2020 ESA Annual Meeting (August 3 - 6)

COS 60 Abstract - The structural sensitivity of competition models: A probabilistic approach to species coexistence

Alba Cervantes-Loreto1, Abigail I. Pastore2, Clement Aldbert3, Michelle Marraffini1, Margie Mayfield4 and Daniel B. Stouffer1, (1)School of Biological Sciences, University of Canterbury, Christchurch, New Zealand, (2)Biology, University of Queensland, Brusbane, QLD, Australia, (3)Mediterranean Institut of Oceaonography, Aix-Marseille University, Marseille, France, (4)University of Queensland, Australia
Background/Question/Methods

Predicting the outcome of competitive interactions among species that share resources is central to our current understanding of diversity maintenance. Among different conceptual frameworks, the interpretation of data via Modern Coexistence Theory (MCT) has become increasingly common, mainly because of its approach to separating mechanisms that drive species coexistence given phenomenological models of competition. In the absence of the knowledge of the resource dynamics, there is a plethora of related mathematical expressions that can quantify competitive ability and lead to predictions of species coexistence. Importantly, these predictions are influenced by parameter sensitivity and structural sensitivity due to the model itself. Structural sensitivity is the extent to which the specification of the mathematical models can lead to qualitatively different outcomes. Nonetheless, the consequences of choosing equivalent biologically plausible competition models or taking into consideration parameter uncertainty have not been explored in the context of MCT or competitive coexistence more generally. Here, we quantify the extent to which model choice and and parametric uncertainty is important for predicting plant coexistence from a set of pairwise competition experiments of annual plants in Western Australia. We studied multiple commonly used models to infer the strength of competition, and fit these models with a Bayesian framework to consider the predictive uncertainty each model brings to the table.

Results/Conclusions

Overall and relative to each other, models that had similar mathematical behavior all captured the per capita effects of increasing neighbor density fairly well. Additionally, all models predicted competitive exclusion when we used the point estimates of their parameter values to test for coexistence. However, when generating the posterior predictions of coexistence outcomes, we found that virtually all models provide varied predictions ranging from the competitive dominance of either of the two species to pairwise coexistence. Moreover, these predictions varied substantially between models in the extent to which they predict coexistence. Our results demonstrate that parameter sensitivity, while important, can become subordinate to structural sensitivity, an outcome that has also been observed in other biological contexts. Our study provides a framework upon which to examine coexistence outcomes in a probabilistic approach and highlights the importance of exploring the predictions of more than one model when testing for coexistence. Finally, we also show that phenomenological models predict different pathways to coexistence with the same data set, even if those models have very similar mathematical behavior.