资料均一化对气温变化趋势及其城市化偏差估计的影响：以北京地区为例

当前的地面气候观测资料普遍存在非气候性因素导致的非均一性,对气候变化监测和研究结论可靠性造成重要影响。结合观测台站的历史沿革数据,使用ACMANT和Pairwise Comparisons方法以及RHtest V4软件,对北京地区20个台站均一化前的月平均气温序列进行了非均一性检验和订正,最后评估了均一化对北京地区气温序列变化趋势及其城市化偏差估算的影响。结果表明：除元数据中记录的断点外,无元数据记录的断点也会对序列的趋势变化造成明显影响,其中乡村站最显著;经过订正,1958—2018年整个北京地区、乡村站以及城市站增温趋势分别为0.27℃/(10 a)、0.10℃/(10 a)和0.32℃/(10 a),较订正前分别上升了0.03℃/(10 a)、0.06℃/(10 a)和0.02℃/(10 a)。利用均一化资料估算,1958—2018年北京观象台的城市化影响为0.24℃/(10 a),城市化贡献率为70.2%,评估结果较前人结论有所降低。可见,在现有的北京地区气温资料序列中,仍可能存在较明显的非均一性和未被记录的断点,对区域平均气温趋势估算具有显著影响。

Effects of data homogenization on the estimates of temperature trend and urbanization bias: taking Beijing area as an example

Inhomogeneity caused by non-climate factors is common in current observation data of surface climate, exerting a great impact on the reliability of climate change monitoring and studies. Based on metadata of observational stations, the non-homogenized (only quality-controlled) monthly mean surface air temperature data series of 20 stations in Beijing were first tested and adjusted for inhomogeneities by the methods of ACMANT and Pairwise Comparisons and RHtestV4 software. The effects of homogenization on the estimates of temperature trend and its urbanization bias were then evaluated. The results show that, in addition to the change points recorded in the metadata, undocumented change points will have a significant effect on the trends of temperature series, especially for the rural stations, which were mostly missed in the previous works of homogenization. The warming trends of the homogenized temperature series for the whole area (all stations), rural stations and urban stations from 1958 to 2018 are 0.27℃/(10 a), 0.10℃/(10 a) and 0.32℃/(10 a), respectively, with an increase of 0.03℃/(10 a), 0.06℃/(10 a) and 0.02℃/(10 a), respectively, compared to the non-homogenized data. Based on the homogenized data, it was estimated that the urbanization effects in the Beijing Observatory data series from 1958 to 2018 is 0.24℃/(10 a), and the urbanization contribution is 70.2%, which are lower than the previous studies of our group. Hence, there are still obvious inhomogeneities and undocumented change points in the existing temperature data in Beijing area, which will have a significant impact on the accurate estimates of regional average temperature change and the urbanization-induced systematic bias of the temperature data series.