Solar radiation on complex underlying surfaces in Xinjiang: A typical arid and semi-arid region in the northwestern China

Abstract

Xinjiang, located in the northwestern China, is a typical arid and semi-arid region characterized by abundant solar energy resources and highly diverse surface types, including oases, deserts, and snow-covered areas. Solar radiation plays a fundamental role in the Earth's energy budget, governing land–atmosphere interactions and driving essential atmospheric processes. Accurate quantification of downward surface solar radiation (DSSR) over such complex terrains is critical for improving regional climate simulations, optimizing water resource allocation, and facilitating the efficient utilization of solar energy. However, the scarcity of ground-based radiation measurements in this region hinders comprehensive solar resource assessments. Consequently, satellite-derived radiation products, such as those from the Clouds and the Earth's Radiant Energy System (CERES) Single Satellite Footprint (SSF) Aqua dataset, serve as valuable substitutes. Nonetheless, uncertainties in DSSR retrievals over complex surface types necessitate systematic validation. Therefore, in this study the downward surface solar radiation (DSSR) derived from satellite for cloud and all sky conditions were compared with ground-based observations.The CERES-derived DSSR showed overall agreement with ground-based observations in temporal and spatial patterns but tended to overestimate radiation, especially in the southern Xinjiang. Larger discrepancies occurred in Kashgar and Hotan, mainly due to dust from the Taklamakan Desert, which affects satellite retrieval accuracy.Under the clear-sky conditions, CERES DSSR data performed better than that under the all-sky conditions, indicating that clouds had a significant impact on CERES DSSR retrieval, especially in Yanqi and Tacheng. A similar effect is observed in Ruoqiang and Hotan, where the typically low cloud cover suggested that inaccuracies may stem from clouds or the misinterpretation of dust as clouds. Six fitting models were optimized, with results showing that the linear model performed best under the both all-sky and clear-sky conditions at the most stations.