Six years ago, we compared the climate sensitivity of 19 atmospheric general circulation models and found a roughly threefold variation among the models; most of this variation was attributed to differences in the models' depictions of cloud feedback. In an update of this comparison, current models showed considerably smaller differences in net cloud feedback, with most producing modest values. There are, however, substantial differences in the feedback components, indicating that the models still have physical disagreements
Abstract. We compare seasonal changes in cloud-radiative forcing (CRF) at the top of the atmosphere from 18 atmospheric general circulation models, and observations from the Earth Radiation Budget Experiment (ERBE). To enhance the CRF signal and suppress interannual variability, we consider only zonal mean quantities for which the extreme months (January and July), as well as the northern and southern hemispheres, have been differenced. Since seasonal variations of the shortwave component of CRF are caused by seasonal changes in both cloudiness and solar irradiance, the latter was removed. In the ERBE data, seasonal changes in CRF are driven primarily by changes in cloud amount. The same conclusion applies to the models. The shortwave component of seasonal CRF is a measure of changes in cloud amount at all altitudes, while the longwave component is more a measure of upper level clouds. Thus important insights into seasonal cloud amount variations of the models have been obtained by comparing both components, as generated by the models, with the satellite data. For example, in 10 of the 18 models the seasonal oscillations of zonal cloud patterns extend too far poleward by one latitudinal grid.• Regional plots of ACRF exhibit substantial interannual variability, a problem that is reduced by addressing only zonal mean ACRF. However, care must be exercised in performing the zonal averaging because of missing clear-sky grid points that are due to cloudiness persisting over some regions throughout an entire month. This will result in missing ACRF values for those grid points, so that if ACRF is zonally averaged, the missing grids result in biases because those grids contain large amounts of clouds. A more accurate procedure is to first evaluate zonal means of F and a, noting for the latter case that averaging the albedo is equivalent to averaging the flux because the monthly mean TOA insolation is effectively constant in the zonal direction. Unlike the clear quantities, there are no missing grid values. Next, Fc and ac are zonally averaged with missing clear-sky grids not being counted in either the numerator or denominator when performing the averaging. This removes the aforementioned bias associated with averaging ACRF, because enhanced cloudiness over the missing clear-sky grid points would not bias the clear-sky averages. Zonal averages of LW and SW ACRF are then evaluated from (3) and (5) using the zonal mean input.
CESS ET AL.' CLOUD-RADIATIVE FORCING COMPARISONS 16,595To be consistent with the GCM simulations, as will be dis-
GCM SimulationsThe GCMs used in the present study are summarized in Table 1 (Figure 4c), which is the sum of the LW and SW components. The same information is summarized in Figure 5 in the form of GCM versus ERBE differences, and the root-
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.