Machine learning-integrated hydrogeochemical and spatial modeling of groundwater quality indices for seawater intrusion and irrigation sustainability in coastal agroecosystems of Skhirat Region, Morocco

Abstract

Study region

Skhirat coastal aquifer, Morocco.

Study focus

This study aimed to evaluate groundwater quality for drinking and irrigation, quantify seawater intrusion (SWI), and explore the added value of machine learning (ML) models for predicting groundwater indices. A total of 30 groundwater samples were collected and analyzed for physicochemical parameters. Hydrogeochemical characteristics were assessed using Piper, Gibbs, and Chadha diagrams. Water Quality Index (WQI), Irrigation Water Quality Index (IWQI), and Saltwater Mixing Index (SMI) were computed. Statistical tools (correlation matrix, PCA, K-means clustering) and GIS-based spatial interpolation were applied. Additionally, Random Forest (RF) and Artificial Neural Networks (ANN) models were tested to estimate groundwater indices and assess predictive performance.

Key findings and implications

Results showed WQI values ranging from 31.58 to 139.28, with 40 % of samples falling into the “poor” to “very poor” categories for drinking. IWQI revealed that 43.3 % of samples were “good,” while 6.7 % were “very poor” for irrigation suitability. SMI values exceeded 1 in 30 % of samples, confirming SWI in northwestern zones. ANN achieved higher accuracy for IWQI prediction (R² = 0.81), while RF performed best for SMI (R² = 0.74). Spatial analysis confirmed that salinization intensified toward the coast. These findings highlight the importance of integrating hydrogeochemical analysis, geospatial mapping, and ML modeling for sustainable groundwater management in Morocco’s coastal agroecosystems.