A Physical Post-Identification Approach for Tropical Cyclone Trackers

Tropical cyclones (TCs) pose a great threat to coastal communities worldwide. TC tracker, which is designed to identify and trace TC-like structure in meteorological data, plays an important role in understanding the variability of TC activity. However, there are still misidentifications when applying these TC tracker algorithms, and thus compromises the accuracy of detection results. Here we present a novel post-identification approach for TC trackers to enhance their accuracy. This approach is developed based on physical connections between TC motions and the ambient environments via environmental steering flow and beta drift. We found that after applying this physical post-identification approach, the False Alarm Rate (FAR) values decline significantly in two commonly used TC trackers, namely CNRM and TRACK algorithms, which were respectively developed to identify cyclonic structure through different environmental indicators. Specifically, the FAR values decreased by 16.3% in CNRM algorithm and 8.4% in TRACK algorithm, while high Probability of Detection (POD) values were maintained in both trackers. Misidentified tracks are primarily characterized by lower intensities and occur at high latitudes. The results indicate that this physical post-identification approach effectively removes misidentified tracks, particularly those corresponding to extra-tropical cyclones, while simultaneously preserving correctly detected tracks. Our findings demonstrate that integrating environmental steering flow and beta drift into post-identification for TC trackers can improve the detection accuracy. Furthermore, our post-identification approach offers potential for better assessing how climate change affects TC activities when applied to the TC tracks identified by tracker algorithms based on climate models.