The U.S. Department of Energy Office of Legacy Management (DOE LM) is investigating options for future management of selected uranium mill tailings disposal cell covers as vegetated, evapotranspiration (ET) covers. ET limits drainage of water through the cell cover profile, while soil structure and drying by plants can increase radon diffusion; therefore, soil water content is a key performance parameter. This study used theoretical simulations to analyze the sensitivity of multi-frequency radar backscatter to soil moisture (SM) at different depths of an in-service DOE LM disposal cell. A machine-learning approach was then developed using Google Earth Engine to integrate multi-source observations and estimate SM across six soil layers from depths of 0-2 m. The model predictors included backscatter observations from satellite Synthetic Aperture Radar, vegetation and temperature products from optical-infrared sensors, and accumulated rainfall data from Daymet. The model was trained using in-situ SM measurements from 2019 and validated using data from 2014-2018 and 2020-2021. The approach produced accurate SM estimates for the six soil layers (R-values from 0.75 to 0.94; RMSE from 0.003 to 0.017 cm³/cm³; bias ~0.00 cm³/cm³). Additionally, the approach captured seasonal SM variability and spatial heterogeneity at 30-m resolution. The machine-learning based multi-source data fusion approach may characterize soil moisture dynamics at DOE LM disposal sites better than in situ measurements alone.