复杂山地的气候变化研究是国内外的热点，也是区域发展的重大需求。地面观测资料的稀缺和空间插值方法的局限是制约山地气候变化研究的突出瓶颈。再分析数据作为观测资料的替代品被应用广泛，但在山区的可信度尚需严格检验，而且再分析资料的空间分辨率仍不能反映微观尺度下的气候变化特征。因此，进行再分析资料空间降尺度有重要科学意义。本研究以天山山区为例，基于网格和地面观测数据，运用多种统计诊断方法系统评估ERA-Interim气温资料的可信度；构建不依赖地面观测站点的气温高程校正模型，将ERA-Interim 0.25°×0.25° 3小时气温数据降尺度到100m网格；根据坡向和坡度信息，运用地形调节统计模型，对降尺度数据进行地形校正；根据植被指数、地表反射率与气温的相关性，构建下垫面校正模型，进一步修正降尺度数据，创建天山山区3小时100m网格气温数据集，以期为山地气候变化研究和降尺度技术推广提供科学参考。

The research of climate change in complex mountain regions is a global hotspot and also a great demand for regional development. Lack of observations and defect of spatial interpolation are the prominent bottleneck problems for mountain climate change research. As the surrogate of observations, reanalysis is widely applied recently. However, the reliability of reanalysis should be precise evaluated at mountain regions. The spatial resolution of reanalysis could not represent the characteristics of micro-scale environment. Therefore, downscaling of reanalysis is of great scientific significance. In this proposal, taking the Tianshan Mountains as an example, temperature data from ERA-Interim reanalysis product is systematically evaluated based on gridded data and observations using multi-methods of statistical diagnostic analysis. A robust temperature elevation correction model which is independent of observations is created based on ERA-Interim data. Then, ERA-Interim 0.25°×0.25°3-hourly temperature data is downscaled to 100m grid. The downscaled temperature is corrected according to aspect and slope information derived from DEM using topographical adjust statistical model. An underlying surface correction model is created according to the correlation between vegetation index, albedo and temperature. The downscaled temperature is further corrected. A temperature dataset for the Tianshan Mountains in high temporal-spatial resolution (3h, 100m) is established. This study tends to provide a scientific reference for mountain climate change research and downscaling technique promotion.