[Pope et al.(2000a)] describe in detail the changes in the HadAM3 GCM when increasing vertical resolution from L19 to L30, so only a brief description will be presented here. In order to make qualitative comparisons with the aqua-planet experiments which will be discussed in section 3.3, the fields displayed from the full GCM will be March monthly means averaged for all 17 years of the AMIPII integrations. Zonal mean cross-sections will be averaged between 60 
E and 150 E to include just the regions of warmest SST. This too is to aid comparison with the aqua-planet experiments.
Figure 5(a) shows the March monthly mean precipitation rate in mm/day from the 30 level full GCM. At this time of year much of the heaviest precipitation in the Indian Ocean and west Pacific is south of the equator. The zonal mean precipitation from the L19 and L30 full GCM runs is shown in fig. 5(b). The main impacts of the resolution change are to weaken the precipitation in the regions where it is heaviest at L19 and to broaden slightly the region of heaviest precipitation in the Tropics. One exception to this is that the heavy precipitation near Madagascar in the southern Indian Ocean becomes even heavier at L30. A large systematic error in the simulation of precipitation at L19 is the under-prediction of precipitation over the Indonesian region (not shown). This error becomes slightly worse at L30.
The zonal mean vertical cross-section of temperature from the 30 level full GCM and the difference between L30 and L19, with negative values indicating that the L30 version is cooler, are shown in fig. 6. The largest differences are in the extra-tropical upper tropopshere and lower stratosphere due to improvements in the position and temperature of the tropopause which is resolved better in the L30 model. [Pope et al.(2000a)] discusses these changes in detail. In the tropical troposphere the L30 version is generally cooler with the largest differences, of about 1.5K, being in the upper troposphere.
Figure 7 shows the zonal mean vertical cross-section of specific humidity from the 30 level full GCM and the difference between L30 and L19, with negative values indicating that the L30 version is drier. The largest differences are between 20 S and 30 S with a maximum at 700 hPa. This is associated with very much drier air over Australia throughout the lower troposphere in L30. The tropical troposphere is generally drier in L30 with a maximum drying occurring at about 500 hPa just south of the equator.
As stated in the previous section, one possibility for the maintenance of the MJO is wave propagation from the extra-tropics through regions of westerly winds in the tropical upper troposphere. Figure 8 shows the zonal mean zonal wind from the L30 full GCM and the difference between this and the L19 configuration. In this case, the zonal averaging includes the entire longitude circle. Changes to the zonal wind between L30 and L19 in the Tropics are small except at around 100 hPa where L30 shows a more westerly flow relative to L19. This indicates that the basic state winds in L30 may be more conducive to the propagation of wave energy from the extra-tropics.
In summary, whilst there are several improvements to the mean climate of the GCM when the vertical resolution is increased, there are also some examples of the simulation worsening. In particular the distribution of precipitation around Indonesia is worse at L30 than L19.