Linking energetic instability to compositional changes in biological communities.

The resilience of an ecological community informs us how it will respond to future environmental disturbances. However, the concept is rarely tested in the context of predicting biodiversity change, particularly at broad spatial and taxonomic scales. Here, we show that measures of instability derived from the resilience of the current state of community compositions greatly improve the predictability of biodiversity change. We applied energy landscape analysis (ELA) to community compositions of both simulated and natural ecosystems of distinctive regions and taxa (birds, fishes, mollusks, and phytoplankton) and estimated the resilience of those systems. We found that a metric of local instability, which represents how current community states are inflated from local optima, explained well the magnitude of species turnover and energy changes from the current to future states for both simulated and real communities. A metric of global instability, which represents a community's tendency to cross-over ridges of local basins of attraction to alternate stable states, also served as a weaker yet significant index of such changes. Our ELA results suggest that quantifying the resilience of real ecosystems is essential for understanding the mechanisms of community dynamics in an effort to improve the prediction of biodiversity change.