Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

  • Matters Arising
  • Published:

One neuron versus deep learning in aftershock prediction

Nature volume 574, pages E1–E3 (2019)Cite this article

Subjects

Matters Arising to this article was published on 02 October 2019

The Original Article was published on 29 August 2018

arising from P. M. R. DeVries et al. Nature https://doi.org/10.1038/s41586-018-0438-y (2018)

Forecasting the spatial distribution of aftershocks in the aftermath of large seismic events is of great importance for improving both our understanding of earthquake triggering and post-disaster management. Recently, DeVries et al.1 attempted to solve this scientific problem by deep learning. Using the area under the curve (AUC) of receiver operating characteristic curves, the authors showed that a deep neural network (DNN) considerably outperformed classical Coulomb stress. Here we first clarify that similar performances had already been obtained (by the same authors, in 2017)2 for various scalar stress metrics, suggesting that deep learning does not actually improve prediction. Second, we reformulate the 2017 results2 using two-parameter logistic regression (that is, one neuron) and obtain the same performance as that of the 13,451-parameter DNN. We further show that a logistic regression based on the measured distance and mainshock average slip (instead of derived stresses) performs better than the DNN. This demonstrates that so far the proposed deep learning strategy does not provide any new insight (predictive or inferential) in this domain.

This is a preview of subscription content, access via your institution

Access options

Buy this article

  • Purchase on Springer Link
  • Instant access to full article PDF
Buy now

Prices may be subject to local taxes which are calculated during checkout

Fig. 1: Prediction of aftershock spatial patterns based on stress features.
Fig. 2: Prediction of aftershock spatial patterns using the distance, r, and the slip, d.

Data availability

The data that support the findings of this study are available from the SRCMOD fault rupture catalogue (http://equake-rc.info/SRCMOD), the International Seismological Centre earthquake catalogue (http://www.isc.ac.uk/iscgem) and from DeVries et al.1 at https://github.com/phoebemrdevries/Learning-aftershock-location-patterns.

Code availability

Original codes by DeVries et al.1 are available at https://github.com/phoebemrdevries/Learning-aftershock-location-patterns. An R code including the distance–slip feature definition and logistic regression training/testing is available from the corresponding authors on request.

References

  1. DeVries, P. M. H., Viégas, F., Wattenberg, M. & Meade, B. J. Deep learning of aftershock patterns following large earthquakes. Nature 560, 632–634 (2018).

    Article  ADS  CAS  Google Scholar 

  2. Meade, B. J., DeVries, P. M. R., Faller, J., Viegas, F. & Wattenberg, M. What is better than Coulomb failure stress? A ranking of scalar static stress triggering mechanisms from 105 mainshock-aftershock pairs. Geophys. Res. Lett. 44, 11,409–11,416 (2017).

    Article  Google Scholar 

  3. Reasenberg, P. A. & Jones, L. M. Earthquake hazard after a mainshock in California. Science 243, 1173–1176 (1989).

    Article  ADS  CAS  Google Scholar 

  4. Reasenberg, P. A. & Jones, L. M. Earthquake aftershocks: update. Science 265, 1251–1252 (1994).

    Article  ADS  CAS  Google Scholar 

  5. Gerstenberger, M. C., Wiemer, S., Jones, L. M. & Reasenberg, P. A. Real-time forecast of tomorrow’s earthquakes in California. Nature 435, 328–331 (2005).

    Article  ADS  CAS  Google Scholar 

  6. Felzer, K. R. & Brodsky, E. E. Decay of aftershock density with distance indicates triggering by dynamic stress. Nature 441, 735–738 (2006).

    Article  ADS  CAS  Google Scholar 

  7. Richards-Dinger, K., Stein, R. S. & Toda, S. Decay of aftershock density with distance does not indicate triggering by dynamic stress. Nature 467, 583–586 (2010).

    Article  ADS  CAS  Google Scholar 

  8. Mignan, A. Utsu aftershock productivity law explained from geometric operations on the permanent static stress field of mainshocks. Nonlinear Process. Geophys. 25, 241–250 (2018).

    Article  ADS  Google Scholar 

  9. Steacy, S., Gerstenberger, M., Williams, C., Rhoades, D. & Christophersen, A. A new hybrid Coulomb/statistical model for forecasting aftershock rates. Geophys. J. Int. 196, 918–923 (2014).

    Article  ADS  Google Scholar 

  10. Cattania, C., Hainzl, S., Wang, L., Roth, F. & Enescu, B. Propagation of Coulomb stress uncertainties in physics-based aftershock models. J. Geophys. Res. Solid Earth 119, 7846–7864 (2014).

    Article  ADS  Google Scholar 

  11. Cattania, C. et al. The forecasting skill of physics-based seismicity models during the 2010–2012 Canterbury, New Zealand, earthquake sequence. Seismol. Res. Lett. 89, 1238–1250 (2018).

    Article  Google Scholar 

  12. LeCun, Y., Bengio, Y. & Hinton, G. Deep learning. Nature 521, 436–444 (2015).

    Article  ADS  CAS  Google Scholar 

  13. Jordan, M. I. & Mitchell, T. M. Machine learning: trends, perspectives, and prospects. Science 349, 255–260 (2015).

    Article  ADS  MathSciNet  CAS  Google Scholar 

  14. Kong, Q. et al. Machine learning in seismology: turning data into insights. Seismol. Res. Lett. 90, 3–14 (2019).

    Article  Google Scholar 

  15. Beroza, G. C. Aftershock forecasts turn to AI. Nature 560, 556–557 (2018).

    Article  ADS  CAS  Google Scholar 

Download references

Author information

Authors and Affiliations

  1. Swiss Federal Institute of Technology, Institute of Geophysics, Zurich, Switzerland

    Arnaud Mignan

  2. Swiss Seismological Service, Zurich, Switzerland

    Arnaud Mignan & Marco Broccardo

  3. Institute of Risk Analysis, Prediction and Management, Academy for Advanced Interdisciplinary Studies, Southern University of Science and Technology, Shenzhen, China

    Arnaud Mignan

  4. Swiss Federal Institute of Technology, Institute of Structural Engineering, Zurich, Switzerland

    Marco Broccardo

Authors
  1. Arnaud Mignan
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Marco Broccardo
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

A.M. and M.B. contributed equally to the design and analysis of this study.

Corresponding authors

Correspondence to Arnaud Mignan or Marco Broccardo.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Mignan, A., Broccardo, M. One neuron versus deep learning in aftershock prediction. Nature 574, E1–E3 (2019). https://doi.org/10.1038/s41586-019-1582-8

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1038/s41586-019-1582-8

This article is cited by

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

Search

Advanced search

Quick links

Nature Briefing AI and Robotics

Sign up for the Nature Briefing: AI and Robotics newsletter — what matters in AI and robotics research, free to your inbox weekly.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing: AI and Robotics