This proposal initiates a new collaboration aimed at improving methods of waveform inversion by including topography in seismic wave modeling and reducing the reliance of these methods on data at extremely low-frequency. The Principal Investigators propose to adapt and improve methods developed for simulating scattering from a diffraction grating. Their approach has extensions to linear elasticity and provides general improvements for solving the inverse problem. Significant mathematical advances are required to develop robust and efficient techniques for this application. The Boundary Perturbation Method extends the Field Expansion approach to an arbitrary number of layers with independent topographies and allows for rapid and accurate simulation of acoustic wave propagation in two dimensions. Extensions to three dimensions and to the general equations of elasticity are required. Moreover, the extension of frequency-independent discretizations based on Geometric Acoustics to multilayered elastic models is a significant and necessary mathematical advance. Additionally, advances in the inverse problem are also necessary to make the technique truly applicable to the seismic imaging problem. The standard method of "full-waveform inversion" requires data at frequencies which are too low to record in practice. The PIs' approach casts the forward problem as the application of a sequence of topography-dependent operators where the interface shapes appear rather explicitly. A number of iteration schemes are proposed for the recovery of these shapes, using re-arrangements of the compositions coupled to standard regularizing techniques from the theory of ill-posed problems.

The propagation properties of seismic waves in layers of sediment are crucial in many technologies including the determination of inner earth properties and structure, earthquake detection and prediction, and hydrocarbon (oil and gas) exploration. In light of its many important applications, it is not surprising that a vast array of numerical and experimental techniques have been brought to bear upon this problem. However, several gaps in understanding and capability still exist. The PIs' address some of these questions through sophisticated numerical simulations which will be validated against both laboratory experiments and field measurements from the Tibetan plateau.

Project Report

This project provides a part of on-going research project into howbest to understand the Earth's subsurface using surface seismic data.When forming an image of the subsurface, we typically compute modeleddata for a model based on an educated guess of what the subsurfacemight look like. We then compare these modeled data to the recordedfield data and see how well our model explains the data. We generallysay that the modeled and recordedd data agree if they are the samewithin some measure of the noise in the data. There are often severalmodels that fit the data equally well, however, and deciding which isbest is a challenging task that involves quite a bit of expertise onthe part of the person interpreting the images. A bettercharacterization of this range of models would both make the job ofthe expert interpreter easier and open up the potential of automaticinterpretation. The particular focus of this project has been on designing methods ofmodeling seismic wave propagation that can be computed quickly usingsimplified models. The advantage of using simplified models is thatit becomes much easier to compute and display things like the range ofpossible models that fit the data. The advantage of models that canbe computed quickly is that it allows us to compare our field datawith modeled data from a larger number of models. This sort of work is applicable to the petroleum industry, who useseismic images both to attempt to locate hydrocarbons and to monitortheir extraction, as well as to the global seismological community whouse seismic images to understand Earth processes like subduction (keyin earthquake studies), volcanos, and water reservoirs. This project has contributed to the training of 6 PhD students, and 1MSc student. Three of these students have now graduated; two areemployed in industry jobs and one in academia.

Agency
National Science Foundation (NSF)
Institute
Division of Mathematical Sciences (DMS)
Application #
1115406
Program Officer
Junping Wang
Project Start
Project End
Budget Start
2011-09-01
Budget End
2014-08-31
Support Year
Fiscal Year
2011
Total Cost
$149,983
Indirect Cost
Name
Massachusetts Institute of Technology
Department
Type
DUNS #
City
Cambridge
State
MA
Country
United States
Zip Code
02139