Open AccessRetrievals for the Rainfall Rate over Land Using Special Sensor Microwave Imager Data during Tropical Cyclones: Comparisons of Scattering Index, Regression, and Support Vector Regression
Author(s) -
Chih-Chiang Wei,
Jinsheng Roan
Publication year - 2012
Publication title -
journal of hydrometeorology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.733
H-Index - 123
eISSN - 1525-755X
pISSN - 1525-7541
DOI - 10.1175/jhm-d-11-0118.1
Subject(s) - typhoon , environmental science , tropical cyclone , special sensor microwave/imager , climatology , meteorology , precipitation , brightness temperature , wind speed , mean squared error , global precipitation measurement , satellite , index (typography) , geography , statistics , microwave , mathematics , computer science , geology , world wide web , telecommunications , aerospace engineering , engineering
Tropical cyclones, also known as typhoons or hurricanes, are among the most devastating events in nature and often strike the western North Pacific region (including the Philippines, Taiwan, Japan, Korea, China, and others). This paper focuses on addressing the rainfall retrieval problem for quantitative precipitation forecast during tropical cyclones. In this study, Special Sensor Microwave Imager (SSM/I) data and Water Resources Agency (WRA) measurements of Taiwan were used to quantitatively estimate precipitation over the Tanshui River basin in northern Taiwan. Various retrievals for the rainfall rate over land are compared by five methods/techniques. They are the single-channel regression, multichannel linear regressions (MLR), scattering index over land approach (SIL), support vector regression (SVR), and the proposed SIL–SVR. This study collected 70 typhoons affecting the studied watershed over the past 12 years (1997–2008). The measurements of the SSM/I satellite comprise the brightness temperatures at 19.35, 22.23, 37.0, and 85.5 GHz. Overall, the results showed the approaches using the SVR and conjoined SVR and SIL performed better than regression and SIL methods according to their performances of the root-mean-square error (RMSE), bias ratio, and equitable threat score (ETS).