SIO Research Candidate Seminar - Evan Um


 
08/29/2017 - 4:00pm

SCRIPPS INSTITUTION OF OCEANOGRAPHY EARTH SECTION RESEARCH CANDIDATE SEMINAR

DATE:          August 29th, Tuesday, 4pm  


LOCATION:     Munk Conference Room
 
SPEAKER:      Evan Um, Ph.D.
            Lawrence Berkeley National Lab
            
TITLE:          Finite Element Modeling and Inversion for Controlled Source Electromagnetic Geophysics
 
ABSTRACT:    
 
Finite element (FE) modeling methods with unstructured meshes increasingly become an integral part of electromagnetic (EM) geophysics because they can accurately and efficiently simulate EM responses to complex multi-scale earth models. For instance, using geometry-conforming tetrahedral meshes, 3D FE methods precisely discretize complex seafloor topography for marine EM modeling. Local refinements of tetrahedral meshes enable us to accurately and efficiently model subtle but important features in a large-scale model (e.g. narrow fault zones in a tectonic-scale earth model, hydraulically-fractured zones in a reservoir-scale subsurface model and hollow steel-cased wells in oil/gas, geothermal and scientific drilling sites)
 
However, the advantages of the FE methods come with extra complications. It is challenging to rapidly generate reliable and efficient tetrahedral meshes for 3D FE modeling. This is a major hurdle to use 3D FE modeling for practical EM inversion. It is also considered difficult to iteratively solve a system of FE equations. The FE system matrix is unstructured and not diagonally dominant. As a result, a simple Jacobi preconditioner used in finite difference solutions does not ensure convergence of FE Krylov solutions. Accordingly, direct solvers are often the method of choice for FE solutions despite their large memory requirements and low parallel scalability. In this talk, I present a simple and effective method for automatically generating 3D tetrahedral meshes for marine controlled-source EM. I also discuss efficient numerical solution strategies for 3D FE problems. These 3D meshing and numerical solution strategies are also effective for onshore geophysical monitoring problems. I introduce new experimental EM methods for monitoring CO2 storage and hydraulic fractures, and discuss their potential use for marine electromagnetic research.
 
 
Faculty Host:  Michael Hedlin (mhedlin@ucsd.edu)
For more information on this event, contact: 
lcosti@ucsd.edu
Event Calendar: 
Location: 
Munk Conference Room