Second author

Ecological food webs are undeniably complex, but whether they are also chaotic has been a matter of debate for ecologists for more than 30 years. Chaotic systems are governed by physical laws so convoluted that long-term predictions of their behaviour are impossible. On page 822, Jef Huisman at the University of Amsterdam in the Netherlands and his colleagues present evidence that plankton food webs are chaotic. He tells Nature why ecology may never be able to make long-term predictions.

There are very few empirical examples of chaos in ecosystems. Why is that?

Because we need long-term studies to prove that ecosystems are chaotic. My co-author Reinhard Heerkloss monitored the abundance of plankton species and nutrient fluxes under controlled temperature and light conditions for 2,300 days. He said it looked as if the system behaved chaotically, but he didn't know how to prove it, so he gave me the data. Plankton have short generation times, so the data covered thousands of generations.

Did you find evidence of the 'butterfly effect'?

Yes. This classic example of chaos theory — that the flap of a butterfly's wings can cause tiny atmospheric changes that result in a tornado elsewhere — describes how slight differences in the initial conditions of a system can change subsequent interactions in that system, leading to large-scale variation. Initially, our plankton species behaved as predicted, but then the abundances of species slowly diverged until, after 15–30 days, we could no longer predict their behaviour. We found the same limit to predictability for all plankton species in the food web — strong evidence that interactions among species cause the ecosystem to behave chaotically.

Did your work resolve any other questions?

Yes. Plankton species compete for several limiting resources. Classical ecological theory suggests that only a handful of species should survive in one place. Yet sampling studies routinely find 60–100 species in a single millilitre of water. Our data show that chaotic fluctuations in species abundance, resulting from species interactions, may create temporal 'windows' for more species to invade and survive in the system.

Will ecology ever be a predictive science?

Predictions in ecology will be similar to those for the weather — short-term predictability is high, whereas long-term forecasts are all but impossible. That said, we need long-term studies to understand all the mechanisms at work when species interact with each other or the environment.