Abstract: The data of sensor geophysical data record(SGDR) of Jason-2 altimeter was collected in China coastal region, and four parameters maximum likelihood estimation method was used for waveform retracking, which can be applied for determinating the mean sea surface.The high frequency correction signals between adjacent observations along the track direction of the altimetry satellite Jason-2 SGDR(sensor geophysical data) record data are correlated near the offshore region. Based on this character, the method for optimal selection Gaussian low pass filter radius is put forward. Firstly, difference calculation is needed between the adjacent observations along the track direction. Secondly, form diffe-rence sequence data sets. Thirdly, determinate the filter radius according to correlation coefficient of the sequences. If the user wants to filter the SGDR data, it is best to select the filter radius of the Gaussian filter equals to 2 km. This method not only can suppress the high frequency error in the SSH(sea surface height), but also can ensure that the data is not excessively smooth. It can be considered as a reference for the establishment of high precision sea surface height model by making full use of satellite altimetry data in offshore region, and also can promote the construction of high precision sea seamless vertical datum.