The representation of subgrid cloud variability and its impact on radiation has been a challenge in general circulation model (GCM) simulations. To improve the representation of cloud and radiative variability and their interactions within a GCM grid, it is essential to understand subgrid cloud structures and their statistics based on long-term cloud and radiation data for various climate regions. In this study, year-long cloud-resolving model (CRM) simulations forced with the Atmospheric Radiation Measurement (ARM) large-scale forcing and prescribed evolving surface albedo were conducted for the year 2000 to document the characteristics of cloud horizontal inhomogeneity and vertical overlap and to evaluate and represent their effects on the radiative fluxes and heating rates over a GCM grid.

The year-long CRM simulations with a prescribed evolving surface albedo allow the investigation of the relationship between the surface albedo, cloud and radiation. It was found that clouds absorb more shortwave radiation at the cloud base due to a high surface albedo in winter, which increases temperature in the low troposphere. This leaded to weaker instability in the low troposphere, so that the amount of low-level clouds decreased. For surface albedo greater than a critical value of 0.35, the upward shortwave flux at the top of the atmosphere (TOA) is positively proportional to the surface albedo when optically thin clouds exist, and is not much affected by the reflection from the cloud top. If optically thick clouds occur and the surface albedo is greater than the critical value, the upward shortwave flux at the TOA is significantly affected by the reflection from of cloud top, but not much affected by the surface albedo. In addition, for a surface albedo larger than the critical value, the downward shortwave flux at the surface is primarily influenced by the surface albedo and the reflection from the cloud base if optically thick clouds occur. However, the downward shortwave flux at the surface is not much affected by the surface albedo when optically thin clouds exist because the reflection on the cloud base is weak.

The year-long cloud statistics from the CRM were evaluated against available observational data at the ARM SGP site. The CRM was able to represent thick mid-level and stratiform clouds in agreement with the observations with overcast and non-precipitating conditions. Both the CRM and observations indicated that the height of ice water content maximum in the vertical column decreases as the ice water path increases. It was found that the vertical distribution of shortwave and longwave radiative heating rates in the troposphere were strongly affected by cloud type that was identified by cloud optical depth and vertical location. Compared to the observational estimates, the CRM-produced non-precipitating clouds had greater longwave cooling in the upper troposphere due to lower altitude of high-level clouds and greater cloud top cooling from optically thick mid-level clouds.

The ARM-validated year-long CRM simulations were used to examine the characteristics of cloud horizontal inhomogeneity and vertical overlap and to evaluate and represent their effects on the domain mean radiative flux and heating rate. The analysis of an inhomogeneity parameter (or reduction factor) defined as a ratio of the logarithmic and linear averages of cloud liquid and ice water paths demonstrated that inhomogeneous clouds more frequently appear in summer than in winter due to the occurrence of different cloud types dominated between two seasons. A parameterization with the reduction factor derived from the year-long CRM simulation captured the dominant impact of cloud inhomogeneity on the shortwave and longwave radiative flux and heating rate. Diagnostic radiation calculations with three overlap assumptions (i.e., maximum, minimum, and random) indicated large biases in the total cloud fractions, domain mean shortwave and longwave radiative fluxes, and radiative heating rates when compared to the CRM simulations. These results suggest the need for a physically-based parameterization that treats the differences of characteristic structure between major cloud types such as convective, anvil and stratiform clouds in order to account the radiative effects of subgrid cloud variability on the domain means.

The original mosaic treatment was developed by modifying a GCM radiation scheme to incorporate the radiative effects of dominant cloud types including convective, anvil and stratiform clouds. It cannot be readily used for different radiation schemes. In this study, a cloud distribution scheme was formulated outside of the radiative transfer scheme, so that it can be applied to any GCM to include the radiative effects of cloud variability in their radiative transfer calculations. The radiation calculation with the cloud distribution scheme improved by the year-long CRM statistics produced domain mean shortwave and longwave radiative fluxes and heating rates comparable to the CRM values in seasonal and annual means, which indicates the cloud distribution scheme represents cloud variability in the much the same way as the CRM does.