基于递归图理论的飓风信号非平稳特征定量分析

Quantification Analysis of Nonstationary Characteristics of Hurricane Utilizing Recurrence Plot Theory

摘要:

为解决传统时频域分析方法无法定量表征风信号的非平稳特征问题，提出采用递归定量分析方法对风场的非平稳特征进行可视化和定量化分析，该方法适用性强且噪声对计算过程的影响较小.首先，采用相空间重构技术和递归图理论对飓风Gustavo (2008)典型时间序列做非平稳性分析；然后，通过引入递归率、确定率、信息熵、对角线平均长度、层状度和捕获时间等6项递归定量分析指标对飓风过境的全过程进行信号特征的定量分析，并给出飓风信号和良态风信号的6项定量指标界限；最后用台风梅花（1109）进行了上述指标的验证.研究发现，飓风过境前期信号具有一定的周期性和平稳性，中期信号的非线性、非平稳性比较突出；6项递归定量分析指标能够客观区分开飓风信号和良态风信号；递归定量分析方法完全适用于风信号系统的研究，能够更全面、定量地描述风信号的非平稳特征.

Abstract:

Wind is a nonstationary stochastic process, but the traditional time domain or frequency domain analytical procedure can’t quantify its nonstationary characteristics. A visual and quantitative signal analytical procedure was put forward, which was the recurrence quantification analytical procedure. In this procedure, there were no specific requirements for the stationary characteristics of system and the procedure was not sensitive to noise. Firstly, the nonstationary characteristics and other characteristics of the typical time series of hurricane Gustavo (2008) were obtained by utilizing phase space reconstruction techniques and recurrence plot theory. Thereafter, 6 recurrence quantification indexes were imported, which were recurrence rate, determinism, entropy, average diagonal line length, laminarity and trapping time and quantifies the signal characteristics of hurricane in the whole process of making landing, and defines one boundary for every index between hurricane signal and normal wind signal. Lastly, the procedure and all the indexes defined are verified by typhoon Muifa (1109). The result shows that the early hurricane has certain periodicity and stationarity; however, the nonlinearity and nonstationary are particularly salient for the middle hurricane; 6 indexes can distinguish the signal characteristics between hurricane and normal wind; recurrence quantification analytical procedure completely suitable for the wind signal researches, and can fully describe the nonstationary characteristics of wind signals.