Estimation Of Lake Turbidity In China Based On Sentinel-3 OLCI Images

Turbidity has a critical impact on light transmission in the water column and often varies in conjunction with total suspended solids,transparency,and colored organic dissolved matter.It can be used both as an optical environmental indicator for water quality monitoring and as an easily measurable proxy for suspended particulate matter concentrations in sediment transport,and turbidity causes changes in underwater light transmission that have important implications for lake primary productivity.This study aimed to investigate the feasibility of using Sentinel-3 OLCI data to map turbidity in Chinese lakes with an area of ≥ 20 km2.Eight hundred and forty-five measured turbidity data collected from lakes in China between 2017 and2021 were matched with Sentinel-3 OLCI imagery for star-ground synchronization.The turbidity dataset was divided into three subsets,two of which were used to build exponential regression,partial least squares regression,support vector machine,K-most proximity,BP neural network,random forest,and XGBoost models,and the other subset was used to validate the algorithm performance.The results show that.(1)XGBoost model is the best(R2 = 0.90,MAE: 9.07 FNU,RMSE: 15.53FNU).Therefore,the XGBoost model was used to estimate the lake turbidity.(2)The turbidity of lakes in China was estimated using OLCI images for 2021.The results showed a large regional variation in turbidity,(turbidity range:0.43-200.22 FNU.NPL(63.46 FNU),EPL(58.66 FNU),IMXL(16.99 FNU),YGPL(11.78 FNU)and TPL(7.46 FNU)).(3)The results of land use showed that different land use types influenced the spatial pattern of lake turbidity.Among them,cropland(56.97 FNU)> wasteland(11.22 FNU)> grassland(8.06 FNU)> forest(4.71 FNU).These results demonstrate the stability of the XGBoost model for estimating turbidity in large-scale observations,which will facilitate the monitoring of lake turbidity and help water quality management and environmental protection.