Friday, July 26

8:30-10:30 AM
## MS45

Resolution for Seismic Inverse Problems: Current Approximations and Uses

Inversion requires mathematicians and geophysicists to estimate model parameters, (for example the velocity of waves traveling through a subsection of the earth) from recorded boundary data. Seismic inverse problems are typically ill-posed and inversion provides model estimates which are only unique in an average sense. Model resolution quantifies the uncertainty in a particular model solution. If G is the forward model relating Model parameters to data, then one defines the model resolution matrix by VVt where UVt is the singular value decomposition of G. Unfortunately, this SVD computation is expensive for realistic (large) inverse problems in seismology. The speakers will describe some of the current approximations to and uses for model resolution.
**Organizer: Susan E. Minkoff **

University of Texas, Austin

**8:30 Computing Tomographic Resolution Matrices Using Arnoldi's Iterative Inversion Algorithm**
- James G. Berryman, Lawrence Livermore Laboratory
**9:00 Resolution of Reflectors in Crosswell Travel Time Tomography**
- Kenneth P. Bube, University of Washington; and Robert T. Langan, Chevron Petroleum Technology Company
**9:30 Model Resolution and Uncertainty in Seismic Inverse Problems: Computing Confidence Intervals with a Reproducing Kernel Approach**
- Jay Pulliam, University of Texas, Austin
**10:00 A Comparison of Seismic Parameter Representations**
- R. Phillip Bording, University of Texas, Austin
**10:30 Information and Uncertainty in Geophysical Inverse Problems**
- Wences Gouveia, Fernando Moraes, and John A. Scales, Colorado School of Mines

*MMD, 5/20/96*