EUVIMEDEUVIMED entdecken
Open Access · CC BY-NC-ND 4.0

Enhanced generalization of a deep learning framework for 1-D magnetotelluric inversion

Mehdi Rahmani Jevinani, Banafsheh Habibian Dehkordi, Ian Ferguson

Annals of Geophysics · 30.06.2026 · 69 (4) · DM453

DOI: 10.4401/ag-9500 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Annals of Geophysics, erschienen am 30.06.2026. Verfasst von Mehdi Rahmani Jevinani und 2 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref). Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Widespread and successful application of deep learning‑based methods to solve inverse problems in Geophysics particularly - in magnetotellurics (MT) – remains limited largely due to poor generalization of trained models to unseen and noisy field data. In order to contribute to addressing this issue, we present a robust one‑dimensional (1‑D) MT inversion framework in which the deep learning model has increased capacity to generalize. First, by employing a randomized conductivity model construction scheme, constrained by geophysical resolution limitations and geologically realistic scenarios, a comprehensive and diverse training dataset representing a wide variety of subsurface conductivity models is generated. To further ensure a representative sampling of the model space, Gaussian noise is added to part of the generated data. Second, a clustering‑based data splitting strategy is introduced as an alternative to the random approach, in which each cluster proportionally contributes to the training, validation, and test subsets, to preserve consistent statistical distribution across them. Qualitative and quantitative evaluations confirm that the combined approach performs robustly and reliably when exposed to a broad range of synthetic data as well as a sample of real field MT data. Resistivity models constructed using a modified convolutional U‑Net architecture, trained with a combination of clean and noisy data generated in the stated manner, are compared to those derived from deterministic methods and show high levels of accuracy.

Zitieren

APA

Jevinani, M. R., Dehkordi, B. H., & Ferguson, I. (2026). Enhanced generalization of a deep learning framework for 1-D magnetotelluric inversion. Annals of Geophysics, 69(4), DM453. https://doi.org/10.4401/ag-9500

Vancouver

Jevinani MR, Dehkordi BH, Ferguson I. Enhanced generalization of a deep learning framework for 1-D magnetotelluric inversion. Annals of Geophysics. 2026;69(4):DM453. doi:10.4401/ag-9500

BibTeX
@article{Jevinani2026,
  author = {Jevinani, Mehdi Rahmani and Dehkordi, Banafsheh Habibian and Ferguson, Ian},
  title = {Enhanced generalization of a deep learning framework for 1-D magnetotelluric inversion},
  journal = {Annals of Geophysics},
  year = {2026},
  volume = {69},
  number = {4},
  pages = {DM453},
  doi = {10.4401/ag-9500}
}
Kostenlos starten →