A new technique called unscented weighted ensemble Kalman filter (UWEnKF) was developed to improve soil moisture predictions by combining different data sources. The filter was tested using soil moisture data from the Yellow River in China. Results showed that UWEnKF outperformed the traditional ensemble Kalman filter (EnKF) in accuracy, especially with larger ensemble sizes and lower assimilation frequencies. UWEnKF was less sensitive to initial conditions and more robust against random noise. Despite being more computationally demanding, UWEnKF is a better choice for enhancing soil moisture predictions by incorporating various data inputs, like satellite observations, at longer assimilation intervals.