基于系统聚类的北京气温变化过程遍历特征分析

Ergodicity analysis based on system clustering for temperature variation process of Beijing

气温变化过程遍历性研究是中长期气象预报的基础.以北京地区气温过程为例,分别研究了北京月最高气温、最低气温、平均气温序列的遍历特征.首先,基于系统聚类分别将各类气温以月为个体进行聚类;其次,以类为单位通过自相关图排除周期性序列,利用ADF检验其余序列的平稳性;再次,对通过平稳性检验的序列进行遍历性分析,包括均值遍历与协方差遍历两方面.研究结果表明,北京6、7、8月份最高气温月际年变化序列有均值遍历性特征,4、10月份最低气温序列则有过程遍历性.从长期趋势看,两类气温序列都围绕其均值上下波动,在4、10月份最低气温的月际年变化序列中,前期气温值对后期气温变化的影响是相对稳定的.最后,还开展了印证性分析和挖掘性分析,从遍历性的角度,佐证了前人的一些研究成果.

The ergodicity analysis of temperature variation is significant for medium and long-term weather forecast. The ergodicity of the month-highest temperature, month-lowest temperature, and month- average temperature in Beijing was discussed in this study. Firstly, each kind of temperature was clustered applying the systematic clustering method by taking a month as the individual. Based on this, autocorrelogram was used to eliminate the periodicity series and ADF （Advanced Dickey ＆ Fuller） was employed to test the stationary of the rest series. And then, ergodicity analysis was carried out to the series that had passed the stationary test, including mean ergodicity and covariance ergodicity. The results obtained from the above steps show that the month-highest temperature series of June, July and August has the mean ergodicity; the lowest temperature series of April and October has the progress ergodicity. For the long period, the two kind temperature series fluctuate around the mean temperature. For the lowest temperature of April and October, the temperature of the former time unit has a relatively stable impact on the latter time unit. At the end of this work, confirmation and further dig analysis also were carried out. The study results of former researchers were validated from the perspective of ergodicity in this study.