STATISTICAL-ANALYSIS OF TROPICAL CLIMATE ANOMALY SIMULATIONS

The statistical analysis of two atmospheric general circulation simula tions using the ECHAM3 GCM in permanent January conditions are present ed. The two simulations utilize different oceanic surface temperatures in the Atlantic as boundary conditions: the cold simulation has SST r epresenting the anomalous cold conditions during the decade 1904-13 wh ile the warn simulation has SST representative for the decade 1951-60 where anomalous warm conditions have been observed. The analysis conce ntrates on the simulated differences between both experiments within t he tropical belt to test the working hypothesis whether changes in the deep tropical heating initiated by the anomalous SST are responsible for the anomalies in the flow and mass field. We present a method whic h extracts the significant and dynamically consistent signal of the to tal difference using a multivariate statistical test based on the ampl itudes of an a-priori specified mode expansion. These expansion modes are defined from a variant of the Matsuno-Gill linearized reduced grav ity model for the tropical atmosphere. The application of the method s hows a clear and well defined tropical signal in the flow and mass fie ld which can be understood as the reponse of the ECHAM3 model to a dee p heating anomaly not in the vicinity of the anomalous SST but on the neighboring continents especially South America and with opposite sign in remote areas between Indonesia and the dateline. The signal can be summarized as an enhancement of the GCM's tropical East-West circulat ion with the ascending branch over South America in the warm simulatio n compared to the cold run.