关键词: 叶绿素a, 透明度, 水体分类, 光谱指数, 水质参数图, 巢湖

Abstract: Chlorophyll-a concentration and transparency,as water quality parameters related to water spectral characteristics,are widely used to remotely assess water quality.In this research,the in-situ water monitoring data and the corresponding four GF/ZY satellite images from January.to September 2016 were used in Chaohu Lake,Anhui Province.K-means as well as SVM methods were used to group water spectrum into three typical types:Chla-dominant,SD-dominant and dual-dominant water,and then the spectral indices NDWC and NDWS were created for Chla and SD estimations.For estimating Chla concentration,NDWC was used for Chla-dominant and dual-dominant water,RVIgreen was used for SD-dominant water.The RMSE was 0.044 3 mg/L which was much lower than those by using the traditional spectral index(NDVI,RMSE=0.053 4 mg/L)and the machine learning algorithm(GDBT,RMSE=0.051 5 mg/L).For estimating the transparency,NDWS was used for SD-dominant and dual-dominant water,G was used for Chla-dominant water.The RMSE was 0.0224 m,which was also lower than those by using the traditional spectral index(NDVI,RMSE=0.030 6 m)and the machine learning algorithm(KNN,RMSE=0.0272 m).The proposed method was applied in remotely sensed images of four periods for mapping water quality of Chaohu Lake in 2016,and the results showed that Chlorophyll-a concentration in the second period(April and June)was the highest while the transparency was the lowest,the first period(March)and the fourth period(September and November)were moderate,while the third period(August)had the lowest Chlorophyll-a concentration and the highest transparency.In general,the water quality in the eastern Lake area was better than that in the western Lake area.

Key words: chlorophyll-a, transparency, water body classification, spectral index, water quality parameter map, Caohu Lake;