Soil moisture (SM) is a fundamental environmental variable for characterizing climate, land surface and atmosphere. In recent years, several SM products have been developed based on remote sensing (RS), land surface model (LSM) or land data assimilation system (LDAS). However, little knowledge is available in understanding spatial patterns of the uncertainty of different SM products and potential regional drivers over the Qinghai-Tibet Plateau (QTP), a complex environment for accurate SM estimation. This paper investigates the sensitivity of the SM uncertainties based on the three-cornered hat (TCH) method and a generalized additive model (GAM) in terms of underlying surface characteristics (sand fraction, soil organic matter, vegetation, land surface temperature, and topography) and near-ground meteorology (precipitation and air temperature) in the third pole environment over the 2015–2018 period. Eleven SM products are involved in this work, including Soil Moisture Active Passive (SMAP), Soil Moisture Ocean Salinity INRA-CESBIO (SMOS-IC), Japan Aerospace Exploration Agency (JAXA), Land Surface Parameter Model (LPRM), Climate Change Initiative - Active/Combined (CCI_A/CCI_C), Global Land Data Assimilation System (GLDAS), European Centre for Medium-Range Weather Forecasts Interim reanalysis (ERA-Interim), Global Land Evaporation Amsterdam Model product a/b (GLEAM_a/GLEAM_b), and Random Forest Soil Moisture (RFSM). Results show that most of the SM products perform well across QTP, while SMOS-IC is strongly affected by radio-frequency interference in this region, JAXA has a relatively higher noise level over QTP, and LPRM has larger relative uncertainties (RUs) in the southeast of QTP. Nonlinear regression analysis demonstrates that variations of RUs in SMOS-IC and JAXA are driven by topography, while LPRM's are mainly controlled by vegetation. In addition, two groups of opposite (positive/negative) effects from topography and vegetation, topography and precipitation, and precipitation and land surface temperature affect the spatial variations of RUs in CCI_A, RFSM, and ERA-Interim, respectively. Meanwhile, more complex relationships are found between multiple surface factors and RUs of different products. In general, the underlying surface factors explain on average 39.41% and 28.34% of the variations in RS and LSM/LDAS SM RUs, respectively. Comparatively, the near-ground meteorology factors have a slightly larger effect on LSM/LDAS products than that on RS products.