成晓裕 张进

摘 要：以我国15个国家级地面气象代表站1961—2010年年平均气温序列为研究对象，设计各站年平均气温序列分别在前、中、末段连续缺测（5年、10年、15年），利用标准序列、多元回归、差值等方法进行序列订正试验，计算原序列与订正后序列气候变化趋势的绝对趋势偏差、均方根误差和平均绝对误差，考察了几种插值方法的相对优劣。结果表明：针对不同订正方法，相同缺测年份下，前段缺测有80%的站用多元回归插值法效果较好，中段和末段缺测分别有60%和53%的站用标准序列插值法效果较好;因为序列趋势变化受两端数据影响较大，所以中段位置插补订正对趋势变化影响最小;一般来说，低海拔地区较高海拔地区的插补订正效果要好，随缺测年份增加偏差在增大。

关键词：连续缺测;插补方法;标准序列法;多元线性回归法;差值法

中图分类号：P412.11 文献标识码：A 文章编号：1003-5168（2021）23-0139-04

Abstract： The annual average temperature series from 1961 to 2010 of 15 national-level surface meteorological representative stations in China are used as the research object.The annual average temperature series of each station are designed to be continuously absent in the front， middle， and end stages （5 years， 10 years， 15 years）， Use standardized， multiple regression， difference and other methods to carry out sequence correction experiment， calculate the absolute trend deviation， root mean square error and average absolute error of the original sequence and the corrected sequence of climate change trend. The relative advantages and disadvantages of several interpolation methods are investigated. The results show that for different correction methods， under the same year of missing test， 80% of the stations with missing test in the front section have better results with multiple interpolation， and 60% and 53% of stations with missing test in the middle and end sections use standard method is good; because the sequence trend change is greatly affected by the data at both ends， the interpolation correction of the middle position has the least impact on the trend change; generally speaking， the effect of the interpolation correction in the low altitude area is better then hight altitude area ， The deviation is increasing with the missing year .

Keywords： continuous missing test;interpolation method;standard sequence method;multiple linear regression;vogel method

氣温是气候变化监测的关键要素。分析气温长期变化趋势需要长时间的连续数据，而气温数据由于仪器故障、台站撤并、非正常人为干预以及不可抗力等因素造成数据缺测的现象比较常见。数据缺测对气候长期变化趋势分析有较大影响[1-2]，不能满足气候变化研究等工作的需求，所以对缺测数据进行适当插补是有意义的。

很多学者对缺测数据的插补进行过研究。屠其璞[3]介绍气温、降水序列插补基本方法。黄嘉佑[4-5]对单月气温序列使用多种回归模型进行缺测数据的恢复研究。……