Enhancing Land Surface Net Radiation Simulations in Arid Regions via Surface Albedo Data Assimilation

Accurate simulation of surface net radiation (Rn) is essential for understanding land-atmosphere interactions and managing agricultural water and energy budgets. This study develops an Ensemble Kalman Filter (EnKF)-based albedo assimilation system, integrated with an energy-balanced Soil–Plant–Atmosphere Continuum (SPAC) model, to improve Rn simulations in the Heihe River Basin, a typical arid and semi-arid agricultural region. The system dynamically updates surface albedo through Bayesian inference, optimizing vegetation and soil reflectance parameters using multi-source observational data. Results demonstrate that the assimilation framework significantly enhances Rn simulation accuracy: the coefficient of determination increased by 16–45%, with the most pronounced improvements observed in alpine meadow and cropland ecosystems. The RMSE decreased by up to 45.7% at the Arou station, and systematic biases in farmland radiation simulations were effectively corrected. The proposed method outperformed single-variable assimilation approaches, particularly in heterogeneous landscapes, and successfully captured diurnal variations in albedo. A novel stratification scheme, which explicitly separates vegetation and soil albedo components, improved physical consistency and sensitivity to leaf area index (LAI) and soil moisture dynamics. These advancements provide a robust foundation for enhancing ecohydrological modeling in arid regions. Future efforts will focus on integrating multi-sensor data fusion and thermal inertia corrections to further improve sub-daily simulation capabilities.