Ecologists have come a considerable way using statistical modelling techniques to predict the long-term coexistence between species. These models rely on calculating competition coefficients between species; parameters based on experimental or field measurements. Though such models have been insightful, they are increasingly considered to be too simplified to be of much use in understanding diverse communities. Notably, these models typically do not incorporate any natural variation. The inherent variability present in all empirical estimates of competition coefficients offers a straight forward approach to improving the accuracy and biological realism of species coexistence predictions. We provide a hierarchical Bayesian framework to incorporate demographic stochasticity from field-based assessments of plant-plant interactions into phenomenological competition models and ask whether it improves our predictions of coexistence. We use annual plant fecundity models to estimate interaction coefficients to parameterise coexistence models, where we use the invasibility criterion to forecast species coexistence for nine annual plant species in a diverse natural system. Using this approach, we calculate whether a species can enter a community at low density and maintain positive population growth in the presence of the resident species (in both monoculture and species mixtures). To further improve on biological realism in this process of predicting coexistence, we allow for facilitation in the annual plant fecundity models and utilise a lottery model framework to restrict simulated population growth based on competition for space.
Results/Conclusions
We found that the inclusion of demographic stochasticity and facilitation revealed flexibility in species’ responses to neighbours and hence coexistence outcomes that is not accounted for in traditional coexistence modelling. For instance, several plant species had coexistence probability distributions overlapping zero, indicating that their chance of coexistence within a monoculture of another species or a community of multiple species is not necessarily a yes or no outcome. The shape of these probability distributions can further inform our expectations of species coexistence, for instance we found species with very tight (certain) distributions but also species with wider (less certain) distributions. This process allows for the potential to attribute these differences in variance to abiotic factors affecting plant performance and hence coexistence such as shade and soil nutrients. This approach has great potential to more effectively operationalise coexistence modelling for use in diverse natural systems, bringing theoretical ecology into the realm of ecosystem management.