Item

Abstract

Solar radiation drives numerous geophysical and biological processes in Antarctica, including sea ice melting, ice sheet mass balance, and the photosynthetic processes of phytoplankton in the polar marine environment. Despite the ability of reanalysis and satellite products to provide valuable insights into global-scale solar radiation, the ground-based solar radiation in the polar region remains poorly understood due to the harsh conditions in Antarctica. This study evaluated the performance of empirical models and machine learning models to establish a long-term daily global solar radiation (DGSR) dataset using meteorological observation data from Zhongshan Station and the Great Wall Station. The machine learning method was found to outperform the empirical formula method, with the random forest (XGBoost) model demonstrating the best performance in terms of DGSR hindcast estimation at the Zhongshan Station (Great Wall Station). The annual DGSR at both stations showed a similar trend, increasing from 1990 to 2004 and then decreasing after that. In addition to clouds and water vapor, poor weather conditions also significantly affect solar radiation at these stations. The high-precision, long-term DGSR dataset can provide a quantitative understanding of Earth's surface radiation balance and validate satellite data for the Antarctic region, advancing our knowledge of Antarctica's role in global climate change and the interactions between snow, ice, and the atmosphere.