The numerical model used throughout this study is The Met. Office Unified Model in its GCM configuration. The model version is HadAM3, the atmospheric component of the Hadley Centre coupled model. This is a newer version of the Hadley Centre climate model than that which is discussed in [Slingo et al.(1996)] and [Sperber et al.(1997)]. Both of these studies showed that this GCM had strong tropical intraseasonal variability. However, since these studies were performed, several changes have been made to the model formulation including a new parametrization of radiation [Edwards and Slingo(1996)], the inclusion of the vertical transport of momentum by the convection scheme [Gregory et al.(1997)] and a new parametrization of soil processes and the exchange of heat and moisture between the Earth's surface and the planetary boundary layer [Cox et al.(1999)]. The combined impact of these and other less major model changes on the GCM simulation are described in more detail by [Pope et al.(2000b)].
Various integrations of this AGCM, with different horizontal and vertical resolutions, have been performed at the Hadley Centre. In this paper we will examine two of these integrations which are identical except for a change in the vertical resolution of the model. The standard number of vertical levels for this GCM is 19, giving a vertical layer thickness of about 100 hPa in mid-troposphere, with higher vertical resolution in the planetary boundary layer and around the tropopause. This is the version of the atmospheric GCM which is used in coupled modelling work at the Hadley Centre. An ensemble of 6 AMIPII integrations with 19 vertical levels has been run at the Hadley Centre. AMIPII is the second Atmospheric Model Intercomparison Project - in these integrations the AGCM is run for 17 years forced with the observed sea surface temperatures from the period 1979-1995. The ensemble members differ only in their initial conditions. The results presented here are from one member of that ensemble but the 6 members all produce very similar mean climatologies and variability so that any single member of the ensemble can be considered to be representative of the performance of the GCM at this resolution. An AMIPII integration with 30 vertical levels has also been performed. The bottom 3 and top 3 levels in this 30 level version remain the same as the 19 level version. The additional levels are inserted evenly between the remaining levels. This leads to a mid-tropospheric layer thickness of about 50 hPa. Figure 1 shows the distribution of vertical levels in these two configurations. The vertical co-ordinate used in this GCM is hybrid eta, giving a transition from pure terrain-following sigma co-ordinates near the Earth's surface to pure pressure levels in the stratosphere. The strength of the MJO in these two integrations is discussed in section 3.1. and the mean climatology of the models is examined in section 3.2.
In order to diagnose the behaviour of tropical convection within the GCM, further integrations were performed using the model in aqua-planet mode with the same sets of vertical levels. This aqua-planet configuration of the Hadley Centre model is described in detail by [Neale(1999)]. A zonally symmetric SST distribution was defined which gave typical Indian Ocean/west Pacific warm pool values on the equator. Incoming solar radiation was fixed at an equinoctial (March) value. The aqua-planet model was integrated for 15 months with both 19 and 30 levels in the vertical. The first 3 months of each integration were discarded from calculations of the model's mean state.
There are several reasons for using an aqua-planet version of the GCM to diagnose the behaviour of convection. The first is that the homogeneity of the model allows us to obtain a large sample of convective events over warm SSTs without needing to perform long integrations. Secondly, the removal of the land areas removes the possibility of circulations forced by land-sea contrasts which may force or interact with convection. Therefore we can be sure that convective events in different geographical locations are subject to the same large scale forcings, giving us a cleaner picture of the behaviour of convection. Thirdly, the aqua-planet GCM provides a more realistic test of the convection scheme than using a single-column model with idealized boundary and large scale forcing functions, but without the complications of the full GCM. The fixing of the solar cycle at a mid-March value further simplifies the analysis of how the convection scheme behaves, and has the useful effect of making the aqua-planet zonal mean state symmetric about the equator. [Neale and Hoskins(2000)] advocate the use of aqua-planet GCMs as a test bed for changes to the formulation of a GCM.
It should be emphasized that we are not looking for an MJO signal in the aqua-planet experiments, but are using them to study the behaviour of general transient convection. It is doubtful whether the aqua-planet model in the configuration described above would produce an MJO due to differences in the basic state winds and the zonal and meridional symmetry of the forcing. Also, since we are only integrating the GCM for 15 months in aqua-planet mode we would not have a long enough time-series to analyse for MJO activity.
The mean climatologies of the aqua-planet L19 and L30 integrations were calculated and compared to see if the changes were in the same sense as the changes in the full GCM when going from 19 to 30 levels. These results will be described in section 3.3. Next, the time evolution of convective cloud and precipitation was studied over a 90 day period in the L19 and L30 aqua-planet runs in order to see how convection evolves and organises on a day-to-day basis over the tropical oceans. This analysis will be described in section 3.4.
All these results will be discussed in section 4 in the context of physical processes that lead to differences between the L19 and L30 integrations. Comparisons will also be made between tropical convection simulated by the model and observations of convection made over the warm pool region during suppressed and active phases of the MJO as part of the Tropical Ocean Global Atmosphere (TOGA) Coupled Ocean-Atmosphere Response Experiment (COARE).