A comprehensive model assessment on a 6-hourly scale from multiple dimensions is critical for model selection of dynamic downscaling to reduce uncertainties. In this study, we evaluated the performance of 13 models involved in the Coupled Model Intercomparison Project-6 (CMIP6) by comparing the simulated meteorological variables at the pressure levels of 850, 500, and 250 hPa, including 6-hourly air temperature, specific humidity, zonal wind, meridional wind, and geopotential height, to those from the ERA5 reanalysis data for the 1979–2014 period. The results indicated that the CMIP6 models could mostly reproduce the spatial pattern of the climatology. Most models underestimated air temperature and geopotential height while overestimating specific humidity and zonal and meridional wind speeds in the upper troposphere. Additionally, the interannual variability of zonal and meridional wind exhibited a relatively better performance but limited ability for specific humidity at 850 hPa. Regarding the annual and diurnal cycles, CMIP6 models reasonably captured the annual cycle shape, while an overestimation was detected in simulating the diurnal amplitude, notably at 250 hPa. Based on a comprehensive rating index and overall rankings, our findings showed that no single model was identified as suitable for the simulation of any variables and regions. The models performed well for five variables, including MPI-ESM1-2-LR, MPI-ESM1-2-HR, TaiESM1, and UKESM1-0-LL, over East Asia. Then, according to the overall performance of the five variables and model accessibility, the optimal model varied by region and shared socioeconomic pathway (SSP) scenario. The multi-model ensemble mean outperformed individual models over almost all regions when it comes to comprehensive performance. The MPI-ESM1-2-LR and MRI-ESM2-0 models were the best two out of 13 models as lateral boundary conditions applied to seven climate scenarios over East Asia. This study provides valuable scientific references for selecting the optimal CMIP6 models for the projection of dynamic downscaling over East Asia and eight subregions.

从多个维度进行 6 小时尺度的综合模型评估对于动态降尺度模型选择以减少不确定性至关重要。在这项研究中，我们通过比较模拟的气象变量在 850、500 和 250 hPa 的压力水平，包括 6 小时气温、特定湿度、纬向风、经向风和位势高度，与 1979-2014 年 ERA5 再分析数据的数据一致。结果表明，CMIP6模型能够大部分再现气候学的空间格局。大多数模型低估了气温和位势高度，同时高估了对流层上层的比湿度以及纬向和经向风速。此外，纬向风和经向风的年际变率表现出相对较好的性能，但在 850 hPa 比湿度的能力有限。关于年和日周期，CMIP6 模型合理地捕捉了年周期形状，而在模拟日振幅时检测到高估，特别是在 250 hPa 时。基于综合评级指数和总体排名，我们的研究结果表明，没有一个模型被确定为适合模拟任何变量和区域。这些模型在东亚的五个变量上表现良好，包括 MPI-ESM1-2-LR、MPI-ESM1-2-HR、TaiESM1 和 UKESM1-0-LL。然后，根据五个变量的整体表现和模型可及性，最优模型因地区和共享社会经济路径（SSP）情景而异。多模型集成意味着在综合性能方面几乎在所有地区都优于单个模型。作为应用于东亚七种气候情景的横向边界条件，MPI-ESM1-2-LR 和 MRI-ESM2-0 模型是 13 个模型中最好的两个。本研究为选择最佳 CMIP6 模型以预测东亚和八个次区域的动态降尺度提供了有价值的科学参考。