Fast Fourier transform of electromagnetic data for computationally expensive kernels

Werthmüller, D., W.A. Mulder, and E.C. Slob, 2021, Fast Fourier transform of electromagnetic data for computationally expensive kernels, Geophysical Journal International, Volume 226, Issue 2, Pages 1336-1347; DOI: 10.1093/gji/ggab171.

Manuscript for Geophysical Journal International paper

Directory contains the LaTeX source of the manuscript as well as the notebooks to reproduce all the figures.

History

1. 2020-06-30: Submitted to Geophysical Journal International.
2. 2020-12-01: Submitted revision I.
3. 2021-04-07: Submitted revision II.
4. 2021-04-21: Manuscript accepted

Requirements and Environment

Required are Python version 3.7 or higher and the modules emg3d, discretize, and SimPEG (for the 3D model comparison). To reproduce the figures with the provided notebooks the modules Jupyter, and matplotlib are required additionally.

We provide an environment.yml to ensure everything can be reproduced.

To create the environment simply run

conda env create -f environment.yml

This will create a new conda environment called article-tdem.

To activate and deactivate the environment run

conda activate article-tdem
conda deactivate

To use this environment in the Jupyter notebook, you have to register it first:

python -m ipykernel install --user --name article-tdem

Then, in Jupyter, you can select it by going to Kernel->Change kernel and select article-tdem.

To completely remove the environment run

conda remove --name article-tdem --all

emg3d requires at least Python 3.7.

Reproducing

Note that the versions in the environment-yaml are fixed to ensure the notebooks will still run in a couple of years. Particularly emg3d<1.0 and discretize<1.0, as both packages have non-backwards compatible changes with their 1.0 release. For up-to-date examples on modelling time-domain data with emg3d consult the documentation.

Each notebook has a table at the end that lists which versions were actually used at the time the results in the article were produced.

Start Jupyter and navigate to the notebooks-folder. The notebooks to reproduce the figures in the article are:

• 01-02-03_Figures_Frequency-Selection.ipynb: Figures 1, 2, and 3.
• 04-05_Figures_Fullspace.ipynb: Figures 4 and 5.
• 06-07_Figures_1D-Model.ipynb: Figures 6 and 7.
• 08_Figures_1D-Model-wrong-x-y.ipynb: Figure 8.
• 09_Figures_Cole-Cole-IP.ipynb: Figure 9.
• 10-11_Figures_3D-Model.ipynb: Figures 10 and 11.

Note

The final edited version is © Geophysical Journal International.