Phys. Rev. Lett. (in the press); preprint at https://arxiv.org/abs/1706.02666
Biodiversity is stunning yet puzzling. How can several species compete for the same nutrient sources and still coexist, rather than having one winner monopolizing the resources? One approach to resolving this sort of paradox is the 'kill the winner' hypothesis: the most successful species is selectively decimated (by, say, a predator) and different prey species take turns at winning. That model has proved insightful, but Chi Xue and Nigel Goldenfeld argue that in its original form it misses important aspects that actually prevent a stable coexistence state. Luckily, they also show how diversity can be restored.
The original model assumes continuous population densities, which is not warranted for finite populations. Xue and Goldenfeld took into account 'demographic stochasticity' — think shot noise for populations — and found that coexistence broke down as species became extinct. But when predators and prey were allowed to coevolve, continuously adapting to one another, then the appearance of fit mutants counteracted the elimination of species. And for sufficiently large mutation rates, winner-alternating and coexisting phases emerged from the interplay between ecological and evolutionary processes.