基于CMIP6动力降尺度对青藏高原降水的评估
Evaluation of Precipitation over the Tibetan Plateau Based on CMIP6 Dynamic Downscaling

为了更好地了解目前数值预报模式对于青藏高原地区降水模拟的准确性，本文采用科学数据银行发布的历史和未来气候动力降尺度偏差校正后的CMIP6全球数据集作为驱动场，驱动WRF区域气候模式对高原进行动力降尺度模拟，得到水平分辨率为50 Km的模拟数据，此后提取其中的积云对流降水RAINC和非对流降水RAINNC数据并将其插值到0.5？ × 0.5？的空间分辨率上，选取国家青藏高原科学数据中心发布的中国地面降水0.5？ × 0.5？格点数据集来作为模式评估的参考数据，对比青藏高原地区WRF模式模拟的降水结果与观测值的差异以评估WRF模式的模拟准确性。结果表明：1) 高原年平均降水量呈现从东南向西北逐渐减小的趋势，WRF区域气候模式对于青藏高原地区的年平均降水量模拟结果偏高，但仍能较好的再现上述空间分布特征。2) 高原地区季节平均降水量分布为：夏季降水最多，春秋次之，冬季最少。WRF模式可以准确模拟出此分布特征，但整体上模拟值较观测值偏高。3) WRF模式对于高原不同地区的模拟准确性不同，根据年降水量随时间的变化特征来看，WRF模式对于降水量较少的地区模拟的差值较小，对于降水量较大的地区模拟差值较大。4) 高原上不同地区WRF模式模拟的降水量与观测数据降水量的相关性存在较大差异，在藏东、藏南、藏西各有一个站点的相关性较好，而藏北部地区的相关性最弱，但整体上都呈现出正相关的趋势。
In order to better understand the accuracy of current numerical prediction models for precipitation simulations on the Tibetan Plateau, this paper uses the CMIP6 global dataset corrected for historical and future climate dynamical downscaling deviations released by the Scientific Data Bank as the driving field to drive the WRF regional climate model to perform dynamical downscaling simulations on the plateau, and obtain simulated data with a horizontal resolution of 50 Km, after which the cumulus clouds are extracted Convective precipitation RAINC and non- convective precipitation RAINNC data are extracted and interpolated to a spatial resolution of 0.5？ × 0.5？, and the 0.5？ × 0.5？ grid point data set of Chinese surface precipitation released by the National Tibetan Plateau Science Data Center is selected as the reference data for model evaluation. model’s simulation accuracy. The results show that: 1) The annual mean precipitation on the plateau shows a decreasing trend from southeast to northwest, and the simulated results of the WRF regional climate model are high for the annual mean precipitation on the Tibetan Plateau region, but still can reproduce the above spatial distribution characteristics well. 2) The distribution of seasonal average precipitation in the plateau area is: the most precipitation in summer, the second in spring and autumn, and the least in winter. WRF model can accurately simulate this distribution, but the simulated value is higher than the observed value as a whole. 3) The simulation accuracy of WRF model for different regions of the plateau is different. According to the variation characteristics of annual precipitation with time, the simulation difference of WRF model for regions with less precipitation is small, and that for regions with large precipitation is large. 4) The correlation between the precipitation simulated by the WRF model and the precipitation from