A Study on Typhoon Center Localization Based on an Improved Spatio-Temporally Consistent Scale-Invariant Feature Transform and Brightness Temperature Perturbations.

Extreme weather events like typhoons have become more frequent due to global climate
change. Current typhoon monitoring methods include manual monitoring, mathematical morpholog-
ical methods, and artificial intelligence. Manual monitoring is accurate but labor-intensive, while AI
offers convenience but requires accuracy improvements. Mathematical morphology methods, such as
brightness temperature perturbation (BTP) and a spatio-temporally consistent (STC) Scale-Invariant
Feature Transform (SIFT), remain mainstream for typhoon positioning. This paper enhances BTP
and STC SIFT methods for application to Fengyun 4A (FY-4A) Advanced Geosynchronous Radia-
tion Imager (AGRI) L1 data, incorporating parallax correction for more accurate surface longitude
and latitude positioning. The applicability of these methods for different typhoon intensities and
monitoring time resolutions is analyzed. Automated monitoring with one-hour observation intervals
in the northwest Pacific region demonstrates high positioning accuracy, reaching 25 km or better
when compared to best path data from the China Meteorological Administration (CMA). For 1 h
remote sensing observations, BTP is more accurate for typhoons at or above typhoon intensity, while
STC SIFT is more accurate for weaker typhoons. In the current era of a high temporal resolution of
typhoon monitoring using geostationary satellites, the method presented in this paper can serve the
national meteorological industry for typhoon monitoring, which is beneficial to national pre-disaster
prevention work as well as global meteorological research.