Characterization of Gas-Charged Porous Media from Joint Inversion of P/S- Wave Attenuation Based on OBS and/or Sonic Log Data

Open Access
Lei, Xiong
Graduate Program:
Energy and Mineral Engineering
Doctor of Philosophy
Document Type:
Date of Defense:
February 19, 2018
Committee Members:
  • Eugene Morgan, Dissertation Advisor
  • Eugene C Morgan, Committee Chair
  • Derek Elsworth, Committee Member
  • Hamid Emami-Meybodi, Committee Member
  • Sridhar Anandakrishnan, Outside Member
  • Q estimation
  • Attenuation inversion
  • Quality factor
  • Bayesian inversion
  • Gas saturation
  • Porosity
  • Permeability
  • OBS
  • Sonic log
Attenuation refers to the exponential decay of wave amplitude with distance. It is caused by energy-conserved factors (scattering or geometric dispersion), and inelastic dissipation (intrinsic attenuation) where energy is converted into heat. The intrinsic attenuation is frequency dependent and of interest to exploration geophysics, including application in wave propagation forward modeling, signal filtering, gas detection, full waveform inversion, and, as focused on in this dissertation, reservoir property estimation. We characterize gas reservoir by intrinsic attenuation inversion. The advantages of seismic attenuation inversion are that attenuation has a stronger relationship to hydraulic properties than velocity, and gas has more pronounced effects in terms of attenuation. The proposed methodology is easily extendable to oil and other types of reservoirs. The foundation of seismic attenuation inversion is the measurement of quality factor, Q, which is inversely proportional to attenuation. However, it is difficult to estimate Q from reflection data due to the presence of noise intervention, which limits its application. Many methods have been proposed for Q estimation mainly for VSP (vertical seismic profile), crosswell, or transmitted data. With this study, we extend those approaches to reflection data. However, the specific techniques to cope with the corresponding issues, comparison of the efficacy for different approaches, and a clear recommendation on which methods are the best to use under which circumstances are rarely presented. The first part of this thesis is dedicated to resolve these issues using synthetic seismic data. We focus on three frequency-domain methods: spectral ratio method (SRM), centroid frequency shift method (CFS), and peak frequency shift method (PFS). They are less affected by scattering interference compared with time-domain methods. For the three frequency-domain methods, five kinds of pre-processing procedures paired with them are tested. We first determine the optimal length of the window function (for seismic signal frequency transformation). Secondly, we find that a traditional FFT coupled with either the SRM or CFS methods works the best and about equally well in terms of Q estimation error under various levels of noise. A close second is a technique that involves the extraction of wavelets from the signal and their subsequent frequency transformation, again coupled with either SRM or CFS. It is noted that this technique is superior when dealing with thin layers because of its stronger capability of wavelet restoration. Additionally, we find that Q tends to be more accurately estimated for layers with higher attenuation. Moreover, the effective-bandwidth coefficients, which control the length of the effective signal participating in the Q estimation, from 0.2 ~ 0.4 are good values. Then, I show that the joint inversion of P- and S-wave quality factor (Qp and Qs) is powerful in characterizing gas-bearing porous media. Compared to the inversion of Qp alone, where a rock physics model giving Qp as an output is inverted for its input parameters (rock and fluid properties), the joint inversion has one more dimension of information, increasing constraints on the model to suppress the occurrence of multiple solutions. Additionally, joint inversion improves the model sensitivity to the input parameters, enhancing its reliability. Moreover, besides porosity, it allows us to invert one more parameter, here gas saturation. In this section, we implement the inversion workflow on the ocean bottom seismometer (OBS) data from Finneidfjord, Norway, where the free-gas accumulation takes place in the sub-seabed. After sensitivity analysis, the efficacy of the inversion for gas saturation and porosity is verified. The nonsensitive parameters are eliminated from the inversion and set as constants, which reduces the complexity of the problem. By using Differential-Evolution MCMC scheme, we efficiently sample the joint posterior of the saturation and porosity. The estimated gas saturation and porosity (modes of the posteriors) agree with previous research in Finneidfjord. So far, we just discuss and invert the porosity and saturation. The next step would be to invert more solvable unknowns by introducing more information. In the final part of the research, we integrate multiple geophysical datasets (OBS and sonic logs) to realize a more advanced joint inversion. Usually, both the compressional and shear wave sonic waveform data has higher frequencies than seismic. At the two different frequencies, we can have two pairs of Qp, Qs. Adding two more dimensions to the inverse problem constrain the inversion even further, thus reducing uncertainty in estimates and improving the number of the solvable parameters. After establishing the workflow, we take the Hydrate Ridge, Oregon margin where there is free gas accompanied beneath the gas hydrate as a practical example to show the validity of a four-parameter inversion. The gas saturation, porosity, permeability, and characteristic (inclusion) size are simultaneously inverted and in good agreement with the literature about the Hydrate Ridge.