The optimal method for water quality parameters retrieval of urban river based on machine learning algorithms using remote sensing images

• DOI: 10.1080/01431161.2023.2209918
• Publication Date: 2023-05-12T12:54:30Z
• Authors: Yizhu Jiang; Jinling Kong; Yanling Zhong; Jingya Zhang; Zijia Zheng; Lizheng Wang; Dingming Liu

Water eutrophication has become one of the prominent problems environmental protection in inland watersheds. Turbidity, total phosphorus (TP) and nitrogen (TN) concentrations are key water quality parameters (WQPs) that reflect level waters. Due to complex interaction effects between different urban rivers, retrieval models still have problem single input features poor applicability. This paper proposed a robust feature selection method based on machine learning utilized Sentinel-2 remote sensing images for Chan Ba rivers Xi'an City. The ReliefF global sensitivity analysis (GSA) methods (ReliefF-GSA) were used select optimal combination from potential dataset. Based combination, Random Forest regression (RFR), LightGBM XGboost constructed three WQPs retrieval, respectively. then invert spatial-temporal variation January 2021 2022 was analysed. results show (1) RelieF-GSA is suitable high-dimensional filtration enables specific retrieval. It revealed BOI index (black odour index) turbidity TN concentration. (2) RFR model found be better than other more appropriate with coefficients determination (R2) 0.90, 0.89 0.81, (3) qualities seasonal characteristics. Turbidity TP showed higher, while concentration relatively low autumn. conclusions this can further provide reference WPQs rivers.