A Comparison of Surface- and Satellite-Derived Cloud Fractions for the ARM SGP
Long, C. N., and Ackerman, T. P., The Pennsylvania State University; Minnis, P., and Smith, W. L., National Aeronautics Space Administration-Langley Research Center
Ninth Atmospheric Radiation Measurement (ARM) Science Team Meeting
Determinations of cloud fractions are essential for radiative energy balance studies. Only satellites afford the global coverage needed to extend these studies to global climate research. However, as with other satellite-derived variables, the accuracy of satellite cloud fraction retrievals must be verified with surface measurements. Previous comparisons have in large part depended on surface observer reports, which include temporal and spatial mismatches and observer subjectivity that add uncertainty to the comparison. We have developed a technique to infer fractional sky cover from surface broadband global and diffuse shortwave measurements. While the absolute accuracy of these retrievals depends on the particular division of clear sky-cloudy sky boundary, comparisons with National Oceanic and Atmospheric Administration/Air Resources Laboratory/Surface Radiation Research Branch (NOAA/ARL/SRRB) Hemispheric Sky Imager data suggest the retrieved sky cover values are accurate to better than a root mean square (RMS) standard deviation of 0.1. The retrievals do not appear to be system or location dependant. The retrieved cloud fractions from collocated radiometer sets at the Atmospheric Radiation Measurement (ARM) Program's Southern Great Plains (SGP) site agree to within an RMS standard deviation of 0.04. Thus, this method affords a consistent, non-subjective estimation of fractional sky cover at high temporal resolution for use in verifying satellite-derived cloud fractions. We present a comparison of satellite-derived cloud fractions with surface-derived sky cover. These preliminary results show, as expected, that point-by-point comparisons yield a wide scatter, which is most probably due to areal coverage differences. However, frequency distribution comparisons between the two data sets for all available data show close agreement in the aggregate.
Note: This is the poster abstract presented at the meeting; an extended version was not provided by the author(s).


