Accurate estimation of acoustic impedance based on spectral inversion

Abstract

A B S T R A C TAcoustic impedance is one of the best attributes for seismic interpretation and reservoir characterisation. We present an approach for estimating acoustic impedance accurately from a band-limited and noisy seismic data. The approach is composed of two stages: inverting for reflectivity from seismic data and then estimating impedance from the reflectivity inverted in the first stage. For the first stage, we achieve a twostep spectral inversion that locates the positions of reflection coefficients in the first step and determines the amplitudes of the reflection coefficients in the second step under the constraints of the positions located in the first step. For the second stage, we construct an iterative impedance estimation algorithm based on reflectivity. In each iteration, the iterative impedance estimation algorithm estimates the absolute acoustic impedance based on an initial acoustic impedance model that is given by summing the high-frequency component of acoustic impedance estimated at the last iteration and a low-frequency component determined in advance using other data. The known low-frequency component is used to restrict the acoustic impedance variation tendency in each iteration. Examples using one-and two-dimensional synthetic and field seismic data show that the approach is flexible and superior to the conventional spectral inversion and recursive inversion methods for generating more accurate acoustic impedance models.Key words: Seismic data, Acoustic impedance, Reflection coefficients, Spectral inversion. I N T R O D U C T I O NAcoustic impedance (AI) represents the rock property of subsurface formations, being closely related to lithology, porosity, and filling matter in pores (Farfour, Yoon and Kim 2015). AI inversion becomes an essential method of quantitatively interpreting seismic data and estimating reservoir properties. Conventional AI inversion methods include the direct inversion, e.g., trace integration (Oldenburg, Scheuer and Levy 1983) and recursive inversion (Lavergne 1975;Lindseth 1979), iterative inversion, e.g., model-based inversion (Cooke and Schneider 1983), sparse spike inversion (Madiba and * E-mail: zhangjz@ouc.edu.cn McMechan 2003), and nonlinear inversion (Zhang, Shang and Yang 2009;Baddari et al. 2010). The trace integration is easily implemented, but it cannot produce the absolute value of the AI. The recursive inversion can reserve well the characteristics of seismic reflection and clearly reflect lithology changes in space, but is restricted by seismic bandwidth and interfered by noises (Sarwar 1991), leading to a vertically low-resolution result. The result usually reflects relative changes of AI, and thus, a given low-frequency component must be merged with it to obtain the associated absolute values (Cooke and Schneider 1983). In addition, the inevitable accumulated errors in the conventional recursive method limit its retrieval accuracy. The model-based inversion suffers from the non-uniqueness problem more