Stage-wise environmental recovery patterns and cross-domain interactions of planktonic bacterial communities in Zhanjiang Bay following typhoon disturbance.

Typhoon events can abruptly alter local biogeochemical cycles, which may strongly influence microbial community structure and ecological interactions. Climate change is increasing the frequency of extreme weather events such as typhoons, yet the consequences of these disturbances for microbial community structure and interspecific interactions remain insufficiently studied. This study investigated the dynamic responses of bacterioplankton communities in Zhanjiang Bay, a typical semi-enclosed bay, before and after Typhoon Prapiroon. By integrating high-throughput sequencing, phylogenetic null models, interpretable machine learning (XGBoost-SHAP), and cross-domain co-occurrence network analysis, we revealed the stage-specific and complex recovery patterns of bacterial communities following typhoon disturbances. The analysis demonstrated that community diversity, initially disrupted by the typhoon, rebounded under the combined influence of both deterministic and stochastic processes. Building on this recovery trajectory, SHAP analysis revealed important associations between bacterioplankton and phytoplankton dynamics, with rare bacterial phyla (such as Deinococcota and Bdellovibrionota) and nitrite concentration contributing strongly to variation in phytoplankton diversity, and showing stronger explanatory power than traditional linear models. Furthermore, network analysis delineated a sequential recovery trajectory: from initial environmental filtering, through a phase of phytoplankton-centered reorganization, to the eventual restructuring of the microbial network towards its pre-typhoon state. Overall, these results provide insights into the recovery pathways and association patterns of microbial communities under extreme climatic disturbances, highlighting the potential ecological importance of rare taxa during the recovery process. By applying machine learning-based association analysis, this study provides a quantitative framework for identifying key variables and potential threshold responses in microbial diversity dynamics, thereby offering insights for assessing ecosystem stability in coastal systems facing increasingly frequent climate extremes.