DL-RMD: A geophysically constrained electromagnetic resistivity model database for deep learning applications
- 1Hydro-Geophysics Group (HGG), Department of Geoscience, Aarhus University, Aarhus C, 8000, Denmark
- 2Aarhus GeoInstruments, Åbyhøj, 8230, Denmark
- 3Department of Electrical and Computer Engineering, Aarhus University, Aarhus N, 8200, Denmark
- 4Aarhus University Centre for Water Technology (WATEC), Aarhus N, 8200, Denmark
- 1Hydro-Geophysics Group (HGG), Department of Geoscience, Aarhus University, Aarhus C, 8000, Denmark
- 2Aarhus GeoInstruments, Åbyhøj, 8230, Denmark
- 3Department of Electrical and Computer Engineering, Aarhus University, Aarhus N, 8200, Denmark
- 4Aarhus University Centre for Water Technology (WATEC), Aarhus N, 8200, Denmark
Abstract. Deep learning algorithms have shown incredible potential in many applications. The success of these data-hungry methods is largely associated with the availability of large-scale data sets, as millions of observations are often required to achieve acceptable performance levels. Recently, there has been an increased interest in applying deep learning methods to geophysical applications where electromagnetic methods are used to map the subsurface geology by observing variations in the electrical resistivity of the subsurface materials. To date, there are no standardized datasets for electromagnetic methods, which hinders the progress, evaluation, benchmarking, and evolution of deep learning algorithms due to data inconsistency. Therefore, we present a large-scale electrical resistivity model database of a wide variety of geologically plausible and geophysically resolvable subsurface structures for the commonly deployed ground-based and airborne electromagnetic systems. The presented database can potentially be used to build surrogate models of well-known processes and aid in labour intensive tasks. The geophysically constrained property of this database will not only achieve enhanced performance and improved generalization but, more importantly, it will incorporate consistency and credibility in deep learning models. We show the effectiveness of the presented database by surrogating the forward modelling process, and urge the geophysical community interested in deep learning for electromagnetic methods to utilize the presented database. The dataset is publically available at https://doi.org/10.5281/zenodo.7260886 (Asif et al., 2022a).
Muhammad Rizwan Asif et al.
Status: final response (author comments only)
-
RC1: 'Comment on essd-2022-345', Anonymous Referee #1, 10 Dec 2022
This manuscript proposed a standardized dataset for deep learning-based electromagnetic methods. The database is geophysically constrained, which produces good accuracy performance and satisfactory generalization and consistency. Overall the paper is very well written, and the data shows its high readiness for the community. I would recommend it get published before some of my concerns are addressed.
Major:
- Evaluation (section 4): This study mainly employed the proposed dataset to train a deep-learning (DL) model and surrogate the forward modeling process and demonstrated that this dataset shows its great performance. The assessment method is rational; however, this section needs additional result comparison with previous relevant DL-based studies. By comparing this proposed data with other DL studies that used limited input data, the authors may demonstrate this proposed database can be treated as the benchmark. Otherwise, it is only another DL experiment for improving the computation efficiency.
Besides, please try to find some weaknesses in the training data utilized by previous DL studies and demonstrate your progress on it after comparison. For example, the introduction is well written but those training sets are not involved after then. Benchmark is a strong word that requires more comprehensive assessment and evidence.
- The assessment section (section 4) needs to provide additional quantified comparisons and descriptions for the figures rather than just some summaries.
Minor:
1) References in Table 1 should be Qin et al. (2019) rather than (Qin et al. 2019), and generally, the table caption is above the table.
2) the dataset is formatted as txt, which caused the code reading speed very slow.
- AC1: 'Reply on RC1', Muhammad Rizwan Asif, 10 Jan 2023
-
RC2: 'Comment on essd-2022-345', Anonymous Referee #2, 12 Dec 2022
General Comments
This paper submitted by Asif et al. describes a geophysically constrained subsurface resistivity model database for electromagnetic systems in a deep learning context. Such datasets and associated analyses are valuable in applying deep learning methods to geophysical applications. So the paper and associated database should be of interest to engineers and those interested in applying deep learning to electromagnetic methods. The manuscript is overall well organized and written, but there are still some shortcomings that need to be addressed before it can be accepted for publication.
Specific comments
- Please enhance the description of data processing in the revised manuscript. This will be important for a full understanding of this dataset.
- My main suggestion relates to Section 4 of this paper. As you stated, the deep learning resistivity model database (DL-RMD) presented in this paper can provide uniformity in benchmarking for DL methods in EM. But Section 4 doesn’t really provide a clear description of the great performance of this dataset. It would be better to compare it with other DL studies that have been published and listed in the Introduction.
- Table 1 – It needs to be greatly improved. The table caption is generally above the table. The references in the Table should be changed to “ Wu et al. (2021a) ”. The table caption should be concise but descriptive.
- Equations – Please check the writing form (e.g. C0). Equation 2 – Suggest revising “log10” to “lg”.
- Figure 4 – Poor quality. It would be better to provide additional descriptions for the figures rather than just summaries. For this figure, only one sentence was used to describe.
- There are lots of abbreviations used in this manuscript, it would be better to add an Appendix. Abbreviations in the title should be avoided. The phrase “depth of investigation” is abbreviated as the “DOI”, this abbreviation is not recommended.
- AC2: 'Reply on RC2', Muhammad Rizwan Asif, 10 Jan 2023
Muhammad Rizwan Asif et al.
Data sets
DL-RMD: A geophysically constrained electromagnetic resistivity model database for deep learning applications (Dataset) Asif, Muhammad Rizwan; Foged, Nikolaj; Bording, Thue; Larsen, Jakob Juul; Christiansen, Anders Vest https://doi.org/10.5281/zenodo.7260886
Muhammad Rizwan Asif et al.
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
259 | 83 | 16 | 358 | 7 | 9 |
- HTML: 259
- PDF: 83
- XML: 16
- Total: 358
- BibTeX: 7
- EndNote: 9
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1