Biodiversity may increase ecosystem resilience. However, we have limited understanding if this holds true for ecosystems that respond to gradual environmental change with abrupt shifts to an alternative state. We used a mathematical model of anoxic-oxic regime shifts and explored how trait diversity in three groups of bacteria influences resilience. We found that trait diversity did not always increase resilience: greater diversity in two of the groups increased but in one group decreased resilience of their preferred ecosystem state. We also found that simultaneous trait diversity in multiple groups often led to reduced or erased diversity effects. Overall, our results suggest that higher diversity can increase resilience but can also promote collapse when diversity occurs in a functional group that negatively influences the state it occurs in. We propose this mechanism as a potential management approach to facilitate the recovery of a desired ecosystem state.

1) Food web models explain and predict the trophic interactions in a food web, and they can infer missing interactions among the organisms. The allometric diet breadth model (ADBM) is a food web model based on the foraging theory. In the ADBM the foraging parameters are allometrically scaled to body sizes of predators and prey. In Petchey et al. (2008), the parameterisation of the ADBM had two limitations: (a) the model parameters were point estimates, and (b) food web connectance was not estimated. 2) The novelty of our current approach is: (a) we consider multiple predictions from the ADBM by parameterising it with approximate Bayesian computation, to estimate parameter distributions and not point estimates. (b) Connectance emerges from the parameterisation, by measuring model fit using the true skill statistic, which takes into account prediction of both the presences and absences of links. 3) We fit the ADBM using approximate Bayesian computation to 16 observed food webs from a wide variety of ecosystems. Connectance was consistently overestimated in the new parameterisation method. In some of the food webs, considerable variation in estimated parameter distributions occurred, and resulted in considerable variation (i.e. uncertainty) in predicted food web structure. 4) We conclude that the observed food web data is likely missing some trophic links that do actually occur, and that the ADBM likely predicts some links that do not exist. The latter could be addressed by accounting in the ADBM for additional traits other than body size. Further work could also address the significance of uncertainty in parameter estimates for predicted food web responses to environmental change.