Saharan dust intrusions influence hydro-climatic conditions across North Africa, but their long-term links with rainfall variability and agricultural moisture stress remain poorly quantified. This study examined these interactions from 1990 to 2024 using multi-source remote sensing datasets and explainable machine learning. Aerosol Optical Depth (AOD), CHIRPS rainfall, and MODIS-derived land surface temperature (LST), NDVI, and evapotranspiration (ET) were integrated in Google Earth Engine to derive a Hybrid Crop Water Stress Index (CWSI). Random Forest and XGBoost models, supported by SHAP interpretation, were used to identify the main controls of moisture stress. Results showed a significant decline in AOD (−0.0051 yr⁻¹, p < 0.001), suggesting reduced dust loading and possible Sahelian vegetation recovery. Rainfall increased weakly but insignificantly (0.22 mm yr⁻¹, p = 0.150), reflecting alternating wet and dry episodes. Hybrid CWSI declined significantly (p = 0.023), indicating reduced moisture stress as increasing NDVI and ET partly offset thermal pressure. Monthly analysis showed strong positive associations between AOD and moisture stress (r = 0.74) and between LST and stress (r = 0.87). XGBoost performed best (R² = 0.326), with SHAP identifying rainfall and AOD as dominant predictors.