英国CRU高分辨率格点资料揭示的20世纪中国气候变化

An Overview of China Climate Change over the 20th Century Using UK UEA/CRU High Resolution Grid Data

中国覆盖比较完整的台站观测始于1951年，1951年之前虽然有一些观测记录，但是残缺不全。所以要建立更长的气候序列就要吸收代用资料，但是代用资料可能与气候要素仅有一定程度的相关，不可能一一对应，因此应用代用资料重建的气候序列有一定的不确定性。英国East Anglia大学的Climatic Research Unit（简称CRU）通过整合已有的若干个知名数据库，重建了一套覆盖完整、高分辨率、且无缺测的月平均地表气候要素数据集，时间范围覆盖1901-2003年，空间为0．5°×0．5°经纬网格覆盖所有陆地。这套资料和中国已有的气候数据相比具有如下优点：第一，中国西部20世纪前半期非常缺少观测，CRU资料尽管包含插值带来的误差，经比较仍可作为有一定信度的参考；第二，中国现有的百年温度序列只是年或季分辨率，而CRU资料达到月分辨率；第三，建立这个序列仅使用观测结果，做统计内插，不包括代用资料所带来的不确定性。因此，CRU的序列与用代用资料补充得到的序列在资料方面有较大不同，比较这两个序列，不仅可以进一步确认中国气候变化的特征，也可以彼此校正。结果表明：（1）CRU资料反映的全国年平均温度年际变化和考虑代用资料重建的序列吻合得很好，相关系数达到0．84；（2）区域尺度上，两者在10个典型分区的气温变率也相当一致，相关整体保持在0．8左右，仅新疆西南部和西藏西北部两者差异较大；（3）CRU资料揭示的中国年总降水量在1951～2000年的变化与160站观测吻合，相关系数达到0．93；（4）CRU资料的中国东部四季降水量和重建资料十分一致，秋季一致性最好，相关0．93；（5）CRU资料和重建的序列比较一致地表现出中国温度和降水年代际变化的主要特征，其给出的20世纪20年代中国大旱和20世纪40年代中国高温的空间分布与作者过去的结论相一致。这表明，作者过去重建的中国气候序列有比较大的可靠性，而CRU资料也提供了新的信息，特别是在20世纪前半期和中国西部。

The operational observations in China started in 1951. The availabity of instrumental observations is poor in the first half of the 20th century. Therefore, it is difficult to use them in discussing long term climate change issue. Although a number of proxy data could be used, the deficiencies are apparent, such as the coarse temporal and spatial resolution, etc. The University of East Anglia recently released the latest version of high resolution grid dataset, named as CRU-TS2.1. It is a completely covered monthly dataset of surface climatic variables. Compared with domestic observations, it has some merits concerning China climate change over the 20th century. Firstly, this dataset provides new information of the climate change over western China before 1950. Although it should be noted that the records are interpolated from observations, the results show significant correlation with the observation in the second half of the 20th century. Secondly, CRU dataset provides monthly mean fields, while domestic centuryscale series is just annual mean or seasonal mean data in the first half of the 20th century. Moreover, there are not any proxy data included in the construction of this dataset, which can bring noise and uncertainty. Hence, comparison between the CRU dataset and domestic observations is the way to verify the characteristics of China climate change during the 20th century and to validate the quality of the dataset in both China and UK. The results are as follows： （1） The interannual temperature variation is identical in both datasets. Their positive correlation coefficient is 0.84. They slightly differ from each other in the 1920s, when CRU underestimates the warming change in China, and therefore overestimates the warming trend of the whole century. （2） Even in the given 10 regional scales, both CRU and domestic data exhibit good consistency, apart from Tibet and Xinjiang areas. About one celsius degree bias is estimated by CRU in Tibet during the 1920s, which is the major difference from the reconstructed record. （3） The seasonal variability of precipitation over eastern China is identical in both series. The highest correlation coefficient （0.93） for 100 years among the four seasons is in autumn, while winter is the season that their correlation is the lowest one （0. 77）. （4） CRU data exhibits appropriate interdecadal variation of temperature and precipitation as shown in China records. Particularly, CRU dataset also presents some primary features before 1951, especially in western China, where and when no observational data can be available. Therefore, CRU high resolution grid data present a more complete picture of climate change in China over the 20th century.