High resolution photography of clouds from the surface and cloud radiative effects

 

Authors

Stephen E. Schwartz — Brookhaven National Laboratory

Clement Li — City College of the City University of New York, Department
Antonio Aguirre — New York City College of Technology, Applied Mathematics

Category

General Topics – Cloud

Description

Principal component analysis of cloud contribution to radiant intensity. Pixelated image is accurately reproduced from single principal component, which is found to vary linearly with Red/(Red + Blue).
Clouds exert important shortwave and longwave effects on Earth's radiation budget that must be accurately characterized by observation and represented in models. Typically scenes are characterized by remote sensing from the surface or space as cloud or as cloud free and results prsented as cloud fraction, the fraction of pixels containing some cloud. Similarly clouds are modeled as present/absent in some fraction of a three or two-dimensional domain. However observationally clouds exhibit rich spatial and intensity structure. Here initial results are presented from high resolution (20 µrad or 20 mm at 1 km) digital photography looking upward from the surface. The field of view, 20 x 30 mrad (roughly 2 x 3 sun diameters), with 16-bit intensity resolution in three colors, yields roughly 1 million independent determinations of cloud contribution to downwelling radiance per photograph. Frequently clouds exhibit considerable spatial and intensity variability even in the small spatial domain examined. Red and Blue intensities are found to be non-orthogonal measures of cloudy and clear-sky contribution to radiant intensity. The ratio RRB = Red/(Red + Blue), commonly used as a discriminant between cloud and clear sky, is high in cloudy regions and low in clear-sky regions. Values of RRB frequently exhibit a rather monomodal histogram rather than bimodal as would be expected for distinct regions of cloudy and clear sky. Cloud fraction determined from RRB is highly sensitive (several tens of percent) to threshold and resolution. Principal component analysis (see Figure) shows that normalized Red and Blue intensities are accurately represented by a single component that is strongly correlated with RRB; thus RRB appears to be more useful as a measure of cloud contribution to radiance than as a discriminant. The spatial variability of clouds on many scales suggests the utility of characterizing clouds as fractal quantities. However differing approaches to determining fractal dimension of clouds within the images obtained here (box counting; slope of power spectrum with frequency) indicate that fractional dimension is likewise not uniquely defined but is dependent on threshold and varies with spatial frequency rather than exhibiting a power-law dependence expected for a fractal quantity. Alternative approaches to characterizing spatial variability are examined.