基于CMIP6模式对青藏高原冬季降雪的时空特征研究
Study on the Temporal and Spatial Characteristics of Snowfall in Winter over Tibetan Plateau Based on the CMIP6 Models

本文利用5个国内的CMIP6 (Coupled Model Intercomparison Project in Phase 6)模式模拟及台站观测的1980~2014年逐月降雪数据，评估了模式对青藏高原(下称高原)冬季降雪的模拟性能，并分析了高原冬季降雪的时空变化特征。结果表明：1) 与观测值相比，5个模式模拟的冬季降雪的空间分布存在一定的差异，但均能模拟出在高原东南部存在大值区，其中FIO-ESM-2-0模式模拟值与观测值的偏差百分率相对较低，表明该模式能够较好地模拟出高原冬季降雪；2) 模拟值与观测值的空间相关系数、均方根误差及标准偏差比的结果不理想，但多模式集合平均后的结果均有所改进，意味着对区域性气候要素来说，多模式集合平均能够提高模拟性能；3) 空间分布趋势来看，观测值表明高原大部分地区冬季降雪存在减少的趋势，尤其在东南部减少更为显著，在时间上也呈现出减少的特征。与其他模式相比，FIO-ESM-2-0模式能够较好地模拟出高原冬季降雪的空间、时间变化趋势。4) 多模式集合平均结果表明高原冬季降雪主要分布在高原东南部及西南部边缘，且在高原东南部冬季降雪呈现出较为显著的减少趋势。
Using the monthly snowfall data from five domestic models of CMIP6 (Coupled Model Intercompari-son Project in Phase 6) and the observation during the period of 1980~2014, the performance of models for simulating the snowfall in winter over the Tibetan Plateau (TP) is projected, and the spatial and temporal characteristics of snowfall in winter over the TP are analyzed. The results show that: 1) Compared with the observation, there are some difference in simulating the snowfall in winter of the five models, however, all the five models can simulate that there is a high value of snowfall over the southeast of TP. Furthermore, FIO-ESM-2-0 model has lower bias percentage with observation, this implies that FIO-ESM-2-0 can better simulate the snowfall in winter; 2) the spatial correlation coefficients, root mean square error and standard deviation ration between the simulation and observation are not satisfactory, while the results are improved after multi-model ensemble mean, which can improve the simulation performance for a regional climatic element; 3) for the spatial distribution trend, the observation indicates that snowfall in winter has a decreasing trend over the most part of the TP, especially over the southeast part of the TP, the decreasing trend is more dramatic. The annual series also indicate a decreasing trend during 1980~2014. Compared with other models, FIO-ESM-2-0 model can better simulate the trend of snowfall in winter over the TP. 4) The multi-model ensemble mean suggests that the snowfall in winter is mainly distributed in the southeastern and southwestern margins of the TP, and it displays a significantly decreasing trend in the southeast part of the TP.