The goal of our study is to directly evaluate treatment of clouds and radiation in an atmospheric global climate model (AGCM) using long-term observations from the Atmospheric Radiation Measurement (ARM) program. In this presentation, we will present a comparison of observations from two ARM sites, one in north central Oklahoma and one at Nauru island in the Tropical Western Pacific region, with the model output from corresponding grid points. Traditional parametric approach of diagnosing cloud and radiation properties from large-scale model fields is not well suited for comparison with observed time series at selected locations. A recently emerging approach called super parameterization has shown promise to bridge the gap. Super parameterization consists of a two-dimensional cloud system resolving model (CSRM) embedded into each grid of the NCAR Community Climate System Model thereby computing cloud properties at a scale that is more consistent with observations. Because the approach is computationally expensive only limited simulations have been carried out.

Two sets of one year long simulations are considered: one using climatological sea surface temperatures (SST) and another using 1999 SST. Each set includes a run with super-parameterization (SP) as well as an AGCM run with traditional or standard (STD) cloud and radiation treatment. Time series of cloud fraction, precipitation intensity, and downwelling solar radiation flux at the surface are statistically analyzed. Nearly all parameters of frequency distributions of these variables from SP run are shown to be more consistent with observation than those from STD model run.

Different temporal and spatial averaging in the simulations and observations imposes limitations on the comparisons and these scale effects will be discussed. Output from the STD run represents statistics for the AGCM grid, which, in our case, is roughly 300 km x 300 km. In contrast, the CSRM domain is 4 km x 256 km and consists of a row of 64 columns, 4 km x 4 km each. One of the benefits of the SP approach is that statistics can be collected for domain-averaged as well as column cloud and radiation properties. The column statistics are representative of scales that are closer to the scales of observations and therefore allow for more direct comparisons.