In this study, we attempted to determine the M2 residual tide with respect to the CSR 3.0 tide model using TOPEX/POSEIDON (T/P) and ERS-1 altimeter data over the Western Pacific Region. The sea surface height (SSH) data were first crossover adjusted to remove biases in satellite orbit and then used to form the residual heights for the recovery of residual tide. Also, a theory of tidal error for averaged SSH was proposed, indicating that a 10 cm error in M 2 tide will be reduced to 0.41 cm for annually averaged T/P SSH and 0.87 cm for annually averaged ERS-1 SSH. The error theory also predicts that data averaging cannot reduce the error due to S 2 tide for the sun-synchronous ERS-1. With the error theory, we formed one mean surface from one year of T/P data and another from 1.5 years of ERS-1 data and the two mean sea surface were used separately as the reference surfaces for T/P and ERS-1. Moreover, a harmonic analysis method, in which the amplitude and phase were modeled by spherical harmonic expansions with transformed spherical coordinates, was adopted to estimate the residual M2 tide. Analysis results indicate that the transformed spherical coordinates not only increase the spatial resolution of the "local" spherical harmonic expansions, but also help to avoid ill-conditioned systems in least-squares. By using data from cycles 2 to 36 of T/P and data from cycles 6 to 15 of ERS-1, we tested with various expansion degrees for the residual M2 recovery. The optimum result is the ERS-1 degree 20 solution which the size of the studied area (5O°×5O°) has a spatial resolution of 1.25°. According to the comparison at selected tide gauge stations, the rms improvement to the CSR 3.0 tide from the ERS-1 degree 20 solution is 1.68 cm, i.e., better than the improvement obtained in other works.