Automated passive microwave tropical cyclone intensity algorithms

Two methods have been developed to extract tropical cyclone intensities from passive microwave digital imagery. Computer vision and neural network techniques have been applied to a large set (~1,500) of Special Sensor Microwave/Imager (SSM/I) images coincident with tropical cyclones around the world. The data set includes systems ranging from tropical depression to Cat 5 or Super Typhoon status.

Tropical cyclone structure is readily apparent when viewing the patterns in SSM/I images (especially 85 GHz channel). Two methods have been trained using a large dependent data set using best track intensities from NHC and JTWC. Tests on independent (withheld) data sets, indicate the algorithms have reached a level near the likely accuracy limits of the best track data sets. Examples for long-lived storms will be used to illustrate the current capabilities using storms from several seasons and basins.