Adjusting Satellite Radiances by Regression with an Orthogonal Transformation to a Prior Estimate

Abstract

Many methods for converting satellite radiances to temperatures require a comparison of observed radiances with radiances calculated from a first guess. Usually, measured values must be tuned to agree with theoretical calculations. A regression in which the calculated radiances are predicted from the observed ones is a method of making the adjustment, but results in unrealistic coefficients. Several modifications to standard regression are tried, and it is shown that rotated regression, a technique developed in this paper, provides the required accuracy with coefficients that satisfy the physical constraints.