Evaluation of the Maximum Cross-Correlation Method of Estimating Sea Surface Velocities from Sequential Satellite Images

Abstract

We evaluate the method of estimating sea surface velocities from sequences of AVHRR and CZCS images using the maximum cross-correlation (MCC) technique. A set of synthetic images is created by advecting an AVHRR-SST field with a QG model velocity field. The MCC method of determining the sea surface velocities is then applied to the synthetic images. The rms differences and vector correlations between the model's velocity field and the field produced by the MCC method are presented. In addition, real AVHRR and CZCS images are used to find the rms difference between the satellite-derived velocity fields and in situ ADCP and hydrographic data. The tests show that AVHRR imagery yields the best results when images are separated by as short a period as possible. The rms errors at 6-h separation are on the order of 0.14 m s?1, growing to ≥0.25 m s?1 at separations of more than 18 h. CZCS images are always separated by 24 h or more, but images with well-defined features may result in rms differences no larger than those produced by AVHRR images separated by 12 and 24 h. The method is most successful when several AVHRR image pairs separated by 12 h or less are available from a short (1?3 day) period and the velocity fields from the individual pairs are averaged to give a single synoptic picture of the current field. Specific examples show some of the reasons for incorrect vectors calculated by the method, and suggestions are made for improvements in the method.