By analyzing the observation data of 242 continuous GPS stations over a period of more than 10 years, we apply the static method and obtain the tidal displacement parameters of the eight major tidal constituents. The comparison with theoretical modeling values demonstrates that the GPS observation accuracy of tidal displacements reaches sub-millimeter level. Comparisons are also analyzed between GPS observations and modeling, and reveal that the seven ocean tide models align well with GPS observations but the DTU10 model. The statistical observation results of tide gauge stations indicate that TPXO9-Atlas is the best ocean model for ocean tide load displacement modeling, and PREM is the most appropriate earth model. Moreover, the residuals between the observed values and the theoretical modeling values for M2 and O1 tides show evident long-wavelength, coherent spatial patterns over the region. The results of 198 inland stations over more than 200 km from the coastline are used to analyze the error of Solid Tide Modeling and other systematic errors. The statistical results show mismatches in the weighted root mean square (WRMS) of inland stations in east, north, and vertical are 0.04 mm, 0.07 mm, and 0.11 mm by M2, while 0.05 mm, 0.07 mm, and 0.11 mm by O1, respectively. The WRMS misfits of 44 coastal stations in the three directions are 0.18 mm, 0.27 mm, 0.59 mm by M2, while 0.10 mm, 0.16 mm, and 0.30 mm by O1, respectively. Apparent residuals exist among some coastal stations, mainly due to the errors of OTL (Ocean Tide Loading) displacement modeling, such as ocean tide model, seawater density and Earth model. To address these issues, we first consider the effects of seawater density and compressibility based on the WOA18 ocean data set, and the results show that the OTL displacement modeling is not sensitive to the variation of seawater density and compressibility in China. Secondly, in the asthenosphere, the anelastic dispersion results in a reduction of the shear modulus of similar to 8.5% (semidiurnal tide), similar to 9.1% (diurnal tide), respectively. The model incorporating the anelastic dissipation effects, which lead to a reduction of up to 0.50 mm and 0.25 mm in the vertical residuals of M2 and O1 tide, respectively. Finally, we evaluate the applicability of four 3D earth models based on GPS observation results, and find the other three earth models can effectively improve the consistency between the GPS observations and the theoretical modeling results except the USTC earth model which is affected by sedimentary layer. Furthermore, the residual misfits between the global crustal model CRUST1.0 and GPS observations reach the smallest values, which indicate that the modeling accuracy of OTL in China is significantly limited by the accuracy of the crustal structure. Based on these results, we conclude that the observation accuracy of GPS is adequate to infer Earth's shallow structures.