Employing Cluster Analysis to Detect Significant Cloud 3D RT Effect Indicators

Abstract

Three-dimensional cloud field morphology contributes to scene-averaged cloud reflectivity, but climate models do not currently incorporate methods of identifying situations where this contribution is substantial. This work represents an effort to identify atmospheric conditions conducive to the formation of cloud field configurations that significantly affect shortwave radiative fluxes. Once identified, these characteristics may form the basis of a parameterization that accounts for radiative impact of complex cloud fields. A k-means clustering algorithm is applied to observed cloud properties taken from the Atmospheric Radiation Measurement Program tropical western Pacific sites to identify specific cloud regimes. Results from a stand-alone stochastic model, which statistically represents shortwave radiative transfer through broken cloud fields, are compared with those of a plane-parallel model. The aggregate scenes in each regime are examined to measure the bias in shortwave flux calculations due to neglected cloud field morphology. The results from the model comparison and cluster analysis suggest that cloud fraction, vertical wind shear, and spacing between cloudy layers are all important indicators of complex cloud field geometry and that these criteria are most often met in cloud regimes characterized by moderate to strong convection. The cluster criteria are applied to output from the Community Climate System Model (version 3.0) and it is found that the presence of persistent high cirrus cloud in model simulations inhibits identification of specific cloud regimes.