The AMSR2 microwave radiometer is the main payload of the GCOM-W1 satellite,launched by the Japan Aerospace Exploration Agency in 2012. Based on the pre-launch information extraction algorithm,the AMSR2 enables remote monitoring of geophysical parameters such as sea surface temperature,wind speed,water vapor,and liquid cloud water content. However,rain alters the properties of atmospheric scattering and absorption,which contaminates the brightness temperatures measured by the microwave radiometer. Therefore,it is difficult to retrieve AMSR2-derived sea surface wind speeds under rainfall conditions. Based on microwave radiative transfer theory,and using AMSR2 L1 brightness temperature data obtained in August 2012 and NCEP reanalysis data,we studied the sensitivity of AMSR2 brightness temperatures to rain and wind speed,from which a channel combination of brightness temperature was established that is insensitive to rainfall,but sensitive to wind speed. Using brightness temperatures obtained with the proposed channel combination as input parameters,in conjunction with HRD wind field data,and adopting multiple linear regression and BP neural network methods,we established an algorithm for hurricane wind speed retrieval under rainfall conditions. The results showed that the standard deviation and relative error of retrievals,obtained using the multiple linear regression algorithm,were 3.1 m/s and 13%,respectively. However,the standard deviation and relative error of retrievals obtained using the BP neural network algorithm were better(2.1 m/s and 8%,respectively). Thus,the results of this paper preliminarily verified the feasibility of using microwave radiometers to extract sea surface wind speeds under rainfall conditions. The AMSR2 microwave radiometer is the main payload of the GCOM-W1 satellite, launched by the Japan Aerospace Exploration Agency in 2012. Based on the pre-launch information extraction algorithm, the AMSR2 enables remote monitoring of geophysical parameters such as sea surface temperature, wind However, rain alters the properties of atmospheric scattering and absorption, which contaminates the brightness temperatures measured by the microwave radiometer. Thus, it is difficult to retrieve AMSR2-derived sea surface wind speeds under rainfall conditions. Based on microwave radiative transfer theory, and using AMSR2 L1 brightness temperature data obtained in August 2012 and NCEP reanalysis data, we studied the sensitivity of AMSR2 brightness temperatures to rain and wind speed, from which a channel combination of brightness temperature was established that is insensitive to rainfall, but sensitive to wind speed. Using brightness temperatures obta ined with the proposed channel combination as input parameters, in conjunction with HRD wind field data, and employing multiple linear regression and BP neural network methods, we established an algorithm for hurricane wind speed retrieval under rainfall conditions. The results showed that the standard deviation and relative error of retrievals, obtained using the multiple linear regression algorithm, were 3.1 m / s and 13%, respectively. However, the standard deviation and relative error of retrievals obtained using the BP neural network algorithm were better (2.1 m / s and 8 %, respectively). The results of this paper preliminarily verified the feasibility of using microwave radiometers to extract sea surface wind speeds under rainfall conditions.