We are in the midst of a global extinction crisis. Biodiversity is in decline as species after species disappear. Some estimates predict that up to 50% of species will be committed to extinction by 2050. Other estimates claim the current rate of extinction may be 10,000 times the background rate. Many ecologists and conservationists have declared the current species decline the sixth great mass extinction.
A recent paper published in the journal Nature argues that our current estimates of species loss are based on a flawed model and tend to overestimate the magnitude of species decline. The paper has received plenty of attention, and has been heavily criticized by ecologists and conservation biologists. The paper is wrong, but it is wrong for the right reasons, and the criticisms it has garnered point to a gaping hole in our understanding of population dynamics.
Before we delve into the gritty details of “Species–area relationships always overestimate extinction rates from habitat loss” we need to understand exactly what we’re trying to measure. In a simplistic sense, species are limited by the amount of area available to them. The more area there is, the more species can inhabit that area. This concept is easily visualized in island systems – bigger islands tend to have more species than smaller islands.
Islands are not exclusively land-masses scattered across the ocean. Island ecosystems emerge whenever essential habitat becomes patchy. An island ecosystem can be a patch of old growth forest surrounded by grassland, a mountain-top, a desert oasis, the bloated carcass of a sunken whale, or any ecosystem upon which endemic species are dependent for survival. Human impacts can also produce islands by dividing and removing habitat. Since human-induced habitat loss is the dominant cause of species extinction, we can determine how many species can be supported by anthropogenic islands and use this to estimate how many species will be lost due to habitat destruction.
But, as Stuart Pimm points out over at National Geographic News Watch:
Imagine this nightmare scenario. Suppose all the forest in eastern North America was cut — except Rock Creek Park. How many species would go extinct? There are several answers.
Suppose all those forests were cut within months — metaphorically “overnight.” If you think that’s unlikely, consider how fast the world’s tropical forests are being cut!
How many species would be extinct “the following morning?” The answer is “not many” — because Rock Creek has many species. That’s the number that the Nature paper calculates. It does so with complex, mathematical formulas, and with considerable elegance. That’s why it took the authors eight years.
It’s not the relevant answer, however.
How many species would eventually become extinct? The answer is very much higher. The populations of the species that survived the initial deforestation elsewhere would eventually die out.
This is known as an extinction debt – even though there are some surviving members of a species after habitat fragmentation (I am using fragmentation instead of destruction because we can assume 100% extinction of endemic species with total habitat destruction, while anything less than total destruction will leave fragmented habitat behind), they are committed to extinction by low population numbers, inbreeding depression, and reduced fitness. The few survivors may persist for several generations, but the the species is functionally extinct. The authors of the Nature paper, however, argue that the extinction debt is largely a result of sampling artifacts. Species-area relationships are calculated forwards in time, based on how often new species are discovered as the search area increases, while extinction/area is inferred by going backwards in time, from the moment the last individual was observed to the original range of the species. To account for this, extinction rates due to habitat loss are calculated using a backwards species-area relationship. The authors developed a model to calculate endemics-area relationship, which considers only species endemic to an area. They argue that their endemics-area relationship is more accurate than species-area, and conclude that backwards species area relationships overestimate the extinction rate.
Criticism has focused on this conclusion. Before we get into those criticisms, we need to talk about what the authors did and did not do. To begin with, they found that SAR (species-area relationship) perfectly mirrors EAR (endemics-area relationship) for populations with true, random distributions throughout a known area. Once they leave this idealized population model, SAR tends towards a higher estimate of extinction rate than EAR, sometimes dramatically so. In one extreme case they found an extinction estimate at least 80% and possibly up to 160% greater than their EAR estimate.
The authors argue that the “extinction” debt is not an actual phenomenon, but a sampling artifact. They call attention to a series of papers that have attempted to predict how many species will go extinct by X year, all of which have been overestimated.
In some aspects, they’re right. At any moment, SAR will overestimate the extinction rate for precisely the same reason that the concept of “extinction debt” exists. Not all species will go extinct as soon as a habitat size is reduced, but they may be slated for extinction in the future, exclusively because of past habitat destruction. Extinction is an ongoing process and the extinction debt adds time to the equation.
The addition of time to the equation leads to my biggest criticism of both this paper and the subsequent responses.
Extinction is not the only process that shapes the dynamics of fragmented populations
The assumption made by the “extinction debt” that fragmented populations will eventually go extinct due to reduced habitat ignores a major concept in island ecology – endemism increases in island ecosystems. Think of Darwin’s famous finches. A small founder population settles on one island, as the population expands and colonizes new islands, each new colony begins to differentiate, in part because there is low migration between island and because each island has slightly different conditions. Eventually we arrive at the present state, where each island has its own endemic finch species.
In human-induced habitat fragmentation, the process is happening in reverse. Instead of colonists founding a new population on an island, new islands are being formed that divide previously connected populations. The end result is an increasingly fragmented system, and increased fragmentation leads to increased speciation.
On one hand, we have decreasing habitat size, which leads to increased extinction rates. On the other hand, we have increasing fragmentation, which leads to increased specitation. These two processes happen over very different time scales, but both play a role in how populations are shaped after habitat destruction. If you want to estimate how many species are going extinct due to habitat loss, you cannot include extinction debt while ignoring speciation surplus, but you can look at absolute extinction rate without the debt, which is what the Nature paper’s authors attempt. Both are valid approaches, but both will profoundly underestimate the real extinction rate for one simple reason.
The vast majority of species alive on Earth do not live on land
Species area relationships have proven relatively effective for estimating terrestrial species abundance, but are not as well-suited to marine ecosystems. Marine habitats, even those with high degrees of endemism, exist in a three-dimensional framework. Take for example a deep-sea coral head. It has an abundance of organisms that rely directly on the coral head, but also a whole community that exists throughout the water column. Remove the coral head, and you don’t just lose the benthic community, but also the pelagic community that relies on it. In effect, what you have is a species-volume relationship, weighted towards the benthos, with greater connectivity among fragmented populations. In this context, estimating extinction rate using either SAR or EAR is inappropriate.
Any estimate of extinction based on species- or endemic-area relationships is going to underestimate the contribution of marine exploitation to the global extinction crisis. While both SAR and EAR have applications in terrestrial conservation, and maybe even for calculating global terrestrial extinction rate, neither provide an accurate reflection of the global extinction rate. And that leads up to my biggest problem with calculating extinction rate.
Number of species is a terrible metric for making conservation decisions
I’ve harped on my issues with species concepts before, so there’s no need to tread on old ground. To quote the closing paragraph of my most recent species concept post:
‘Species’ is a categorical problem. It relates to how we perceive and interact with the world we currently live in. While it is essential to understand not just what a species is, but under what criteria that species was determined, these distinctions should not be at the heart of conservation. Biodiversity is not measured in number of species alone and the current Biodiversity Crisis is much bigger than simply losing N species.
If you’re interested, you should give Conservation and the Concept of Species in a Biodiversity Crisis a read, but the short answer is that we should be looking at ecosystem function, not just raw species counts to set conservation and management priorities.
It’s unfortunate that certain news organizations have taken this paper and run with headlines which imply species declines are not happening. That is a gross misinterpretation of what this paper is about, and, as even the authors say in their closing statement:
These results might receive a mixed reaction from the conservation community. On the one hand, the good news is that all extinction rate estimates based on the backward SAR method are overestimates. Because it is derived from sample areas of first contact with each species, the backward SAR method makes the previously unrecognized assumption that any loss whatsoever of population due to habitat loss commits a species to extinction, which clearly is not true. On the other hand, there is likely to be concern that these results could jeopardize conservation efforts and be falsely construed in some quarters to imply that habitat loss is not a problem. Nothing could be further from the truth. There is no doubt whatsoever that the Millennium Ecosystem Assessment has correctly identified habitat loss as the primary threat to conserving the Earth’s biodiversity, and the sixth mass extinction might already be upon us or imminent.
(He and Hubbell 2011)
The Global Extinction Crisis is real, habitat destruction is the leading cause of species loss, and we are woefully unequipped to measure the true extent of the problem.
UPDATE: Bob O’Hara takes a deeper look at the underlying mathematical principles behind this paper.
He F, & Hubbell SP (2011). Species-area relationships always overestimate extinction rates from habitat loss. Nature, 473 (7347), 368-71 PMID: 21593870
Neigel, J. (2003). SPECIES–AREA RELATIONSHIPS AND MARINE CONSERVATION Ecological Applications, 13 (sp1), 138-145 DOI: 10.1890/1051-0761(2003)013[0138:SARAMC]2.0.CO;2
This is definitely the best response to that paper I’ve seen so far. The problem of estimating species loss has a lot in common with fisheries management. Unfortunately with modeling and incomplete ecological information in both cases we’re often reduced to educated guesses.
Read New Scientist early March 2011 article on Mass Extinctions. We have had 5. 2 were really bad – one of them caused by K2. The reef systems took 15 million years to recover. There was no coal layer for that period due to the mass die off of vegetation (that creates coal). It said as you say, we are creating the 6th mass extinctin NOW. Lists (examples) of what we have killed off already. With the acidification of the oceans just imagine….. our Australian Barrier Reefs are already adversely being affected by the acidification of our oceans.
I’m writing this paper up now, after I spent the afternoon working out what the hell it was doing. Basically, they didn’t understand that if you take a circle out of an area of Euclidean space, you’re not left with a circle (the surface of a globe is not Euclidean, BTW).
I’m eagerly awaiting your write-up. It goes without saying that you have a much better grasp of the mathematical framework than me.