Based on the space-borne SAR(synthetic aperture radar) data and model data, a variational fusion method is presented to implement the fusion of SAR data and numerical weather prediction model data, and to improve the accuracy of sea surface wind field and operational level. First, two-dimensional continuous wavelet transform is applied to extract wind direction of high precision wind stripe in SAR image, and geophysical model function is used to calculate meridional and zonal components of sea surface wind field. Then the Kriging interpolation method is used to interpolate the wind speed of numerical weather prediction model to the coverage area of SAR sea surface wind field, resulting in that the SAR wind speed observation operator is obtained and the cost function of SAR wind field combined with model wind field is established. The variational method is adopted to solve the analysis wind field. Finally, the sea surface wind field results after fusion are obtained. Simulation results show that through the variational fusion, the sea surface wind speed and wind direction results are closer to the ideal values, especially in the coverage area of SAR wind field. The ENVISAT/ASAR sounding data and the spatial temporally matched model wind data of ECMWF(European Center for Mediumrange Weather Forecasts) are selected to carry out the case verification, and the results show that the sea surface wind field after fusion is closer to the buoy observed result than the model wind field, so it is confirmed that the variational fusion method is effective.