Key points are not available for this paper at this time.
An automated procedure has been developed for determining cloud motion from geosynchronous satellite pictures based on the use of cross correlation. The speed required for use in a real-time operational system is attained by application of the fast Fourier Transform as a computation algorithm in determining the cross-correlation coefficients. The input data consists of a pair of pictures from the Applications Technology Satellite ATS-I taken 24 main apart and mapped on a Mercator projection. Relative mapping errors are automatically corrected by matching common landmarks on the two pictures. A sample of 300 vectors determined from low-level clouds was compared with those obtained by manual methods in a non-operational environment. Speeds agreed within 10 kt in 82% of the cases and directions within 30° in 72% of the sample. Comparisons of 300 vectors with those obtained from a manual operational procedure showed similar results for direction but indicated that the automated procedure was better at resolving differences in speed. With multiple cloud layers, the ability to determine the individual motions is dependent upon the structure of the pattern in each layer as well as the difference in speed between the layers and the amount and opacity of the upper layer. Test results to date indicate that a combination of the manual and automated techniques provide the best operational solution to obtaining cloud motion vectors from geosynchronous satellite data.
Leese et al. (Mon,) studied this question.