Capturing Spatiotemporal and Subgrid Variability in Global Land Surface Albedo Parameterization

Accurate surface albedo parameterization is critical for modeling Earth’s energy balance, yet many schemes rely on static look-up tables or semi-empirical formulations that fail to capture spatiotemporal variations and complex radiative interactions. This study develops a physics-informed machine learning parameterization using 19 years (2003–2021) of MODIS BRDF data to predict direct and diffuse albedo in visible and near-infrared bands across major land cover categories (LCCs). The framework leverages ten biogeophysical predictors, including solar geometry, vegetation state, soil moisture, soil texture, topography, and background climate. It comprises a dynamic component resolving physical spatiotemporal patterns and a static correction, the Surface Albedo Localization Factor (SALF), which accounts for subgrid heterogeneity, together explaining most observed variability. The parameterization shows strong agreement with MODIS albedo (overall R² = 0.81, MAPE = 0.08), with performance ranging from R² = 0.72 in diffuse visible to R² = 0.89 in diffuse near-infrared, and generally higher accuracy in the near-infrared part. It performs well across diverse LCCs, including grasslands, shrublands, croplands, and challenging barren regions where empirical methods underperform. SALF improves accuracy across all albedo parts (average R² increase of 0.11 and MAPE reduction of 0.025). Feature importance analysis highlights solar zenith angle and leaf area index as dominant dynamic drivers, with soil texture and topography shaping static variability. Counterfactual experiments confirm biophysically consistent albedo responses, enhancing interpretability and model trust. This framework offers a physically grounded alternative to empirical schemes and holds strong potential for integration into Earth system models to improve surface energy exchange representation. This repository hosts the machine learning models for land surface albedo parameterization, along with the Python code used for model development, testing, and data analysis associated with the manuscript. It also includes all required static input files to run the parameterization framework. For detailed instructions on usage and data structure, please refer to the accompanying README file.