An Investigation of the Burger Distribution to Characterize Cloudiness

Abstract

The Burger distribution, which is characterized by two parameters, mean cloud amount and scale distance, is evaluated in terms of its usefulness for representing cloud amount frequency distributions, and as a means of rescaling these frequency distributions to areal extents of the sky other than those from which the input observations were derived. It is found that the Burger distribution performs almost as well (rms errors of ?3% absolute frequency) as the beta distribution (rms errors of ?2% absolute frequency) when the conventional method of calculating mean cloud amount is employed. The Burger distribution performs as well as the beta distribution when the calculation of mean cloud is correlated to take into account observing practice. The advantages of the Burger distribution include the prediction of nonzero values for clear and overcast conditions and the potential for areal extent of sky rescaling of cloud amount frequency distributions. It is found that scaling down (e.g., from conventional surface observations of the full-sky dome to a near-zenith view commensurate with a nadir satellite retrieval) is highly successful with rms errors similar to those obtained in fitting to the observations. However, scaling up from zenith to full-dome views is less successful with rms errors up to 11% (absolute frequency).