基于随机森林法的镇江市地表水质反演模型构建  

Inversion Modeling of Surface Water Quality in Zhenjiang City Based on Random Forest Method

作　　者：罗刚 郭仁庆 刘春乐 井潇溪 LUO Gang;GUO Ren-qing;LIU Chun-le;JING Xiao-xi(Jiangsu Zhenjiang Environmental Monitoring Center,Zhengjiang,Jiangsu 212000,China;Jiangsu Environmental Monitoring Center,Nanjing 210019,China;Jiangsu Hongwan Weipeng Information Technology Co.,Ltd.,Zhengjiang,Jiangsu 212000,China)

机构地区：[1]江苏省镇江环境监测中心,江苏镇江212000 [2]江苏省环境监测中心,南京210019 [3]江苏虹湾威鹏信息技术有限公司,江苏镇江212000

出　　处：《四川环境》2025年第2期141-148,共8页Sichuan Environment

基　　金：2023年度江苏省环境监测科研基金(2321);2024年度镇江市社会发展指导性科技计划项目(Fz2024113);2025年度江苏省镇江环境监测中心科研基金(ZJ2503)。

摘　　要：旨在通过构建基于随机森林法的遥感反演模型,利用哨兵2号卫星遥感影像数据,对镇江市汛期前后地表水体中的总磷(TP)和总氮(TN)浓度变化进行监测。研究采用了随机森林算法结合多光谱数据,通过比较汛期前后的水体光谱特征,构建了反演模型,并使用现场手工监测数据进行模型验证。结果显示,汛期后镇江市92.3%水域面积的水体中总磷和84.6%水域面积的水体中总氮浓度显著上升,其中吕城断面总磷指标浓度上升40.9%,北山水库总氮指标浓度上升64.1%。模型验证表明,遥感反演结果与手工监测结果的相对误差总磷浓度不超过23.4%,总氮浓度不超过26.5%,证实了该技术在水质反演应用上的可行性。本研究的意义在于为水质监测与管理提供了一种快速、可靠的数据支持方法,有助于汛期水质保护和改善措施的制定与实施。In this study,a remote sensing inversion model based on the random forest method was constructed to monitor the changes of total phosphorus(TP)and total nitrogen(TN)concentrations in surface water before and after the flood season in Zhenjiang City by using the remote sensing image data of Sentinel-2.A random forest algorithm was used in this study in combinationwith multispectral data,an inversion model was built by comparing the spectral characteristics of the water bodies before and after the flood season,and on-site manual monitoring data was used for model validation.The results showed that the total phosphorus in 92.3%of the water bodies and the total nitrogen in 84.6%of the water bodies in the watershed area of Zhengjiang City increased significantly after the flood season,with the concentration of total phosphorus in the Lvcheng section increasing by 40.9%and the concentration of total nitrogen in the Beishan Reservoir increasing by 64.1%.The model validation showed that the relative errors between the remote sensing inversion results and the manual monitoring results were no more than 23.4%for the concentration of total phosphorus and 26.5%for the concentration of total nitrogen,which confirmed the feasibility of this technique in the application of water quality inversion.The significance of this study is to provide a fast and reliable data support method for water quality monitoring and management,which can help the development and implementation of water quality protection and improvement measures during flood seasons.

关 键 词：卫星遥感 随机森林法 多光谱 汛期 水质反演 

分 类 号：X87[环境科学与工程—环境工程]