Understanding the frequency distribution of BRF through 3-D radiative transfer simulations and its applications on satellite cloud detection

The accuracy of cloud detection directly impacts the quality of most satellite geophysical products used in weather, climate and environmental research. All operational cloud detection algorithms employ the threshold method that assumes cloudy and clear pixels can be distinguished by comparing their grey level values of an observed quantity such as the BRF. This assumption indicates that the frequency distribution of the BRF field--the histogram--has some sort of discontinuity, yet the opposite is observed over a wide range of space and time scales. Understanding the continuous nature of the BRF histogram is crucial to satellite cloud detection.

In this study, the 3-D model SHDOM (Spherical Harmonics Discrete Ordinate Method) is adopted to simulate the radiative transfer through predefined heterogeneous cloud fields to generate synthetic satellite images. The following questions will be addressed in our presentation: under what circumstances does the overlap between the distributions of clear pixels and cloudy pixels occur? What are the roles of surface, atmospheric molecules, aerosol, satellite spatial resolution, sun-view geometry, and cloud heterogeneity in the formation of a continuous histogram? What are the uncertainties of commonly used thresholding algorithms?