Inference of Deep Ocean Structure from Upper-Ocean Measurements

Abstract

Many in situ oceanographic measurements sample only the upper few hundred meters of the ocean. It is therefore important to understand to what extent these measurements may be used to infer the structure in the deep ocean. In this paper the use of empirical orthogonal functions (EOFs), normal modes, and linear regression for vertical extrapolation are examined. Results of trials using historical data from the northwest Atlantic are presented and compared with the predictions of a simple theory. When using EOFs or normal modes, the accuracy of the extrapolation is dependent upon the structure of the basis functions used. Increasing the number of basis functions does not always improve the results. The most accurate extrapolation is obtained by using linear regression. Undulating conductivity?temperature?depth profilers typically measure down to 500-m depth. For the dataset used in this study, it is found that if a 500-m-deep profile is extrapolated to 700-m depth, the expected errors are about one-third of the standard deviation of temperature at that depth.