Predicting river water quality: An imposing engagement between machine learning and the QUAL2Kw models (case study: Aji-Chai, river, Iran)

Abstract

Rivers play an essential role in supplying high-quality water to diverse sectors. Understanding water quality indicators and systematic monitoring is crucial for water resources management and macro-level decision-making. In this context, the forthcoming article delves into the simulation of three crucial parameters, namely EC, SAR, and TDS, through a reach of 106 km length along the Aji-Chai River, Iran, encompassing stations from Markid, Khajeh, Akhola, and Serin Dizj. This simulation employs three advanced machine learning models: SVM, GEP, and MLP, in conjunction with the QUAL2Kw mathematical simulator. The study meticulously evaluates the performance of these models using four key indices: RMSE, MAE, R2, and DDR. The calculated results unequivocally establish the superiority of the SVM in simulating three essential water quality parameters across all stations. This is supported by consistently high R2 and DDR values, along with low RMSE and MAE values. While the mathematical model used in this study showed reasonable accuracy in simulating the parameters under investigation, it consistently performed less effectively than the SVM model. In summary, the SVM model with specific parameters (C = 68.5, ε = 4.55, and γ = 205) emerges as the optimal choice for accurately simulating river water quality parameters based on the conducted study.

Keywords

• Water surface quality
• Evaluation process
• Artificial intelligence
• Aji-Chai river