近半个世纪中国区域历史气温网格数据集的建立

CONSTRUCTION OF THE GRIDDED HISTORIC TEMPERATURE DATASET OVER CHINA DURING THE RECENT HALF CENTURY

气候序列长度不一致以及由此导致的空间抽样误差在气候变化检测研究中常常产生影响,因此往往需要把全球或者区域尺度气候数据转化为格点气候数据集。文中分别采用修改后的一级差分方法和普通Kriging方法,把中国大陆(共约728个站,不含港、澳、台地区)气象台站1951年1月—2004年12月经过质量控制和均一化的历史气温数据转化为2.5°×2.5°经纬度网格化数据集。对比分析表明:应用上述两种方法,(1)格点化带来的空间抽样误差较小;(2)格点化后的格点气温序列和站点序列相关程度很高;(3)不同的方法建立的格点数据集在序列的相关性、距平场的相似性描述方面均非常一致。二者的相关性和相似性也表明文中所建立的格点数据集是比较合理的。最后利用格点化后的气温数据集,分别采取距平平均方法和一级差分方法,对近54年中国气温变化趋势进行了更为准确的估计。不同的计算方法对中国区域气温增暖幅度的研究表明,整个中国区域内近50年气温表现出一致的增暖特点,其中以北方地区(西北、东北)气温增暖幅度最大,个别地区达到近0.6℃/(10a),而最低的长江中、上游地区,年均气温增暖幅度较小,还不到0.1℃/(10a),全国平均的年平均气温近54年来增暖速率约为0.23—0.25℃/(10a),这和其他专家得出的结论是基本一致的,而由于文中对数据集进行了较为系统的质量控制,使得估计结果更为精确、可信。

The detection of large-area averaged temperature changes using long-term in situ observations is always hampered by the spatial sampling errors caused by the different lengths of series. Usually, people need interpolate these stations＇ temperature series onto grid boxes/points on regional or global scales. Based on the quality controlled and homogenized observational monthly mean surface air temperature data series over China's Mainland （the number of the stations added up to 728, and the stations of HongKong, Taiwan and Macao were excluded in this study）, traditional Kriging methods and modified First Difference Method （FDM） were used in this paper to grid the monthly historic temperature dataset during Jan 1951 - Dec 2004 over China's Mainland by 2.5° × 2.5° resolutions, and much validation work was performed later to confirm the quality of the grid dataset. Comparison analysis shows no matter by which method the grid dataset is constructed ： （ 1 ） the spatial sampling errors of each grid box/point are relatively small, which means the interpolation is relatively accurate for a local region; （2） the correlations between gridded series and corresponding station series are very high, which means the gridded series did not change the original inter-annual and decadal regional climate change in station series; and （3） the anomaly distribution shows very high consistencies even by using different interpolating methods. Of course, these high correlations and consistencies between two different methods can also prove the rationality of the two versions of gridded datasets. In addition, the linear trend of surface air temperature changes over China during recent 54 years were also analyzed with the two versions of gridded datasets constructed by First Difference Method （FDM） and widely used Climate Anomaly Method （CAM）, similar results were achieved. During recent 50 years （1952- 2004）, the annual temperature over China shows consistently warming trends, with the most significant warming trend in Northern China, especially in Northeast China and Northwest China. It was near 0.6 ℃/10a in the maximum trend grid box in Northwest China, but much weaker in the upper-middle reaches of the Changjiang River Basin （less than 0.1℃/10a in the lowest trend region）. The whole country-averaged linear trend during the 54 years was about 0.23 -0.25 ℃/10a, and for seasonal means, the wintertime temperature＇s warming trend was the largest, while the summer＇s was the smallest. All these results are coinci- dent with many earlier researches, but our evaluation in this study seems more reliable and accurate due to the facts that the dataset we used is systematically quality controlled.