Item

Abstract

In Part I (Storch and Xu, 1990) the Principal Oscillation Pattern (POP) analysis of 200 mb equatorial velocity potential leads to the definition of a bivariate (POP—) index of the tropical 30- to 60—day oscillation. Using the POP prediction scheme this index is predictable for a few days in advance. In this Part II, the prediction of the equatorial velocity potential field, made by the POP method and made by two GCMs, is investigated. The POP index forecast can incorporate skillful forecasts of the equatorial velocity potential (x) field. The 0.50 level of the ensemble correlation skill score is passed after 7 days. If there is a strong signal in the initial state, useful forecasts of more than 20 days are sometimes possible. General— ly, the forecasts are best in northern winter (DJF) and when the initial signal is strong. Apart from day 0 and day 1, the POP forecast scores higher skills than persistence. Only if the 30— to 60—day oscillation is weak in the initial state does the POP forecast fail. Parallel to the POP forecast scheme, two GCMs are considered in terms of their ability to predict the equatorial velocity potential. The NCAR T31 GCM uses regular NMC analyses as initial conditions, and the CNRM T42 uses the initialized analyses of ECMWF. In a series of 15 experimental wintertime forecasts the NCAR T31 GCM appears to be quite skillful in predicting the equatorial x—field, and in particular the 30— to 60—day oscillation. Its skill, however, is less than that of the POP scheme. The CNRM T42 GCM, which was tested with 9 wintertime forecasts, seems not to be able to predict the regular development associated with the tropical 30— to 60—day oscillation. The power of the POP index in explaining the equatorial x—field is a measure of the strength and dominance of the 30— to 60—day oscillation. In the case of the NCAR T31 GCM, this measure at day 0 is an a-priori indicator of the GCM‘s skill in predicting the equatorial velocity potential field.