- Seismic Interferometry
- RESULTS:
- The book "Seismic Interferometry" is completed published by Cambridge Press (http://www.cambridge.org/catalogue/catalogue.asp?isbn=9780521871242).
- A new interferometric method is devised to extrapolate marine SSP and OBS receivers to be outside the recording array. Results with 2D and 3D data suggest some of the following possibilities: wide-angle data can be collected from narrow recording arrays, the recording aperture can be more than doubled in conventional OBS surveys, and we can possibly overcome problems in shooting seismic in towns where vibrators are forbidden. There are still some problems in the matching filter for extrapolation.
- Least squares redatuming. Yanwei Xue presents his successful results for least squares datuming (LSD) of synthetic data. The LSD results are noticeably more accurate than the conventional datuming results. He also applied this technique to the VSP->SSP transform and the results show noticeable improvement.
- Interferometric surface wave elimination using curvelet filters. Curvelet filters provide a more efficient means for eliminating predicted surface waves than our adaptive-subtraction method.
- Source estimation by correlation migration. Correlation migration algorithm is developed to locate the source location from crosscorrelated passive seismic data. A formula is developed for estimating the vertical resolution of the migration image. This formula explains the degradation of vertical resolution of the correlation migration scheme, and suggests that a good resolution at depth requires a large recording aperture.

- IMPLICATIONS: Seismic Interferometry book is first of its kind and has MATLAB codes to improve the reader's understanding and also test ideas. Interferometric extrapolation opens up the door to increasing aperture of short receiver arrays. Perhaps redesign of OBS surveys is in order. Phase-encoded modeling is possible means to perform multisource FD modeling and still get separate Green's functions efficiently. Opens up new doors for target-oriented RTM and waveform inversion.
- PROBLEMS: Need better matching filter approach for extrapolation.
- 2009 GOALS: 3D field data tests for interferometric extrapolation and interpolation; develop better image matching filter for extrapolation.

- RESULTS:
- Modeling
- RESULTS:
- Phase-encoded modeling. Theory of phase-encoded modeling to efficiently generate Greens functions for migration and inversion. Simulations suggest that this method can generate early arrival waveforms for any desired Green's function using just two finite-difference simulations with long duration. New trick for speeding up simulations is developed.
- Forward modeling with a $1500. GPU processor show a 20x speedup compared to a single INTEL processor.
- Mapping signal processing problem from medicine to geophysics for Hidden Markov Models.

- IMPLICATION: Order of magnitude speedup by GPUs and Phase encoded modeling.
- PROBLEMS: Need further development, both efforts are in initial stages.
- 2009 GOALS: Library of GPU modeling and migration codes. Possible GPU based waveform inversion codes. Full speed development of phase-encoded modeling, migration, and inversion, and further development of Hidden Markov Models.

- RESULTS:
- Migration
- RESULTS:
- 3D least squares migration (LSM) and iterative migration deconvolution (IMD) of synthetic and field data, with deblurring filters. The convergence rate of these iterative methods is accelerated by factors between 2 and 5 if deblurring filter is used.
- 3D tests of Bottom-Up reverse time migration (BUM) successfully tested on 3D field data.
- LSM successfully applied to synthetic VSP data.

- IMPLICATION: BUM is the way to go for target oriented RTM. Deblurring filters are cost-effective preconditioners and can be used to improve migration image quality.
- PROBLEMS: Need to test deblurring on 3D field data.
- 2009 GOALS: Deblurring and iterative MD and LSM tested on field data. Can we use deblurring to speed up waveform inversion.

- RESULTS:
- Tomography
- RESULT:
- Several results were achieved: a). Multi-scale waveform tomography successfully tested on marine data. b). Test of multi-scale tomography shows accurate results when applied to test model supplied by a sponsor. Initial results were not accurate because the parameters for the topography were not accurate. c). 2D migration and forward modeling with a $1500. GPU processor show a 20x speedup compared to a single INTEL processor. d). Accounting for attenuation effects in Friendswood crosswell field data shows a slightly improved tomogram compared to standard crosswell tomography. e). Theory for phase-encoded wave equation inversion. Potential for significant speedup.

- IMPLICATIONS: Multi-scale tomography appears to be the right way to go in x-t waveform inversion. GPU computing may give order of magnitude speedup in waveform inversion.
- PROBLEMS: Problems in getting correct topography parameters fixed. WI still requires many iterations. We have standard algorithm for estimating attenuation.
- 2009 GOALS: Multi-scale tomography on land data. Possibly start work on elastic tomography. Can deblurring filters act as effective preconditioners? Can phase-encoded WI increase computational efficiency of WI?

- RESULT:
- Imaging Passive Seismic Data
- RESULTS:
- Several results are reported: a). Passive hammer drill data are deconvolved to give estimate of partial Green's function for a point source at depth. This is an inexpensive VSP data set. b). Resolution formulas for estimating source location using crosscorrelation migration.

- IMPLICATION: New understanding of capabilities and limitations of crosscorrelation migration. Demonstration of some possibilities of turning noise into signal.
- PROBLEMS: Insufficient data. Not enough students to work in this area. Xiang Xiao worked with Aramco in Dhahran for 5 months but results are proprietary.
- 2009 GOALS: Full tests on extensive Aramco passive seismic data recently given to us. Wyoming "Signal form Noise" field test in Wyoming scheduled for February, 2009.

- RESULTS:
- Multisource Data
- RESULTS:
- Local adaptive filter used to separate multisource data, with acceptable results for data that is a blend of 5 shot gathers.

- IMPLICATION: Potential for acquiring multisource data if this method can be further improved.
- PROBLEMS: Method fails when data is a blend of 10 shot gathers.
- 2009 GOALS: Full tests on 3D data. Improve local adaptive filter.

- RESULTS:
- Summary of Salient Goals
- Library of GPU modeling and migration codes to be given to 2009 sponsors.
- Full tests on range of topics for phase encoded modeling, migration, and imaging.
- Full development of Hidden Markov Models and applications to geophysics.
- Further improvements on matching filter (or another approach) for improving interferometric extrapolation. 3D data tests.
- Tests of crosscorrelation source imaging on 3D Aramco data.
- Wyoming field test for getting signal from noise.
- Strong effort on waveform inversion (using Q adjustments) of land data. Perhaps elastic inversion tests.
- 3D preconditioned LSM applied to field data. LSM applied to VSP field data.
- Finish up LSD applied to VSP and SSP data.