A parameterization of the cloudiness associated with cumulus convection; Evaluation using TOGA COARE data

A new parameterization of the cloudiness associated with cumulus convection is proposed for use in climate models. It is based upon the idea that the convection scheme predicts the local concentration of condensed water (the in-cloud water content) produced at the subgrid scale, and that a statistic al cloud scheme predicts how this condensed water is spatially distributed within the domain. The cloud scheme uses a probability distribution functio n (PDF) of the total water whose variance and skewness coefficient are diag nosed from the amount of condensed water produced at the subgrid scale by c umulus convection and at the large scale by supersaturation, from the degre e of saturation of the environment, and from the lower bound of the total w ater distribution that is taken equal to zero. This parameterization is used in a single-column model forced by the Tropic al Ocean Global Atmosphere Coupled Ocean-Atmosphere Response Experiment (TO GA COARE) data, and including the cumulus convection scheme of Emanuel whos e humidity prediction has been optimized using these data. Simulations are carried out during the 120 days of operation of the TOGA COARE intensive ob servation period. The model is able to reproduce some of the main character istics of the cloudiness observed over the warm pool. This includes the occ urrence of different populations of clouds (shallow, midlevel, and deep con vective), a minimum cloud cover between 600 and 800 hPa, some relationship between the distribution of cloud tops and the presence of stable atmospher ic layers, the formation of long-lasting upper-tropospheric anvils associat ed with the maturation of the convective cloud systems, and the presence of an extensive layer of thin cirrus clouds just below the tropopause. Nevert heless, shallow-level clouds are likely to be underestimated. The behavior of the predicted cloud fields is consistent with some statistical features suggested by cloud-resolving model simulations of tropical cloud systems ov er oceans. The radiative fluxes calculated interactively by the model from the predicted profiles of humidity, temperature, and clouds are in reasonab le agreement with satellite data. Sea surface temperatures predicted by the model using its own radiative and turbulent fluxes calculated at the ocean surface differ from observations by a few tenths of a degree. Sensitivity tests show that the performance of the cloudiness parameterizat ion does not critically depend upon the choice of the PDF. On the other han d, they show that the prediction of radiative fluxes is improved when the s tatistical moments of the PDF are predicted from both large-scale variables and subgrid-scale convective activity rather than from large-scale variabl es only.