GP005-06
Open-Source Landscape for Three-Dimensional Controlled-Source Electromagnetic Modeling

Monday, 14 December 2020: 10:37
Virtual
Dieter Werthmuller, Delft University of Technology, Delft, Netherlands, Raphael Rochlitz, Leibniz Institute for Applied Geophysics, Geoelectrics and Electromagnetics, Hannover, Germany, Octavio Castillo Reyes, Barcelona Supercomputing Center, Barcelona, Spain and Lindsey Justine Heagy, University of California Berkeley, Department of Statistics, Berkeley, CA, United States
Abstract:
The open-source landscape of three-dimensional modeling for controlled source electromagnetic (CSEM) data has changed dramatically within the last five years. While modeling large-scale CSEM data was previously the privilege of large exploration companies or academic consortia, nowadays it is possible for everyone. This is not only thanks to open-source codes, but also to increased computing power which makes it affordable for anyone to run 3D simulations.

We combined forces between four open-source CSEM codes to showcase the capabilities of open-source CSEM modelers today. The four codes under consideration are custEM, PETGEM, SimPEG, and emg3d. The former two are finite-element (FE) codes using unstructured tetrahedral meshes, and the latter two are finite-volume (FV) codes using structured octree or tensor meshes. Besides layered and block models we also compare them for the realistic, open-source resistivity model Marlim R3D and its accompanying CSEM data.

Comparing results of 3D CSEM computations are a crucial aspect of code verification. The common way to verify codes is to compare results to analytical or semi-analytical solutions, which only exist for very simplistic models. The only method to verify complicated, realistic models is to compare the results of different codes. Our comparisons show that all four codes yield the same result within acceptable errors, independent of the chosen finite element method (FE or FV) and of the applied discretization.

Our comparison of four codes not only verifies all four codes with each other, but further, this project has motivated improvements in each of the code-bases and demonstrates how collaboration between (open-source) projects is beneficial and advances all participants. We hope that our efforts will help to ensure the reliability of CSEM simulations, not only of the presented open-source codes, but also for other open-source or proprietary projects.