Water quality estimates using machine learning techniques in an experimental watershed

• DOI: 10.2166/hydro.2024.132
• Publication Date: 2024-11-05T16:34:45Z
• Authors: David Costa; Yared Bayissa; Kargean Vianna Ba...; Mariana Dias Vill...; Arun Bawa; Jader Lugon Junior; Antônio J. Silva ...; Raghavan Srinivasan
• Keywords: 01 natural sciences; 0105 earth and related environmental sciences

ABSTRACT This study aims to identify the best machine learning (ML) approach predict concentrations of biochemical oxygen demand (BOD), nitrate, and phosphate. Four ML techniques including Decision tree, Random Forest, Gradient Boosting XGBoost were compared estimate water quality parameters based on biophysical (i.e., population, basin area, river slope, level, stream flow), physicochemical properties conductivity, turbidity, pH, temperature, dissolved oxygen) input parameters. The innovation lies in combination on-the-spot variables with additional characteristics watershed. model performances evaluated using coefficient determination (R2), Nash-Sutcliffe efficiency (NSE), Root Mean Squared Error (RMSE) Kling-Gupta Efficiency (KGE) coefficient. robust five-fold cross-validation, along hyperparameter tuning, achieved R2 values 0.71, 0.66, 0.69 for phosphate, BOD; NSE 0.67, 0.65, 0.62, KGE 0.64, 0.75, 0.60, respectively. yielded good results, showcasing superior performance when considering all analysis performed, but his was closely match by other algorithms. overall modeling design approach, which includes careful consideration data preprocessing, dataset splitting, statistical evaluation metrics, feature analysis, curve are just as important algorithm selection.