Data-Driven Water Management Through Machine Learning for Predicting Total Dissolved Solids in Treated Wastewater Used for Irrigation

Regions facing water scarcity use treated wastewater (TWW) to conserve the available freshwater resources mainly for irrigation while ensuring safety and quality standards that can assure public health and prevent any adverse impacts on crops or the environment. Since total dissolved solids (TDS) in irrigation water can have several adverse effects on crops based on their sensitivity, therefore in the current study, the TWW samples collected from an agricultural field near the Al-Qatif region located in the Eastern Province of Saudi Arabia were assessed, pre-processed and fed into supervised machine learning (ML) techniques, namely Multiple Linear Regression (MLR) and Support Vector Regression (SVR) to predict TDS. Based on the most suitable inputs assessed in the lab through an IoT-enabled natural time monitoring system and a multi-probe device, two different model combinations were selected for each ML approach, i.e., MLR-M1, MLR-M2, SVR-M1, and SVR-M2. Four statistical assessment measures, namely correlation coefficient (R), coefficient of determination (R2), mean square error (MSE), and root mean square error (RMSE), were used to assess prediction performance. During the training phase, MLR-M1 and MLR-M2 demonstrated an impressive R2 value of 0.99 and 0.99, respectively. In contrast, SVR-M1 attained an R2 of 0.98, and SVR-M2 reached an R2 of 0.98. Transitioning to the testing phase, MLR-M1 maintained high predictive accuracy with an R2 of 0.96, and MLR-M2 also exhibited strong performance with an R2 of 0.96. Meanwhile, SVR-M1 and SVR-M2 showed slightly lower R2 values, with 0.87 and 0.87, respectively. The MLR models outscored others when their overall performance was examined across all datasets. It is most likely a result of its more vital generalization skills. However, exploring a more comprehensive range of ML algorithms and water quality datasets in future studies holds promise for further refinement and enhanced generalization capabilities.