Abstract There is an urgent need to evaluate the accuracy of reanalysis datasets to improve the reliability of climate research, hydrology, ecology, engineering work, and so on. The ground surface temperature (GST) is also regarded as a significant upper boundary condition when simulating permafrost distribution. In this study, the newest Modern-Era Retrospective Analysis for the Research and Applications Version 2 (MERRA2) reanalysis GST from the National Aeronautics and Space Administration's (NASA) Global Modeling and Assimilation Office (GMAO) is compared with GST data from 69 meteorological stations over the Qinghai-Tibet Plateau (QTP). Then, elevation correction of the MERRA2 GST was performed for different altitudes on seasonal scales. The correction indicates that the accuracy of raw MERRA2 GST data has been greatly improved after the elevation calibration. The decreased proportions of root mean square error (RMSE) for calibrated GST in winter, autumn, summer, and spring are 88.4, 92.8, 95.7, and 92.8%, respectively. The improvements in the mean bias error (MBE) in winter, autumn, summer, and spring are 88.4, 92.8, 95.7, and 92.8%, respectively. Most of the QTP presents a statistically increasing tendency of 0.2 to 0.5 °C/10a for GST at a significance level of 0.01 from 1980 to 2013. Then, the elevation corrected MERRA2 GST was applied in the surface frost number semi-physical permafrost model. The simulated results show that the permafrost area on the QTP is approximately 1.02 × 106 km2 in the early 21st century, excluding lakes and glaciers. The permafrost degradation was characterized by the disappearance of permafrost which mainly occurred in the eastern QTP.

中文翻译：

---

高程修正MERRA2再分析地表温度在青藏高原多年冻土模型中的应用

摘要 为提高气候研究、水文、生态、工程等工作的可靠性，迫切需要对再分析数据集的准确性进行评估。在模拟永久冻土分布时，地表温度 (GST) 也被认为是一个重要的上边界条件。在这项研究中，美国国家航空航天局 (NASA) 全球建模和同化办公室 (GMAO) 最新的现代时代回顾性分析研究和应用版本 2 (MERRA2) 再分析 GST 与 69 个气象站的 GST 数据进行了比较青藏高原（QTP）。然后，在季节性尺度上对不同海拔高度进行 MERRA2 GST 的海拔校正。校正表明，经过高程校准后，原始 MERRA2 GST 数据的准确性得到了很大提高。冬季、秋季、夏季和春季校准 GST 的均方根误差 (RMSE) 降低比例分别为 88.4、92.8、95.7 和 92.8%。冬季、秋季、夏季和春季的平均偏差误差 (MBE) 改进分别为 88.4%、92.8%、95.7% 和 92.8%。从1980年到2013年，大部分青藏高原的GST在0.01的显着性水平上呈现出0.2至0.5°C/10a的统计增加趋势。然后，将高程校正的MERRA2 GST应用于地表霜冻数半物理多年冻土模型。模拟结果表明，21世纪初青藏高原多年冻土面积约为1.02×106平方公里，不包括湖泊和冰川。