The Madingley Ecosystem Model
This week, scientists from Cambridge have unveiled a way to model on a computer, how Earth's different ecosystems, the networks of plants and animals that depend upon each other, actually work.
This will enable researchers to understand better how human activities are likely to influence the world around us, and how to better conserve what we have.
Mike Harfoot is one of the creators of the system and he's based at the United Nations Environment Programme-World Conservation Monitoring Centre in Cambridge and also, Microsoft Research in Cambridge
Chris - First of all, one question, one word - why?
Mike - Yeah, so I think you can separate that into two components really, you hit the nail on the head in your introduction. So until now, or essentially, at the present, we have a very limited ability to be able to predict the future of the living Earth, and therefore, to understand the consequences of the human actions in the present and how they might affect the biological world in the future. Another component to that is there is more of an academic one, but really, it's interesting to know, is it at all possible. Can we predict the biological world and what parts of it can and can't we predict?
Chris - How have you gone about doing this?
Mike - We've taken a relatively novel approach in that we're trying to model every different type of organism on Earth and model them at the level of the individual. So, how that individual feeds, reproduces, why it dies, and how it moves around the environment. So, we have lots and lots of different types of these organisms that essentially we throw into the model, and we allow them to run around interacting, eating each other and running away from each other. And then we're interested in - when you do all that, the properties of those collections of organisms, how do they compare to the real world really.
Chris - Essentially, you have a field with grass, you put cows in the field with grass, you then put humans in and they farm the cows. They eat the cows, the cows having eaten the grass. There are bugs living in the cows. And so, you build up this sort of layers and layers of things living on and in, and around each other and depending on each other.
Mike - That's right. I guess my caveat to that would be that we don't have any humans in there at the minute, which is relatively important in the present day. So, we're really modelling I guess a pristine world without human influence. But you're right. We have plants and then we have small organisms that might eating the plants and then large organism might be eating those.
Chris - Do we know enough about all of those relationships in order to build an accurate model?
Mike - I think that's the interesting thing, that actually, the field of ecology knows quite a lot about different parts of it. So, we know an awful lot about how fast organisms of different sizes burn energy essentially to survive. We know quite a lot about how different organisms move around their environment in order to find prey - be that plant or animal species - and how fast they can eat those organisms.
So, we know quite a lot I think of fundamental facts about ecology. What we've been trying to do is encode all that into our model. There's another question which is, having put all that together, how do we know if what we predict looks like the real world. And that, we've found a bit more challenging. So, in our paper that was published this week. We pulled together essentially as much data as we could find on properties of the organisms and the collections of organisms that we predict in the model; be that at the level of individuals. So, how fast those individuals are growing in the model, how fast they're dying or reproducing. But then equally, when we look at those collections of organisms in a particular place, how many of those organisms are very small and how many are very large. So, we have a certain amount of information from the real world on that kind of stuff and broadly, the model recreates those patterns.
Chris - So, this is validating the model. It's saying, "Let's take real world data. Let's give the same problem to our model and see if it arrives at a solution resembling the real world. If it does, we must be doing something right."
Mike - Indeed, yes. I guess what we've done today is relatively crude in terms of the form of methods of evaluation and that partly reflects the amount of data that we can get, and also, the stage of the model. But yes, broadly speaking, we evaluated the model and it looks like it's a feasible way forward which is quite exciting.
Chris - The fact you haven't put the humans in, is that going to be the next step then so that you can then begin to ask things like, "I want to build a housing estate here. What influence is that going to have on the local ecosystem?"
Mike - Indeed, yes. So, I think there's a sort of many fold sets of research that derive from I guess where we are now with a proof of concept that the framework might work. One of which as you've eluded to is much more rigorously evaluating what it can and can't do. Another one is definitely including humans or the impact of humans and evaluating, so, what might the future hold, given that we've got a model now to predict it. And that might be at a small scale as you say, sort of a housing estate level. It might be a large scale, so, what happens if fisheries lead to a collapse of fish production in the world? What does that mean for how much food we need to grow on land and where we might do that without collapsing terrestrial ecosystems at the same time?