MODTRAN deals only with infrared light, and at least the version we are using includes only low level cloud types.
Above is a screen shot of the menu. It can be seen that the highest altitude is three kilometers. This is unfortunate. The atmosphere radiates at the Planck curve like a brick below one kilometer. Significant deviations from the Planck curve begin at about a kilometer in the CO2 bands, and two kilometers in the water bands.
The extent of MODTRAN cloud limitation can be seen above. Oh well, we will just have to work within the MODTRAN limitations and parse out the CERES deep convective data to get a direct comparison.
The CERES data has Cloud Area Fraction by latitude, and when latitude is sorted to our prior zonal criteria, you get the result above. The general decline in boreal cloud fraction is interesting. It shows a steep decline from the millennium to 2012, and then a partial recovery. Tropical cloud fraction possibly shows a small increase, and mid latitude fraction seems steady.
What we must do to compare CERES and MODTRAN is multiply this cloud area fraction by the percentage of cloud types in the MODTRAN menu. For that, we need cloud type data, and fortunately ISCCP has this data from 1983 to 2009.
It can be seen above that the period shows a steady decline in cumulus, interrupted by the eruption of Mt. Pinatubo in June 1991 shown as a vertical line. Interesting that other cloud types were not impressed with the volcano. The CERES data we will be comparing with MODTRAN begins in 2001. Over the 2001-2010 period we see a decline and increase in annual variation of stratocumulus; a modest increase and plateau in altocumulus; and an increase and increase in annual variability of Nimbostratus. This annual variability seems peculiar, and we may return to this. Stratus clouds seem not to have changed much over the entire period.
Notice that the low level clouds have the lowest percent coverage globally, at about 5%.
We realized at the beginning that this exercise of comparing CERES and MODTRAN would be difficult. We are integrating data from three different and uncoordinated formats (CERES, MODTRAN, ISCCP). The differences in formats and data ranges allows only nine years (2001 to 2009) of direct comparison, as the ISCCP cloud type data ends in 2009. Differences in ranges of latitude between the different formats had to be normalized, adding uncertainty.
MODTRAN ignores deep convective clouds, which are imbedded in the CERES coverage data. Since MODTRAN is the source of the radiance data by cloud type, it is not possible to subtract deep convective clouds from CERES coverage rigorously. ISCCP cloud types are for daylight only, and CERES data is day and night.
Nevertheless, the cloud integrated MODTRAN result is interesting.
We can see that MODTRAN predicts far more variation than CERES measures. Any analysis of trend over nine years would be overreaching, particularly as MODTRAN variation wildly exceeds any apparent trend.
The 2006/7 spike predicted by MODTRAN results from modest increases is several cloud types (notably altocumulus in the IPCC graph above), probably from the weak El Nino of those years.
We can recall from earlier in this series that the 2006/7 period showed high total precipitable water without any corresponding increase in atmospheric temperature. It seems that MODTRAN falls victim to this discrepancy.
The MODTRAN global prediction is heavily influenced by the tropics, which cover 50% of the planet as defined here.
Mid Latitude Winters show the 2006/7 spike far more than summers.
Boreal summers and winters differ little and show little response 2006/7.
At the end of this seven post series, we cannot say that MODTRAN integrated for clouds shows the same divergence in trend from CERES that the two show for clear skies. For clear skies, MODTRAN predicts a steady decrease in LW radiation to space, while CERES measurements show a small increase.
If cloud type data could be found after 2009, this project could be extended, and perhaps reveal a divergence. All we can say at this point is that MODTRAN overreacts to the clouds, which according to CERES cover ~65% of the planet at any given time.