Abstract
An experimental colocation and statistical regression scheme is used to verify the hypothesis that large consistent mean errors in the cloudy oceanic satellite temperature retrievals north of 30°N from TIROS-N are due to a continental bias in the statistical colocation base. Sea surface water temperature is a useful predictor for these conditions. Significant improvement in these retrievals is possible. However, much of this improvement requires the presence in the colocation base of the, mid-ocean radiosondes from the ocean weather ships.
