Modeling the diurnal change of cloud cover using multiple global models and ISCCP/TRMM data sets

A recent study of Song and Smith (J. Climate, 2006) demonstrated the usefulness of the TRMM data sets for the extraction of the phase and amplitudes of the diurnal mode for precipitation over the Tropics. The phase of tropical precipitation and cloud varies geographically beyond the land-ocean distributions. A combinations of TRMM and the ISCCP data sets provides a unique opportunity to examine the collective behavior of precipitation and clouds on the diurnal time scales.

Modeling/forecasting of the diurnal change is a complex issue since the forcing for the diurnal change can arise from radiative transfers (including effects of clouds), planetary boundary layer physics, shallow and deep convections and the surface energy balance. We have examined the errors for the diurnal change from a suite of multimodels. After carrying out some 360 forecast experiments on the time scale of 1 through 5 days of forecasts, we noted that the models have a great deficiency in matching the phase and amplitude of the diurnal change with the TRMM or ISCCP based 'observed' estimates. These errors appeared to largely systematic from one model to the next. Thus it is possible to design a multimodel superensemble specifically towards improving the diurnal change of forecasts. Drastic improvements on the geographical distributions of the phase and amplitude of the diurnal mode of clouds and precipitation were possible from the deployment of the multimodel superensemble. This study also addresses several other issues such as possible sources of model errors, relationship of the phase errors between clouds and precipitation and the lack of a clear separation between land versus ocean in the distribution of phase and amplitudes of the diurnal forcing. This goes beyond the conventional wisdom of afternoon cloud/precipitation activity over land areas and early morning hours over the oceans.