Nowcasting Methods for Satellite Imagery
Abstract
This report documents the development and testing of nowcasting methods for satellite imagery. Techniques for identifying, tracking, and forecasting features of interest through contours of brightness temperature in geostationary IR imagery were evaluated in a study covering over 200 features from all four seasons. Results indicate that because contours frequently split, merge, or dissipate, contour based techniques are successful in only a minority of cases. The most skillful contour extrapolation technique is movable persistence, which simply extrapolates the position of the contour based on the past movement of its centroid position. Its verification scores are significantly better than persistence for forecasts from 1/2 to 3 hours in advance. As an alternative to the contour-based schemes, a cross-correlation technique for predicting short-term changes in pixel grayshade values (brightness temperatures) was developed and tested. Pixel-based verification of the cross-correlation forecasts show an improvement over persistence out to 2 1/ 2 hours (the longest time period studied for this technique). Contour forecasts derived from the cross-correlation are significantly more skillful than those of the best contour extrapolation technique (movable persistence) out to 1 hour, and no worse out to 2 1/2 hours.... Satellite imagery, Nowcasting, Cross- correlation, Image processing.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 27, 1993
- Accession Number
- ADA267294
Entities
People
- Lawrence W. Knowlton
- Thomas C. Hamill
- Thomas Nehrkorn
Organizations
- Atmospheric and Environmental Research, Inc