Abstract
This paper investigated the relation between the coefficients of an estimation function for the sea surface temperature and the standard deviation of errors by the split window method of remote sensing. "A set of error ellipses" is defined to a validation data set. It describes contours of the standard deviation of errors in the γ-β plain where γ and β are the coefficient pair of the estimation function. The, standard deviation of errors remains small when a coefficient pair exists near the center of error ellipses along their long axes. By using Mutsu Bay observation data sets, the coefficient pairs of published estimation functions were examined and a sub-optimal coefficient pair is proposed to provide smaller, but not necessarily smallest, standard deviations of errors to various observation data sets.
