基于高分一号影像的沙河集水库水质遥感反演  

Remote Sensing Inversion of Water Quality in Shaheji Reservoir Based on GF-1 Image

作　　者：郭坤 李虎 陈冬花 谢以梅 冯武涛 GUO Kun;LI Hu;CHEN Dong-hua;XIE Yi-mei;FENG Wu-tao(School of Geography and Tourism,Anhui Normal University,Wuhu 241002,China;Anhui Province High-Resolution Earth Observation System Data Products and Application Software R&D Center,Chuzhou 239000,China)

机构地区：[1]安徽师范大学地理与旅游学院,安徽芜湖241003 [2]安徽省高分辨对地观测系统数据产品与应用软件研发中心,安徽滁州239000

出　　处：《安徽师范大学学报（自然科学版）》2023年第3期250-258,共9页Journal of Anhui Normal University(Natural Science)

基　　金：安徽省科技重大专项(202003a06020002);芜湖市重点研发项目(2020ms1-3).

摘　　要：沙河集水库是滁州市重要的水源供应地,关系着市民的饮水安全问题,水质的变化监测受到了政府的高度重视。通过遥感影像实现水质参数反演,对水库水质监测和保护工作具有非常重要的理论意义。因此,利用2021年6月和7月高分一号影像和同步的实测水质数据,建立了基于多元回归和BP神经网络的叶绿素a浓度和浊度模型,并进行精度评价,对研究区定量反演,以探求水库水质分布情况。结论如下:从模型精度来看,叶绿素和浊度进行多元回归模拟的RMSE分别为0.745、0.216,R^(2)为0.694、0.844,平均相对误差为13.19%、7.09%,BP神经网络模拟的RMSE分别为0.103、0.19,R^(2)为0.865、0.958,平均相对误差为2.5%、7.11%,两者对沙河集水库水质反演都具有可行性,但BP神经网络要优于多元回归;从水质遥感反演来看,2021年6月沙河集水库的叶绿素a浓度在2.2~5.6ug/l之间,7月叶绿素a浓度在1.8~6.7ug/l之间,均处于中营养状态,为Ⅱ类水质,6月浊度在0~4NTU之间,均值为1.16NTU,7月浊度值集中0.5~4.6NTU之间,库区均值为2.46NTU,总体达到了饮用水源标准。Shaheji Reservoir is an important source of water supply in Chuzhou City,and it is related to the safety of drinking water for citizens.The monitoring of water quality changes has received great attention from the gov‐ernment.The inversion of water quality parameters through remote sensing images has very important theoretical significance for the monitoring and protection of reservoir water quality.Therefore,the chl-a concentration and turbidity model based on multiple regression and BP neural network was established using GF-1 images in June and July 2021 and the synchronized measured water quality data,to explore the distribution of water quality in reservoirs.Conclusion:in terms of model accuracy,the RMSE of multiple regression simulation of chl-a and tur‐bidity are 0.745 and 0.216 respectively,and R^(2)is 0.694 and 0.844 respectively,the average relative error is13.19%,7.09%;The RMSE simulated by BP neural network is 0.103 and 0.19 respectively,and the R^(2)is 0.865and 0.958,the average relative error is 2.5%,7.11%.Both of them are feasible for the inversion of water quality of Shaheji Reservoir,but BP neural network is better than multiple regression;from the perspective of water qual‐ity remote sensing inversion,the chl-a concentration of Shaheji Reservoir in June 2021 is between 2.2~5.6ug/l,the chl-a concentration in July is between 1.8~6.7ug/l,both of which are in the mesotrophic state,and the water quality is Class II.The turbidity in June is between 0~4NTU,with an average value of 1.16NTU,and the turbidi‐ty in July is between 0.5~4.6 NTU,the average value of the reservoir area is 2.46NTU,which generally meets the drinking water source standard.

关 键 词：沙河集水库 高分一号 叶绿素A浓度 浊度 遥感反演 

分 类 号：X824[环境科学与工程—环境工程]