WIND SPEED RETRIEVALS FROM MULTI-SENSOR SATELLITE DATA FROM HURRICANES AND TROPICAL CYCLONES Yijun He 1 , Biao Zhang 1,2 , Hui Shen 1 and William Perrie 3 1 Institute of oceanology, Chinese Academy of Sciences, Qingdao 266071, China 2 Graduate School of the Chinese Academy of Sciences, Beijing, 100039, China 3 Bedford Institute of Oceanography, Dartmouth, Nova Scotia, B2Y 4A2 Canada heyj@ms.qdio.ac.cn 1. INTRODUCTION It is well known that tropical cyclones (TCs) pose an increasingly serious threat to coastal areas, for example in China and USA. Millions of people live and vacation along the coastline and the rate of construction of homes and businesses in coastal areas continues to increase. The accuracy and lead-time for forecasts of tropical cyclone tracks and intensities must improve in order to protect the threat to lives and property. Surface winds are a key factor in improved forecasts and analyses of the weather patterns associated with TCs and hurricanes, their tracks and intensities. Spaceborne wind scatterometry, such as available from ERS-1/2, may provide a means to measure the ocean surface wind fields. However, winds tend to be underestimated within TCs because the CMOD4 formulation overestimates the normalized radar cross section (NRCS) of the ocean surface for high wind speeds. To improve the geophysical model functions (GMFs) for high wind speeds, Donnelly et al.[1] developed a new model, CMOD4HW, which incorporates a reduction in sensitivity. These results were used in the derivation of the currently operational CMOD5 GMF. New GMFs were also presented by Carswell et al. [2] at C- and Ku-band and V polarization for wind speeds ranging from 15 to 55 m/s. Recently, a new set of geophysical model functions at C- and Ku band were developed by Fernandez et al. [3] from airborne ocean backscatter measurements at C- and Ku-band wavelengths and H and V polarizations at multiple incidence angles, for ocean surface winds ranging from 25 to 65 m/s. A particular problem for tropical cyclones is that there is always rain. Therefore the effect of the rain rate on the NRCS must be considered, particularly for high frequency sensors, eg. Ku band scatterometer and radiometer. Yueh et al.[4] improved the QuikSCAT Geophysical model function for tropical cyclones using co-located rain rates from SSM/I data. For SSM/I or AMSR data, the effect of precipitation on the brightness temperature must be considered. 2. METHOD AND RESULTS In this presentation, we describe a new model for retrieving wind speed from SSM/I data, which includes the effort of precipitation on the brightness temperature. We collected SAR, QuikSCAT and SSM/I data from several hurricanes: Katrina