Semi-arid regions are particularly vulnerable to climate change, characterized by rising temperatures, altered precipitation regimes, and more frequent droughts. These factors intensify water scarcity and pose significant challenges to agricultural sustainability. To address this challenges, efficient irrigation management is essential to alleviate crop water stress and maintain yields. This study applies the Interactive Canopy Radiative Exchange (ICARE) soil–vegetation–atmosphere transfer (SVAT) model to simulate water and energy fluxes over four winter wheat fields in Morocco's Tensift Basin. Model calibration was performed on a reference field using the Non-dominated Sorting Genetic Algorithm II (NSGA-II), with a focus on a root zone soil moisture (RZSM)-constrained approach to enhance simulation accuracy. Model performance was evaluated across key phenological stages; initial, development, mid-season, and late season by assessing RZSM, surface energy balance components, and evapotranspiration (ET). The model demonstrated strong predictive capability, with optimal simulations achieving RMSE (R²) values of 0.01 m³/m³ (0.88) for RZSM and 0.46 mm/day (0.83) for ET. Simulations also showed high agreement with observed values: 40 W/m² (0.96) for net radiation, 21 W/m² (0.86) for ground heat flux, and 37 W/m² (0.73) for sensible heat flux. A clear phase-dependent behavior was observed, with the RZSM-constrained calibration yielding particularly accurate results during the development and mid-season stages, when plant transpiration is dominant. Validation conducted on three additional fields confirmed the robustness and transferability of the calibrated model, reinforcing its potential as a decision-support tool for improving irrigation efficiency in semi-arid agricultural systems. Overall, this study advances the understanding of soil–plant–atmosphere interactions and offers practical insights for sustainable water resource management under changing climate conditions.