I’m lagging a little behind the news, but maybe it’s also an advantage to write about this issue after the first smoke has cleared. The topic is a paper by Fangliang He and Stephen Hubbell that appeared in Nature two weeks ago and that has sparked considerable controversy. Basically, the authors claim that “conventional” extinction estimates based on species-area relationships (SAR) are valid only if species are randomly distributed in space, in which ase SAR and the actually relevant endemics-area relationship (EAR) are identical (see Fig. below for an illustration). As they further point out, EAR estimates get smaller than SAR estimates with increasing spatial clustering (which is a sensible estimate for most species). Hence, they conclude that most SAR based extinction numbers overestimate species extinction rates.
The paper was received by the media with the foreseeable incomprehension (see e.g. the German article in the Sueddeutsche Zeitung) and with the also foreseeable reduction of the papers content to the fact that the “extinction problem was overestimated”. The Voice of America, for example, titles “Scientists: Species Extinction Rates Are Greatly Overestimated”.
The issue several people have taken with the paper is not the fact that SAR and EAR differ as such. In fact, I believe that the phenomenon that SAR and EAR can differ depending on the degree of spatial clustering can already be seen in Green and Ostling, 2003 [ADDITION 08.07.11: and also in Smith, 2010], although they do not presented in in the same context as He and Hubbell. Thus, the fact that EAR will yield different results than the EAR is, I believe, credible (but see Bob O’Haras blog post, which I hadn’t time to study in detail yet), and the paper is in this respect an important addition to the literature. I would add, though, that I found it somewhat unsatisfactory that the authors did consider clustering of species distributions, but not alternative geometries of habitat destruction. Alternative spatial distributions of habitat destruction would create additional need for corrections. For example, if area is lost in small bits that are completely randomly distributed in space, the issue of spatial clustering is probably entirely inconsequential.
The spark of anger, however, seems to be the confidence with which the paper promotes the premise of the EAR being the “true” predictor for extinction, and alternatives being wrong. This is most prominently maybe in the title which states that “Species-area relationships always overestimate extinction rates from habitat loss”, but also in a press briefing organized by Nature, in which Stephen Hubbell stated according to the New York Times that
“This is welcome news in that we have bought a little time for saving species […?] But we have to redo a whole lot of research that was done incorrectly.”
[ADDITION 08.07.11: an interview with Stephen Hubbel]
From what I read, there are two main lines of criticism against this general claim. Firstly, it is not unanimously supported that the EAR is really the relevant measure for how many species go extinct due to a reduction of area. EAR measures those species that are “immediately” killed by habitat destruction, SAR more a long term equilibrium of species abundance given a certain amount of habitat area, which includes the fact that many species need a significantly larger area than the “last individual” to maintain a viable population. The truth within our typical time horizon for decision making is probably somewhere in between. As Michael Rosenzweig, who suggested according to the New York Times to reject the paper, put it:
Endemic extinctions are like money borrowed from a loan shark, but the species-area relationship also describes, say, your car loan and 30-year mortgage
Secondly, the question is to which extend current methods of estimating species extinction rates really suffer from the claimed overprediction. As a matter of fact, many estimates of extinction rates are actually not based on area losses, which obviously neglect other effects such as pollution, fragmentation or overexploitation. According to the BBC News, Jean Christophe Vie, IUCN’s species programme deputy director states that:
We (IUCN assessors) do not use this system between area and species because we know there are flaws.
We have explicit details in our guidelines that to estimate extinction is not something we should do; for example, we know that species are not evenly distributed in ecosystems; habitat loss is not the only threat.
But even authors that have used area-based extinction estimates claim that their studies do not suffer from the overestimation problem, for example Stuart Pimm who has responed quite strongly in a blog article. Also Chris Thomas, whose 2004 Nature paper was also among the culprits identified by He and Hubbell, replied according to CBC News that
his paper actually uses the method He and Hubbell propose, estimating that five to 16 per cent of species would lose 100 per cent of their “climactically suitable” habitat by 2050. The number was published alongside the estimate, based on the species-area method, that 18 to 35 per cent of species could go extinct by 2050.
“We published the result seven years before He and Hubbell guessed this might be the case,” he said. “It is a pity the authors did not realize this.”
I’m quite sure there will be a few replies in the next issues of Nature, so watch out for follow-ups.