On the maintenance and initiation of the intraseasonal oscillation in the NCEP/NCAR and ECMWF reanalyses and in the GLA and UKMO AMIP simulations

Julian Intraseasonal (Madden-Julian) oscillations are a dominant model of tropical variability (Madden and 1971, 1972). Satellite derived outgoing longwave radiation (OLR) and reanalyses from NCEP/NCAR and ECMWF are used as verification data in a study of intraseasonal variability in the Goddard Laboratory for Atmospheres (GLA) and the United Kingdom Meteorological Office (UKMO) atmospheric general circulation models. Sling0 et al. (1996) indicated that no model was able to capture the dominance of the intraseasonal oscillation (IO) found in the ECMWF/JDP analyses. However, in the case of the GLA and UKMO AMIP integrations, when a clear eastward propagating signal is evident, the period of the oscillation is realistic.Therefore, in order to show the models in their best light, we examine the November-May period during which these models exhibited their strongest&most coherent IO`s. 1987/88 from observations and the reanalyses will be compared with 1986/87 from GLA and 1980/81 from UKMO. Case studies are important since specific processes/mechanisms may be evident which might otherwise be obscured by cornpositing over many years (e.g., Matthews et al. 1996). During the active phase of the IO, convection migrates from the Indian Ocean to the western/central Pacific Ocean, and into the SPCZ. To demonstrate this, we have calculated an IO index to be used for lagged correlation analysis. This pentad averaged time series is constructed from 20-100 day bandpass filtered 200hPa velocity potential over the region 1OO{degrees}- 140{degrees}E, lO{degrees}N- 10{degrees} S from the NCEP/NCAR reanalysis (not shown; the IO index from the ECMWF reanalysis is virtually identical with the NCEP/NCAR IO index [correlation coefficient=0.987]). This region was chosen since this is where the diabatic heating associated with the IO is greatest. This IO index is then correlated with pentad averaged OLR at various time lags. Convection first arises over the western Indian Ocean on day -15. Through day 0 the convective envelope matures quickly, dominating the eastern Indian Ocean, the Maritime continent and much of Australia. Subsequently, the extent of the convection decreases, with the strongest enhancement located in the SPCZ. The simulated convection, particularly in the GLA model, is most realistic over the western/central Pacific Ocean and the SPCZ. However, both models fail to simulate IO related convection over the Indian Ocean and the propagation eastward into the west Pacific. The maintenance and initiation of the intraseasonal oscillation has also been investigated. Evaporative wind feedback hypothesizes that evaporation to the east of the convection is fundamental for maintaining the eastward migration of the convection. To examine the viability of this hypothesis we have correlated the IO index with 20-100 day bandpass filtered latent heat flux from NCEP/NCAR reanalysis and ECWMF reanalysis. Both reanalyses indicate that evaporation is enhanced to the west of the convection, particularly from day -5 onward, with both reanalyses exhibiting virtually identical lag correlation patterns. This result indicates that evaporative wind feedback is not the dominant process by which the eastward propagation of the intraseasonal oscillation is maintained. Correlations of the simulated IO indexes with filtered latent heat flux from the GLA and UKMO integrations are also shown. In the GLA simulation, enhanced evaporation tends to develop in-place over the west Pacific warm pool, while in the UISMO simulation westward propagation of enhanced evaporation is evident. Thus, the models do not simulate the processes suggested by the reanalyses that occur during the eastward propagation of the IO. While our results suggest a wave-CISK type mechanism , the contribution due to frictional convergence is not apparent. It is suggested that lack of an interactive ocean may be associated with the models systematic failure to simulate the eastward transition of convection and the latent heat flux from the Indian Ocean into the western Pacific Ocean. Examination of observed SST and its relationship to the active phase of the intraseasonal oscillation suggests that air-sea interaction may be important during the course of the evolution of the IO. To explore this in more detail, we show the correlation of the IO index with filtered observed SST (and skin temperature over land from the ECMWF reanalysis) for the 1987/88 case study