Reliable soil moisture monitoring and prediction remain challenging due to spatial heterogeneity, temporal gaps in satellite observations, and limitations of individual sensors. This talk explores complementary approaches to addressing these challenges, including large-scale data assimilation to integrate satellite observations with land-surface models, short-term change-detection methods to capture rapid soil moisture dynamics, and AI-based time-series reconstruction to provide more continuous estimates. At the local scale, drone-based SAR tomography offers new opportunities to investigate radar scattering within vegetation and improve soil moisture retrieval. Together, these approaches highlight the potential of combining Earth observation, physical modelling, and artificial intelligence for improved soil moisture monitoring across scales.