Assessment of Dynamic NMME Models for Meteorological Drought Forecasting in Western Regions of Iran

Abstract

Abstract This study assessed the performance of North American multimodel ensemble (NMME) dynamic systems in forecasting meteorological drought within the western and southwestern watersheds of Iran. Without suitable observational data in this region, the global Precipitation Climatology Centre (GPCC) precipitation and Climatic Research Unit (CRU) temperature datasets served as the foundation for comparative analysis. The standardized precipitation evapotranspiration index (SPEI) was employed for drought evaluation. The findings indicated that longer forecast horizons significantly reduced model accuracy. Furthermore, the assessment of drought predictability based on SPEI revealed that both CanCM3 and CanCM4 models could predict seasonal drought variations, particularly in the northern regions, with a correlation coefficient (CC) exceeding 0.93 at a forecast horizon of 0.5 months. While both models performed similarly at the watershed level in terms of root mean square error (RMSE), the CanCM4 model displayed a higher characteristic stability index (CSI) correlation (above 0.08) than CanCM3 in diagnosing drought. Seasonal variations were evident, with better drought predictions in northern regions during spring and more noticeable model performance differences between northern and southern regions in summer. The evaluation of forecasting capability in both hindcast and forecast periods showed no significant disparity between the models, albeit the CanCM4 model exhibited superior performance in some instances. These results provide valuable insights for water resource planners, enabling more effective decision-making in drought adaptation.