A stochastic method for convective storm identification, tracking and nowcasting

Abstract The convective storm identification, tracking and nowcasting method is one of the important nowcasting methodologies against severe convective weather. In severe convective cases, such as storm shape or rapid velocity changes, existing methods are apt to provide unsatisfied storm identification, tracking and nowcasting results. To overcome these difficulties, this paper proposes a novel approach to identify, track and short-term forecast (nowcast) of convective storms. A mathematical morphology-based storm identification method is adopted which can identify storm cells accurately in a cluster of storms. As for the difficult tracking problem, sequential Monte Carlo (SMC) method is utilized to simplify the tracking process. It is not only inherently suitable for handling complicated splits and mergers, but also capable of handling the case of storm-missing detection. In order to provide more accurate forecast of a storm position, this method takes the advantages of the cross-correlation method. The qualitative and quantitative evaluations show the efficiency and robustness of the proposed approach.