ScarpLearn : Artificial Intelligence for the Geomorphology of Normal Faults

Normal faults, often responsible for destructive earthquakes, leave characteristic scarps in the landscape. Measuring their height helps to better understand the past activity of faults, but this task remains time-consuming, subjective, and difficult to reproduce.



To simplify and make this task more reliable, researchers developed ScarpLearn, a method based on deep learning. The tool uses convolutional neural networks (CNNs) trained on simulated data to automatically estimate scarp heights, while also providing an assessment of uncertainties.

Tested on faults in Mexico (Trans-Mexican Volcanic Belt) and the Malawi Rift, ScarpLearn shows several advantages compared to conventional methods :

  • a considerable time saving,
  • more reproducible and accurate results,
  • improved management of uncertainties.

Beyond these case studies, ScarpLearn illustrates the potential of artificial intelligence for morphotectonic analysis. This approach paves the way for a more detailed understanding of normal faults and the seismic hazards they pose.


The full study is published in SEISMICA.

References

Arrowsmith, J. Ramón, Pollard, D. D., & Rhodes, D. D. (1996).
Hillslope Development in Areas of Active Tectonics.
Journal of Geophysical Research : Solid Earth, 101(B3), 6255–6275.

Hodge, M., Biggs, J., Fagereng, AAke, Elliott, A., Mdala, H., & Mphepo, F. (2019).
A Semi-Automated Algorithm to Quantify Scarp Morphology (SPARTA) : Application to Normal Faults in Southern Malawi.
Solid Earth, 10(1), 27–57.

Mattéo et al., 2021
L. Mattéo, I. Manighetti, Y. Tarabalka, J.-M. Gaucel, M. van den Ende, A. Mercier, O. Tasar, N. Girard, F. Leclerc, T. Giampetro, S. Dominguez, J. Malavieille
Automatic Fault Mapping in Remote Optical Images and Topographic Data With Deep Learning
Journal of Geophysical Research : Solid Earth , 126(4) (2021)

Núñez Meneses, A., Lacan, P., Zúñiga, F. R., Audin, L., Ortuño, M., Rosas Elguera, J., León-Loya, R., & Márquez, V. (2021).
First Paleoseismological Results in the Epicentral Area of the Sixteenth Century Ameca Earthquake, Jalisco – México.
Journal of South American Earth Sciences, 107, 103121.

Pousse-Beltran et al., 2022
L. Pousse-Beltran, L. Benedetti, J. Fleury, P. Boncio, V. Guillou, B. Pace, M. Rizza, I. Puliti, A. Socquet
36Cl Exposure Dating of Glacial Features to Constrain the Slip Rate along the Mt. Vettore Fault (Central Apennines, Italy)
Geomorphology, 108302 (2022)

Scientific contacts :

  • Léa Pousse-Beltran – Researcher, ISTerre – IRD
  • Théo Lallemand – Researcher, ISTerre
  • Laurence Audin – Researcher, ISTerre – IRD
  • Pierre Lacan – Researcher, UNAM (Universidad Nacional Autónoma de México)
  • Andres David Nunez-Meneses – Researcher, UNAM (Universidad Nacional Autónoma de México)
  • Sophie Giffard-Roisin – Researcher, ISTerre – IRD

Funding

  • Agence Nationale de la Recherche under the France 2030 programme, reference ANR-23-IACL-0006.
  • MIAI@Grenoble Alpes (ANR19-P3IA-0003)
  • GRICAD infrastructure
  • ISTerre part of Labex OSUG@2020 (Investissements d’avenir – ANR10 LABX56)
  • CNES R&T Call 2022 ”Hybridation des donnees” Nº34500075632
  • PNTS program of INSU CNRS
  • PAPIIT grant IN108220 and IG101823 awarded to Pierre Lacan
  • France-Mexico collaborative project SEP-CONACYT-ANUIES-ECOS Nº321193 and the IGCP-669 Ollin Project of UNESCOIUGS.
  • Universidad Nacional Autónoma de México under PASPA – DGAPA grant (P. Lacan’s academic stay)
  • DINAMIS program