Integrating several analytical methods to assess strength of ecological processes behind metacommunity assembly

Abstract

Understanding processes and mechanisms of how species assemble in a metacommunity is crucial for illuminating the factors that contribute to the maintenance of biodiversity and developing management decisions. Ecologists have proposed a number of analytical methods for identifying the effects of various ecological processes, but there is no consensus on the best approach. Our study extends the existing framework which synthesizes multiple analytical methods and incorporates community data across space and time to understand the underlying ecological processes. We extended this framework by 1) including null‐model‐based analytical methods; 2) defining metacommunity archetypes that illustrate extreme cases of metacommunities, to observe how well they can be distinguished by different summary statistics, 3) applying the extended framework to real‐world vegetation data from a subtropical forest and interpreting the results, and 4) discussing the potential advantages, limitations, and future directions of applying this framework. We used a process‐based metacommunity simulation model to generate a simulated community dataset and applied random forest (RF) approach to estimate the strength of ecological processes in the process‐based model by considering the summary statistics calculated by the analytical methods as predictors. We also quantified the performance of the trained RF and applied it to estimate the strength of ecological processes in Fushan Forest Dynamics Plot. Our results demonstrate the framework's flexibility in incorporating different analytical methods and its generality to be applied to different community systems. We highlight its theoretical values in evaluating the performance of different statistics or indices in identifying ecological processes and its practical values in assessing the strength of ecological processes underlying real‐world metacommunities. Future improvements should focus on synthesizing statistics that capture specific signals of ecological processes and evaluating the robustness of estimation against dataset complexity and incompleteness.