Characterization of Tropical Cyclone Intensity Using Microwave Imagery

In the absence of wind speed data from aircraft reconnaissance of tropical cyclones (TCs), analysts rely on remote sensing tools to estimate TC intensity. For over 30 years, the Dvorak technique has been applied to estimate intensity using visible and infrared (IR) satellite imagery, but struggles in pinpointing the strength of weaker storms with maximum winds less than 90 knots. Furthermore, the radiative effects of high clouds can obscure the TC structure below. Microwave imagery highlights areas of precipitation and deep convection revealing different patterns than visible and IR imagery. This study explores application of machine learning algorithms to identify patterns in microwave imagery to infer storm intensity, particularly focusing on weaker storms where other analysis methods struggle. An analysis of 91 GHz Special Sensor Microwave Imager/Sensor (SSMI/S) imagery onboard various Defense Meteorological Satellite Program (DMSP) assets from August 2005 to 2017 is presented. Incorporating pattern recognition methods into the current analysis process at the Joint Typhoon Warning Center has the potential to significantly improve TC intensity estimates across all basins of responsibility.