从冬季平均温度、 温度日变化及日较差等方面入手， 基于2021年CLDAS逐小时产品评估了ECMWF全球高分辨率确定性数值预报产品对我国西南环横断山区复杂地形区近地面2 m温度的预报能力， 并通过区分高地形区（川西高原）和低地形区（四川盆地南部）， 对比了不同地形区近地面2 m温度预报的偏差特征。结果表明： （1）ECMWF模式可合理预报我国西南环横断山区冬季平均2 m温度的空间分布特征， 但偏差分布与地形高度有关， 随着地形高度的增加， 预报偏差呈增大趋势。（2）ECMWF模式很好再现了西南环横断山区冬季温度的日变化特征， 峰值时刻出现在14:00（北京时）； 各时刻温度的预报偏差在不同地形高度存在差异， 川西高原和横断山区的最大负偏差出现在下午， 四川盆地南部的最大负偏差出现在早晨。同时， 高地形区各时刻的预报偏差均高于低地形区。（3）ECMWF模式对日内各时刻不同地形处2 m温度的空间分布均有合理预报， 但偏差存在日变化特征。特别是在横断山区高地形区， 其在各时刻有不同的冷暖偏差特征。（4）在环横断山区温度日较差预报偏差较大的区域（大致为昆明准静止锋线发生频次较高的区域）， 模式对于温度日较差较大的日数， 其2 m温度的预报偏差要大于日较差较小的日数， 且在该区域内， 温度日较差的预报偏差相对不稳定。

Based on the hourly product of CLDAS （CMA Land Data Assimilation System） in 2021， this study is to evaluate the prediction capacity of the global high-resolution deterministic numerical prediction product ECMWF （European Center for Medium Weather Forecasting） for winter mean near-surface 2 m temperature of complex terrain region around the Hengduan mountains in southwest China by starting from winter average temperature， daily variation， and diurnal temperature range.And this study compares the temperature deviation characteristics of near-surface 2 m temperature in different topographic regions by distinguishing between high terrain region （the Western Sichuan Plateau） and low terrain region （the southern Sichuan Basin）.The results show that： （1） The ECMWF model can reasonably predict the spatial distribution characteristics of the winter mean near-surface 2 m temperature around the Hengduan mountains in southwest China， but the deviation distribution is related to the terrain height.With the increase of the terrain height， the prediction deviation tends to increase.（2） The ECMWF model well reproduces the daily variation characteristics of winter mean near-surface 2 m temperature around the Hengduan mountains in southwest China， with the peak time appearing at 14:00 （Beijing Time）.The prediction deviation of temperature at various times varies at different terrain heights.The maximum negative deviation of the western Sichuan Plateau and the Hengduan mountain regions occurs in the afternoon， while the maximum negative deviation of the south Sichuan Basin occurs in the morning.At the same time， the prediction deviation at each moment in high terrain areas is greater than the prediction deviation at each moment in low terrain areas.（3） The ECMWF model can reasonably predict for the spatial distribution of winter mean near-surface 2 m temperature over different terrain at various times during the day， but the deviations have diurnal variation characteristics.Especially in the high terrain region of the Hengduan mountain regions， there are different characteristics of cold and warm deviations at various times.（4） The area with large forecast bias of diurnal temperature range is generally the area with frequent Quasi-stationary front activities in Kunming.For the days （A total of 90 days， from December 1， 2021 to February 28， 2022） with large diurnal temperature range， the prediction deviation of winter mean near-surface 2 m temperature in this area is greater than the days with small diurnal temperature range.What’s more， the prediction deviation of diurnal temperature range is relatively unstable in the area with large forecast bias of diurnal temperature range.