气温时间序列的均值变点检验的探究——以北京、上海和广州为例.pdfVIP

摘 要 日常生活中，人们的工作、学习等生活各方面均受到气候变化的影响。伴随着全球变 暖这一趋势越发明显，气候变化尤其是气温变化越来越被人们所重视并成为研究的热点问 题。通过总结国外已有的对于长气候序列中变点的检验方法，分析不同检验方法对时间序 列分布假设的要求，结合三大城市气温序列数据特点，确定最优的变点检验方法。主要研 究结果如下：第一、近60 年来，北京市气温升温趋势明显，上海和广州的气温波动较为 平稳；第二、因研究总体个数不足，三大城市年均气温序列不服从正态分布特点，SNHT 不适用于该气温序列的变点检验。PMFT 检验方法由于软件的不可得性，在实证方面有很 大的限制。CUSUM 检验统计量对研究样本的分布的限制性较低，该方法适用于检验三大 城市年均气温序列；第三、运用 CUSUM 方法对三大城市日均气温序列进行研究，结果 表明北京市的年均气温序列存在明显的变点，变点位置出现在1988 年。其他两大城市的 年均气温序列未出现变点，气温年均值统计特征未发生显著性的变化。这一结果与气象元 数据的记录相吻合，上海和广州自建站以来未发生过基站迁移事件。 关键词：气温均值序列；变点；CUSUM 检验统计量 I 万方数据 Abstract With global warming tendency increasing prominent, the temperature change has become a key factor on the impact of climate change. The change tendency and regularity of meteorological elements can be identified through the analyzing the historical meteorological records. Different test methods for long sequences of climate changepoint are summarized firstly. The assumes of the corresponding time series distribution for different methods are compared sequently, and a reliable method is chosen for the analyzing temperature mean series finally. The results obtained are listed as follows: First the temperature mean of Beijing is rising from 1952 to 2014 ,but the data of Shanghai and Guangzhou is relatively stable. Second, the average annual temperature of this three cities are not distributed normally in the past 60 years within an inadequate overall number, therefore SNHT is not suitable for testing its temperature changepoint. Third, CUSUM test statistic is applied on the three cities‘ temperature mean series for changepoint test .The result shows there is a changepoint in Beijing‘s temperature m