Three different boundary layer cloud models are incorporated into the ECHAM5 general circulation model (GCM) and compared to CloudSat and CALIPSO satellite observations. The first boundary layer model builds upon the standard Tiedtke (1989) parameterization for shallow convection with an adapted convective trigger; the second is a bulk parameterization of the effects of transient shallow cumulus clouds; and lastly the Dual Mass Flux (DMF) scheme adjusted to better represent shallow convection. The three schemes improved (Sub)Tropical oceanic low-level cloud cover, however, the fraction of low-level cloud cover remains underestimated compared to CALIPSO observations. The representation of precipitation was improved by all schemes as they reduced the frequency of light intensity events <0.01 mm d-1, which were found to dominate the radar reflectivity histograms as well as be the greatest source of differences between ECHAM5 and CloudSat radar reflectivity histograms. For both lidar and radar diagnostics, the differences amongst the schemes are smaller than the differences compared to observations. While the DMF approach remains experimental, as its top-of-atmosphere radiative balance has not been retuned, it shows the most promise in producing nonprecipitating boundary layer clouds. With its internally consistent boundary layer scheme that uses the same bimodal joint distribution with a diffusive and an updraft component for clouds and turbulent transport, the ECHAM5_DMF produces the most realistic boundary layer depth as indicated by the cloud field. In addition, it reduced the frequency of large-scale precipitation intensities of <0.01 mm d-1 the greatest.

An investigation of the impact of the subgrid-scale variability of cloud liquid water on the autoconversion process as parameterized in a general circulation model is presented in this paper. For this purpose, a prognostic statistical probability density distribution (PDF) of the subgrid scale variability of cloud water is incorporated in a continuous autoconversion parameterization. Thus, the revised autoconversion rate is calculated by an integral of the autoconversion equation over the PDF of total water mixing ratio from the saturation vapor mixing ratio to the maximum of total water mixing ratio. An evaluation of the new autoconversion parameterization is carried out by means of one year simulations with the ECHAM5 climate model. The results indicate that the new autoconversion scheme causes an increase of the frequency of occurrence of high autoconversion rates and a decrease of low ones compared to the original scheme. This expected result is due to the emphasis on areas of high cloud liquid water in the new approach, and the non-linearity of the autoconversion with respect to liquid water mixing ratio. A similar trend as in the autoconversion is observed in the accretion process resulting from the coupling of both processes. As a consequence of the altered autoconversion, large-scale surface precipitation also shows a shift of occurrence from lower to higher rates. The vertically integrated cloud liquid water estimated by the model shows slight improvements compared to satellite data. Most importantly, the artificial tuning factor for autoconversion in the continuous parameterization could be reduced by almost an order of magnitude using the revised parameterization.