Stem cells and cancer cells have enough molecular similarities that the former can be used to trigger immunity against the latter.
A proposal to simulate all of Earth’s ecosystems is exposing a rift between small and big ecology.
January 28, 2013|
CaelioFor most of its history, ecology has been dominated by small-scale field work and studies of specific ecosystems. Now, some ecologists are calling upon their peers to think bigger, and to bring the field into the era of Big Science. And it does not come much bigger than computational ecologist Drew Purves’s suggestion of modeling all life on Earth.
Purves, a researcher at Microsoft’s computer science research center in Cambridge, United Kingdom, is building a prototype that simulates entire ecosystems across the planet. Known as the Madingley model, Purves’s simulation captures the life, growth, migration, interactions, and death of individual creatures, and the flow of energy and nutrients between them.
Rather than modeling specific species, the Madingley model places them into broad bins, like carnivore or herbivore, diurnal or nocturnal, birds or mammals. And rather than simulating every individual—an impossibly large task for modern computing—the model uses “cohorts” to represent similar clusters of individuals, such as fish shoals.
Purves hopes that this and other “general ecosystem models” (GEMs) will provide a better understanding of how different ecosystems fit together, and how their properties “bubble up from the underlying principles,” he said. They may also help policy-makers to decide how best to conserve our fragile planet, by showing the effects of hunting, invasive species, and changing climates.
It sound ambitious, perhaps unrealistically so. As Bob Paine, a marine biologist from the University of Washington, said, “it’s mission impossible.” But Purves argues that we have enough data to get started, such as ratios of predators to prey in specific niches, and the metabolic rates and life-spans of many species. “The ecological community has a tendency to assume that it isn’t possible and we want to challenge that,” he said. “You don’t know until you try.”
He points to the climate change community as an example of how big models can change the way research is done. “There was a lot of skepticism about whether there’d be a global climate model just until the first one was built,” he said. Now, such models are the status quo in climate science and regularly feed into the work of the Intergovernmental Panel on Climate Change. Purves hopes his large-scale model will bring the same value to ecology, and recently laid out his views in a comment piece in Nature. But not everyone is convinced. “I don’t hold big ecology as being terribly useful for solving the world’s problems,” Paine said. “And in a zero-sum financial game, it represents a raid on the national treasury.”
Global models hinge on the interactions between species, Paine argued, and discovering those interactions depends on the small-scale, muddy-boots ecology he has long championed. For example, in the 1960s, Paine coined the concept of a keystone species—one that has a disproportionate influence on its ecosystem—after throwing starfish off a beach and showing that mussels would take over. These interactions are unpredictable, and discovering them “requires intense biological studies, often of single species,” said Paine.
He also rails against the approach of aggregating species into “gross, clumsy units.” Not all birds or carnivores or fliers behave in the same way, he said. Some species of starfish act as keystones but not others, and even those will influence their communities to different degrees depending on what part of the world they are in. Conversely, species from disparate categories can fulfil the same roles. “You can have a big crab in the Pacific that occupies the same niche as a British earthworm,” Paine said. “That’s invisible in the [Madingley] model and takes a lot of work to discover.”
Josh Tewksbury, an ecologist at the Luc Hoffmann Institute in Switzerland takes a more pragmatic view. Although he is pessimistic that Purves’s approach will inform many policy decisions, Tewksbury believes it could promote collaborative data collection and open-access data sharing—areas where ecologists are lagging behind other fields.
“Academic ecologists are a privileged lot,” he said. “We get to follow our ideas and our passions to the most spectacular places on the planet. We might consider spending a bit more time deciding what data we should collect when we go there, and making sure that what we bring back from those adventures is available to all.” And pausing to assess current methodologies doesn’t mean the end of field work, he added. “It may mean we end up doing more of that.”
Purves agrees. He supports small-scale studies, but he thinks that a global perspective will be useful in pinpointing gaps in our knowledge and galvanizing field ecologists to address them. “Again, in the climate community, people are studying topics because those topics have been shown to matter in these large models,” he said. “We hope that we give people a context.”
January 29, 2013
Perhaps this could make some contribution. But since ecology is massively stochastic, a massive chaotic system, it's like the weather. Like the weather, it has to take geography into account. It has to be built on top of the global climate modeler because weather and ocean circulation have such radical effects on ecosystems.
Meteorology computing is founded on gathering better and better field information that is collated in near real-time. Modeling depends on fairly fine grained simulation.
Since ecology is a more complex system than weather, with a great deal of unknown territory, the results of such modeling cannot be what it claims. It is impossible that such a giant model could be even close to what it claims for reasons that are easily proven using theorems.
The proper name for this project is "Very Large Scale Daisyworld Model". I just hope that its limitations will be properly understood.
January 29, 2013
I agree that this is ambitious and probably unrealizable given today's techonology and stores of ecological interaction data. To continue the weather metaphor, it has taken decades to be able to predict patterns more than a few days hence with any sort of confidence. Sure, gains have been made in part because of improved understanding of system interactions and better information gathering. But a huge part of the improvement has been due to faster and more parallelized supercomputers. These bring much higher resolution to the model's spatial parameters, and important small-scale interactions can now influence the final solutions.
Similarly, the power of these ecological models could depend just as much on the quantization of inputs. So I think it's unfair to compare the proposed ecological model to current weather models, since the potential for both will undoubtable improve in the coming decade.
That being said, I think BPH is right that the incredible amount of unknown territory could very well result in a model that poorly represents reality. But I don't think that's necessarily a bad thing. What if, after feeding in all that is known about an ecosystem, the model does not spit out anything close to reality in any of its stochastic iterations? Awesome! It can tell us that something funny and unexpected is going on and possibly give some cluse as to what it is. Maybe it woul be enough to warrant a field study that could change understanding of ecology, at least in that particular domain (spatial, interspecies, seasonal,...).
I'm not convinced that such a model would be possible without becoming an overly complicated and unmanageable immensity of special cases and inflexibility and losing all utility as a result. But I think it's worth a shot.
(Does anyone know if there's a comprehensive database of species distribution and interaction that has a good interface to mine data?)
January 29, 2013
Even this is not the perfect simulation, it represent the start of integrating the microbiome influence in this wonderful resonant cavity.
January 30, 2013
“You don’t know until you try" says Purves and that is true of everything: including years and years of painstaking, "muddy-boots" kind of ecology that leads to gathering concrete data that alone can test the biggest or the smallest of models.
Working in palynology and paleoecology, I well appreciate the importance of data and models to advance the science; I also know, for example, that the large scale Global BIOME model for reconstructing vegetation (only) all over the planet actually failed in places like tropical India and what may actually work is smaaler scale "regional to local" models.
so in an ideal world, both aspects would be well appreciated and funding to work on either would be equally easy to get - but it is sadly not so; in the past decade or so, much of the funding available is strongly biased towards supporting efforts at modelling (whatever data exists) and ignoring the fact that in terms of ecosystems data what we really know "can be held literally in the palm of ones hand, while what we dont know is as large as the planet earth" (rough translation of a proverb in Tamil). In essence, yes models are important, even big ones but in the global geopolitical context of funding for science, my vote would be for the smaller scale, muddy boots ecology : )