OOS 10-10
A predictive community ecology model for extrapolating from field observations to novel conditions
Predictive models of community assembly are important, but rare. Existing methods can partition the relative contributions to ecosystem properties of individual species and of interactions among species based on experimentally assembled communities, and can also predict the relative abundance of species in communities based on observed pairwise interactions between species. However, such models describe outcomes only for the species combinations and environmental conditions used to parameterize them. This presents a problem for applying community models to novel real-world systems, both because a dauntingly large number of potential combinations of species would have to be tested, and because environmental conditions vary greatly across space and time. Here, we introduce a mechanistic resource competition model that can be fit with data measured only in monocultures. Because the model focuses on interactions between species and resources, rather than direct interactions among species, it accounts for changes in interactions that result from changes in community composition without the need to refit the model, and is able to capture changes in species responses across a wide range of environmental conditions.
Results/Conclusions
We fit this model to monoculture grassland plots at the Cedar Creek LTER site, and test its predictions against observations in experimental mixtures of 2-16 grassland species. We find that this method accurately predicts the observed resource availability, species composition, and species- and plot-level aboveground biomass in experimental polycultures, and also successfully reproduces community responses to experimental changes in resource availability. Because these predictions are accurate for conditions beyond those found in the monocultures used to parameterize the model, these results suggest that this method could be a tractable tool for predicting community responses to environmental change, and for engineering diverse species mixtures to meet desired criteria for species richness, biomass production, and other ecosystem properties. This is particularly promising because the method can be parameterized for many systems based on existing data or small novel experiments. Potential applications therefore range from conservation management, to designing new combinations of polyculture crops, mixed-species biofuels, or production forests.