Critical roles of the atmospheric feedback to ocean in the monsoon rainfall predictability

Climate predictability is model dependent. Assessment of the multi-model ensemble simulations made by 11 atmospheric general circulation models (GCMs) show a common feature: no skills in modeling seasonal rainfall variations over the Asian-Pacific monsoon convergence zone during the summers of 1997 and 1998 when the unprecedented 1997 El Nino occurred. Does this failure suggest that all models are poor or that the strategy used in climate simulation has deficiencies? We address this issue through analysis of observations and numerical experiments with a couple GCM. Over the Asian-Pacific monsoon convergence zone (a.k.a. Intertropical convergence zone), the observed summer mean SST and rainfall anomalies tend to show a significant negative correlation. This negative correlation reaches maximum when the rainfall leads SST by one-month (using monthly mean data), suggesting that the atmosphere plays an active role in changing local SSTs over the monsoon convergence zone. The ECHAM model, when couples with an ocean model, can realistically reproduced this negative correlation. In contrast, the ECHAM model alone simulation forced by the same daily SST generated by the coupled model yields a positive rainfall-SST correlation, suggesting that in the forced model simulation, the SST anomalies determine the atmospheric response. These results suggest that the atmosphere-warm ocean interaction plays an essential role in realistic modeling (or predicting) summer monsoon rainfall. The atmospheric precipitation anomalies cannot be realistically reproduced by AGCM alone with observed SST as a forcing. Therefore, improvement of an atmospheric model based on its stand-alone performance is a poor strategy. Similarly, the two-tier approach for seasonal mean rainfall prediction may work in many regions of the world but not in the major summer monsoon convergence zones, where the local air-sea interaction rather than the ocean memory provides a source of predictability.