ENSO-related climate variability can reshape marine communities, yet community-level responses remain poorly understood in subtropical ecosystems. We examined spring fish community structure, environmental responses, and residual species associations across contrasting ENSO-related climate conditions in the historically exploited the Beibu Gulf. Using bottom-trawl data from six spring surveys conducted in 2015, 2018, 2020, 2022, 2024, and 2025, together with remotely sensed environmental variables, hierarchical modelling of species communities (HMSC) was used to compare alternative model structures and quantify climate-dependent community responses. The model including a quadratic sea surface temperature term (SST 2 ) and random effects provided the best balance between predictive performance and ecological interpretability. Sea surface temperature (SST) was the dominant environmental driver across all climate-year types, with its total explained variance peaking in El Niño years (53.7%). Dissolved oxygen (DO) and water transparency (SDD) also influenced several species, although their effects were weaker and more species-specific. After accounting for measured environmental variables and random effects, residual covariance structures differed markedly among climate-year types, with residual association networks being denser and more cohesive in La Niña years but simpler and weaker in El Niño years. These findings indicate that spring fish community structure and residual association patterns vary substantially among ENSO-related climate-year types, accompanied by differences in the relative importance of measured environmental conditions and latent community structure. This study provides a quantitative basis for climate-adaptive fisheries management and ecosystem security in the Beibu Gulf.