Deep reinforcement learning for inverting earthquake focal mechanism and its potential application to marine earthquakes

Abstract

Earthquake data are one of the key means by which to explore our planet. At a large scale, the layered structure of the Earth is revealed by the seismic waves of natural earthquakes that go deep into its inner core. At a local scale, seismology for exploration has successfully been employed to discover massive fossil energies. As the volume of recorded seismic data becomes greater, intelligent methods for processing such a volume of data are eagerly anticipated. In particular, earthquake focal mechanisms are important for assessing the severity of tsunamis, characterizing seismogenic faults, and investigating the stress perturbations that follow a major earthquake. Here, we report a novel deep reinforcement learning method for inverting the earthquake focal mechanism. Unlike more typical deep learning applications, which require a large training dataset, a deep reinforcement learning system learns by itself. We demonstrate the validity and efficacy of the proposed deep reinforcement learning method by applying it to the Mw 7.1 mainshock of the Ridgecrest earthquakes in southern California. In the foreseeable future, deep learning technologies may greatly contribute to our understanding of the oceanographic process. The proposed method may help us understand the mechanism of marine earthquakes.

Keywords

Deep learning / Deep reinforcement learning / Earthquake focal mechanism / Nonlinear inversion

References

[1] Adeyemi T, Sharma J, Tabjula J. Monitoring and characterization of gas migration in oil-based mud using fiberoptic DAS and DTS. SPE J, 2023, 28(6): 2802-2816,

[2] Attary N, Van de Lindt JW, Barbosa AR, Cox DT, Unnikrishnan VU. Performance-based tsunami engineering for risk assessment of structures subjected to multi-hazards: tsunami following earthquake. J Earthq Eng, 2021, 25(10): 2065-2084,

[3] Barbu A, Zhu SC. Monte Carlo Methods. Options and Derivatives Programming in C++23, 2020 Singapore Springer 225-239

[4] Bergen KJ, Johnson PA, de Hoop MV, Beroza GC. Machine learning for data-driven discovery in solid Earth geoscience. Science, 2019, 363(6433): eaau0323,

[5] Hardebeck JL, Shearer PM. Using S/P amplitude ratios to constrain the focal mechanisms of small earthquakes. Bull Seismol Soc Amer, 2003, 93(6): 2434-2444,

[6] Jones GL, Qin Q. Markov chain Monte Carlo in practice. Annu Rev Stat Application, 2022, 9: 557-578,

[7] Kong QK, Trugman DT, Ross ZE, Bianco MJ, Meade BJ, Gerstoft P. Machine learning in seismology: turning data into insights. Seismol Res Lett, 2019, 90(1): 3-14,

[8] Kuang WH, Yuan CC, Zhang J. Real-time determination of earthquake focal mechanism via deep learning. Nat Commun, 2021, 12(1): 1432,

[9] Kuang WH, Yuan CC, Zhang J. Network-based earthquake magnitude determination via deep learning. Seismol Res Lett, 2021, 92(4): 2245-2254,

[10] Kuang WH, Yuan CC, Zhang J, Zhang W. Relative focal mechanism inversion and its application to Ridgecrest sequence. Seismol Res Lett, 2023, 94(1): 305-317,

[11] Kuang WH, Yuan CC, Zou ZH, Zhang J, Zhang W. Autonomous earthquake location via deep reinforcement learning. Seismol Res Lett, 2024, 95(1): 367-377,

[12] Kuang WH, Zhang J, Zhang W. A novel deep-learning image condition for locating earthquake. Geophys J Int, 2023, 235(3): 2168-2178,

[13] Kuna VM, Nábělek JL. Seismic crustal imaging using fin whale songs. Science, 2021, 371(6530): 731-735,

[14] Li L, Wang W, Yu Z, Chen Y. CREDIT–X1local: a reference dataset for machine learning seismology from ChinArray in Southwest China. Earthq Sci, 2023, 37(2): 139-157,

[15] Mnih V, Kavukcuoglu K, Silver D, Rusu AA, Veness J, Bellemare MG, et al.. Human-level control through deep reinforcement learning. Nature, 2015, 518(7540): 529-533,

[16] Mousavi SM, Beroza GC. A machine-learning approach for earthquake magnitude estimation. Geophys Res Lett, 2020, 47(1): e2019GL085976,

[17] Nanjo KZ. Were changes in stress state responsible for the 2019 Ridgecrest, California, earthquakes?. Nat Commun, 2020, 11(1): 3082,

[18] Papazachos BC, Dimitriu PP. Tsunamis in and near Greece and their relation to the earthquake focal mechanisms. Nat Hazards, 1991, 4: 161-170,

[19] Sambridge M. Geophysical inversion with a neighbourhood algorithm–I. Searching a parameter space. Geophys J Int, 1999, 138(2): 479-494,

[20] Shen J, Zhu T. DAS with telecommunication fibre-optic cable in urban areas can record storm-induced seismic noise. Geophys J Int, 2023, 235(3): 2122-2136,

[21] Silver D, Huang A, Maddison CJ, Guez A, Sifre L, van den Driessche G, et al.. Mastering the game of Go with deep neural networks and tree search. Nature, 2016, 529(7587): 484-489,

[22] Silver D, Schrittwieser J, Simonyan K, Antonoglou I, Huang A, Guez A, et al.. Mastering the game of Go without human knowledge. Nature, 2017, 550(7676): 354-359,

[23] Sugawara D. Numerical modeling of tsunami: advances and future challenges after the 2011 Tohoku earthquake and tsunami. Earth-Sci Rev, 2021, 214,

[24] Sumy DF, McBride SK, von Hillebrandt-Andrade C, Kohler MD, Orcutt J, Kodaira S, et al.. Long-term ocean observing coupled with community engagement improves tsunami early warning. Oceanography, 2021, 34(4): 70-77,

[25] Sutton RS, Barto AG (1998) The reinforcement learning problem. In: Reinforcement learning: an introduction. MIT Press, Cambridge, MA, pp 51–85

[26] Tan Y, Helmberger D. A new method for determining small earthquake source parameters using short-period P waves. Bull Seismol Soc Amer, 2007, 97(4): 1176-1195,

[27] Toda S, Stein RS, Beroza GC, Marsan D. Aftershocks halted by static stress shadows. Nat Geosci, 2012, 5(6): 410-413,

[28] Wang X, Zhan ZW. Seismotectonics and fault geometries of the 2019 Ridgecrest sequence: insight from aftershock moment tensor catalog using 3-D Green’s functions. J Geophys Res-Solid Earth, 2020, 125(5): e2020JB019577,

[29] Williams EF, Zhan ZW, Martins HF, Fernández-Ruiz MR, Martín-López S, González-Herráez M, et al.. Surface gravity wave interferometry and ocean current monitoring with ocean-bottom DAS. J Geophys Res-Oceans, 2022, 127(5): e2021JC018375,

[30] Zhang X, Zhang J, Yuan CC, Liu S, Chen ZB, Li WP. Locating induced earthquakes with a network of seismic stations in Oklahoma via a deep learning method. Sci Rep, 2020, 10: 1941,

[31] Zhao LS, Helmberger DV. Source estimation from broad-band regional seismograms. Bull Seismol Soc Amer, 1994, 84(1): 91-104

[32] Zhao M, Xiao ZW, Chen S, Fang LH. DiTing: a large-scale Chinese seismic benchmark dataset for artificial intelligence in seismology. Earthq Sci, 2023, 36(2): 84-94,

[33] Zhu L, Helmberger DV. Advancement in source estimation techniques using broadband regional seismograms. Bull Seismol Soc Amer, 1996, 86(5): 1634-1641,

[34] Zhu L, Rivera LA. A note on the dynamic and static displacements from a point source in multilayered media. Geophys J Int, 2002, 148(3): 619-627,

[35] Zhu WQ, Beroza GC. PhaseNet: a deep-neural-network-based seismic arrival-time picking method. Geophys J Int, 2018, 216(1): 261-273

Funding

National Natural Science Foundation of China(No. 42104047)