Automatic storm(s) identification in high resolution, short range, X-band radar images

Long range weather radar observations are regularly used for short term forecasting and severe weather warning systems. Here, data acquired by portable and short range X-band mini weather radar are used, in order to try to forecast storms which are moving within the radar visibility area. Key step in developing an adequate storm forecasting system is the storm identification, because the overall system accuracy strictly depends on it. A storm is a contiguous region within a meteorological radar map where radar reflectivity is greater than a certain threshold and its area is greater than a certain areal threshold. Most of the storm identification techniques are manual or semi-automated and many of them are based on single or multi-level threshold(s). These techniques are facing the problems of choosing a suitable threshold value, and the one related to the so called “false merger problem”. Moreover they are hardly able to indentify sub-storms within a cluster of storms. To cope with these issues, an automated multi-level thersholding technique is introduced where the initial threshold is automatically calculated by using two different techniques. After that, multi-leveling of threshold solves the problem of identifying sub-storms within the cluster of storms. The false merger problem has been instead solved by exploiting the mathematical morphology erosion concept. Storms are approximated and modeled by ellipses and their properties such as centroids, area, and major and minor axis length are calculated. Results show that all of the problems faced by previous systems in storm identification phase seem to be mitigated. Preliminary results about storm tracking and forecasting and preliminary are also presented in this paper.