Soil Moisture Prediction: A Review of Models, Principles, Applications and Future Needs

Abstract

Soil moisture is a fundamental variable controlling hydrologic partitioning, land–atmosphere exchange, and biogeochemical cycling. Although widely characterized through observations, remote sensing, and data-driven approaches, the representation of soil water fluxes and subsurface processes in predictive models remains less systematically synthesized. Here, we present a structured review of widely used predictive models, focusing on their process formulations, spatial discretization, and treatment of soil water fluxes across scales. The synthesis reveals consistent trade-offs among physical realism, scalability, data intensity, and computational efficiency. Conceptual models based on tipping-bucket approaches simplify soil water movement as threshold-driven storage processes and underrepresent transient redistribution and deep soil moisture dynamics. Storage-routing models introduce flux-based redistribution but rely on empirical parameterization that dampen transient flux dynamics. Physically based models resolve hydraulic gradients and provide the highest process fidelity but require extensive data and computational resources. Across scales, the absence or simplification of preferential flow and lateral subsurface flow constitutes the dominant structural limitation, producing systematic smoothing of soil moisture variability. The review provides a comparative basis for model selection and highlights critical directions for improving the representation of subsurface hydrologic dynamics in predictive frameworks.