Analysis of aquaplanet tropics

We are trying to find a straightforward explanation for why the NCAR CAM and GFDL AM have such different climate sensitivity. From previous analysis, we can say with some confidence that tropical low clouds are playing a leading role in determining the climate sensitivity, and the fraction of low cloud in the tropics changes in opposite senses for the two models under the same prescribed climate change. Now we would like to know why the cloud fraction changes differently between the GCMs.

To get at this, I started looking at the distribution of low cloud amount (LCA) in the tropics for the two models. One way to do this is just to look at the zonal mean distribution, which I show for the QOBS configuration at all latitudes below.


	NCAR CAM T42
NCAR CAM T85	NCAR CAM T42	GFDL AM

The region we have focused on so far is the Tropics, which is about 30S to 30N in each of these cases. The upper panels show the base QOBS configuration in blue, and the SST+2 case in red (dashed). The lower panels show the difference (SST+2 - base), colored relative to a baseline at zero. This is the absolute change, not the fractional change, so a change from 60% cloudiness to 66% appears as a change of 6 (%) instead of the fractional change (10%).

Again here we see that the GFDL tends to have strong decreases in LCA in the tropics, while both versions of the NCAR CAM show increases, though there are some regions of decreased LCA in the lower resolution version in the deep tropics.

To get another perspective on this, I made histograms of LCA at each latitude, and made a contour plot of the results. These plots simply contour the fraction of the data in LCA-latitude space. I use different colors for the base and SST+2 cases, and then take the difference for the lower panel. The way to look at these plots is to think of them as data-density contours. So if we look at the upper panel of each plot, there is a 10% contour line. That contour shows the LCA and latitudes where at least 10% of the data points (grid points or tropical area) reside. For example, the GFDL experiments (3rd column below), no bin interval of LCA contains more than 20% of the data at a given latitude, so there is a single contour line at 10% for each model run. In the base case, the blue curve, the peak of the LCA distribution increases from about 20% cloud cover in the deep tropics to about 35% cloud cover in the near tropics, and at all those latitudes the peak is somewhere between 10% and 20% of the data. That means at a given latitude, there is a fairly wide distribution of LCA, centered on a value of about 35% (for most of the tropics). The lower panels show the SST+2 distribution minus the base distribution, i.e, the red curves minus the blue curves. Here we see that the deep tropics tend to have a shift in the distribution to less low cloud coverage, especially evident in the T42 CAM and AM. In the near tropical latitudes though, both NCAR configurations show an increase in cloudiness while the GFDL AM shows a decrease. Note the scales differ for each column... something to be adjusted later, and fixing this will basically wipe out the noisy deep tropical stuff in CAM at T85, where the deep tropics hardly change at all. Most noteworthy from these figures is that the clearest differences between models are outside of the deep tropics, in subsidence regimes (20-30N for CAM, 15-25N for AM).


	NCAR CAM T42
NCAR CAM T85		GFDL AM

One possible explanation for the different cloud response is a change in the large-scale static stability. To investigate this, I made the same kind of plots for lower tropospheric stability (potential temperature at 700 hPa minus that at 1000 hPa, LTS).

NCAR CAM T85	NCAR CAM T42	GFDL AM

The zonal means are fairly similar. In the Tropics, each configuration experiences a modest increase in LTS around the equator. This is probably due to stronger subsidence associated with increased deep convection straddling the equator. This is consistent with the meridional streamfunction, which shows the T42 version of CAM with a stronger vertical circulation as the ITCZs separate with higher SST while the T85 version has a weaker equatorial response because the ITCZs are not well separated. Both the CAM at T42 and the AM show regions around 15 degrees latitude becoming slightly less stable, arising from a shift in the location of the ITCZs.

NCAR CAM T85	NCAR CAM T42	GFDL AM

Looking now at the distribution of LTS, we see there is again a well defined peak, especially in the deep tropics. The distribution is narrow compared to that for LCA, as can be seen by the number of contour lines in the upper panels. The deep tropics are characterized by LTS around 15 K, with base cases less stable by 1-2 K. At latitudes greater than about 15 degrees, the change is greatly diminished, though there is some indication of a systematic increase instability at latitudes poleward of 25 degrees in the GFDL AM. This pattern of increased stability in the deep tropics and no change in stability in the near-tropics does not correspond with the changes in LCA. Thus is seems unlikely that stability is driving the cloud response.