Time-frequency analysis of seismic data using synchrosqueezing wavelet transform

Introduction

Time-frequency (TF) analysis solves the problems of identifying and quantifying the oscillatory components presented in the signal, which has been exclusively utilized by the field of exploration geophysics over the past two decades (Partyka et al., 1998; Liu et al., 2011; Reine et al., 2009; Castagna et al., 2003). One of the most common use of TF analysis is that deeper channels are usually stronger at lower frequency and the shallower flank of the channel has stronger amplitudes at higher frequencies. In addition, TF analysis can be used to estimating attenuation, pore-pressure prediction, and seismic unconformities, and some implementation of seismic chronostratigraphy (Lin et al., 2013). Many approaches have been proposed for TF analysis, such as wavelet transform (WT), shot-time Fourier transform (STFT), Wigner-Ville distribution (WVD), S transform and recently proposed local attributes based TF (Liu et al., 2011). All of the mentioned approaches have their own advantages, however, they all have limited resolution either in time or in frequency.

The empirical mode decomposition (EMD) (Huang et al., 1998) algorithm can separate a signal into locally-constant frequency components, and have been shown to have a high resolution both in time and frequency with some types of extensions, like ensemble empirical mode decomposition (EEMD) and complete ensemble empirical mode decomposition (CEEMD). However, the EMD algorithm is still remaining heuristic because of the lack of mathematical support. The newly proposed synchrosqueezing wavelet transform (SSWT) capture the flavor and philosophy of the EMD approach, but with a mathematical way in constructing the components (Daubechies et al., 2011). Because of the high-resolution property of SSWT, it is becoming more and more popular for characterizing non-stationary property in signal analysis field recently. In the exploration field, SSWT has been successfully used for removing ground rolls (Shang et al., 2013). In this abstract, we use one benchmark non-stationary synthetic model for showing SSWT's high resolution both in time and frequency compared with other two robust TF decomposition approaches. We also applied SSWT onto two field data examples and show its potential in detecting anomalies of high-frequency attenuation and detecting deep-layer weak signal.

Time-frequency analysis of seismic data using synchrosqueezing wavelet transform