Environmental responses and interspecific associations of marine fishes revealed by HMSC: A case study in the offshore waters of southern Zhejiang, China

Accurately predicting fish spatial distributions is essential for fisheries management and ecosystem conservation. However, organism-environment relationships are often complex and dynamic. Using summer survey data collected between 2017 and 2023 in the offshore waters of southern Zhejiang, China, we applied Hierarchical Modeling of Species Communities (HMSC) framework to compare five candidate models that differed in environmental variables and the inclusion of a spatial random effect. Model comparison identified the model combining both linear and quadratic terms for seawater temperature, a linear term for salinity, and a spatial random effect as optimal. The result indicates that allowing a nonlinear temperature response and accounting for spatial dependence improved model fitting over the other four specifications. Variance partitioning attributed 59.4 % of the explained variation to the spatial random effect, 35.7 % to temperature, and 4.9 % to salinity, underscoring the dominant role of spatial processes in shaping fish community patterns. After controlling for environmental covariates and spatial structure, residual species associations grouped eight dominant species into two assemblages: (1) small demersal predators, including Harpadon nehereus , Larimichthys polyactis , Atrobucca nibe , and Pennahia argentata ; and (2) pelagic species, comprising Decapterus maruadsi , Psenopsis anomala , Trichiurus lepturus , and Pampus echinogaster . Species within the same assemblage showed significant positive residual correlations, whereas correlations between assemblages were generally moderate to strongly negative, revealing contrasting habitat use and trophic niches. These findings provide a scientific basis to support fisheries management, the assessment of stock enhancement programs, and ecosystem conservation in the offshore waters of southern Zhejiang. • HMSC with nonlinear temperature and spatial effects best predicts fish distributions. • Spatial random effect accounts for 59 % of explainable variation in fish community. • Residual correlations split eight dominant fishes into small demersal and open-water predators.