Modeling Cloudy and Clear Interval Length Probabilities Using Space Shuttle Imagery

摘要

Interval length probabilities provide an alternative to other characterizations of cloudy and clear regions as viewed from atop the atmosphere. This work attempts to accurately model these probabilities using very high resolution space shuttle orbiter images. Probabilities extracted from these images are compared with three model representations. Metric and congruent statistical methods based on absolute deviations are used to determine model goodness-of-fit. An exponential model is shown to exhibit the least error of the three. Further examination shows that the parameters used to fit the exponential model to observed probabilities can be obtained from the cloud field itself (in the form of mean cloudy and mean clear interval lengths). These mean values are determined for image fractions as small as 1/32 and used to predict probabilities for the entire image.