Imagery Randomized Block Analysis (IRBA) Applied to the Verification of Cloud Edge Detectors

Abstract

A recently developed statistical method is described for evaluating the effectiveness of numerical models. This method is applied to the verification of the effectiveness of satellite imagery cloud edge detectors. Digital satellite data are objectively modeled as cloud/no-cloud imagery using three different edge detection techniques. The imagery resulting from each technique is then quantitatively compared to a verification image through a recently developed imagery analysis method termed Imagery Randomized Block Analysis. The method also provides an agreement measure for objectively choosing a situationally effective technique among several competitors.