Integration of Noise and Coda Correlation Data into Kinematic and Waveform Inversions With Microearthquake Data for 3D Velocity Structure, Earthquake Locations, and Moment Tensors in Geothermal Reservoirs Geothermal Project
Last modified on July 22, 2011.
|Project Title||Integration of Noise and Coda Correlation Data into Kinematic and Waveform Inversions With Microearthquake Data for 3D Velocity Structure, Earthquake Locations, and Moment Tensors in Geothermal Reservoirs|
|Project Type / Topic 1||Recovery Act: Enhanced Geothermal Systems Component Research and Development/Analysis|
|Project Type / Topic 2||Induced Seismicity|
|Project Description|| This project will focus on using microearthquakes (MEQ) and noise correlation Green’s functions (NCF) obtained from MEQs and ambient noise to image the physical properties of geothermal reservoirs.
MEQ and ambient noise data recorded by seismographic networks at the Coso geothermal field and the Paradox Valley injection well would be used in the inversions to demonstrate the benefits of processing and using the NCF data and the implementation of full waveform inversion in terms of improved resolution of 3D velocity structure and MEQ locations and focal mechanisms. In these regions of the western U.S., waveforms contain deterministic information about 3D velocity structure and focal mechanisms to maximum frequencies of 3-10 Hz. Currently, it is only feasible to use large-scale supercomputing facilities to calculate synthetic waveform Green’s functions to such high frequencies over the scales of typical geothermal reservoirs. Extending the maximum frequency of synthetic Green’s function calculations from 1-2 Hz in full waveform inversions for 3D velocity structure to 3-6 Hz requires an increase in computation time of 4096. Graphic processing units (GPU) can now provide teraflop computational capabilities in single inexpensive graphics cards. Co-Pi Liu has developed an MPI parallel implementation of 3D finite difference viscoelastic RGF calculations that reduces memory requirements by using variable mesh separates into two regions, with finer grid spacing employed near the surface where velocities are lowest.
The Paradox Valley deep injection well began continuously monitored several years prior to injection and provides an excellent opportunity to understand relationships between fluid injection, MEQ occurrence, and fluid pressure propagation and flow. The Coso data provides a test case to evaluate thermal and production effects in the Abaqus modeling. Drs. Roeloffs and Denlinger will use Abaqus to develop testable physical models of the causal mechanisms of induced seismicity at Paradox and Coso.
|Objectives||Simultaneously invert body- and surface-wave arrival and travel-time data and body-wave cross-correlation time-difference data to image three-dimensional (3D) variations of physical properties of geothermal reservoirs, particularly 3D variations of P- and S-wave velocities. It will be a joint collaboration between co-PI Liu of Reclamation and Prof. Feng of Virginia Tech to optimize the RGF and objective functional calculations for GPUs to provide a desktop terra-computing platform for conducting full waveform inversions of MEQ and NCF data for 3D P- and S-wave velocity structure.|
|Awardees (Company / Institution)||William Lettis & Associates, Inc.|
|Partner 1||Virginia Tech|
|Partner 2||Bureau of Reclamation|
|Partner 3||United States Geological Survey|
|Funding Opportunity Announcement||DE-FOA-0000075|
|DOE Funding Level (total award amount)||$708,000.00|
|Awardee Cost Share||$194,852.00|
|Total Project Cost||$902,852.00|
|Principal Investigator(s)||Dr. Daniel R.H. O’Connell, William Lettis & Associates, Inc.|
|Other Principal Investigators|| Prof. Wu-chun Feng, Department of Computer Science at Virginia Tech; Drs. Lisa Block and Pengcheng Liu, Bureau of Reclamation; and Drs. Roeloffs and
Denlinger, the USGS.
|Targets / Milestones|| MEQ arrival time data poorly constrain velocity structure at depths shallower than the shallowest concentrations of MEQs because body-wave ray paths are steep and confined to the vicinity of seismographic stations. Short-period surface waves and P- and S-wave travel times can be extracted from NCFs from ambient noise and MEQ coda recordings from all station pairs in a seismographic network to provide improved constraints on shallow P- and S-wave velocity structure within geothermal reservoirs. The inversion would proceed in two stages. The first inversions would use kinematic body- and surface-wave data. The second inversions would employ full waveform inversion of body- and surface wave data using 3D finite-difference viscoelastic reciprocity Green’s functions (RGF) with a quasi-Newton trust region algorithm, both developed by co-PI Liu.
Project results would include: 1) improved resolution of 3D variations of reservoir properties, 2) better understanding of MEQ evolution and it relation to fluid properties and evolution, 3) development of full-waveform inversion tools to image 3D crustal velocity structure that run efficiently on low-cost computer platforms to assist in initial exploration and project development and optimization, and 4) an Abaqus modeling approach.
|Location of Project||Blacksburg, VA, Denver, CO|
|Impacts||If successful, the project would provide improved resolution of 3D variations of reservoir properties, better understanding of microearthquake evolution, and development of full waveform inversion tools to image 3D crustal velocity structure.|
|Funding Source||American Recovery and Reinvestment Act of 2009|
|References||EERE Geothermal Technologies Programs|