Soil moisture (SM) is critical for various hydro-meteorological applications. Land surface models (LSMs) can produce global spatio-temporal continuous SM estimates. Recently, NASA and ECMWF released GLDAS-2.1 and ERA5-Land datasets, respectively, which contain newly produced LSM-based global SM products, and these have not been thoroughly evaluated in China. To better understand the two products, we decomposed them into SM climatology (i.e., mean seasonal cycle) and SM anomaly (i.e., short-term variability) components and evaluated them separately in China. In particular, the evaluation was conducted considering ground-based SM observations obtained from 1411 stations and two remotely sensed SM products. The following key results were obtained: (a) In the SM climatology evaluation, ERA5-Land showed a larger bias in (semi-) humid areas (0.06 m³/m³ on an average), while GLDAS-2.1 was generally unbiased. GLDAS-2.1 showed higher temporal precision (temporal mean R = 0.47 [-]) than ERA5-Land (temporal mean R = 0.17 [-]) in northern arid areas, while ERA5-Land exhibited better performance (temporal mean R = 0.64 [-]) than GLDAS-2.1 (temporal mean R = 0.34 [-]) in southern humid areas. (b) For the SM anomaly evaluation, ERA5-Land and GLDAS-2.1 performed similarly, and ERA5-Land (temporal mean R = 0.45 [-]) marginally outperformed GLDAS-2.1 (temporal mean R = 0.40 [-]). (c) For the raw SM, GLDAS-2.1 and ERA5-Land had higher temporal precision in the northern and southern areas, respectively, which are mostly determined by their SM climatology. Our findings highlight the important role of SM climatology and provide an important reference for improving the aforementioned SM products.