Abstract
For three types of the estimation functions for sea surface temperature by the split window method, the error statistics were described in an unified manner by using the statistics of the validation match-up data set. The mean of sqnared errors is described as the sum of the residual variance in the regression function and the squared difference between the estimared values by the specified function and the regression function. The error variances were compared among the estimation functions, and those results were confirmed by the simulations using the Mutsu Bay match-up data set of NOAA-9/AVHRR.
