Abstract. Soil moisture can be obtained from in-situ measurements, satellite observations, and model simulations. This study evaluates different methods of combining model, satellite, and in-situ soil moisture data to provide an accurate and spatially-continuous soil moisture product. Three independent soil moisture datasets are used, including an in situ-based product that uses regression kriging (RK) with precipitation, SMAP L4 soil moisture, and model-simulated soil moisture from the Noah model as part of the North American Land Data Assimilation System. Triple collocation (TC), relative error variance (REV), and RK were used to estimate the error variance of each parent dataset, based on which the least squares weighting (LSW) was applied to blend the parent datasets. These results were also compared with that using simple average (AVE). The results indicated no significant differences between blended soil moisture datasets using errors estimated from TC, REV or RK. Moreover, the LSW did not outperform AVE. The SMAP L4 data have a significant negative bias (−18 %) comparing with in-situ measurements, and in-situ measurements are valuable for improving the accuracy of hybrid results. In addition, datasets using anomalies and percentiles have smaller errors than using volumetric water content, mainly due to the reduced bias. Finally, the in situ-based soil moisture and the simple-averaged product from in situ-based and Noah soil moisture are the two optimal datasets for soil moisture mapping. The in situ-based product performs better when the sample density is high, while the simple-averaged product performs better when the station density is low, or measurement sites are less representative.