Journal of Arid Meteorology ›› 2023, Vol. 41 ›› Issue (5): 792-801.DOI: 10.11755/j.issn.1006-7639(2023)-05-0792

• Technical Reports • Previous Articles     Next Articles

The application of three interpolation methods to temperature in southwestern China

GAI Changsong1(), CAO Lijuan2(), YANG Yuanyan1,3   

  1. 1. CMA Key Open Laboratory of Transforming Climate Resources to Economy, Chongqing Meteorological Information and Technology Support Centre, Chongqing 401147, China
    2. National Meteorological Information Centre, Beijing 100081, China
    3. Chongqing Institute of Meteorological Sciences, Chongqing 401147, China
  • Received:2022-10-25 Revised:2023-05-25 Online:2023-10-31 Published:2023-11-03

三种气温插补方法在中国西南地区的应用分析

盖长松1(), 曹丽娟2(), 阳园燕1,3   

  1. 1.中国气象局气候资源经济转化重点开放实验室,重庆市气象信息与技术保障中心,重庆 401147
    2.国家气象信息中心,北京 100081
    3.重庆市气象科学研究所,重庆 401147
  • 通讯作者: 曹丽娟(1980—),女,正高级工程师,主要从事气候资料分析与气候变化研究。E-mail: caolj@cma.gov.cn
  • 作者简介:盖长松(1977—),男,高级工程师,主要从事气象观测数据质量控制研究。E-mail: gai_cs@163.com
  • 基金资助:
    重庆市气象部门业务技术攻关项目“基于融合实况产品与机器学习的温/湿度要素质量控制与数据插补技术研究”(YWJSGG-202209);中国气象局创新发展专项“全球关键气候要素长序列数据产品研制”(CXFZ2023J049)

Abstract:

The interpolation of meteorological observation data is an important technique to improve data integrity and recover authenticity of missing values. The applicability analysis of three interpolation methods namely standardized series, spatial regression and random forest to daily mean temperature series in five major climatic divisions and monthly mean temperature series at two centennial stations in Qianwei and Beibei is carried out in order to improve the accuracy of temperature interpolation in southwestern China, and four test indicators including mean absolute error, root mean square error and the proportion of samples (P0.8 and P0.5) with the bias between the interpolation value and the observation within ±0.8 ℃ and ±0.5 ℃ are used to evaluate. The results show that three interpolation methods are better in interpolating daily mean temperature in five climatic zones and monthly mean temperature at two centennial stations in southwestern China, among them the spatial regression method has the highest accuracy and the best applicability, and its interpolation accuracy is higher than those of other two methods in five climatic zones. The P0.8 test indicator of daily mean temperature interpolated by the spatial regression method reaches about 0.90 in Sichuan Basin with a relatively flat topography, and it reaches more than 0.60 in mountainous area in southwestern Sichuan and northern Yunnan with a most rugged topography, which indicates that the terrain has obvious influence on the accuracy of temperature interpolation. The optimal numbers of reference stations can effectively reduce interpolation errors, and the interpolation errors of more than 95% samples at centennial stations can be controlled within ±0.5 ℃.

Key words: standardized series, spatial regression, random forest, temperature data interpolation, southwestern China, centennial station

摘要:

观测数据的插补是提升气象数据完整性、恢复缺失数据真实性的重要手段。本文采用标准序列、空间回归和随机森林3种插补方法,对中国西南地区5个主要气候区地面气象观测站日平均气温序列数据以及犍为和北碚2个百年站月平均气温序列数据进行插补试验,并选用平均绝对误差、均方根误差以及插补值与观测值偏差分别在±0.8 ℃和±0.5 ℃区间的样本占比(P0.8P0.5)等4项指标对插补结果进行评估。结果表明:3种方法对中国西南地区5个气候区站点气温日均值与2个百年站气温月均值插补效果较好,且空间回归方法的插补精度高、适用性最好,在5个气候区的插补精度都高于其他两种方法,在地形较为平坦的四川盆地P0.8约0.90,在地形最为崎岖的川西南滇北山地P0.8也在0.60以上,地形对气温插补精度影响明显。选取最优参考站数可有效降低插补误差,2个百年站95%以上的月平均气温样本插补误差可控制在±0.5 ℃以内。

关键词: 标准序列, 空间回归, 随机森林, 气温数据插补, 中国西南地区, 百年站

CLC Number: