Community state transitions during assembly differ under priority effects and environmental filtering

Abstract A major goal of community ecology is to understand the processes by which species assemble from the regional species pool to form local communities. The assembly process often leads to transitions between community states driven by two fundamental processes: environmental filtering and priority effects. Environmental filters can determine which species can successfully establish and persist in the patch and promote a single stable equilibrium, while priority effects often lead to alternative stable states due to the stochasticity in assembly history. Despite their important yet often contrary roles in determining community compositions, how the interaction between environmental filtering and priority effects affect community assembly is unclear. Here, we explore the interaction between environmental filtering and priority effects in community assembly with a seasonal metacommunity model. We found that priority effects reduced alpha diversity and accelerated community stabilization on larger patches by enabling early-arriving species to dominate, whereas environmental filtering promoted species coexistence and delayed the emergence of early warning signals. In contrast, the strength of priority effects declines as habitat loss intensifies environmental filtering on smaller patches. These findings enhance our understanding of how priority effects and environmental filtering jointly influence community composition during assembly, providing a process-based foundation for predicting biodiversity patterns under global change. Code Files The file Community_dynamic_Models.R contains all functions used to construct the models and run the simulations. This script contains all functions used to: Construct seasonal metacommunity models; Implement priority effects and environmental filtering; Simulate population dynamics across patch networks; Generate outputs related to biodiversity metrics and community state transitions. All simulations were conducted using R version 4.4.2. AcknowledgmentsWe thank the members of Si lab at East China Normal University for their feedback on the project. This study was supported by the National Key Research and Development Program of China (no. 2024YFF1307602) and the National Nature Science Foundation of China (no. 32311520284 and 32371590).