Rolling the dice on species extinction?

Climate change and species extinction, two phrases that seem to be on everyone’s mind. But opinions diverge and even if the majority of us can no longer deny climate change – as the signing of the Kyoto agreement by most countries shows – its real dimension and impact on species extinction is still very controversial. But now scientists from Oxford University’s Biodiversity Research Group and colleagues decided to test our capacity to see the future by…going back to the past. And the conclusion is that the most commonly used models to predict species extinction are basically not that good. But not all is bad news.

Where are we going to be in 100 years’ time? The scientific results that reach the public vary so much that we can no longer know what to believe and many times it’s simply our political choices that define our ecological opinion. We are not challenging scientists’ integrity, but how accurate are their forecasting models? The problem is that we cannot go to the future to test their predictions.

But now Miguel B. Araújo, Robert J. Whittaker, Richard J. Ladle and Markus Erhard from the Oxford University’s Biodiversity Research Group, the London’s Natural History Museum Biodiversity Research Group and the Institute for Meteorology and Climate Research in Germany, in a paper just published online in the journal of Global Ecology and Biogeography might have found a solution by approaching the problem in a very different way.

In fact, the team of scientists decided that instead of trying to predict the future why not test the models by going back to the past instead? By using available and very complete population and distribution data on one hundred and sixty one 161 species of British birds during two distinct time periods (period 1 or T1= 1967-1972 and period 2 or T2= 1987-1991), Araújo and colleagues were able to test the accuracy of sixteen of the most widely used models of species evolution. They used the different models to predict what would happen to the British birds’ species from T1 to T2 by using the available species data on T1 together with known climate variation of those twenty years. Subsequently, the results obtained by the different models were compared with the real figures observed in T2. The approach is ingeniously simple but, nevertheless, very informative.

The models tested are climate envelope models. Each species can only survive on a range of particular climates (what is called the species’ climate envelope). The models use this information to predict whether a species will have a tendency to grow or disappear as consequence of a particular climate change.

But when Araújo and colleagues tested the most widely used climate envelope models to predict what would happen to British birds from T1 to T2 , to their surprise, the predicted numbers were totally different from what has happened in reality.

In fact, for 90% of the species tested, the models could not even agree if the species were going to expand or shrink under the given climate scenario. For the remaining 10% of the species, where all the models managed to agree whether the species would shrink or expand, only in half of the predictions the direction was correct. This means that in 5% of the species tested all the sixteen models came to the wrong conclusion by predicting that a species would expand when in fact it shrank, or vice versa.

As one of co-authors, Richard Ladle, says, “It would be just as accurate and a lot less hassle just to toss a coin”.

But not everything is dark; Araújo and colleagues might have found an alternative solution by using what is called a “consensus model”. A consensus model is a mathematical model, which, in this case, finds a projection that reflects the central tendency found by the different climate envelope models used. In fact, Araújo and colleagues show that if the alternative models are used to find a consensus projection, the predictions obtained could become as much as 75% accurate.

But since the consensus projection depends, nevertheless, of other projections what is clear is that scientists need to improve their models’ accuracy in order to have the capacity to predict something that actually resembles reality.

As Richard Ladle says “If we don’t improve our forecasting soon then not only will the climate sceptics find it easy to criticize climate change research, but we will be left making decisions about the future of the planet based on guesswork”.