An intelligent tropical cyclone eye fix system using motion field analysis

摘要

Tropical cyclones (TCs) are weather systems with vast destructive power Accurate location of their circulation centers, or "eyes", is thus important to forecasters. However, the eye fix process is often done manually in practice. While multiple factors are considered in the process, with subjective elements in these methods, forecasters could disagree. This paper describes a TC eye fix system that uses a novel motion field structure analysis method. It can handle TCs without well-defined structure that are partially out of the image. The system also adapts user inputs and past results to improve its accuracy. Implemented on a commodity desktop computer, the system can process about 5 images per minute, giving an average error of about 0.16 degrees in latitude/longitude on Mercator projected map for TCs that are completely inside the radar image. This is well within the relative error of about 0.3-0.4 degrees given by different TC warning centers. This TC eye fix system is useful in giving an objective TC center location in contrast to traditional manual analysis.