This paper proposes the retrieval method of ocean wave spectrum for airborne radar observations at small incidence angles,which is slightly modified from the method developed by Hauser.Firstly,it makes use of integration method to estimate total mean square slope instead of fitting method,which aims to reduce the affects of fluctuations superposed on normalized radar cross-section by integration.Secondly,for eliminating the noise spectrum contained in signal spectrum,the method considers the signal spectrum in certain look direction without any long wave components as the assumed noise spectrum,which would be subtracted from signal spectrum in any look direction for linear wave spectrum retrieval.Estimated ν from the integration method are lower than the one from fitting method and have a standard deviation of 0.004 between them approximately.The assumed noise spectrum energy almost has no big variations along with the wave number and is slightly lower to the high wave number part of signal spectrum in any look direction,which follows that the assumption makes sense.The retrieved directional spectra are compared with the buoy records in terms of peak wavelength,peak direction and the significant wave height.Comparisons show that the retrieved peak wavelength and significant wave height are slightly higher than the buoy records but don’t differs significantly (error less than 10%).For peak direction,the swell waves in first case basically propagate in the wind direction 6 hours ago and the wind-generated waves in second case also propagate in the wind direction,but the 180° ambiguity remains.Results show that the modified method can carry out the retrieval of directional wave spectrum. This paper proposes the retrieval method of ocean wave spectrum for airborne radar observations at small incidence angles, which is slightly modified from the method developed by Hauser. Firstly, it makes use of integration method to estimate total mean square slope instead of fitting method, which aims to reduce the affects of fluctuations superposed on normalized radar cross-section by integration. Secondarily, for eliminating the noise spectrum contained in signal spectrum, the method considers the signal spectrum in certain look direction without any long wave components as the assumed noise spectrum, which would be subtracted from signal spectrum in any look direction for linear wave spectrum retrieval. Estimated ν from the integration method are lower than the one from fitting method and have a standard deviation of 0.004 between them approximately. assumed noise spectrum energy almost has no big variations along with the wave number and is slightly lower to the high wave number part of signal spectrum in any look direction, which follows that the assumption makes sense sense the retrieved directional spectra are compared with the buoy records in terms of peak wavelength, peak direction and the significant wave height. Comparisons show that the peak width and significant wave height are slightly higher than the buoy records but do not not significantly more (error less than 10%). For peak direction, the swell waves in first case basic propagate in the wind direction 6 hours ago and the wind-generated waves in second case also propagate in the wind direction, but the 180 ° ambiguity remains. Results show that the modified method can carry out the retrieval of directional wave spectrum.