Atlas home || Conferences | Abstracts | about Atlas


Inverse Problems and Quantification of Uncertainty

IMA Workshop

April 22, 2002

Minneapolis, MN, USA

Mathematics

Host: Institute for Mathematics and its Applications
Homepage: http://www.ima.umn.edu/geoscience/spring/g9.html
Email: staff@ima.umn.edu

Organizers: William W.Symes, Philip Stark, John Scales, C. Deutsch

Description:
All indirect inference of parameters and system states in the Earth sciences is subject to uncertainty. The model description of the physical processes underlying the inference is necessarily oversimplified, data are measured with limited accuracy, and simulation methods are necessarily approximate. Each of these causes contributes to the overall uncertainty of Earth property estimates. Model parameterization affects the level of ambiguity, and trades off in many problems with explanatory power. The complexity of the Earth system imposes other limitations: many features of the subsurface have an aggregate effect on the data, and estimation of these subresolution aspects of models is subject to great ambiguity.

Many approaches have been proposed to quantify this uncertainty, including linear sensitivity analysis, Bayesian PDF estimation, minimax, construction of solutions that are extremal in some sense, and many others. Some of these techniques are even used. This workshop will explore the capabilities and shortcomings of current methodology in a wide variety of contexts. The juxtaposition of many different applications in which roughly the same uncertainty questions arise is meant to provoke much needed progress towards better understanding of the information content of geophysical data.

Keywords: tomography---whole Earth, crustal, crosswell, ocean acoustic; imaging/inversion--crustal, exploration/exploitation, near surface multispectral imaging, remote sensing gravity, magnetics, electromagnetics; helioseismology; earthquake source mechanisms

Date received: January 31, 2001, revised September 09, 2002


© 2008 Atlas Conferences Inc.