Non‐random correlation of species dynamics in tropical tree communities
The importance of environmental stochasticity for tropical tree dynamics has been recently stressed by several studies. This has spurred the development of a ‘time‐averaged neutral model’ of community dynamics by Kalyuzhny and colleagues that extends the neutral model by incorporating environmental stochasticity. We here show that this framework can be used to assess the presence of non‐random correlations between species dynamics. Indeed, the time‐averaged neutral model makes the simplifying assumption that species responses to environmental variation are uncorrelated. We therefore propose to use this model as a null hypothesis against which observed community dynamics can be compared. This study makes five contributions. First, we describe a novel time‐averaged neutral model of community dynamics that is close to, but more flexible than the one previously proposed by Kalyuzhny and colleagues. Second, we develop an inference method based on approximate Bayesian computation (ABC) and demonstrate the identifiability of the model parameters from community time series data. Third, we develop a test of the significance of environmental stochasticity, and a method to quantify its contribution to population variance. Fourth, we develop a test of non‐random correlation between species dynamics. Fifth, we apply these developments to three datasets of tropical tree dynamics. We evidence both a strong contribution of environmental stochasticity to population variance in the three datasets, and a non‐random correlation of species dynamics in one of them. We finally discuss the implications of these results for the modelling of tropical tree community dynamics.