The Skill of RegCM-4 in Forecasting Iran's Precipitation: A Basin-Scale Intra-Seasonal to Seasonal Analysis

As industries like agriculture and energy become increasingly reliant on weather information, there is a growing need for accurate medium to long-term precipitation forecasts. This study examined the skill of the Regional Climate Model v4.7 (RegCM4), forced by the CFSv2, in forecasting intra-seasonal to seasonal precipitation over seven specific basins in Iran. The analysis covered the period 2000-2019 and considered lead times of one to three months (L1 to L3). Deterministic and categorical statistical metrics were used to evaluate the model's skill. Results indicate significant regional variability in model performance, suggesting that climatic factors influence the model's accuracy. Deterministic metrics reveal that the RegCM4-CFSv2 model is better at forecasting precipitation for basins along the southern Caspian Sea coastal plains than for basins in central Iran. While categorical metrics exhibited the highest proficiency in forecasting the correct precipitation category for basins in the coastal plains of the Persian Gulf and the lowest proficiency for basins in northwestern Iran. Overall, the model exhibits reasonable skill, particularly at shorter lead times (e.g. L1). At the L1, the average Kling-Gupta efficiency (KGE) and correlation coefficient (CC) are around 0.3, and the relative root mean square error (RMSE) is approximately 28%. Performance weakens with increasing lead time, with significant deterioration at 3-month lead time (L3). Categorical metrics like probability of detection (POD) and threat score (TS) indicate moderate accuracy, especially for ‘below-normal (BN)’ and ‘above-normal (AN)’ categories. However, the model is prone to false alarms, particularly in the ‘normal (N)’ precipitation category. Results highlight the model's ability to detect upper and lower tercile categories (33rd and 66th percentiles), making it a valuable tool for developing effective flood or drought management strategies. Overall, the model shows promise for monthly precipitation forecasts but has limitations for long-range (beyond one month) prediction.