Low-frequency oscillations appearing in three GCM seasonal cycle integrations are compared with the analyses of the European Centre for Medium-Range Weather Forecasts (ECMWF). All three models have the same resolution: 4 degrees latitude by 5 degrees longitude, with 9 levels. The GLAS GCM simulates a realistic eastward propagation of the 30-60 day oscillation in the tropical upperlevel divergent flow. The eastward travelling planetary scale structure becomes more stationary over the Indonesian region and accelerates over the central Pacific, as observed. In the GLA GCM, the oscillation propagates into the higher latitudes of both hemispheres as the waves leave the convective region. The presence of the eastward propagating oscillation is not obvious in the UCLA GCM. The wavenumber-frequency spectra of the 200 mb velocity potential reveal that all the GCMs have a significantly weaker signal for eastward propagation in the 30-60 day range than the analyses. The spectrum for the GLAS GCM is dominated by 20-60 day periods, while the GLA GCM has a spectral peak around 20-30 days. There is a weak eastward propagating peak near 15 days in the UCLA GCM. The dominant phase speeds and the different vertical structures of the heating profiles in the GCMs are in general agreement with current theory involving the positive feedback between latent heating and moist static stability.
The composited patterns of the observations indicate that in the tropics a Kelvin wave-type structure is dominant near the center of the oscillation. The simulated winds are fairly realistic, although the meridional component is too strong, especially in the GLA GCM. The vertical structures of the zonal wind component and moisture suggest that a mobile wave-CISK (Lau and Peng, 1987) is an important mechanism in maintaining the intraseasonal oscillation in these GCMs. The vertical distribution of the moisture field further suggests that evaporation-wind feedback (Neelin, et al., 1987) may play a role in maintaining the eastward propagating tropical waves. The differences in the structure of the oscillation in the GLAS GCM and GLA GCM appear to be a consequence of the different numerical schemes used. The GCMs have preferred zones for diabatic heating, with a turn-on heating occurring when the rising branch of the intraseasonal oscillation passes over these convective regions.
All three GCMs fail to capture the detailed evolution in the different stages of the development and decay of the oscillation. The results suggest that an improvement in the boundary layer moisture processes may be crucial for a better simulation of the oscillation.