The continuing trend of advances in computing power leads to the possibility of many developments in atmospheric modelling. One such development has been the use of general circulation models (GCMs) with higher resolution. This trend has tended to concentrate on increases in the horizontal resolution of the models (for example [Boyle(1993)], [Phillips et al.(1995)], [Williamson et al.(1995)] and [Gualdi et al.(1997)]). However, as early as 1989, [Lindzen and Fox-Rabinowitz(1989)] were emphasizing the need for the vertical resolution of a GCM to be consistent with the horizontal resolution in order for various atmospheric waves to be correctly represented. More recently, work by [Tompkins and Emanuel(2000)] has shown that the vertical distribution of water vapour in GCMs can be very sensitive to the vertical resolution of the model. Their work questions the ability of the current generation of GCMs to represent correctly some of the physical processes which determine the distribution of water vapour in the atmosphere due to the rather coarse vertical resolution of the models, particularly in the mid-troposphere. At the same time, observational studies of tropical convection have shown the importance of cumulus congestus clouds, with tops around the melting level, in moistening the mid-troposphere prior to periods of deep cumulonimbus convection [Johnson and Lin(1997), Johnson et al.(1999)]. The formation of these congestus clouds depends crucially on physical processes occurring around the melting level - a region with typically rather poor vertical resolution in the current generation of GCMs.
This paper will describe a study of the behaviour of tropical convection in a GCM when the vertical resolution is doubled in the free troposphere and around the tropopause. The results will be discussed in the context of observations of tropical convection and the implications for the simulation of the Madden-Julian Oscillation (MJO) by GCMs will also be considered. In a related study, [Pope et al.(2000a)] discuss the impact of the increased vertical resolution on the distribution of water vapour.
The motivation for the work described in this paper is our intention to use a coupled GCM to investigate the interaction between atmospheric variability associated with the MJO and the state of the tropical Pacific ocean prior to and during El Niño events. This work will be presented in future publications. However, in order to perform this work we require an atmospheric component of the coupled model system which can reproduce the correct degree of intraseaonal variability associated with the MJO, and this is the subject of the present study.
It has been suggested that in order for a GCM to correctly simulate the MJO, coupling to an ocean model may be necessary [Sperber et al.(1997), Flatau et al.(1997)], and observational studies do suggest that the MJO is, at least in part, a coupled phenomenon. In a study of 15 years of observed data covering the Indian Ocean, the Maritime Continent and the Pacific Ocean, [Woolnough et al.(2000)] show that coherent variations in SST at intraseasonal timescales show correlations with variations in winds and cloudiness associated with the MJO. Several other studies [Zhang(1997), Flatau et al.(1997), Hendon and Glick(1997)] show similar relationships using data from shorter periods or covering more limited areas. These studies all suggest that there may be a feedback between the variations in SST and the enhanced convection which acts to maintain the eastwards propagation of the MJO. However, some atmospheric GCMs (AGCMs) forced with slowly varying SSTs with no intraseasonal variations have also been shown to produce at least some features of the MJO whereas others show very little variability at all on intraseasonal timescales in the tropics [Slingo et al.(1996)]. Some studies have shown that the representation of tropical intraseasonal variability by individual GCMs is very sensitive to the parametrization of physical processes. [Slingo et al.(1994)] show that the intraseasonal variability of a GCM can change dramatically by using a different convective parametrization, and [Inness and Gregory(1997)] show that the inclusion of the vertical transport of momentum by the convection scheme can considerably weaken the upper tropospheric signal of the MJO in a GCM, possibly due to changes in the basic state winds in tropical latitudes.
All these studies suggest that the formulation of the AGCM, and in particular the convective parametrization, is crucial to the simulation of tropical variability in general and the MJO in particular. So in order for a coupled GCM to have the best chance of simulating the MJO and the associated interactions between atmosphere and ocean, it seems clear that the atmospheric component must be able to represent the physical processes that lead to the organization of convection on the correct timescales. If the AGCM does not reproduce these processes then simply coupling it to an ocean model will not lead to an improvement in the simulation of the MJO. However, if an AGCM does reproduce some of the features of the observed MJO there is evidence that coupling to an ocean can lead to an improved MJO simulation [Waliser et al.(1999)].