On the Splitting and Merger in Storm Tracking Using Radar

Storm identification and tracking using radar are basic and essential parts of severe weather warning researches and operations. Large numbers of studies have showed that errors in nowcasting could be introduced in all the phases including identification, tracking and extrapolation. Recently, the storm cell identification and extrapolation algorithms are relatively well-developed, but in tracking phase splitting and merger of storms have been recognized as one of main difficulties limiting the accuracy of severe weather warning. In this paper the occurrence of splitting/merger is judged, and under which condition an algorithm to calculate storm speed and direction is given as follows: Firstly, identification based on TITAN is followed by erosion and dilation processing referring to mathematical morphology. Secondly, a joint method composed of maximum overlapping and cost function based on centroids distance, storm speed and area changing rate is used in storm tracking. Thirdly, occurrence of splitting/merger is confirmed when more than one storms are matched to one storm, then another identifying with higher threshold for the bigger storm is made to isolate more small storms, and one more tracking is implemented, so more one-on-one pairs present. Repeat above processes until no splitting/merger confirmed. And finally a simply way based on the displacement of centroids is used to calculate storm move speed and direction. Actually this re-identification method proposed to deal with the splitting/merger implies an assumption that storm would maintain its original structure shortly after splitting or merging, though good performances are showed in practice with radar volume scanning data with about 6 minutes interval.