شبیه سازی زلزله سونامیگنیک با استفاده از قوانین اصطکاکی مشتق شده تجربی Tsunamigenic earthquake simulations using experimentally derived friction laws

1. Introduction Seismological, geodetic, and tsunami observations have shown that subduction zones are complex systems where the properties of earthquake rupture vary with depth (Lay et al., 2012). For example, earthquake duration normalised for event size has been observed to decrease with depth; this recurrent feature has been attributed to depth varying shear modulus and/or stress drop for individual earthquakes (Bilek and Lay, 1999; Bilek et al., 2016; Geist and Bilek, 2001). Depth variation in subduction ruptures is, for example, evident when comparing the different historical earthquakes that occurred off the Pacific coast of Tohoku region in Japan * Corresponding author. E-mail address: shane.murphy@ifremer.fr (S. Murphy). (Fig. 1). A number of major (Mw 7–7.9) thrust earthquakes mostly slipped within a depth range of 10–40 km. These events involved individual patches of concentrated slip implying the breaking of at least one prominent, high stress asperity (Shao et al., 2011; Yamanaka and Kikuchi, 2004). Conversely, the 1896 Meiji event (M 8.2–8.4), likely involved slip primarily at the base of the shallow accretionary wedge or beneath it. This earthquake produced a disproportionately large tsunami relative to its moment magnitude, possibly making it a potential ‘tsunami earthquake’ (Kanamori, 1972). The great Mw 9.0 2011 Tohoku earthquake nucleated at ∼20–25 km depth, and produced slip at traditionally expected depths while also realising a substantial amount of slip all the way to the trench (i.e., at less than 10 km depth) (Chu et al., 2011; Ide et al., 2011; Romano et al., 2014). Numerical models of the dynamic rupture process have successfully described either individual types of earthquakes, for example the Tohoku event (Kozdon and Dunham, 2013; Noda and Lapusta, 2013), or both thrust and tsunami earthquakes in the same model (Mitsui and Yagi, 2013). Numerical models coupled with the rateand-state friction law have been used to reproduce full seismic cycles for subduction environments. However, this comes at the expense of either failing to account for geometry/free surface effects and inhomogeneity in the material surrounding the fault (Cubas et al., 2015; Noda and Lapusta, 2013), or by simplifying wave propagation to static stress changes on the fault plane (Shibazaki et al., 2011). Fully dynamic simulations including a free surface and variable geometry have tended to focus on specific rupture features of the Tohoku earthquake such as the slip in the trench or long period guided wave propagation in the ocean (Hirono et al., 2016; Huang et al., 2013; Kozdon and Dunham, 2014). Depth dependent changes in frictional parameters have been tested using rate-andstate models for the 2011 Tohoku (Kozdon and Dunham, 2013). However, to our knowledge, no numerical model has been able to reproduce a range of different observed earthquakes types (e.g. Fig. 1) while at the same time accounting for the fault geometry and complex structure as proposed here. The focus of this study is to provide a simple yet geologically consistent model that reconciles the different observed earthquake types with fault properties from independent theoretical and laboratory studies. We focus our investigation on the aspect of rupture dynamics due to depth-dependent frictional conditions focusing on a specific time window of the seismic cycle including the subseismic frictional properties of the fault materials (den Hartog et al., 2012). Hence, the friction law parameters were chosen based on available geological and geophysical constraints. On the other Fig. 2. Numerical model set up. a) Structural model used in the numerical simulations, black line denotes the subduction interface. b) Variation of frictional coeffi- cients with depth: the dark and light grey boxes denote the clay-rich and crystalline rock frictional coefficients; the white is the transition between the two materials.