Climate change and African vertebrates

Last year I spent a very happy evening in Cape Town enjoying some of the local specialities with a colleague and a visiting student. Or at least, that's what I thought - poor Raquel now tells me I was giving her a hard time... Still, good practice for her eventual defence of her thesis I hope. Anyway, she pointed me in the direction of the paper she was writing at the time that's now out and attempts to describe what's going to happen to some 2723 species of African vertebrates as the climate changes over the next several decades. Now, despite climate change being a huge conservation issue and one of my main research interests (and climate/weather being one of my 10 things to talk about), we've not talked much about it here on the blog before, so the chance to discuss what might happen to 2723 species across the continent as a whole is an ideal opportunity to start!

First, let me explain what Raquel and her colleagues have done. Essentially, they've said that we know that climate is an important driver of species distributions - coldness, wetness, hotness and everything all act to limit distributions, a concept that will be familiar to all. From that understanding Raquel set out to identify, for each one of the species, which climate variables are important and in what way - what the upper limit for temperature is for a given species, say. She made use of a suite of methods called bioclimate envelope models (and a whole range of other names too) that seek to identify the limits in a whole range of ways, using the observed distribution as the basis. So if a species is only found in areas where the temperature is never below 0^oC or above 20^oC, they assume these are the thermal limits in this species. Lots of different methods have been developed to do this, and once you've got a good idea of the limits using these methods (that idea of the limits we'll call a 'model') you can combine it with output from global climate models that predict what the climate will be like in the future to project where suitable climate will be for that organism (assuming the climate preferences don't change). Now, there are lots of methods to make these models, and also lots of different global climate models (not to mention alternative scenarios about how the world might respond to climate change), so one of the things Raquel was interested in knowing is how much variation is there between different model choices and, in particular, whether some sort of methods for combining the projections from each models into a since 'ensemble' model can be useful.

In doing so, of course, she mapped the changes in distributions of each one of the 2723 species involved which, if you're like me, is the first thing you'll want to know about. In fact, the authors don't give us all the individual predictions, but combine them to give us measures of 'species turnover' across the continent. It's not quite clear from the paper exactly how this is calculated (which is a shame, because it's another subject I'm interested in, and some methods are much better than others), but they say it's some combination of both the species that are predicted to colonise an area and the species that are predicted to go extinct in an area (presumably with those also not expected to change). So, what's the rub? Well, a subset of the results (from the more moderate climate change scenario) are pasted on the left here. Red areas are where there's high turnover (100%, whatever that means), blue where there's little or none. Rows are different species groups, columns (with the heading "Cluster 1, 2, etc.") are from different groupings of the climate change model predictions (those in cluster 1 and 2 seem more moderate than those in cluster 3). So you can see there's lots of turnover predicted for Namibia and the Sahel regions - both very arid areas. All the taxonomic groups show broadly similar patterns (a set of patterns that, sadly, weren't investigated, and something I'll come back to). And most of the variation shown here is between climate models.

Now, perhaps the most important bit of climate change isn't the movements per se, but the local extinctions. So Raquel and co also plot just the proportion of species that remain in an area - i.e. they're species that are there are the moment and are predicted to remain there in the future (right). This time the columns are the different climate change scenarios and, rather counter-intuitively I think, the redder the area the lower the local extinction rate. IE, in the dark blue areas, everything currently present int hat area is expected to be lost by 2080, whilst in the red areas only just over 50% of species are expected to be lost. (Don't forget, other species might come in to replace those lost.). So again we see massive extinction in Namibia and the Sahel in all groups, and greater stability in places like Tanzania and, perhaps, the Ethiopian highlands. But even here Raquel and co are saying more than 50% of our current species will have gone by 2080. [EDIT: That's not quite right, sorry! This is effectively a species richness map, not a useful map of climate space loss - see Raquel's comment below and the discussion soon to appear in a new post...]

With thanks to xkcd

Now, before I tell you what I make of all this, I should point out that Raquel and colleagues frame their paper in terms of understanding how to generate 'ensemble' models from sets of predictions. It's not something I think is necessarily a wise idea (I've published a bit about it here if you're really interested) and what I like most about this paper is that they've shown both that different ways of generating ensembles can make very large differences to your results, but also that they've not always reduced the complexity to single combinations. I don't think any of them really cover my main objection though, which is rather nicely illustrated by the problem with any averaging method in this area and shown simply on the left! I should probably also confess that whilst my wife complains I'm generally very critical of anything, I'm know to be particularly critical of these methods, having published some stuff on their failings here and here.

Usambara Pygmy-chameleon. History, not climate, must explain this distirbution.

So, what do I think overall? Well, I think there are a few minor things I'd quibble with - I'm not sure the choice of climate variables are the most meaningful options out for Africa (there's nothing about dry season length in there, for example, which I think is crucial). I also really disliked reading things like "method X gave very accurate predictions". Now I think I know what they mean - accurate in the sense that the variability associated with the predictions was relatively small - doesn't mean it's right, just that the predicted variation was small (precise predictions can be very wrong!). But not in the sense that any normal person who reads it might think - where accurate prediction means, well, it's telling you what really will happen. It's a technical issue I got a bit heated about here. And, of course, I have to confess I still don't think most species distributions are mainly driven by climate anyway - I don't think these methods are particularly good at determining when a distribution is associated with climate, and when it's not. They start from the premise that all species are, ultimately, climate limited and, when tested, that doesn't seem to be the case (or at least, that's what I suggested in that paper). And if there's anywhere that's going to be wrong, it's going to be Africa because, as regular readers know, the savanna biome is very special and there's a lot of savanna in Africa! What's special about the savanna biome? Well, importantly here is that it's distribution isn't mainly climate driven. Climate is important, but so too (as much, or more) is fire. And so too, is history - many of the species that differ from mountain to mountain are there simply as isolated forms no longer connected, but probably perfectly happy to live in any mountain and certainly not limited to that one mountain along by climate. Not to mention a whole suite of other issues.

Burchall's Coucal: limited by competiton with White-browed?

Put all of these things together, and add a little gut feeling (because I think science at times should be tempered with common sense) and I have to say I find it extraordinarily unlikely that there's going to be nowhere in the whole of Africa where more than half the current vertebrate species have died off. Maybe I'm an eternal optimist, but I really can't see that degree of change happening over the next 70 years. That's not to say that I don't think human driven climate change is one of the biggest threats to biodiversity we currently face, but I really don't think the estimates here are plausible. Because, ultimately, I don't believe the assumptions that underpin these models. There are important points in this paper, but I don't think the overall view of impacts will stand up to scrutiny.

EDIT: There's a problem with my interpretation of the second graph here that Raquel notes in her comment below. We've had some discussions about the paper by email and I'll post a second set of thoughts on this soon

Reference:

Garcia, R., Burgess, N., Cabeza, M., Rahbek, C., & Araújo, M. (2011). Exploring consensus in 21st century projections of climatically suitable areas for African vertebrates Global Change Biology DOI: 10.1111/j.1365-2486.2011.02605.x