The MJO in an AGCM with three different cumulus parameterization schemes

The Madden–Julian oscillation (MJO) is simulated using an AGCM with three different cumulus parameterization schemes: a moist convective adjustment (MCA) scheme, the Zhang–McFarlane (ZM) mass-flux scheme, and the Tiedtke scheme. Results show that the simulated MJO is highly dependent on the cumulus parameterization used. Among the three cumulus parameterizations, only the MCA scheme produces MJO features similar to observations, including the reasonable spatial distribution, intraseasonal time scales and eastward propagation. Meanwhile, the amplitude is too large and the eastward propagation speed too fast than observations and the relationship between precipitation and low-level wind anomaly is unrealistic with enhanced convection occurring within easterly anomalies instead of westerly anomalies as in observations. The over-dependence of precipitation on boundary convergence produced by the MCA scheme is presumably responsible for this unrealistic phase relation in the simulation. The other two schemes produce very poor simulations of the MJO: spectral power of westward propagation is larger than that of eastward propagation in zonal wind and precipitation, indicating a westward propagation of the intraseasonal variability.The mean state and vertical profile of diabatic heating are perhaps responsible for the differences in these simulations. The MCA scheme produces relatively realistic climate background. When either ZM or Tiedtke scheme is used, the observed extension of westerly winds from the western Pacific to the dateline is missing and precipitation over the equatorial region and SPCZ is dramatically underestimated. In addition, diabatic heating produced by both ZM and Tiedtke schemes are very weak and nearly uniform with height. The heating profile produced by the MCA scheme has a middle-heavy structure with much larger magnitude than those produced by the other two schemes. In addition, a very unrealistic boundary layer heating maximum produced by the MCA scheme induces too strong surface convergence, which perhaps contributes to the too strong intraseasonal variability in the simulation.