The Relationship between Satellite Measured Convective Bursts and Tropical Cyclone Intensification

Abstract

The relationship between the mean temperature of the top of the cloud canopies and the future maximum winds of Atlantic Ocean tropical cyclones is analyzed. The area-average cloud top temperatures from 309 observations of 12 tropical cyclones which occurred during 1974?79 were compiled from infrared measurements made by Geostationary Operational Environment Satellites. Maximum winds were obtained from best track records. The satellite measurements showed that prolonged surges of intense convection developed in the near region surrounding the depression centers before the maximum winds initially increased. Subsequent weakening of the convection occurred but was frequently followed by new surges of intense convection. It was found that when these prolonged surges of intense convection 1asted for 9 or more hours, and the filtered (6-h running mean) area-average cloud top temperature within 222 km of the tropical cyclone centers was 238 K or less, that the maximum winds of the tropical cyclones increased by 5 m s?1 or more within 24 h later, 71 % of the time. However, when intense convection was not present, similar maximum wind increases occurred only 37% of the time. The future maximum winds were compared with both the filtered area-average cloud top temperatures measured during the strong convective surges and the storm's intensities at the filtered temperature times using multiple linear regression. The correlation was found to be 0.771 for moderate/strong storms (storm intensity of 26 m s?1 or more) and 0.610 for weak storms (stores intensity of less than 26 m s?1). The relationships are statistically significant at the 0.0005 and 0.05 levels, respectively, and the lag time is near 24 h. The standard error of the regression is 5.7 and 6.2 m s?1, respectively. Statistical tests made to determine the quality of expected performance suggest that predictive equations will yield maximum wind intensities within 3 and 4 m s?1, respectively, of the standard error of the regression 95% of the time. In an independent test, the standard deviation of the error of the predicted maximum winds of moderate/strong storms was 8 m s?1, or well within the expected bounds.