Three-dimensional models for MT code testing: Preliminary results and comparison

Category Geophysic
Group GSI.IR
Location International Geological Congress,oslo 2008
Author Miensopust, Marion P.۱; Queralt, Pilar۲; Jones, Alan G.
Holding Date 08 September 2008

In the recent years the importance of three-dimensional modeling of magnetotelluric (MT) data increased and various 3D forward and inversion codes became available. These codes use different formulations of the problem (finite differences, finite elements or integral equations), various orientations of the coordinate system and either the plus or the minus sign convention for the time dependence. Nevertheless the resulting impedance values for all these codes should be the same once they are converted to a common sign convention and coordinate system.
To compare the results of the various codes we designed two different models with three-dimensional resistivity structures embedded in a homogeneous subsurface. One fundamental model requirement was to generate impedances which have significant values in the diagonal elements which are not negligible. Unlike the one-dimensional and two-dimensional modeling of magnetotelluric data for the three-dimensional case these diagonal impedance elements carry information about the resistivity structure.
One of the ground models was used to compare the various forward algorithms which are the base for the different inversion codes. This comparison was followed by an inversion of the retrieved data set - inversion of a known structure. We calculated the responses ourselves but also sent this exercise to the programmers of the codes and other users within the context of a workshop on 3D inversion of magnetotelluric data. The second (secret) model was used to generate the forward responses, which were send to the same persons for inversion - inversion of an unknown structure.
In this paper will present the two used resistivity models as well as the preliminary results and the comparison for these three parts of code testing - first forward modeling, second inversion of a known structure and finally inversion of an unknown structure.