Assessing the importance of unresolved cloud-radiation interactions in atmospheric global climate models using the multiscale modelling framework

The sensitivity of a global climate model (GCM) to the inclusion of unresolved interactions between clouds and radiation was studied. The GCM used in this study was the Multiscale Modelling Framework (MMF), which uses a cloud system resolving model (CSRM) as a parametrization of cloud processes. Four sensitivity experiments were performed with the MMF. Within the CSRMs, either local or horizontally averaged radiative heating rates were used and the GCM used either biased or unbiased radiative heating rates and fluxes. For simulations over a single season, the MMF was sensitive to the inclusion of local cloud-radiation interactions within the CSRMs and the MMF was less sensitive to GCM receiving unbiased radiative heating rates and fluxes. This suggests that future GCM parametrizations may need to include unresolved cloud-radiation interactions.; Because the sensitivity experiments require that the MMF model clouds reasonably, cloud fields from the MMF were compared with data from the International Satellite Cloud Climatology Project. The MMF generated similar zonal mean cloud fractions, similar histograms of cloud-top exposed to space and similar relative standard deviations of cloud optical thickness. However, clouds generated within the MMF were more optically thick.; Within the MMF local radiative fluxes and heating rates were computed using the independent column approximation. To assess biases in radiative heating rates and fluxes due to using the ICA, shortwave and longwave two-dimensional Monte Carlo models were applied to nine instantaneous MMF cloud fields. Although there were zonal mean differences between ICA and two-dimensional Monte Carlo shortwave radiative heating rates and fluxes, the magnitude of these differences were generally smaller than differences between ICA and assuming clouds to be maximum-random overlapped and plane-parallel homogeneous. At two locations, one in a polar location and the other in a tropical location, CSRM fields were used to assess differences between local ICA and two-dimensional Monte Carlo radiative heating rates. These differences were less then local radiative heating rate differences due to using domain mean rather than ICA radiative heating rates.