The Madingley Model
The Madingley Model simulates how the structure and function of ecosystems at global scales emerges from the underlying ecology of individual organisms. Think, simulating the fate of every single organism on Earth*, from the smallest to the largest, on land and in the sea, for tens or hundreds of years. This gives you some idea of what The Madingley Model is about, and why it’s been a challenging thing to put together!
* at present, every multicellular organism **, and not those in the soil. Well, you have to start somewhere.
** except we do have some unicellular plankton!
The Madingley Model grew out of discussions between UNEP-WCMC, and the CEES group at Microsoft Research. We asked the question: how could we best contribute to international policy making pertaining to biodiversity and ecosystem function? It seemed to us that these international discussion suffered from the lack of a functional model of ecosystems. Without such a model, how could we rationally make sense of the various pressures to which ecosystems are subject, devise metrics to measure the impacts of these pressures, develop strategies to mitigate these impacts, or prioritize scientific research to help with all this? As an analogy, consider the central role of ‘global circulation models’ (GCMs) in the carbon-climate debate. These models predict the potential impacts of various greenhouse gas emissions scenarios. But they also help in making sense of satellite data, in prioritizing research in biogeosciences, etc. This is why we describe The Madingley Model as a ‘GEM’ – a global ecosystem model, of a kind that we hope to see more of in coming years.
The UNEP-WCMC and CEES partnership seemed perfectly suited to developing a GEM. CEES members have lots experience in ecological modelling: various taxa, various processes, various scales, and using a variety of computational and statistical approaches. Whereas, UNEP-WCMC has lots of experience in the world of international conservation policy: data, metrics, reports, and policy and international negotiations. Combing these strengths, we hope to build a model that is both ecological defensible, and useful for international policy.
Our ultimate goal however, is not to develop a perfect model – but to trigger the creation of a set of competing models of this sort. We hope that, in providing a proof-of-principle model, other ecologists will be encouraged to suggest improvements to, adaptations of, or indeed complete replacements for, The Madingley Model.
Since setting ourselves this ambitious goal, we (particularly, in order of appearance, Tim Newbold, Mike Harfoot, Derek Tittensor), have put in several years of research, experimenting with different model formulations, coding, debugging. But even more than this, we’ve done a lot of reading around and discussing. How long would it take Mike to eat a salmon? A fox to eat a rabbit? A ladybird to find an aphid? Why do ecto and endotherms coexist? Why don’t omnivores win out everywhere? What defines my ‘ideal prey’? What to assume about dispersal, and movement more generally? How much do thrips weigh...
So does it work?
We are very happy to say that The Madingley Model now reproduces key features of ecosystem structure at global scales, without any ‘top down’ fixes to ensure such a result. Simulating the fate of all those individuals – and how they grow, eat each other, die, reproduce, and disperse – what emerges in silico are ecosystems that are, broadly speaking, consistent with what know about the real ones. For example, we get reasonable predictions for the total biomass of plants, herbivores, and carnivores; reasonable distributions of body size; reasonable rates of aggregate energy flow (e.g. from herbivores to carnivores); reasonable geographic gradients, and reasonable differences between land and sea (e.g. inverse pyramids of biomass in some parts of the ocean, but not on land).
As stated above, our intention is not to claim that we have it all sown up. Far from it! We need better sub-models for every process, which itself would be greatly helped by way more data on ecosystem structure (something that is rarely surveyed in itself). We also need whole new modules (e.g. for the soil, for microbes more generally, for coastal regions), new metrics of ecosystem health (and data on the same) and new applications and tests of this, and similar, models. Indeed, we need, and hope to see, many more models of this kind, which will at some point completely eclipse the Madingley model in terms of ecological realism predictive ability, and usefulness. But in the meantime, we hope you don’t mind too much if we take just a little pride in the fact that The Madingley Model works at all.