多种地表温度资料在青藏高原的适用性及其主要影响因子研究

Study on the Applicability and Main Influencing Factors of Various Surface Temperature Data in the Qinghai-Xizang Platea

摘要: 本文使用2001~2016年站点观测资料评估三套地表温度再分析产品(ERA5、JRA-55和MERRA-2)在青藏高原(简称“高原”)的适用性，并分析三个气象因子(海拔、NDVI和积雪覆盖率)对高原地表温度的影响程度。基于偏差、均方根误差和相关系数指标综合比较各再分析资料对观测值的模拟情况，结果表明：MERRA-2在青藏高原适用性最好，与地表温度相关性最显著，而ERA5和JRA-55在高原适用性不佳，对地表温度偏差较大，相关性弱。通过EOF分析得到青藏高原全域地表温度呈现随年份上升的趋势，三种再分析资料对于地表温度的时空变化均与观测资料所得的结果有所差异，其中仅有MERRA-2在第二模态中青藏高原大部分区域(除西南部分区域外)的地表温度呈现随年份上升的趋势。在全年和冬季两个尺度上，海拔和积雪覆盖率因子对地表温度均为负影响，而NDVI因子对地表温度的影响随季节变化，冬季为负影响(总效果系数为−0.117)，全年尺度下为正影响(0.134)，三个再分析地表温度对三个因子的响应情况与观测地表温度相差不大，其中综合比较MERRA-2的响应效果与观测资料最为接近。

Abstract: This paper uses station observations from 2001~2016 to assess the applicability of three surface temperature reanalysis data (ERA5, JRA-55, and MERRA-2) on the Qinghai-Xizang Platea (referred to as the “Plateau”) and to analyze the influence of three meteorological factors (elevation, NDVI, and snow cover) on surface temperatures. Based on the deviation, root-mean-square error and correlation coefficient statistical indexes and the in-situ observation data, the accuracy of the reanalysis data is comprehensively analyzed. Our findings indicate that MERRA-2 has the best applicability and shows the most significant correlation with the surface temperature, while ERA5 and JRA-55 have poor applicability on the plateau, large bias and weak correlation with the surface temperature. The EOF analysis shows that the surface temperature of the whole Plateau has been increasing over the years. However, the spatial and temporal variations of the surface temperature of three reanalysis data are different from those obtained from observations, with only MERRA-2 showing an increasing trend with years in most areas of the Plateau in the second mode (except for part of the southwestern part of the Plateau). At both the year-round and winter scales, the elevation and snow cover factors have a negative effect on surface temperature, while the effect of the NDVI factor on surface temperature varies seasonally. In winter, it has a negative effect in winter (total effect coefficient of −0.117), while it has a positive effect at the year-round scale (0.134). Furthermore, we found that the response of the three reanalyzed surface temperatures to the three factors is similar to that of the observed surface temperatures, with MERRA-2 showing the closest response effect to observations.