DescriptionThis work represents an effort to improve climate model treatment of shortwave radiative transfer through cloudy atmosphere. Results from a stochastic model, which statistically represents radiative transfer through broken cloud fields using a Markovian distribution of observed cloud chord lengths, are compared against those of a plane-parallel model for a variety of cloudy scenes observed in the tropics over a four year period in an effort to reduce the error in radiative transfer calculations due to neglected cloud field morphology. A k-means clustering algorithm is applied to observed cloud optical and dynamical properties in order to identify the presence of specific cloud regimes, which in turn are categorized using large-scale simulated climate variables. The results from the model comparison and cluster analysis are used to develop criteria by which to identify situations where the stochastic model outperforms a traditional plane-parallel model. These criteria are applied to output from a climate model resulting in the identification of some of the issues involved in applying a parameterization developed with observations to model-generated cloud fields.