Biodiversity plays an important role in regulating and supporting ecosystem functions and services, whereas in recent years, multiple pressures, including overfishing, habitat destruction and climate change, have led to severe declines of biodiversity in marine and coastal ecosystems. Consequently, the biodiversity-ecosystem functioning (BEF) relationship has become one of the research focuses in the field of biodiversity conservation. Previous literatures have accounted for the BEF in the marine ecosystems, and suggested that biodiversity can sustain ecological functions through complementary effects and other approaches; however, relevant studies neglected the interactions among different trophic levels and their heterogeneous responses to biodiversity. In this study, we investigated the spatial and temporal distribution of the biomass and species diversity of the fishery biomes in Haizhou Bay, based on the bottom trawl surveys in the area in autumn from 2013 to 2023. Species distribution models and structural equation models (SEM) were used to examine the relationships between environmental factors, species diversity (richness and evenness) and biomass at different trophic levels (the species were divided into high, medium, and low levels according to their feeding habits). Among them, SEM was used to consider the correlations between multiple variables simultaneously and to intuitively reflect the direction and intensity of their possible effects. This provided an effective tool to delineate the interspecific interactions at different trophic levels, as well as a new perspective for the study of BEF mechanisms. The results revealed that the biomass of different trophic levels showed significant interannual fluctuations in Haizhou Bay, and their spatial distribution patterns also differed substantially. Species richness showed strong interannual fluctuations, whereas the changes in evenness were relatively small over time. The SEM model suggested that species diversity had substantially different influences on the biomass of different trophic levels. Specifically, richness had a significant positive correlation with the biomass of the middle trophic level, which reflected the complementary effect of ecological niches. Species evenness showed a negative correlation with the biomass of the middle and high trophic levels, consistent with the selection effects in the community. The biomass of low and high trophic levels was less responsive to the diversity indicators, and only evenness was found to correlate with the biomass of high trophic level. The diverged responses to biodiversity indicated that ecological niche complementarity and selection effects worked differently for different trophic levels. Additionally, significant correlations were found between different trophic levels, and environmental factors influenced the biomass through both direct and indirect pathways. That is, significant positive correlations were found between bottom temperature and high-trophic-level biomass and between salinity and low-trophic-level biomass, where there were negative correlations between the other environmental factors and biomass. This study reveals the complexity of the BEF relationships in marine ecosystems as noted by the differential effects of biodiversity on different trophic levels of the community, which emphasized that the trophic structure of the ecosystem should be taken into account when exploring the BEF relationships in marine ecosystems. In future studies, care should be taken not to treat the total community biomass in a simplified way, but to carry out analyses with full considerations of ecosystem structures, trophic-level compositions, and spatial and temporal scales, and to incorporate interactions among species to reveal the relationship between species diversity and biomass. The results may promote the scientific understanding on the BEF relationships in marine ecosystems, and provide theoretical support for the conservation of biodiversity and the collaborative management in Haizhou Bay. Further studies were suggested to focus on comparing different methodologies for SEMs, such as including latent variables and time lag effects in the model structure to explore the effects of time-continuous effects on biomass, in order to provide in-depth investigations of the BEF relationship in marine ecosystems.