Predicting peak summer monsoon precipitation over Pakistan in ECMWF SEAS5 and North American Multimodel Ensemble

Abstract

AbstractThe potential and actual forecast skill of peak summer monsoon precipitation (July–August [JA]) over the core summer monsoon region of Pakistan (CSMRP: 68°–76°E, 30°–36°N) is investigated. The predictions from the latest version of the European Centre for Medium‐Range Weather Forecast System5 (SEAS5) and the North American Multimodel Ensemble (NMME) initialized in the beginning of June, May, and April are utilized. The potential skill is estimated by signal‐to‐noise ratio and perfect model correlation. The forecast skill is evaluated using the temporal anomaly correlation as well as the ranked probability skill score for probabilistic forecasts. The SEAS5 generally captures the observed climatological mean and variability patterns of peak summer monsoon precipitation over the region, with some biases mainly located over the complex topography zones. Observed SST anomalies in the El Niño–Southern Oscillation (ENSO) region are negatively correlated with observed peak summer monsoon precipitation over the region. The ENSO–precipitation connection is weakly reproduced in SEAS5 at Lead‐1 and fades gradually with increasing lead time. The potentially predictable component (signal) is small compared to the unpredictable part (noise). Nonetheless, statistically significant potential skill is found over the CSMRP. However, the forecast skill of peak summer monsoon precipitation over the region is lower than the potential skill in SEAS5. A multimodel ensemble average of five models shows little benefit in predicting peak monsoon precipitation over Pakistan. Extended spatial averaging has some impact on the skill; however, temporal averaging has no effect on the skill of the summer monsoon precipitation in SEAS5. Together these results show limitations of the current seasonal prediction models and demonstrate their inadequate performance in predicting peak summer monsoon precipitation over Pakistan.