Sentinel-3 OLCI observations of Chinese lake turbidity using machine learning algorithms
| 第一作者： |
Li, Yong |
| 英文第一作者： |
Li, Yong |
| 联系作者： |
Li, Sijia |
| 英文联系作者： |
Li, Sijia |
| 发表年度： |
2023 |
| 卷： |
622 |
| 摘要： |
Turbidity has a substantial effect on light propagation in water and is an optical indicator for monitoring water quality. The Ocean and Land Colour Instrument (OLCI) installed in the Sentinel-3 satellite could help to monitor turbidity variations due to its narrow spectral bandwidth and high revisit time. This makes us develop suitable turbidity algorithms for mapping the dynamics of lake turbidity using OLCI imagery. In this study, a total of 1081 samples were collected from 73 typical lakes across China during 2017-2022. Here, we developed and proposed applicable turbidity algorithms i.e., SR (simple regression), PLSR (partial least squares regression), SVR (support vector regression), BP (backpropagation neural network), KNN (K-nearest neighbor), RF (random forest), and XGBoost (extreme gradient boosting), which integrate a broad scale dataset of turbidity ranging from 0.15 NTU to 262.57 NTU using OLCI imagery. The performance results of developed turbidity algorithms demonstrated that RF obtained performed best with good agreements of measured-and derived-fitting (validation dataset: R2 = 0.92, MAPE = 54.32%, RMSE = 12.65 NTU). This allowed us to estimate turbidity concentration in Chinese lakes with a surface area>20 km2. The averaged turbidity level of lake groups in five lake regions was inves-tigated with a decreased tendency as Northeast Plain Lake Region (NPL, 63.31 NTU) > Eastern Plain Lake Region (EPL, 58.30 NTU) > Inner Mongolia-Xinjiang Lake Region (IMXL, 16.64 NTU) > Yunnan-Guizhou Plateau Lake Region (YGPL, 11.63 NTU) > Tibetan Plateau Lake Region (TPL, 7.05 NTU). Further, we analyze the distribu-tions of lake turbidity considering different land use types, and the results showed a decreasing tendency as cropland (56.97 NTU) > barren land (11.22 NTU) > grassland (8.06 NTU) > forest (4.71 NTU). Our results demonstrated and highlighted the quantification of lake turbidity concentrations using OLCI imagery and the stability of the RF algorithm, which can generate turbidity data for large-scale monitoring and decision-making related to environmental protection. |
| 刊物名称： |
JOURNAL OF HYDROLOGY |
| 参与作者： |
Li,Y(Li,Yong)[1],[2];Li,SJ(Li,Sijia)[1],[2];Song,KS(Song,Kaishan)[1];Liu,G(Liu,Ge)[1];Wen,ZD(Wen,Zhidan)[1];Fang,Chong[1];Shang,Yingxin[1];Lyu,Lili[1],[3];Zhang,Lele[1] |
吉公网安备22017302000214号
