Based on the Waveform data from Jason-2 and SARAL/AltiKa satellites, a new method of eliminating the gross error of altimetry data was developed. By eliminating the sea surface height that was not in the predetermined trajectory, the gross error in each segment of altimetry data was eliminated according to the sea surface height in each cycle of each Pass file, in order to improve the usability of satellite data in coastal areas. Through crossover adjustment, the time-varying signals for the radial orbit error and sea level were further weakened. The discrete sea surface height with high accuracy, which was obtained with the remove-restore technique, was gridded using the radial basis function method. A mean sea surface height model with grid resolutions of 2. 5′×2. 5′ was established. The root mean square error between the sea surface height data from the established model and the data from tidal stations is ±0. 017 m, and the standard deviation between the established model and MSS-CNES-CLS11 was ±0. 070 m. The results show that the established mean sea surface height model for Zhejiang coastal areas is reliable.