Skeleton Based Hook Echo Detection in Radar Reflectivity Data

Automatic detection of severe weather events is of great interest to meteorologists. In this paper, we describe and evaluate a method to identify hook echoes automatically in radar reflectivity data. The method is based on the skeletonization of 2D shapes which are used to describe the shape of storms. We use 4 skeleton shape features: curvature, curve orientation, thickness variation and boundary proximity, and two shape features of the 2D object: southwest localization and size to detect hook echoes. We use these features to design a hook echo algorithm. We evaluate this algorithm on hook echoes detected in several radar datasets, which are verified against those manually identified by an human expert.