The accurate assessment of profile soil moisture for spatial domains is usually difficult due to the associated costs, strong spatial-temporal variability, and nonlinear relationship between surface and profile moisture. Here we attempted to use observation operators built by Cumulative Distribution Frequency (CDF) matching method to directly upscale surface observations to profile soil moisture based on multi-station in situ measurements from the Soil and Climate Analysis Network (SCAN). We first analyzed the effects of temporal resolution (hourly, daily and weekly) and data length (half year in non-growing season, half year in growing season, one year, two years and four years) on the performance of observation operators. The results showed that temporal resolution had a negligible influence on the performance of observation operators. However, data length significantly changed the prediction accuracy of observation operators; prediction errors decreased as data length increased from half year (non-growing season) to two years, but accuracy did not further improve at longer interval. A dataset with a two-year duration was therefore used to test the robustness of observation operators in three primary climates (humid continental, humid subtropical and semiarid) of the continental USA, with the popular exponential filter employed as a reference approach. The results indicated that observation operators generally performed better than exponential filter method in both calibration and validation periods. This suggests that observation operators are a robust statistical tool for upscaling soil moisture from surface to profile. The findings here may be applied in the prediction of profile soil moisture from surface measurements obtained via various means, including remote sensing techniques.