An evaluation of East Asian summer monsoon seasonal forecast with NMME hindcast data

Abilities of three models (CFSv2, CanCM3 and CanCM4) in North American Multimodel Ensemble for seasonal forecast of East Asian summer monsoon (EASM) were evaluated with their 29-year hindcast data (1982-2010). Three models and their ensemble mean can forecast the climatological features of monsoon circulation and precipitation well, although underestimation exists in forecasting the upper-level circulation systems and low-level southwesterly over Indian continent and the western North Pacific (WNP) subtropical high. For the intra-seasonal migration of monsoon precipitation, bias of weaker precipitation magnitude is universal as well as the southward shifted WNP summer monsoon (WNPSM) rain-belt. CanCM3 and CanCM4 forecast earlier-than-observed monsoon onset and retreat. The year-to-year variation of EASM and its associated low-level winds and precipitation anomalies are generally well-forecasted with correlation coefficients from the MME against the observation of 0.84 for WNPSMI and 0.61 for EASMI. Forecast skills generally decrease as the lead increases. The skill for EASMI is not as good as WNPSMI, in which only CFSv2 forecasts well at leads up to 6 months, whilst the skill from CanCM3 (CanCM4) decreases rapidly when the lead increases to 2 months (3 months). The failure of CanCM3 is mainly attributed to the poor forecast of the relationship of EASMI with the El Nino-Southern Oscillation and northern Indian Ocean-WNP sea surface temperature anomaly. Finally, sources of the forecast error (FE) for monsoon precipitation were analyzed. Generally speaking, FE is much larger than predictability error (PE), indicating that model deficiency plays a dominant role in limiting EASM precipitation forecast. However, PE cannot be neglected from 0-month lead to 2-month lead over the areas such as the tropical western Pacific in CFSv2, the India continent in CanCM3 and the Tibetan Plateau in CanCM4. As lead time increases, FE dose not remarkably change while PE reduces significantly.