According to ecologist Mark Urban, advanced climate models run on supercomputers might be 'the greatest, fanciest thing on Earth'. But if we’re not putting in the right data, those models aren't going to help us truly understand the threat we face.
Ellen Glover Hello and welcome to Science Node podcast, an audio series that explores the ways big data and high performance computing are changing the world. Megabytes, gigabytes, terabytes, petabytes of data are being collected every day by researchers everywhere. In this podcast we break that down. We’re looking not only at the latest scientific discoveries, but the multi-billion dollar industry that’s backing them up. Literally. One of the hottest areas of research right now is climate change. Rising oceans, rising temperatures, rising piles of garbage...we’re covering it all in this four-part series. Welcome to the Heat Wav.
Ellen Glover High performance computers have become essential to studying climate change because they allow scientists to gather incredible amounts of data and then simulate reality. However, an international team of biologists argue that many climate prediction simulations aren’t as accurate as they should be. The group’s leader, Mark Urban, an ecologist out of the University of Connecticut, explains that this is largely due to the fact that these simulations are made through statistics. They use correlations to make predictions about the world’s plants and animals.
Mark Urban This is okay as long as there isn’t a large part of biology that’s going to play a role in these responses.
Ellen Glover Urban says, essentially, scientists look at the particular climate a species lives in right now, and then they predict where the climate will be in the future. All the while, they’re assuming the species will remain in the same space as the climate changes.
Mark Urban But, you know, a lot could happen along the way. We don’t want to treat a mouse the same as an elephant or a fish or a tree, right? We know that organisms can experience the environment in many different ways and have many different approaches to dealing with that environmental change.
Ellen Glover A lot can happen in the wild, things a remote supercomputer can’t always anticipate. And, of course, not all animals living in one ecosystem act the same way. That’s why Urban and his colleagues suggest a mechanistic approach. This means scientists need to start actually going into the field and collecting information that’s unique to specific plants and animal species. Like birth and death rates. Or movement and genetic variation.
Mark Urban It’s sort of like creating a computer game for biologists, right? We can put in all of these different behaviors and all these different mechanisms, and we can create some sort of a landscape that might be realistic, based on maps, and then we can let those organisms sort of do their thing across that landscape. We can throw all sorts of things at them like climate change, and we can alter their ability to move across the landscape, and we can alter their birth and death rates. We can play this game many, many, many, many times and start to get some understanding about what will happen in the future. Also, species interaction. So, one of the things that the correlative models don’t account for in general is the degree to which organisms interact with other species and how those species might be reacting to climate change, as well.
Ellen Glover These inaccurate models may not impose a big problem right now or even 20 years from now. Because these simulations are trained to correlate climate with problems that exist right now.
Mark Urban Where we really expect to see the difference is in the long term, so going out 40, 50, 60, 100 years, when those correlations break down. So, for example, let’s say that a species is currently limited in their range by a predator species. Well, because we’re working across space, there’s probably also going to be some sort of climatic variation. So the model is going to say oh, yes, it’s, say, rainfall that determines range, whereas maybe in reality, it’s predators. So the model might work for a little while but then, once the predator moves differently and the climate variables are still there, then suddenly you start to see a difference in what the model predicts and what nature is actually doing. We face that same problem in biology that, if we’re just going to fit statistical correlations, we may be fitting to the wrong things, and then we’re going to be really surprised in the future when those correlations break down and perhaps the real mechanism is revealed. But, by then, it might be too late. So, you know, in some ways we need to get back to boots and binoculars and go out in the field and understand these things and record these processes to study organisms.
Ellen Glover The more personal approach that Urban suggests sounds great. But there’s an estimated eight to ten million different species of organisms in the world. Where do you begin with something like that?
Mark Urban So we’re certainly in a sort of triage situation where we’re sitting on the battlefield and we’re in the medical tent and there are already cases lined up at the door. So we’ve got to choose those cases pretty carefully. One way is just to choose those species that are sort of representative of all the variation that’s out there. You can sort of imagine a bird that flies and is a predator. You might take some small insect, you might take something that’s aquatic. So you choose from different ecosystems, different types of traits that they have. Then, the hope is that by modeling them you also know what happens to the species that have similar traits.
Ellen Glover Another approach is to look at those species that have the biggest impact on the rest of the ecosystem.
Mark Urban This is an approach in which you would identify those so-called multipliers of climate change, we’ve called them biotic multipliers. In particular, it often ends up being the top predators. Top predators tend to be really sensitive to small climate changes, either positive or negative. And, because they are the big eaters in the system, they often control the number of prey species in the system. Those effects can even trickle all the way down to the base plant species.
Ellen Glover Like many other areas of climate research, analyzing the ways the world’s animals and plants are responding to rising temperatures is a big undertaking. Urban and his colleagues argue that there needs to be a more coordinated effort to not only get the work done, but to get it done right. Specifically, this involves groups like the Intergovernmental Panel on Climate Change and the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services. Both international groups are concerned with climate change and biodiversity.
Mark Urban Right now, there isn’t really an organization. Individual research groups will create a model or will model a certain organism or ecosystem, but that work’s not really coordinated, it’s not really funded internationally. It’s really effective to have many groups coordinated, working on similar issues. I think to make some real progress here we need to have some sort of overriding group that can coordinate those efforts.
Ellen Glover As you might be noticing, monitoring and analyzing how climate change influences any given ecosystem is extremely complicated because, hey, mother nature is fickle. There’s hardly ever a black and white reason for the way she acts.
Mark Urban We’d like to think that we know a lot about the natural world, but unfortunately that’s not true. We actually know very little about most organisms. There are some key organisms, usually they’re worth some money. Things like salmon or trees for lumber or sort of food crops, we know a lot about those species. But we know very little about most wild plants and animals. Often not enough to even put into one of these models because we’re missing some key piece of information. So there’s a lot of complications and, for sure, biology is complicated, but the great thing is that today we have this computing power which allows us to incorporate a lot of these interesting realities and try to understand how they’re going to affect future responses. So the models are only as good as the data. There’s the saying ‘crap in, crap out.’ We need to have good data to go into these models. They can be the fanciest, greatest thing on Earth, but if we’re not actually putting the right data in, then the models aren’t going to help you.
Ellen Glover Although Urban and his colleagues may take issue with some of the ways they are used, they see supercomputers as a part of the solution. The simulations this technology is capable of making, allows scientists to test their theories. As a result, they can develop a deeper understanding of a given species.
Mark Urban There’s really an opportunity for this sort of cycle of creating predictions, seeing where the uncertainties are, seeing if they match reality—maybe we can even try to predict some things and see how well they do—and then refine them. The great thing about today is that we’ve got these amazing computers that can allow us to simulate a couple million individuals of, say, frogs in a complex landscape with all sorts of interesting genetics and movement decisions they can make. The hard part is actually having the data to do that in a way that’s particular to an individual species.
Ellen Glover While the system isn’t perfect, amazing strides are being taken with the help of these supercomputers, too. Tune in next time to learn how one microscopic discovery could eat away at the world’s excess plastic. This episode of Heat Wav was brought to you by Science Node, an online magazine developed in collaboration with organizations in the US and Europe to bring information exploring the real-world impact of advanced computing and networks to experts and non-experts alike. Our featured sponsor today was the University of Connecticut. For more information about the research discussed today, you can find the actual paper on sciencemag.org – our direct link will be on the website. I’m your host, Ellen Glover. See you next time.