Initiation of single-cell thunderstorms has often been associated with boundary-layer convergence lines (Purdom, 1982-ref; Wilson and Schreiber, 1986- ref). MIGFA, a Machine Intelligent Gust Front Algorithm developed at MIT/Lincoln Laboratory (Delanoy and Troxel, 1993-ref), is being utilized to automatically detect and extrapolate radar observed convergence lines. This is done by first constructing individual interest images (e.g. a map of numerical values ranging from 0 to 1 that indicates the presence of radial convergence features) using functional template correlation (FTC). FTC is a generalized matched filter incorporating aspects of fuzzy set theory. A combined interest image, representing a single pixel-map of evidence of where convergence lines are likely, is produced by combining individual interest images at the pixel level. MIGFA is able to detect and track the most probable locations of convergence from these combined interest images. Tracking is done by establishing point-by-point correspondence between consecutive scans. Prior history of a MIGFA detection is used to make predictions of where the convergence line will be in the future. The MIGFA prediction and convergence boundary type (e.g. stationary, moving or colliding) are used to generate an initial "first-guess" forecast box which indicates areas where thunderstorms are likely to form.
Examples of MIGFA detections and predictions:
Go to TITAN
Go to Cu cloud detection algorithmReturn to Diagram
