Measuring and Interpreting Seismic Attenuation
The project goals are to:
- Perform, for the first time, laboratory measurements of the seismic velocities and attenuation across the entire seismic band of frequencies on rocks that have been well characterized (work to be performed by CSM).
- Provide, for the first time, detailed explanations and models (both analytical and numerical) for the lab measurements that can be used to relate the material properties of interest to the oil industry to the seismic measures (work to be performed by Lawrence Berkeley National Laboratory, or LBNL).
- Invert field data to obtain frequency-dependent attenuation and dispersion properties and relate such field measurements to the underlying rock properties of interest (work to be performed by the University of California at Berkeley and LBNL).
Lawrence Berkeley National Laboratory, Berkeley, CA
Colorado School of Mines, Golden, CO
University of California at Berkeley, Berkeley, CA
Chevron Corp., San Ramon CA
Heterogeneity of the elastic properties of rocks at the scale of millimeter-to-centimeter (so-called “mesoscopic scales”) is responsible for a highly variable fluid pressure response across these same scales when the rock is squeezed by a seismic wave. The subsequent fluid flow due to such wave-induced fluid pressure gradients provides significant amounts of seismic attenuation.
The project has resulted in the development of highly efficient poroelastic finite-difference modeling codes that allow the mesoscopic-scale fluid flow induced inside of a rock by a passing wave to be numerically determined. Researchers then calculate the attenuation (1/Q) and seismic velocity for a given synthetic rock sample. This has led to significant insight both into the mechanism responsible for seismic attenuation and into what new information measuring attenuation can bring.
Among many results so far obtained, the project performers have demonstrated that when the mesoscopic heterogeneity is distributed as a self-affine fractal having a Hurst exponent b, the attenuation as a function of frequency f follows the power law Q(f) = f b.
The most important question this project asks is: What is the benefit of obtaining seismic Q from seismic data? To answer this question, one must have a physical model that adequately explains the data. Project results after 10 months of modeling seem to indicate that wave-induced fluid flow over mesoscopic-length scales is the mechanism that shows the most promise. This model will allow researchers to answer the question of what information is in Q. A key step will be next year, when the model predictions are compared directly with data measurements occurring at CSM. At just 10 months into the project, the project performers do not want to answer the posed question until more research has been performed.
This project is a collaboration among CSM (laboratory experiments), LBNL (theoretical and numerical model development), Cal-Berkeley (seismic forward modeling), and Chevron (providing field data for Q analysis).
Approaching the end of this first year of effort, the LBNL theoretical and numerical modeling has advanced as it should have.
Current Status (July 2007)
The project is in the second of a 3-year funding effort with the subcontract job for CSM.
This project is a subcontract to DOE project number DE-FC26-04NT15505. The project is not funded in FY 07
Project Start: December 14, 2004
Project End: December, 2006
Anticipated DOE Contribution: $206,000
Performer Contribution: $0
NETL - Purna Halder (email@example.com or 918-699-2083)
LBNL - Steven Pride (firstname.lastname@example.org or 510-495-2823)
CSM - Michael L. Batzle (email@example.com or 303-384-2067)
Three journal articles based on these numerical experiments are in the final stages of preparation.
Compressional attenuation and dispersion as numerically determined using LBNL's finite-difference solutions to the poroelastic governing equations (symbols) for the various geometries are shown in the top panel. Also shown (solid lines of second panel) are the analytical results of Pride et al. (2004), obtained without fit parameters to the same geometries.
Corresponding to the more-complicated geometries having multiple length-scales present, the Pride et al. (2004) predictions for each individual-length scale are simply averaged to produce the solid curve labeled "mean." It is seen that such a simple average does an adequate job explaining the numerical data.