CWP Seminars - 2013 Spring

CWP C-Team members will lead a session on effective designs for slides and presentations.

Team leader: Prof. Dave Hale
Members: Stefan Compton, Simon Luo, Andrew Muñoz, Xinming Wu

We study point-source radiation in homogeneous attenuative anisotropic media using asymptotic methods. Formally, the analysis is similar to elastic media, except that many of the relevant quantities are complex in attenuative media. We will discuss the relevant steps in the seminar.

Steam members will present a brief explanation of his or her research topics and problems.

Team leader: Prof. Roel Snieder
Members: Farhad Bazargani, Michael Behm, Filippo Broggini, Johannes Douma, Pablo García López, Chinaemerem Kanu, Ali Knaak, Myoung Jae Kwon, Nori Nakata, Ernst Niederleithinger, Satyan Singh

Using two-way operators for image domain migration velocity analysis can lead into artifacts in complex models. Sharp boundaries in the velocity model introduces backscattered waves in the adjoint and state wavefields. The correlation between forward and backscattered waves from state and adjoint state wavefields produces coherent artifacts (cross-talk). We propose a filter step when correlating state and adjoint state variables which preserves the low-wavenumber components of the gradient and removes the cross-talk events. This filter allows the full use the two-way operator by taking advantage of reflected, backscattered and diving waves kinematic information.

In addition to briefly introducing Ikon Science to CWP faculty and students, Dr. Ehsan Naeini, senior research geoscientist at Ikon Science, will explain some of the key techniques used by Ikon Science in quantitative reservoir characterisation. During his presentation, Dr. Naeini will also highlight potential research and development topics, along with the latest developments in Ikon Science's R&D related to seismic processing.

Integrating well logs and seismic data has become an industry standard for interpretation workflows. Seismic data is recorded and commonly interpreted in vertical two-way time, and well logs, measured in depth, must be tied to seismic using a time-depth curve. However, well ties commonly contain a large amount of uncertainty due to unrealistic synthetic seismogram generation and manual correlations of synthetic seismograms to seismic traces. We improve well ties though by generating synthetic seismograms with realistic properties such as internal multiples, free surface multiples, and attenuation. Then, using dynamic warping, we automatically tie synthetic seismograms to recorded seismic, and we obtain updated time-depth curves and updated velocity curves. We also attempt to further improve well ties by using a 2D seismic image and multiple wells.

In attenuative media, ray-based modeling methods are efficient and easy to implement. In Kirchhoff modeling, we need to compute tables of Green's functions (source-diffractor + diffractor-receiver). This can be done using a host of "two-point" ray tracing methods. Another alternative is to perform a weighted summation of Gaussian beams - this is the approach that I use. I will show examples where I reconstruct Green's functions in unbounded homogeneous and heterogeneous media and scattered data from layered media.

Blind deconvolution refers to simultaneous estimation of the source wavelet and the transfer function (e.g. reflectivity series). I present a new approach for multichannel recordings to obtain crustal reflectivity below a local deployment from teleseismic body waves. Compared to conventional deconvolution, the blind deconvolution algorithm provides clearer results in the presence of white noise. Application to field data allow to recover a shallow basement reflector in ca. 4 km depth as well as the crust - mantle boundary (~40 km depth).

Apparent displacements between migrated images obtained from a baseline and monitor survey can be estimated using local correlations in the image domain. These displacements carry information about the changes in the model parameters in the subsurface due to reservoir production activities. I will show how we can invert this displacement and reconstruct the relative model error between the baseline and the monitor model.

A well-known difficulty in waveform inversion is the nonlinearity of the objective function, which can result in convergence to erroneous models given a poor starting model. This difficulty can be attributed in part to the fact that the waveform inversion residual, e.g., the difference between observed and simulated data, contains the error due to both the smooth background model which affects the kinematics of wave propagation, as well the rough reflectivity model which affects the amplitudes.

To address this issue, I propose and investigate two alternatives to the waveform inversion residual that can be used to better separate these sources of error. The first is a weighted sum of the waveform residual and a traveltime-like residual, and the second is a simple approximation to the waveform inversion residual. Tests on synthetic models demonstrate that these alternatives can improve the convergence properties of the inversion.