Estimating Tropical Cyclone Size in the Northwestern Pacific from Geostationary Satellite Infrared Images

Thirty-year (1980–2009) tropical cyclone (TC) images from geostationary satellite (GOES, Meteosat, GMS, MTSAT and FY2) infrared sensors covering the Northwestern Pacific were used to build a TC size dataset based on objective models. The models are based on a correlation between the size of TCs, defined as the mean azimuth radius of 34 kt surface winds (R34) and the brightness temperature radial profiles derived from satellite imagery. Using satellite images between 2001 and 2009, we obtained 16,548 matchup samples and found the correlation to be positive in the TC’s inner core region (in the annulus field 64 km from the TC center) and negative in its outer region (in the annulus field 100–250 km from the TC center). Then, we performed a stepwise regression to select the dominant variables and derived the associated coefficients for the objective models. Independent validation against best track archives shows the median estimation error to be between 27 and 65 km, which are not significantly different to other satellite series data. Finally, we applied the models to 721 TCs and made 13,726 measurements of TC size. The difference of mean TC size derived from our models, and also that from the US Joint Typhoon Warning Center (JTWC) best track archives is 19 km. The developed database is valuable in the research fields of TC structure, climatology, and the initialization of forecasting models.