Estimation of Tropical Cyclone Intensity using TRMM/TMI Brightness Temperature Data with Asymmetric Components

Abstract

Estimation of tropical cyclone (TC) intensity by using satellite observations is essential for operational TC warnings in the western North Pacific basin where reconnaissance aircraft observations are not conducted. The Japan Meteorological Agency (JMA) typically uses a method of estimating the maximum wind speed of a TC based on the brightness temperature at the 10, 19, 21, 37, and 85-GHz channels of the TRMM Microwave Imager (TMI). In the original method, parameters for concentric circles and annular regions within 2° latitude from the TC center were calculated to represent the TC structure. To improve the estimation, parameters for the four quadrants on the forward, backward, left, and right sides of the TC center relative to the TC motion were added to represent asymmetric components of the TC structure. These parameters were calculated for TC cases from 1998 through 2008 in the western North Pacific basin, and k-means clustering was applied to the parameters to classify the TC cases into 10 clusters. Then a regression equation for the estimation of the TC intensity was computed for each cluster. Several selected parameters and the maximum wind speed in the best track data of the JMA were set for the explanatory variables and the explained variable, respectively, for each regression equation.
The improved estimation method, based on the TC cases from 1998 through 2008, was applied to TCs from 2009 through 2012 to validate the estimation of maximum wind speed. The root mean square error derived from all validated cases was 6.26 m s^-1. The estimates in the clusters for TCs that had relatively asymmetric structures were mostly improved in comparison with those based on the original estimation method. However, in several clusters for TCs that had relatively symmetric structure, the estimation errors were larger than those of the original method. This suggests that the improvement of the estimation method in the present study is limited because of several factors, including the fact that the TC maximum wind speed in the best track data is mainly based on the Dvorak analysis and thus has a certain amount of error, as well as the uncertainty of the TC position determined by using interpolation of the 6-hourly best track data.