The Assessment of Machine Learning Algorithms for Predicting Irrigation Water Quality : A Comparative Study

The growing world population is increasing the demand for food, and climate change is causing erratic rainfall patterns. This has led to greater reliance on irrigation, especially in the Sahel and arid regions, to sustain food production. As freshwater resources become scarce, farmers are utilizing various water sources to keep their crops irrigated, aiming to ensure food availability and profitability. However, the rise in industrial activity is increasing pollution, which affects water quality, making its monitoring time-consuming and extensive. The aim of this study is to develop a machine learning approach for predicting irrigation water quality. To achieve this, a large dataset consisting of 1750 water samples was curated. The data were preprocessed, and Sodium Adsorption Ratio (SAR) and Irrigation Water Quality Index (IWQI) were computed. Five machine learning models (XGBoost, K-Nearest Neighbors, Support Vector Machine and Random Forest) were trained using the following parameters: Sodium(Na), Calcium(Ca²⁺), Bicarbonate (HCO3^-), Electical Conductivity (EC), and SAR. The results revealed that XGBoost outperformed the other algorithms, achieving a mean absolute error (MAE) of 0.90, a mean squared error (MSE) of 3.26, a root mean squared error (RMSE) of 1.81, and an R² of 0.96. The use of machine learning algorithms in predicting irrigation water quality is essential for farmers and crop planning, as it can save costs and time while ensuring healthy food production.