中国水产科学2026，Vol.33Issue(6)：1-16,16.DOI:10.12264/JFSC2026-0136
基于可解释机器学习的浙江近岸春季龙头鱼时空分布与环境因子关系研究
Study on the relationship between spatiotemporal distribution of spring Harpadon nehereus and environmental factors in the coastal waters of Zhejiang based on interpretable machine learning
摘要
Abstract
Harpadon nehereus represents a commercially significant fish species widely distributed in the coastal waters of the East China Sea.While its distribution pattern is known to be strongly modulated by marine environmental fluctuations,the spatiotemporal dynamics of its spring resources and the underlying environmental driving mechanisms in Zhejiang coastal waters remain poorly characterized.Based on bottom trawl survey data collected annually in April across the period 2014-2024 in the coastal waters of Zhejiang,the present study incorporated a suite of environmental variables including sea surface temperature(SST),sea surface salinity(SSS),bottom sea temperature(BST),bottom sea salinity(BSS),surface chlorophyll-a concentration(Schla)and water depth,alongside spatiotemporal variables of year,longitude and latitude.Six distinct models,namely the generalized additive model(GAM),random forest(RF),boosted regression tree(BRT),eXtreme Gradient Boosting(XGBoost),artificial neural network(ANN)and deep neural network(DNN),were constructed to predict the spatial distribution of H.nehereus biomass,with model performance systematically assessed via repeated five-fold cross-validation.The results revealed that high-biomass stations of H.nehereus in spring exhibited pronounced spatial aggregation during 2014-2024,forming predominantly localized patches or belt-shaped distributions along the Zhejiang nearshore continental shelf,with the core distribution area confined to 122.0°～123.0°E and 29.0°～30.2°N.The annual cumulative biomass index peaked in 2022 and reached its minimum in 2023.Among the six competing models,random forest(RF)delivered the most satisfactory overall predictive performance.Variable importance analysis demonstrated that longitude,depth and bottom sea salinity(BSS)contributed disproportionately to model outputs,with longitude primarily capturing the inherent spatial structure of resource distribution.Further habitat suitability prediction indicated that the central and northern coastal waters of Zhejiang serve as critical spring habitats and resource hotspots for H.nehereus.These findings advance our understanding of the spring resource distribution patterns and associated environmental driving mechanisms of H.nehereus in Zhejiang coastal waters,and provide a robust scientific foundation for nearshore fishery resource management and targeted habitat conservation efforts.关键词
龙头鱼/浙江近岸/时空分布/机器学习/栖息地适宜性Key words
Harpadon nehereus/Zhejiang coastal waters/spatiotemporal distribution/machine learning/habitat suitability分类
农业科技引用本文复制引用
董书奇,解明阳,董钇江,薛利建,周永东,张洪亮,陈峰,朱泽瑞,朱文斌..基于可解释机器学习的浙江近岸春季龙头鱼时空分布与环境因子关系研究[J].中国水产科学,2026,33(6):1-16,16.基金项目
国家重点研发计划项目(2024YFD2400703). （2024YFD2400703）
