CWP Seminars  2015 Spring
CWP seminars discuss topics pertaining to our broad areas of research interests. These seminars are led by CWP faculty, students and, on occasion, by guest presenters. CWP seminars are held every Monday at 4 p.m. in the Green Center on the Colorado School of Mines campus. Click here to see previous CWP Seminars.
Note: To view weekly seminar schedules of individual CWP teams, click a link below:
CTeam seminars 
Steam seminars 
Spring 2015 CWP seminars
Date  Speaker(s)  Title  Abstract 
3/23  Datadomain tomography using localcorrelations 
Wavefield tomography in the data domain is usually formulated by using the data difference as the misfit criterion. The data difference however, is very susceptible to cycle skipping which leads to convergence into local minima. To overcome the strong nonlinearity, a multiscale approach to the inversion is often needed. The success of this approach relies on the lowfrequency content of the data. Here, we define the data domain tomographymisfit criterion using local (windowed) correlations. Less correlation crosstalk leads to cleaner adjoint sources, and hence, cleaner gradients. We further improve the gradients using a penalty function that is consistent with the bandwidth of the seismic data, which is more realistic than linear penalty functions that annihilate infinite bandwidth data. 

3/16  Moving faults while unfaulting 3D seismic images 
Unfaulting seismic images to correlate seismic reflectors across faults is helpful for seismic interpretation, especially for seismic horizon extraction. Most unfaulting methods assume the fault geometry does not change when unfaulting a seismic image. This assumption is not valid for images with multiple faults, especially intersecting faults, and often yields unnecessary distortions in the corresponding unfaulted images. I will discuss two methods to compute vector shifts that move fault blocks and faults themselves simultaneously to obtain an unfaulted image with minimal distortions. 

3/9  No seminar  CSM Spring Break  
3/2 
Nori Nakata 
Ambientnoise bodywave tomography at Long Beach, California 
In contrast to surfacewave extraction, bodywave extraction is much more difficult from ambientnoise fields because of smaller energy of body waves and strict stationary locations. To characterize elastic media even in the isotropic case, however, we need to determine both shear and compressional moduli. Therefore, I develop a signalprocessing technique to extract P diving waves from ambient seismic noise and use them for tomography to estimate 3D Pwave velocities. The technique is a postcorrelation process and contains a waveform matching filter and a noise suppression filter. This is the first body wave tomography result obtained from only ambient noise recorded at the ground surface. The estimated velocity structure has higher resolution than using surface waves. 
2/23  Anisotropy signature in RTM extended images for VTI media 
I will discuss the defocusing in RTM extended images due to inaccuracy in an anisotropy model. I will show the influence of the anellipticity parameter eta, and the Thomsen coefficient delta. I will discuss the sensitivity of differential semblance optimization (DSO) and stackpower objective functions to eta and delta. Using a model with a single homogeneous layer, I will show DSO sensitivity kernels and gradients for the anellipticity parameter eta.


2/16  No seminar  Presidents' Day holiday  
2/9  Gradient computation and regularization for elastic waveform tomography 
Elastic waveform tomography uses multicomponent waveform recordings and elastic wave equation modeling to reconstruct models of subsurface properties. Equations needed for elastic waveform tomography to reconstruct P and Swave velocity models are derived using the adjointstate method. Synthetic tests demonstrate that, for a given acquisition geometry, P and Swaves in general will update different areas of the subsurface. Moreover, certain areas might be updated by only one wave mode, and, as a result, the velocity model corresponding to the missing wave mode will not be well constrained. To better constrain the inversion, we introduce a regularization term that penalizes unrealistic relations between Vp and Vs. 

Anisotropic waveform tomography with pseudoacoustic equations 
Multiparameter seismic waveform inversion especially with anisotropic parameters is intriguing but challenging due to the computational cost and the huge nullspace introduced by the imperfect model parameterizations. I will briefly discuss the theory and the implementation of the gradients using the wellknown adjointstate method. Some computational cost issues associated with the modeling and adjoint modeling will be addressed. I will also illustrate the sensitivity kernels of different parameters using some synthetic models. If the time allows, I will introduce one possible set of inequality constraints to the inversion which can also be treated as the regularization term. 

2/2  No seminar  CSM Department of Geophysics presentation  
1/26  Walking in the footsteps of Michael Berry 
I will give an overview of research that I did during my sabbatical. This includes superresolution without evanescent waves, seismic waves as a Foucault pendulum (or not), interferometry in caustics, and timelapse changes. 

1/19  Internships with our sponsors and hackathon fun 
BHP Billiton Petroleum Geoscience Internship ExxonMobil Exploration Company Geophysics Internship SEG 2014 Geophysics Hackathon


1/12  1st CWP seminar of the Spring 2015 semester (CWP administrative topics) 
Previous CWP Seminars
2014 

2013 

2012 