Keystone species play a critical role in sustaining ecosystem structure and function. Thus, accurately identifying keystone species is essential for effective biodiversity conservation. This study investigates the benthic ecosystem of Yanpu Bay’s mangroves, utilizing stable isotope analysis in combination with Bayesian mixture models and ecological network analysis to characterize trophic relationships and topological network structures, with the aim of identifying keystone species within the community. The benthic food web in this study comprised 96 connections and 27 nodes. Among them, Scartelaos histophorus preyed on eight benthos species, constituting 18.51% of the total prey sources in food web. Sedimentary organic matter (SOM) was identified as a critical food source, sustaining 17 consumer species, 62.96% of the total species recorded in the community. Quantitative analysis using criticality indices and key player problem indices identified Cerithidea cingulate, Littorinopsis scabra, Periophthalmus magnuspinnatus, S. histophorus, Bostrychus sinensis, and Metaplax longipes as keystone species. The identification of these keystone species provides valuable insights for developing targeted biodiversity conservation strategies and offers a robust scientific foundation for the restoration and sustainable management of the mangrove benthic food web.

The fish community structure of the Zhoushan Fishing Ground is undergoing change due to overfishing, climate variability, and other anthropogenic stressors. To investigate community-level environmental responses and interspecific associations in this region, we used 11 consecutive years (2014–2024) of spring bottom trawl survey data from the Zhoushan Fishing Ground and integrated environmental covariates to build a two-part hurdle model within the Hierarchical Modelling of Species Communities (HMSC) framework. The results showed that the spatial random effect had the highest contribution (41%), followed by the interannual trend (18%), indicating that community occurrence patterns are primarily shaped by the superposition of stable spatial structuring and long-term change. Depth was significant for more species, whereas salinity was significant for the fewest. Residual correlations further revealed that the focal fish species could be partitioned into two assemblages with one linking species. Meanwhile, within the two-part hurdle model, the direction and significance of responses to the same covariate were not always consistent, supporting that species occurrence probability and positive biomass are governed by different ecological processes. Overall, this study provides a transferable quantitative framework for community assessment in coastal fishing grounds and offers a more operational chain of evidence for ecosystem-based fisheries management.