Identifying the State Dependence of Effective Material Properties in a Simplified Hydrologic Hillslope Model

Abstract Accurate representation of closure relationships, such as subsurface storage‐discharge, is essential for hydrologic modeling at the hillslope scale, but remains challenging due to the nonlinear state‐dependent nature of soil hydraulic behavior. We investigate the state dependence of such parameters and closure relationships at the hillslope scale by developing a data assimilation framework that couples a 2D Boussinesq groundwater model with a simplified explicit unsaturated zone reservoir, avoiding the need to solve the full 1D Richards equation. A sequential particle filter (PF) is applied to dynamically update model states and parameters using observed seepage discharge and groundwater depth from the Landscape Evolution Observatory (LEO), the world's largest controlled artificial hillslope facility. Compared to a time‐invariant calibrated model (Kling–Gupta Efficiency, KGE = 0.85), the PF achieves significantly higher KGEs of 0.94, 0.94, and 0.96 for ensembles of 25, 100, and 400 particles, respectively, demonstrating robustness and computational efficiency. Correlation analysis reveals that parameter evolution is strongly state‐dependent: saturated hydraulic conductivity shows consistent negative correlation with groundwater levels (up to , ), reflecting compensatory behavior for unmodeled lateral unsaturated flow; van Genuchten parameters and evolve in opposing directions to regulate retention behavior; and specific yield , derived from , , and pressure head, acts as a dynamic closure variable that declines with rising water table. These results provide a demonstration of temporally evolving parameters and closure relationships in a first‐order catchment and underscore the ability of PFs to reveal emergent hydrologic behavior while preserving model tractability.