Hydro-agricultural applications often require surface soil moisture (SM) information at high spatial resolutions. In this study, daily spatial patterns of SM at a spatial resolution of 1 km over the Babao River Basin in northwestern China were mapped using a Bayesian-based upscaling algorithm, which upscaled point-scale measurements to the grid-scale (1 km) by retrieving SM information using Moderate Resolution Imaging Spectroradiometer (MODIS)-derived land surface temperature (LST) and topography data (including aspect and elevation data) and in situ measurements from a wireless sensor network (WSN). First, the time series of pixel-scale (1 km) representative SM information was retrieved from in situ measurements of SM, topography data, and LST. Second, Bayesian linear regression was used to calibrate the relationship between the representative SM and the WSN measurements. Last, the calibrated relationship was used to upscale a network of in situ measured SM to map spatially continuous SM at a high resolution. The upscaled SM data were evaluated against ground-based SM measurements with satisfactory accuracy—the overall correlation coefficient (r), slope, and unbiased root mean square difference (ubRMSD) values were 0.82, 0.61, and 0.025 m3/m3, respectively. Moreover, when accounting for topography, the proposed upscaling algorithm outperformed the algorithm based only on SM derived from LST (r = 0.80, slope = 0.31, and ubRMSD = 0.033 m3/m3). Notably, the proposed upscaling algorithm was able to capture the dynamics of SM under extreme dry and wet conditions. In conclusion, the proposed upscaled method can provide accurate high-resolution SM estimates for hydro-agricultural applications.