Estimation of daily net radiation based on the continuous meteorological data observed from different underlying surfaces
摘要
Net radiation (Rn) drives the exchange of sensible and latent heat fluxes between the surface and atmosphere. Previous studies have mostly focused on cross-site validation or long-term calibration of individual Rn models, yet few have conducted a coupled analysis of key variables, model types, and underlying surface conditions, or quantified the specific impacts of environmental factors on model performance. In this study, eight empirical and physics-based Rn models that incorporate incoming shortwave radiation (Rs), albedo (α), air and surface temperature (Ta and Ts), clearness index (CI), actual water vapor pressure (ea), relative humidity (RH) and NDVI were calibrated and evaluated based on the meteorological data from 2020 to 2023 under vegetation cover, snow cover, and bare soil surface in Northeast China. The results showed that the calibrated models performed best under vegetation cover, with R2 > 0.93, and RMSE < 31 W·m−2. In contrast, the Rn models performed worst under snow cover, with R2 ranging from 0.60 to 0.95, and RMSE ranging from 25.38 to 64.90 W·m−2. The empirical models that contained α, CI, Ta, Ts, ea, RH and NDVI were superior to the physics-based models. Moreover, elevated values of Rs, Ta, Ts, RH, NDVI and cloudiness were found to enhance model accuracy, while the increase in α had negative impacts on the Rn estimation. This study clarifies the positive and negative driving mechanisms of environmental factors on model accuracy, providing clear guidance for model parameter optimization. It proposes integrating a dynamic parameterization module of α into Rn models to enhance estimation accuracy across diverse underlying surfaces.