时间序列转折突变点检测的线性函数方法

根据转折突变的定义，在引入线性半截多项式的前提下，提出检测时间序列转折突变点的一种新方法。此方法不仅能找出时间序列中存在的多个转折突变点，而且使得检测到的突变点通过统计显著性检验。应用此方法对北半球1851—1990年年平均气温距平及其11年滑动平均序列，1901—1999年昆明5月降水及其11年滑动平均的降水序列分别进行了检测。发现北半球年平均气温距平在1879、1889、1939和1973年附近出现了年际转折突变，在1878、1888和1941年以及1972年出现了年代际尺度的转折突变；昆明5月雨量不存在年际转折突变点，但昆明5月雨量在1918、1966以及1978年附近出现了年代际尺度转折突变。

The Linear Function Method for Detecting Transit Jump Point of a Time Series

After introducing the linear step polynomial, a new method for detecting transit jump point of a time series was pointed out. The significant transit jumps in time series can be partly detected with this method. The new method was used to detect the transit jumps temperature anomaly series in Northern Hemisphere, and its series of 11 year running average, the May rainfall series of Kunming meteorological observatory and its series of 11 year running mean, respectively. The results indicated that the annual mean temperature anomaly in Northern Hemisphere, occurred interannual transit jumps in 1879, 1889, 1939 and 1973, and that of transit jumps of inter-decade scale in 1878, 1941, and 1972; Because the May rainfall of Kunming can not be detected any transit jump point but it's rainfall occurred transit jumps of interannual scale.