Shortrange Forecasting through Extrapolation of Satellite Imagery Patterns. Part II. Testing Motion Vector Techniques.
Abstract
An effort is underway at Air Force Geophysics Laboratory to develop automated procedures to make short-range (0-6 hr) terminal weather forecasts using GOES imagery data. A simple approach is to extrapolate the cloud patterns using motion vectors derived from a comparison of successive images. This report describes a test of candidate motion vector techniques using twelve cases of six successive images in a variety of weather conditions. Included in the techniques were two that track brightness centers, three that use cross-covariance, and two using winds aloft. All were compared against persistence (no motion, no change). For all time periods and all thresholds, a binary covariance technique had the highest scores but the techniques using winds aloft were very close. Also, no technique was much better than persistence. There is evidence that most (perhaps 75 percent) of the total changes occurring are not due to simple motion of the cloud patterns but due to more complex processes. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 10, 1979
- Accession Number
- ADA086862
Entities
People
- H. Stuart Muench
Organizations
- Air Force Research Laboratory