[S-SS03] New trends in data acquisition, analysis and interpretation of seismicity

[SSS03-06] Mechanisms of deep earthquakes unraveled thanks to unsupervised machine learning

*Thomas P. Ferrand¹, Gilbert L. Mao^2,3, Jiaqi Li⁴, Min Chen^2,5 (1.Institute of Geological Sciences,Free University Berlin, Germany, 2.Department of Computational Mathematics, Science & Engineering, Michigan State University, East Lansing, MI, USA, 3.Georgia Institute of Technology, Atlanta, GA, USA, 4.Department of Earth, Planetary & Space Sciences, University of California, Los Angeles, CA, USA, 5.Department of Earth & Environmental Sciences, Michigan State University, East Lansing, MI, USA)

Keywords:Machine Learning, Deep earthquakes, Subduction, b value, Metastable olivine wedge, Rupture mechanisms

Although transformational faulting in the rim of the metastable olivine wedge is hypothesized as a triggering mechanism of deep-focus earthquakes, there is no direct evidence of such rim. Variations of the b value – slope of the Gutenberg-Richter distribution – have been used to decipher triggering and rupture mechanisms of deep earthquakes. However, detection limits prevent full understanding of these mechanisms. Using the Japan Meteorological Agency catalog, we estimate b values of deep earthquakes in the northwestern Pacific Plate, clustered in four regions with unsupervised machine learning. The b-value analysis of Honshu and Izu deep seismicity reveals a kink at magnitude 3.7-3.8, where the b value abruptly changes from 1.4-1.7 to 0.6-0.7. The anomalously high b values for small earthquakes highlight enhanced transformational faulting, likely catalyzed by deep hydrous defects coinciding with the unstable rim of the metastable olivine wedge, the thickness of which we estimate at ≈ 1 km.