Short Term Forecasting of Cloud and Precipitation
Abstract
A methodology for real-time operations has been developed for the short-term forecasting of cloud and precipitation fields. Pattern recognition techniques are employed to extract useful features from the data field and extrapolation techniques are used to project these features into the future. To reduce computational load, contours defined by directional codes are used to delineate features. These contours are subdivided and attributes such as length, location, and location of each segment are determined. Segment matching is performed for successive observations and attribute changes are monitored over time. Several techniques for the forecasting of attributes have been explored, and an exponential smoothing filter and a linear trend adaptive smoothing filter have been chosen as most appropriate. Currently analysis is performed on a minicomputer and image processor system utilizing radar reflectivity data. Refinement of these techniques and extension into a more comprehensive short term forecasting program is planned. Keywords: Now casting, Pattern recognition, Extrapolation, Forecasting, Radar, Data filtering, Contour, Segmentation.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 15, 1988
- Accession Number
- ADA212692
Entities
People
- A. Borne
- D. Egerton
- F. I. Harris
- P. A. Sadoski
Organizations
- Air Force Research Laboratory