Area of Occupancy (AOO) is defined (on page 48) as “the area within its ‘extent of occurrence,’ which is occupied by a taxon excluding cases of vagrancy.”

A careful read of our methods leaves no question that our “model” range maps are equivalent to IUCN’s EOOs, not their AOOs. Why? Because the IUCN EOOs and our model range maps both include areas where the particular species has been observed, but also areas where they have not been observed, but could inhabit. The AOO includes only the area “currently occupied by the taxon.”

Second, Hilton-Taylor incorrectly states that the IUCN threat status of a species can only be estimated using the AOO as its baseline, to which other sub-criteria are applied. In Table 2.1 on page 14 of the Guidelines for Using the IUCN Red List Categories and Criteria, Version 13 (March 2017) (at the time of our study, Version 12 (February 2016) was made available to us), one is given a choice of either using the EOO or AOO (see Criterion B1 and B2). Indeed, following those guidelines and the IUCN definitions of EOO and AOO, we compared our range maps to the EOOs published by the IUCN for the species examined. It is these comparisons that have led us to conclude that the IUCN EOOs are inflated and that by their criterion (B1), we should flag at least 10 species for potential uplisting of their threat statuses.

White-bellied blue flycatcher (Cyornis pallipes). Photo by Bopanna Pattada.

Third, let’s look more closely at how IUCN has estimated EOOs. At the time of our study, the IUCN EOO reported for each of the species we examined was exactly equivalent to the area defined by the BirdLife International (BLI) range maps. Since then, IUCN has applied the Minimum Convex Polygon (MCP) method, which they define as the minimum boundary around all of the BLI sub-range maps of a species. This technique converts the fragmented BLI sub-range maps for a species into a contiguous MCP, whose area can be estimated. If one goes to the IUCN website now, one finds that the EOOs for three of the species we studied are still equal to the BLI range maps, EOOs for 10 species we studied are equal to the MCPs drawn around the BLI range maps, and EOOs for five species we studied are of unknown data origin (that is, they are neither equal to the BLI range maps nor the MCP around those range maps). This inconsistency, of course, frustrates attempts to make direct comparisons.

Fourth, since, by the IUCN’s own definitions, it is clear that our “model” ranges are equivalent to their EOO and not their AOO, we can now compare our model range values with their EOO values. Doing this, one can see that our model range values are a fraction of their EOOs as they reported them when we did the study (i.e. equal to the BLI range maps) and even more so as they calculate them now (i.e. equal to the MCP drawn around the BLI sub-range maps). Further, if we take our model range values and compare them to the EOO threshold values in Table 2.1 on page 14 of the Guidelines for Using the IUCN Red List Categories and Criteria, Version 13 (March 2017), Criterion B, there is no question that our numbers point to a potential uplisting of at least 10 of the 18 species examined.

For illustration purposes, let’s take Hilton-Taylor’s example, the black and orange flycatcher (Ficedula nigrorufa). The EOO listed by IUCN is 21,400 square kilometers (a little over 8,200 square miles), equal to the BLI range map. If we instead use what IUCN hopes to apply uniformly, the minimum convex polygon or MCP, the EOO, using IUCN’s published method of estimation, is 53,070 km2 (close to 20,500 square miles). Our model range for this species, analogous to the IUCN EOO, is 9,300 km2 (about 3,590 square miles). If one now goes back to Table 2.1 and uses the thresholds listed for EOO, potentially uplisting this species to Vulnerable appears to be the prudent course of action.

Having said all of this, we agree with Hilton-Taylor that our publication does not mention information relevant to IUCN’s sub-criteria for Criterion B. That is why we clearly state that these 10 species deserve a potential uplisting of their threat levels, as most of the supplementary information regarding habitat loss, fragmentation, and number of mature individuals is presently lacking for most of the bird species that we examined. In the same way that we have developed an automated approach to determine more accurate range maps to compare to Criterion B1, based on available geo-referenced data, we are now working on ways to automate the assessment of the IUCN sub-criteria (B (a), (b), (c)), using available geo-referenced data on habitat fragmentation, habitat quality, and population density. The results of this new research will move us to a more data-driven, comprehensive assessment of Criterion B and its sub-criteria, changing the “potential” uplisting to a “real” uplisting, or not.

The species we have flagged in the research published in Biological Conservation should be marked with a yellow warning light, awaiting the results of our new research effort. And IUCN should be encouraged by our efforts to use available geo-referenced data to estimate the Criterion B sub-criteria in an automated way, because, along with our automated methods for more accurately estimating range maps, it will allow for a more accurate estimate of threat status for the endemic bird species of the Western Ghats, and any other taxa for which detailed, geo-referenced data are available.

Finally, we believe that Hilton-Taylor has mistakenly compared our work to Ocampo-Peñuela et al. (2016), Li et al. (2016), and Tracewski et al. (2016). These studies use worldwide data sets on deforestation to arrive at global estimates of the number of species whose threat status may have been underestimated. While these studies provide broad scale estimates and might even be helpful in pinpointing geographical areas of concern, they are not intended to be, nor do they take the place of, more detailed studies like our own. The above studies employ a few data layers (e.g. forest cover, altitude) to an expansive geographical space, whereas our approach takes into account 27 environmental covariates that truly drive the distribution of the species examined in a much more limited area. Hence, our range maps are more intensely data-driven, based as they are on densely distributed geo-referenced citizen science and ecological and geophysical data within a machine-learning framework, which is very different from the papers Hilton-Taylor cites.

It is clear that the IUCN does try to ensure that its criteria for determining threat status can be uniformly applied across all species. It is also clear that IUCN has applied this valuable service for a long time, for which we should all be grateful. However, if we are using inaccurate range data in the first place, then we are all likely to find ourselves designating the wrong threat status for many species. The approach we have used takes advantage of the enormous amounts of freely available geo-referenced data to build much more accurate range maps based on expert filtered citizen science sightings, a plethora of high resolution ecological and geophysical data from the Western Ghats, and well-tested statistical tools. Since these data can now be found for many places in the world and for many taxonomic groups, we call on IUCN to embrace this approach. We would certainly be more than willing to work with them to implement it.

Nilgiri shortwing (Sholicola major). Photo by Bopanna Pattada.