Comparison of Two Automatic Identification Algorithms for Cyclones Affecting the Changjiang River–Huaihe River Valleys

Abstract

In this study, two commonly used automated methods of detecting cyclones in the lower troposphere were compared with respect to various features of cyclone activity. The first method is based on the neighbor cyclone center point (NCP), while the second method is the cyclone area algorithm (CAA), which relies on the detection of the outermost enclosed contour to identify the horizontal structure of a cyclone. We obtained climatologies of cyclones that affected the Changjiang River–Huaihe River Valleys (CHV) of China (derived from ERA-Interim data for 1979–2015) and compared their structures. We found that the distribution of the track and the cyclogenesis locations of influential cyclones (ICs) showed a consistent spatial pattern between the NCP and CAA. However, there were still notable differences between the statistical features of cyclone activity derived by the NCP and CAA: (1) Only <46% of cyclones shared the same cyclone center between these two schemes. (2) ICs derived from the CAA typically had longer lifetimes and travel distances, with stronger central intensities than those from the NCP. (3) The track of ICs by the CAA with high resolution was consistent with that of ICs by the low-resolution CAA as well as the low-resolution NCP. However, compared to other methods, the high-resolution NCP presented large deviations during the early cyclone stage. The involvement of open systems in the NCP resulted in weaker cyclone intensities and increased uncertainty in cyclone tracking. On the other hand, more cyclones with stronger intensities and longer lifetimes coming from the midlatitudes were detected using the CAA. In addition, the short-lifetime ICs (<18 h) found using the CAA were active (39%) in the CHV, and were typically excluded by the NCP. These ICs had comparable center intensity and showed a good correlation with the occurrence of simultaneous rainfall events.