A physically based scheme for the treatment of stratiform clouds and precipitation in large‐scale models. II: Comparison of modelled and observed climatological fields

Abstract

AbstractFields from two experiments performed with 18‐level versions of the Commonwealth Scientific and Industrial Organization (CSIRO) global climate model (GCM) are compared with observed fields, focusing on quantities related to clouds and precipitation. The first experiment (denoted by PROG) employed a new prognostic treatment of stratiform clouds and precipitation, while the second experiment (denoted by DIAG18) employed a diagnostic treatment, similar to that used in the standard 9‐level CSIRO GCM. The main findings are as follows. Global‐mean quantities agree well with observations, although the global cloudiness in both runs is a little lower than observed values. Zonal‐mean fields generally show good to very good agreement with observations, particularly in the PROG run, where marked improvements in the cloudiness and long‐wave cloud radiative forcing (LWCF) at high latitudes are noted. The PROG run has cloud‐liquid‐water paths (LWPs) that are larger over mid‐latitude oceans than those from satellite retrievals. Geographical distributions of precipitation, cloudiness, LWCF and SWCF (short‐wave cloud radiative forcing) from both runs are generally in reasonable agreement with observations. Overall, the cloudiness and LWCF are somewhat more realistic in the PROG run, the SWCF is slightly more realistic in the DIAG18 run, and the precipitation is not greatly affected by the change of cloud scheme. Problems affecting both runs to some degree are: deficient cloudiness in the subtropics, and to a lesser extent in mid‐latitudes;deficient SWCF in mid‐latitudes, with a tendency towards excessive SWCF at low latitudes;deficient LWCF over land, mainly in the tropics and northern mid‐latitudes;excessive precipitation, cloudiness and cloud radiative forcing in the tropical western Pacific Ocean in July.The reasons for the above findings are investigated, in part, via the use of sensitivity tests. The improved high‐latitude cloudiness in the PROG run results from (a) replacement of a cloudiness parametrization based on relative humidity with one based on a generalized relative humidity that includes the contribution from cloud water, and (b) inclusion of the effect of frozen precipitation processes on the cloud fraction. The improved LWCF is primarily the result of more realistic treatment of cloud emissivity in the prognostic cloud scheme. The excessive LWPs over mid‐latitude oceans in the PROG run can be corrected by a modest reduction in the critical cloud droplet radius that controls the onset of autoconversion. The deficient cloudiness in the subtropics and mid‐latitudes (typical of current GCMs) can be improved by simple changes to the critical relative humidities used to control the onset of cloud formation or by an increase of vertical resolution, but this improvement comes at the cost of excessive cloudiness in the tropics. The errors in the modelled SWCF (also typical of current GCMs) suggest that there is a systematic latitudinal bias in the calculation of cloud–radiation interactions, such as the effect of solar zenith angle. The deficient LWCF over land is related to a deficiency of high cloud. The vigorous circulation in July over the tropical western Pacific is much more prominent in these 18‐level simulations than in the standard 9‐level version of the model, and is related to aspects of the model other than the cloud treatment.