Using data from the A-Train satellites, we investigate the distribution of clouds and their microphysical and radiative properties in Southeast Asia during the summer monsoon. We find an approximate balance in the top of the atmosphere (TOA) cloud radiative effect, which is largely due to commonly occurring cirrus layers that warm the atmosphere, and less frequent deep layer clouds, which produce a strong cooling at the surface. The distribution of cloud ice water path (IWP) in these layers, obtained from the 2C-ICE CloudSat data product, is highly skewed with a mean value of 170 g m-2 and a median of 16 g m-2. We evaluate the fraction of the total IWP observed by CloudSat and CALIPSO individually and find that both instruments are necessary for describing the overall IWP statistics and particularly the values that are most important to cirrus radiative impact. In examining how cirrus cloud radiative effects at the TOA vary as a function of IWP, we find that cirrus with IWP less than 200 g m-2 produce a net warming. Weighting the distribution of radiative effect by the frequency of occurrence of IWP values, we find that cirrus with IWP around 20 g m-2 contribute most to heating at the TOA. We conclude that the mean IWP is a poor diagnostic of radiative impact. We suggest that climate model intercomparisons with data should focus on the median IWP because that statistic is more descriptive of the cirrus that contribute most to the radiative impacts of tropical ice clouds.
Given these findings, we use the A-Train observations to address the issues of IWP occurrence and high cloud forcing in a global climate model (GCM). Our goal is to determine whether the clouds that heat the upper troposphere in the model are the same genre of clouds that heat the upper troposphere in the real atmosphere. First, we define a cloud radiative kernel that's a function of IWP to determine whether the modeled ice clouds produce similar shortwave and longwave radiative effects at the TOA. Then we will evaluate whether the model (Community Atmosphere Model version 5) produces a distribution of IWP that is similar to the IWP derived from the A-Train. This methodology will allow us to investigate how GCMs with large differences in global mean IWP can produce similar estimates of a positive high cloud feedback.