Summary
AVA/AVO inversion works as an important tool in elastic parameters estimation, which can be implemented under statistical such as Bayesian or deterministic inversion scheme. In this study, an AVA inversion strategy in combined time and frequency domain (AVA-TF) is proposed under Bayesian inversion scheme to enhance the resolution of elastic parameters including P-wave and Swave velocities and density with amplitude variation with incident angle seismic traces. The objective function of AVA-TF is initially yielded with Bayesian inference by combining seismic information in both time and frequency domain. Cauchy and Gaussian probability distributions are utilized for prior information of model parameters and likelihood function, respectively. The elastic parameters are further estimated by solving the initial objective function with additive model constraint. Synthetic examples demonstrate that the frequency spectrums of the estimated elastic parameters with the proposed AVA-TF strategy are much wider than those with conventional AVO inversion only in time domain, which verifies the advantage of the proposed inversion strategy in enhancing resolution of estimated parameters. Furthermore, synthetic examples with added Gaussian noise demonstrate the robustness of the proposed approach. From real data examples, we can see that more details of model parameters can be yielded with the proposed AVA-TF inversion strategy.
Introduction
Elastic parameters underground are usually estimated by utilizing pre-stack seismic data with abundant amplitude information, which can be implemented in an AVO inversion (Zong et al, 2012) or EVA (Elastic impedance variation incident angle) inversion scheme (Zong et al, 2013). However, besides the amplitude information, frequency or even phase information implicit in pre-stack seismic data has increasingly drawn more attention recently (Wu et al, 2014).
The objective function of AVO/AVA inversion can be established in deterministic or statistical such as Bayesian inference approach. Comparing to the deterministic approach, the uncertainty of model parameters or incomplete observed data information can be easier taken into consideration with Bayesian inference.
The AVO/AVA inversion approach is extended to combined time and frequency domain in this study based on the study of Bayesian AVO inversion approach in time domain by Zong et al (2012). The likelihood function is extend by combing the differences between observed and synthetic seismic data in both time and frequency domain. Cauchy distribution is utilized for prior distribution of model parameters. The elastic parameters are further estimated by solving the initial objective function with additive model constraint. Synthetic and real examples verify the advantage of the proposed AVA-TF inversion strategy in enhancing resolution of estimated parameters.
