This study aims to analyze the differences in acoustic density of fishery resources between surface and bottom layers (surface mixed layer and bottom cold water layer) in the northern South China Sea and to explore the relationship between these differences and 41 abiotic factors. This research provides a scientific basis for the effective management and conservation of fishery resources in the northern South China Sea. The northern offshore area of the South China Sea is a crucial traditional fishing ground and an important spawning and feeding ground for marine fish. In recent years, fishery resources in this region have shown significant declines in age, size, and quality, attracting significant attention from both academia and fishery management authorities. Fishery acoustic methods were employed, using a Simrad EY60 split-beam scientific echosounder to collect acoustic data in the northern South China Sea. Acoustic data were analyzed using the Echoview fishery acoustic data processing system to calculate the acoustic density (NASC) of surface and bottom layers. Extreme Gradient Boosting (XGBoost) and Random Forest algorithms were utilized to model the influence of 41 abiotic factors on the differences in acoustic density and to assess the importance of these factors. Results indicated that the bottom layer had significantly higher acoustic density than the surface layer, with mean values of 106.00 m²/nmi² and 43.39 m²/nmi², respectively. Both XGBoost and Random Forest models performed similarly, with temperature factors (bottom 2 m temperature, surface-bottom temperature difference, and surface 2 m temperature) and water depth identified as the most critical factors affecting acoustic density differences. The negative value region, where surface density exceeds bottom density, is primarily distributed around Hainan Island. The study concluded that temperature and water depth are the primary factors influencing the distribution differences of fishery resources, while human activities may also contribute by altering the concentrations of factors such as phosphate and chlorophyll. Additionally, the discussion highlights the implications of these findings for fisheries management, suggesting that targeted measures to monitor and regulate temperature and nutrient levels could significantly improve resource sustainability. The analysis underscores the importance of incorporating advanced machine learning algorithms in marine resource assessment to enhance the accuracy and reliability of environmental impact evaluations. These findings provide vital scientific insights for the management and conservation of fishery resources in the northern South China Sea, offering a comprehensive understanding of the environmental factors that drive spatial distribution patterns in marine ecosystems. This research thus lays a foundation for future studies aiming to mitigate the impacts of climate change and human activities on marine biodiversity and resource availability.