摘要
鄱阳湖是中国第一大淡水湖和国际重要湿地,对区域经济发展和生态文明建设都具有非常重要的作用。近年来受气候变化及流域经济发展影响,其水质持续逼近轻度富营养,局部水华发生风险较高。为保护鄱阳湖水生态环境,探索适用于鄱阳湖的大尺度水质遥感监测方法至关重要。文章以鄱阳湖为实验区域,结合2018年7月和2019年8月两次鄱阳湖丰水期的实测水质数据和"高分一号"卫星影像,基于梯度提升决策树算法构建水质参数反演模型,反演了高锰酸盐指数、总磷、总氮、透明度、叶绿素a、悬浮泥沙等6种水质参数。对反演算法的输入波段和参数配置进行了调试与优化,以均方根误差和决定系数作为精度评价指标,测试了该算法对各水质参数反演的精度和速度,结果表明,梯度提升决策树算法反演各水质参数的精度较高且速度较快,对多数水质参数反演的决定系数在0.8以上,具有实用价值,能够实现对内陆复杂水体水质的高精度遥感监测。
Poyang Lake is the largest fresh water lake in China and an important wetland in the world,which plays an important role in economic development and ecological civilization construction.In recent years,due to the impact of climate change and economic development of the basin,its water quality continues to approach light eutrophication level,and the risk of algal bloom becomes higher.In order to protect the ecological environment of Poyang Lake,it is very important to explore a suitable method to monitor the water quality of Poyang Lake using large-scale remote sensing technology.In this paper,taking Poyang Lake as the experimental area,combined with the measured data of water quality parameters of Poyang Lake in July 2018 and August 2019 and the remote sensing images of GF-1 satellites,several retrieval models of water quality parameters were constructed based on gradient boosting decision tree algorithm,and six water quality parameters including permanganate index,total phosphorus,total nitrogen,transparency,chlorophyll-a concentration and sediment content were retrieved.Taking root-mean-square error(RMSE)and decision coefficient(R2)as the accuracy measures,the input band and parameter configuration of the retrieval models were debugged and optimized,and the accuracy and speed of the retrieval algorithms for each water quality parameter are tested.The result shows that the gradient boosting decision tree algorithm has high accuracy and fast speed in retrieving various water quality parameters,and the decision coefficients in retrieving most water quality parameters are more than 0.8,which means this algorithm has practical value and can realize remote sensing monitoring with high accuracy water quality for complex inland water body.
作者
李怡静
孙晓敏
郭玉银
刘发根
周冠华
徐崇斌
刘亮
LI Yijing;SUN Xiaomin;GUO Yuyin;LIU Fagen;ZHOU Guanhua;XU Chongbin;LIU Liang(School of Instrumentation and Optoelectronic Engineering,Beihang University,Beijing 100191,China;Beijing Institute of Space Mechanics&Electricity,Beijing 100094,China;Hydrology Bureau of Poyang Lake,Jiujiang 332803,China)
出处
《航天返回与遥感》
CSCD
2020年第6期90-102,共13页
Spacecraft Recovery & Remote Sensing
基金
国家自然科学基金项目(41971320)
江西省水利厅科技项目遥感技术在鄱阳湖水质生态监测中的应用(201820TG07)。
关键词
遥感反演
机器学习
梯度提升决策树算法
“高分一号”卫星数据
水环境
鄱阳湖
remote sensing retrieval
machine learning method
gradient boosting decision tree algorithm
GF-1 satellite data
water environment
Poyang Lake
