Metacommunity ecology was developed to advance our understanding of the local and regional processes that interact to shape community assembly. However, empirical metacommunity studies have largely focused on snapshot analyses that insufficiently describe the varying processes that act at the metacommunity scale. Estuarine assemblages are particularly underrepresented in empirical metacommunity studies despite their importance to global ecosystems. In this study, we analyzed long-term monitoring data using a Bayesian joint-species modeling framework to infer assembly processes in estuarine metacommunities over local and regional scales. We showed that metacommunity processes varied between local and regional analyses: local analyses provided support for species sorting among heterogeneous habitat patches, while regional analyses revealed spatiotemporal patterns of community assembly related to dispersal. Our work not only emphasizes the need to consider scale of analysis when investigating empirical metacommunities, but also highlights the importance of long-term ecological datasets for describing metacommunity processes.