Efficacy of GIS-based AHP and data-driven intelligent machine learning algorithms for irrigation water quality prediction in an agricultural-mine district within the Lower Benue Trough, Nigeria.

Agricultural productivity can be impaired by poor irrigation water quality. Therefore, adequate vulnerability assessment and identification of the most influential water quality parameters for accurate prediction becomes crucial for enhanced water resource management and sustainability. In this study, the geographical information system (GIS), analytical hierarchy process (AHP) technique, and machine learning models were integrated to assess and predict the irrigation water quality (IWQ) suitability of the Okurumutet-Iyamitet agricultural-mine district. To achieve this, six water quality criteria were reclassified into four major hazard groups (permeability and infiltration hazard, salinity hazard, specific ion toxicity, and mixed effects) based on their sensitivity on crop yield. The normalized weights of the criteria were computed using the AHP pairwise comparison matrix. Eight thematic maps based on IWQ parameters (electrical conductivity, total dissolved solids, sodium adsorption ratio, permeability index, soluble sodium percentage, magnesium hazard, hardness, and pH) were generated and rasterized in the ArcGIS environment to generate an irrigation suitability map of the area using the weighted sum technique. The derived IWQ map showed that the water in 28.2% of the area is suitable for irrigation, 43.7% is moderately suitable, and 28.1% is unsuitable, with the irrigation water quality deteriorating in the central-southeastern direction. Two machine learning models-multilayer perceptron neural networks (MLP-NNs) and multilinear regression (MLR)-were integrated and validated to predict the IWQ parameters. The coefficient of determination (R) for MLR and MLP-NN ranged from 0.513 to 0.858 and 0.526 to 0.861 respectively. Based on the results of all the metrics, the MLP-NN showed higher performance accuracy than the MLR. From the results of MLP-NN sensitivity analysis, HCO, Cl, Mg, and SO were identified to have the highest influence on the irrigation water quality of the area. This study showed that the integration of GIS-AHP and machine learning can serve as efficient and rapid decision-making tools in irrigation water quality monitoring and prediction.