两种低频信息的关联和形成差异的原因

Causes for Correlation and Inconsistency Between Two Kinds of Low-Frequent Signals

在变振幅的数据分析中 ,低频信号可以反映在低频分析结果中 ,也可以体现在高频信号的振幅变化中。本文将其分别称为滤波低频信号和变幅低频信号。两种低频信号有很好的相关性 ,也存在明显差别。研究结果表明 ,滤波低频信号反映了过程整体性的上升或下降 ,变幅低频信号反映了极大值与极小值之间差异的缩小或加大。两种低频信号的差异多源于原始信号极大值与极小值变化过程之间的关系。在本文的例子中 ,冬季气温变化占优势时两种低频信号的滑动相关系数为正值 ;夏季气温变化占优势是导致两种低频信号差异的根本原因 ,在物理上表现为一定发生了强大的异常过程 ,打破了气温正常变化的秩序。大连站的气温变化主要受制于北方的冷空气过程 ,但当厄尔尼诺事件发生时 ,热带的暖过程大大加强 ,导致两种低频信号发生明显的差异。从 1965～ 2 0 0 3年 ,滑动相关系数共出现 6次负时段 ,都对应着厄尔尼诺事件。因此 ,在低频过程的研究中 ,需要用滤波低频信号研究整体性的低频变化。但是 ,在希望获取导致整体变化的主要成分时需要使用变幅低频信号。另外 ,在揭示异常过程时需要用滑动相关系数来发现两种低频信号的差异 ,提取异常信息。

In the analysis of data with amplitude variation, low-frequency signals may appear as decomposed results, referred to hereafter as filtered low-frequency signals (FLSs). Low-frequency signals may also be embodied in the amplitude variation of high-frequency signals, which are called amplitude-varied low-frequency signals (ALSs). The correlation between the two kinds of low-frequency signals is quite good, but with obvious inconsistency between them sometimes. It is indicated in this paper that an FLS embodies an average increase or decrease of the original signal, whereas an ALS expresses the decrease or increase of the difference between the extrema. Obvious inconsistency between the two kinds of low-frequency signals is determined by the relationship of the maxima and minima. In the example of the surface air temperature (SAT) in Dalian, the extrema are extreme air temperatures in winter and summer. When the variation of winter air temperature is dominant, the running correlation coefficient is positive. When the summer air temperature variation is dominant, the difference of the two kinds of low-frequency signals becomes obvious, a powerful anomalous process must have occurred to destroy the normal process. The SAT in Dalian is mainly controlled by the cold system of the frigid zone. However, when in El Nina events appear, the process in the tropic zone enhances and the inconsistency between the FLS and ALS occurs. There are six negative time periods in the running correlation coefficient in our example from 1965 to 2003, and all of them coincide with the periods of El Nino processes. It is suggested that the FLS is effective in expressing an average low frequency process, whereas the ALS be used to reveal the exact cause of the average variation. The correlation analysis of FLS and ALS is helpful in extracting information on anomalous events.