The advent of the geosynchronous satellite with 1-km resolution has presented meteorologistis with the opportunity to routinely track and forecast smaller scale weather phenomena than hitherto was possible. The enormous data rate, however, precludes a completely manual system of data interpretation and forecasting, so to fully utilize this opportunity we must use computers. Following examples set by forecasters using digitized radar data, we are developing techniques to make guidance forecasts through automatic extrapolation of satellite imagery. Candidate techniques for extracting motion vectors from satellite imagery have been selected and programmed. This report describes the techniques and presents results of a preliminary test using simulated motions of selected satellite video images.

An attractive approach to short-range forecasting is to determine cloud motion from a sequence of satellite images and extrapolate the patterns and associated weather into the future. Objective motion vector techniques are available and the forecast procedure can be accomplished by computer. This approach is being evaluated at AFGL and this report presents results of testing motion vector techniques. Tracking and covariance techniques were compared along with winds aloft and persistence (no motion) as controls. A covariance technique had top score, but only slightly better than persistence. Complicating factors and implications to forecasting are discussed. (Author).

The advent of the geosynchronous satellite with 1-km resolution has presented meteorologistis with the opportunity to routinely track and forecast smaller scale weather phenomena than hitherto was possible. The enormous data rate, however, precludes a completely manual system of data interpretation and forecasting, so to fully utilize this opportunity we must use computers. Following examples set by forecasters using digitized radar data, we are developing techniques to make guidance forecasts through automatic extrapolation of satellite imagery. Candidate techniques for extracting motion vectors from satellite imagery have been selected and programmed. This report describes the techniques and presents results of a preliminary test using simulated motions of selected satellite video images.