Richard Neale and Brian
Hoskins
Department of Meteorology
University of Reading
The atmosphere and ocean components of the climate system are intimately linked through exchanges of heat, momentum and water. The large scale properties of the atmosphere change on much larger time-scales than that of the ocean tend to so a useful approach to understanding the circulation of the atmosphere is to examine its response to specified SST. The tropical atmosphere in particular is subject to forcing from the underlying SST and many tropical circulations are intimately related to SST e.g. ENSO. The Inter Tropical Convergence Zone (ITCZ) is seen in satellite pictures as a narrow cloudy band in the tropics and reflects the intense cumulus convection which results partly in response to the underlying SST field. The exact relation between SST and the resulting strength and position of the ITCZ is not always clear.
Figure 1: Annual mean precipitation (from the Xie Arkin precipitation climatology).
Figure 1 shows the climatological average precipitation with the maximum values in the tropics marking the ITCZ which is predominantly north of the equator throughout most of the tropics over the oceans. There is also a tendency to form a split ITCZ over the west Pacific warm pool region indicating that there is not always a precipitation maximum over the warmest SST.
In order to investigate the nature of the ITCZ as a response purely to SST forcing an `aqua-planet' version of the UK Met Office Unified Model (UM) is used. This allows a clean ITCZ signal to be obtained without the many complicating effects of land. The model has the following characteristics:
· Grid point atmosphere only model 3.75 ° long by 2.5 ° lat.
· Prescribed SST, entirely water covered surface.
· Perpetual March insolation conditions.
· 12 month long integration.
Three initial integrations are performed with an SST distribution that is symmetric about the equator and longitudinally invariant (figure 2(a)), and assigned the names control, peaked and flat. The different profiles are designed to investigate the contrasting role of SST and meridional SST gradient in driving tropical circulations.
Figure 2(a) : SST profiles used to force each experiment (K).
Figure 2(b) Zonal mean precipitation for each experiment (mm/day).
The zonal mean precipitation from each of the three experiments is shown in figure 2(b) and demonstrates that the three SST distributions result in both qualitatively and quantitatively different precipitation patterns. Examining the rainfall together with the mean meridional circulation (figure 3) it is clear that the stronger near equatorial SST gradients (peaked and control case) give rise to the strongest Hadley circulation and most intense ITCZ rainfall in a narrow belt centred on the equator.
Figure 3: Mean meridional circulation for axisymmetric experiments (kg/s).
In the presence of weak near equator SST gradients (flat case) there is no preferred location for a tropical precipitation maximum and the Hadley circulation is non-existent. According to Held and Hou (1980) the Hadley circulation exists in order to alter the equator to pole temperature gradient such that the associated zonal mean thermal wind in the tropics does not exceed the zonal mean wind maximum consistent with angular momentum conservation. The greater the difference between the imposed latitudinal temperature gradient and the maximum equilibrium temperature gradient due to angular momentum constraints, the stronger the Hadley circulation has to be in order to adjust the temperature to this equilibrium gradient such as in the control and peaked case. If the imposed temperature gradient does not exceed this equilibrium temperature gradient then there is no dynamical adjustment required and therefore no Hadley circulation exists as in the third flat case. The present results therefore agree with the simple modelling studies of Held & Hou (1980).
Further aqua-planet experiments have been performed to investigate the response to SST anomalies. Two experiments are highlighted here. The first experiment was forced with the control SST profile but with the addition of an SST anomaly maximising at 3K on the equator and stretching 15 ° N and ° S, and 30 ° E and W. The second experiment was also forced with the control profile but this time a zonal wave number one SST anomaly of amplitude 3K was added so that the SST varied between 303K and 297K along the equator.
Figure 4: Precipitation for the confined anomoly case, control removed (mm/day) and the regions used for 6 hrly OLR sampling.
The anomaly was damped to zero poleward of 30 ° N and S. The distribution of precipitation difference from the control (figure 4) shows an increase over the anomaly region maximised 5 ° to the west of the anomaly maximum. Also of note is a suppression of rainfall throughout the rest of the tropics, as high as 6 mm/day especially in the region of Rossby wave response immediately to the west of the anomaly region. The different strengths of convection over the tropics can be shown by compositing the 6 hourly averaged OLR values at each grid point within each region shown in figure 4.
Figure 5: Distribution of the 6 hrly averaged OLR for each grid point in the latitude band 5 ° N to 5 ° S split into different regions and compared with the distribution for the control case (W/m-2).
The distribution of cloud top OLR for each region is given in figure 5 and clearly shows that the convection is deeper over the anomaly region with lower cloud top OLR, and much shallower and often absent particularly in the Rossby region where the ocean OLR signal (~290Wm-2) is frequently seen. The convection depth is suppressed throughout the whole tropics away from the anomaly when compared to the control case.
Figure 6: Precipitation in the wave number 1 case (mm/day). Black is convective, blue is dynamic rainfall.
Figure 6 shows the time-averaged precipitation in response to the wave number 1 SST with distinct patterns local to the SST maximum in the tropics and a remote response in the mid-latitudes and further downstream in the tropics where there is considerable suppression of precipitation.
Figure 7 shows the zonal mean wind for the control case and the two anomaly experiments. The control case has easterlies throughout the depth of the troposphere over the equator whereas in both anomaly cases there are westerlies through a depth of the troposphere. In the wave number 1 case these westerlies exceed 20ms-1 and extend through half the depth of the troposphere.
Figure 7: Zonal mean wind for (a) control case (b) confined anomaly case and (c) wave number 1 anomaly case.
In order to identify a possible generation mechanism for these westerlies we shall examine the zonal mean zonal momentum equation with the effects of stationary eddies included (given below). The blue term is the acceleration of the zonal wind due to mean horizontal convergence of eddy momentum and the green term is the acceleration due to vertical convergence of eddy momentum.
Figure 8 shows the magnitude of these two stationary eddy fluxes for the two anomaly experiments.
Figure 8: Top: Confined anomaly, Bottom Wave No 1 anomaly,
Left: Zonal mean [U V], Right: Zonal mean [U W]
In the upper troposphere the main contribution to the equatorial westerlies comes from the horizontal eddy momentum flux convergence which is stronger in the wave number one case. This is consistent with the equatorial westerlies driven by the propagation away from the equator of Rossby waves generated by the anomaly heating. Lower in the troposphere there is a significant contribution to the westerly acceleration from the vertical eddy momentum flux convergence suggesting the existence of vertical orientated stationary eddies which transport westerly momentum into the tropical mid-troposphere. These large scale wave number one stationary eddies can be seen in the model wind fields.
A climate model, particularly in its tropics, contains a very complicated interaction of dynamics and the parameterisation processes of surface fluxes, shallow and deep convection and radiation. There is currently nothing between a single column test and a full GCM run, and little means of investigating the interaction of processes and the effect on this interaction of changes in parameterisation schemes. It is proposed that the idealised SST aqua-planet experiments described should be developed as the basis for such a test of GCMs.