Key points are not available for this paper at this time.
Accurately estimating daily shortwave downward radiation (SWDR) in mountainous terrain remains challenging due to complex topography and limited satellite observations. This study proposes a novel approach combining a physically terrain-corrected radiative transfer model with machine learning to efficiently estimate daily SWDR from only a few hourly or instantaneous satellite measurements. Among the tested models, the ShortWave Daily Topographic Radiation based on 3 Observations method (SWDaily-3O) demonstrated the best performance, significantly outperforming traditional flat surface assumptions. Over the rugged areas of the Tibetan Plateau, SWDaily-3O reduced the multi-year average root mean square error (RMSE) from 23.62 W/m2 to 5.57 W/m2, decreased the mean bias error (MBE) from 20.04 W/m2 to 4.35 W/m2, and improved the coefficient of determination (R2) from 0.833 to 0.991 relative to uncorrected SWDR. The model maintained high accuracy across various slope conditions and spatial resolutions (1-km and 5-km), and performed consistently with both instantaneous and hourly inputs. This strategy provides an accurate, computationally efficient, and scalable solution for daily SWDR estimation in complex mountainous regions, offering valuable support for mountain radiation studies and remote sensing applications.
Xian et al. (Mon,) studied this question.