基于GRNN模型的流域沉积物中有效态氮、磷含量预测研究  

Concentration Prediction of Available Nitrogen and Phosphorus in Catchment Sediments Based on GRNN Model

作　　者：曹文斌 袁旭音[1] 章海燕 唐黎 汪宜敏 朱宁远 CAO Wen-bin;YUAN Xu-yin;ZHANG Hai-yan;TANG Li;WANG Yi-min;ZHU Ning-yuan(College of Environment,Hohai University,Nanjing 210098,China)

机构地区：[1]河海大学环境学院,江苏南京210098

出　　处：《环境科技》2023年第2期1-6,共6页Environmental Science and Technology

基　　金：国家自然科学基金项目(41372345,4602540);科技部国际合作专项(2102DFA60830)。

摘　　要：因沉积物中有效氮(AN)和有效磷(AP)的分布对流域水环境有重要影响,故研究其空间分布对评估和管理流域水环境质量有一定价值。基于GRNN模型和SHAP解释模型,选取多种土壤参数、自然环境与人类活动指标,以太湖流域数据集作为模型训练集,预测洪泽湖流域和钱塘江流域沉积物中AN,AP的含量。结果表明:以耕地为主的洪泽湖流域沉积物中AN的含量相对较高;钱塘江下游因城镇密集而导致沉积物中AP的含量较高。SHAP解释模型分析表明:沉积物中AN的含量主要受农业活动影响(除滨岸带土壤氮含量指标外);而沉积物中除磷肥指标外,生活污水排放量是AP含量主要影响因素。Since the distribution of available nitrogen(AN)and available phosphorus(AP)in sediments has a significant effect on the water environment in catchment,it is of great value to study its spatial distribution for the assessment and management of water environmental quality in catchment.Based on GRNN model and SHAP interpretation model,soil parameters,natural parameters and anthropogenic parameters were selected as input parameters,and data set of the Taihu catchment was used as model training set,to predict the content of AN and AP in sediments of the Hongze Lake catchment and the Qiantang River catchment.The results showed that the sediment AN content in the Hongze Lake catchment was relatively high,which is mainly induced by cultivated land.The sediment AP content in the lower reaches of the Qiantang River catchment was higher due to discharge from the dense towns.Based on the SHAP interpretation model,the distribution of sediment AN content is mainly influenced by agricultural activities(except soil nitrogen content),and the domestic sewage discharge is the main influencing factor except phosphate fertilizer for the distribution of sediment AP content.

关 键 词：沉积物 N P有效态 广义回归神经网络(GRNN) SHAP解释模型 流域 

分 类 号：X5[环境科学与工程—环境工程]