Abstract:Accurate subsurface soil moisture (SM) estimation is critical for vegetation growth, drought monitoring, and climate change mitigation, yet remains a significant challenge. Previous data assimilation (DA) approaches are limited to only surface SM assimilation. In this study, we utilized the proxy subsurface SM estimated via the exponential filter method (ExpF) as another assimilation variable in our 2‐dimensional DA. Meanwhile, the dual updating DA scheme was implemented to simultaneously update model parameters and states. The two DA pathways were incorporated into the proposed framework (DA_1E_D), which enhanced the subsurface SM accuracy, with the effects of 2‐dimensional assimilation being more significant. Under 2‐dimensional DA, the information transfer between layers was more accurately characterized, leading to overall improvements with unbiased root‐mean‐square error (ubRMSE) reductions of 0.015 and 0.005 m3 · m−3, and Kling–Gupta efficiency (KGE) increases of 0.248 and 0.067 for surface and subsurface SM, respectively, across five SM networks. The soil thickness (d2) and hydraulic conductivity exponent (expt2) are the most influential parameters affecting subsurface SM dynamics through model propagation. DA_1E_D also outperformed ExpF in subsurface SM accuracy, particularly in SM networks with weak surface‐subsurface correlation, achieving an average ubRMSE reduction of 0.003 m3 · m−3 and an average KGE increase of 0.202. It was also applied to Soil Moisture Active Passive data at regional scale, demonstrating significant improvements. The model surface‐subsurface SM coupling was adjusted toward the actual coupling after subsurface assimilation and dual updating. This study may provide new insights into the diagnosis and refinements of the model representation of surface‐subsurface processes.

Improving Subsurface Soil Moisture Estimation Using a 2‐dimensional Data Assimilation Framework Incorporated with a Dual State‐Parameter Scheme