Three-dimensional voxel based magnetic inversion is a well-established and powerful tool for geophysical interpretation. The codes to run these inversions have been available for over twenty years, however for much of that time their use was limited to academic researchers and industry experts. Recent innovations in cloud-based computing have brought these codes to the masses, but as the cliched saying goes "With great power comes great responsibility"; or more directly "Just because your inversion converged doesn’t mean you should believe the result." The non-uniqueness of geophysical inversion is a well-known fact but before we accept a given result as one of the infinite set of possible models it is important to make sure that the code has behaved as expected. While many of the methods used to achieve this goal are well known within the geophysical inversion community, they are not always easily accessible to new users. This paper endeavors to fill in some of the gaps by formulating a two-step process focusing on the input to and output from the inversion. The first step ensures that the forward modelling process is as accurate as possible. This is accomplished by pre-processing the data based on the survey specifications and the voxel model parameters. The second step analyses the output from the code in order to verify that the inversion process has converged in the manner we expect. This is a qualitative process that involves reviewing the convergence curves, the predicted data and the recovered model. The guidelines are illustrated with examples using synthetic and field data. While this process will not remove non-uniqueness from 3D magnetic inversion, it should help ensure that the suite of models being compared are the best possible candidates for interpretation.

Presentation Date: Wednesday, October 17, 2018

Start Time: 1:50:00 PM

Location: 213B (Anaheim Convention Center)

Presentation Type: Oral

This content is only available via PDF.
You can access this article if you purchase or spend a download.