Assessment of regional seasonal rainfall predictability using the CPTEC/COLA atmospheric GCM

This is a study of the annual and interannual variability of regional rainfall produced by the Center for Weather Forecasts and Climate Studies/Center for Ocean, Land and Atmospheric Studies (CPTEC/COLA) atmospheric global climate model. An evaluation is made of a 9-member ensemble run of the model forced by observed global sea surface temperature (SST) anomalies for the 10-year period 1982-1991. The Brier skill score and, Relative Operating Characteristics (ROC) are used to assess the predictability of rainfall and to validate rainfall simulations, in several regions world wide. In general, the annual cycle of precipitation is well simulated by the model for several continental and oceanic regions in the tropics and mid latitudes. Interannual variability of rainfall during the peak rainy season is realistically simulated in Northeast Brazil, Amazonia, central Chile, and southern Argentina-Uruguay, Eastern Africa, and tropical Pacific regions, where the model shows good skill. Some regions, such as northwest Peru-Ecuador, and southern Brazil exhibit a realistic simulation of rainfall anomalies associated with extreme El Nino warming conditions, while in years with neutral or La Nina conditions, the agreement between observed and simulated rainfall anomalies is not always present. In the monsoon regions of the world and in southern Africa, even though the model reproduces the annual cycle of rainfall, the skill of the model is low for the simulation of the interannual variability. This is indicative of mechanisms other than the external SST forcing, such as the effect of land-surface moisture and snow feedbacks or the representation of sub-grid scale processes, indicating the important role of factors other than external boundary forcing. The model captures the well-known signatures of rainfall anomalies of El Nino in 1982-83 and 1986-87, indicating its sensitivity to strong external forcing. In normal years, internal climate variability can affect the predictability of climate in some regions, especially in monsoon areas of the world.