Methodological dependencies of cloud radiative forcing for the Canadian Climate Centre second‐generation general circulation model

Clear‐sky fluxes at the top of the atmosphere (TOA) are essential for calculating cloud radiative forcings (CRF). Both sets of quantities have been computed by three different methods using data from the Canadian Climate Centre general circulation model (CCC‐GCM) (simulations of July 1985 and January 1986). The methods differ in the way they compute monthly mean clear‐sky fluxes at the TOA ( ). The first method, known as method 2, is ideal for intercomparing GCMs, but does not sample clear skies as satellites do. The other two methods compute by avoiding days with minimum and average cloud amounts that exceed a threshold. These methods are referred to as method 3 and 4, respectively. Thus, while methods 3 and 4 attempt to mimic satellite samplings, they depend on GCM parameterizations of cloud amount. It is shown that for the CCC‐GCM, methods 3 and 4 result in vastly different samplings and that the disparities are perhaps attributable to problems with the GCM's clouds. Globally averaged differences in shortwave CRF (or ) between methods 2, 3, and 4 are less than 1 W m −2 and are therefore relatively unimportant. For longwave (LW) CRF, however, globally averaged differences between methods 4 and 2 are ∼3 W m −2 . These values are achieved by computing for method 4 using only fluxes for days with diurnally averaged cloud amounts less than 0.5. The zonally averaged fraction of days used in this method approximates closely corresponding values obtained by Earth Radiation Budget Experiment analysis. More important than global averages, local differences in LW CRF between methods 4 and 2 exceed ±25 W m −2 . Positive extremes occur over the west central Pacific and are associated with methods 4's elimination of very moist days with low atmospheric transmittances. This leads to larger values of and hence LW CRFs. Negative extremes occur over Antarctica because the clearest days have the lowest surface temperatures. This results in both smaller values of and LW CRFs. The geographic distributions and magnitudes of LW CRF differences between method 4 and method 2 resemble strongly corresponding fields for the European Centre for Medium‐Range Weather Forecasts and the National Center for Atmospheric Research community climate model 2 GCMs.