IMAGE
Consortium
Inversion and Modelling of Applied Geophysical Electromagnetic data
General description and sponsors
The 3-year IMAGE project begins Sept. 1,
1999. The research is sponsored by NSERC
through its Collaborative Research and Development (CRD) program,
and the following companies:
-
Newmont
-
Rio Tinto
-
Falconbridge
-
Placer Dome
-
Anglo American
-
INCO
-
MIM
-
Cominco
-
AGIP
-
Muskox Minerals
-
Billeton
This newly founded consortium
involves a major new research collaboration between the
Geophysical Inversion Facility
(UBC-GIF)
and
the
Scientific Computation and Visualization Group
(SCV)
at the University of British Columbia.
The primary focus is on forward modelling and
inversion of 3D (three-dimensional)
geophysical electromagnetic data.
There are a host of computational and geophysical challenges
associated with solving the resulting mathematical models.
These include numerical methods for inverse problems, fast
simulation methods for the
3D electromagnetic equations, handling sources and
boundary conditions on infinite domains, and parallel computing.
People
- Professors
- Research Manager
- Programmer
- Outreach coordinator
- Research Associates
- Visiting Research Scientists
- Students
- Affiliates
- Yaoguo Li (Now at Colorado School of Mines)
- Jim Varah (Professor of Computer Scienceat UBC)
Publications
The inverse problem
-
E. Haber and U. Ascher (Jan., 2001):
Preconditioned all-at-once methods for large,
sparse parameter estimation problems.
-
U. Ascher and E. Haber (Oct., 2000):
Grid refinement and scaling for distributed parameter
estimation problems.
-
E. Haber, U. Ascher and D. Oldenburg,
Inverse Problems, 16 (2000), 1263-1280:
On optimization techniques for solving nonlinear inverse problems.
-
E. Haber and D. Oldenburg,
Computational Geoscience, 5 (2000), 41-63:
A GCV based method for nonlinear ill-posed problems.
Financial support
We can provide financial support and are looking for students and postdoctoral
fellows
for this project. Contact
ascher@cs.ubc.ca
or doug@geop.ubc.ca
if interested.
Goals for a 3-year research program
-
Develop fast and accurate 3D (three-dimensional) forward
modelling of EM data using
vector and scalar potentials.
-
Use new techniques in non-linear optimization to solve the
3D inverse problem by working
with the differential equations rather than the integral equations.
-
Concentrate upon three generic data sets for frequency domain data:
-
MT (Magnetotellurics)
-
CSAMT (Controlled source audio frequency magnetotellurics)
-
Surface and borehole data from a loop source.
-
Develop a 3D forward modelling code for time domain (TEM) data.
-
Attempt the inversion of 3D TEM data.
Research Priorities
Research priorities are determined partly by considering
what are important problems, and partly
by what data sets are available to the research group.
At the consortium meeting on May 9, 1999,
we identified the following:
-
Priority I: 3D Forward modelling and inversion of CSAMT data.
-
Priority II: 3D Forward modelling and inversion of surface
loop source with surface and
borehole data.
-
Priority III: 3D Forward modelling and inversion of magnetotelluric data.
-
Priority IV: 3D Forward modelling of time domain data.
IMAGE Research Proposal Summary
We propose to develop methodologies and practical computer codes to invert geophysical
electromagnetic observations to recover
a three-dimensional (3D) distribution of electrical
conductivity.
A new electromagnetic forward-modelling algorithm will
solve for the low-frequency electric and
magnetic fields over an arbitrary 3D conductivity structure.
Vector and scalar potentials will be
used rather than solving for the fields directly.
Preliminary work shows this to have many
advantages over existing approaches both in computational
efficiency and robustness. The inverse
problem will be solved by combining traditional non-linear
optimization techniques with new
techniques based on the differential equations themselves
rather than the usual integral equations.
By coupling the inverse and forward modelling directly,
we plan to develop fast, flexible and robust
algorithms.
Although our inverse procedures will be applicable to all
low-frequency electromagnetic data, we will
be concentrating upon three data types that are both generic and frequently used:
frequency-domain magnetotelluric (MT) data,
controlled-source audio-frequency magnetotelluric
(CSAMT) data, and both surface and drill-hole data from a loop source.
We will also develop software
for forward modelling time-domain electromagnetic (TEM) data. The sponsors
will supply the data on which the inversion codes will be tested,
and the codes will be made
available to the sponsors at the conclusion of the three-year period.