A Template-based Pattern-recognition Algorithm for the Removal of Bright Band from Cartesian Radar Data

This paper describes an algorithm designed to detect and remove areas of brightband from radar reflectivity data which has been remapped onto a Cartesian projection. In these gridded data, the vertical profiles in areas of brightband have the distinctive shape of enhanced reflectivity within a band relative to the layers above and below it. The location of the enhanced reflectivity in the vertical varies with the height and thickness of the band and the location of the grid point within the Cartesian volume. The detection portion of the algorithm uses a fuzzy logic approach. First, regions of possible brightband are identified by computing the goodness of fit between the measured vertical profile and a series of `templates' which are intended to match that vertical profile. The user can specify up to three templates, each of which is tried in turn. The template fit test is done at all CAPPI levels so that bands of varying height may be identified. A fuzzy-logic interest field is computed for each template by assigning a high interest value to vertical profiles for which the given template fits well. The best-fit height is also recorded. The maximum of the three template interest fields is adopted as the interest for that point. The algorithm searches for contiguous regions in which this interest value is reasonably uniform and the height of best fit does not vary significantly. This is done using a texture field of the interest values. If the texture field is `speckled' the uniformity is low, and vice versa. Even fairly small regions of brightband may be identified using this texture field method. Those which are too small to be physically-based are disregarded. Data bands that have been identified as brightband are then removed by interpolating between the reflectivity above and below the band.