ABSTRACT 

This study evaluates the potential of two models within the North American Multi-Model Ensemble (NMME) system, i.e., CanCM3 and CanCM4, for improving drought risk management through reliable prediction. By employing the Standardized Precipitation Evapotranspiration Index (SPEI) and gridded datasets (GPCC and CRU), this study assesses their drought forecast capabilities across four semi-arid to arid basins in Iran. The results reveal that both models effectively capture drought events at short lead times (0.5 months), achieving correlation coefficients exceeding 0.93. The performance decline at longer lead times (3.5 months) is less severe in spring and autumn, maintaining correlations of >0.6 compared to summer. A Critical Success Index (CSI) analysis further highlights the models' skill in detecting summer drought events at a 1.5-month lead time (CSI >0.94), underscoring their utility for critical agricultural and water resource planning. Seasonal analysis shows CanCM4 outperforming CanCM3, particularly regarding CSI and correlation stability. These findings offer a novel contribution to understanding the applicability of CanCM3 and CanCM4 for drought forecast purposes in arid and semi-arid basins and underline their value for enhancing drought early warning systems and supporting efficient resource allocation to mitigate drought impacts.