The scenarios presented in this paper are accompanied by detailed input files and model information (see Appendix  A). (a, b) Set-up of the subduction physical model and (c) final geometry of the subduction zone at the approximate time of the slip event used to initialize the earthquake model in Scenario C (Section 4). The S-wave speed for the basalt around the fault is 3.2 km s–1. Its scalability enables large and long dynamic rupture models. 2008). In both scenarios, surface waves continue to propagate until the predefined end of the model run, which is set to t = 120 s for the blind rupture and at t = 124 s for the surface-breaching rupture. Characteristics for the blind (Scenario A), surface-breaching (Scenario B) and subduction-initialized (Scenario C) dynamic earthquake rupture models. (See sketch below for parts.) \end{eqnarray}$$, Dynamic rupture modeling of the transition from thrust to strike-slip motion in the 2002 Denali fault earthquake, alaska, Verification of a 3D fully-coupled earthquake and tsunami model, Numerical tsunami simulation including elastic loading and seawater density stratification, Supershear tsunamis and insights from the, A rupture model of the 2011 off the Pacific coast of Tohoku Earthquake, Dynamic rupture simulation reproduces spontaneous multifault rupture and arrest during the 2016, rupture velocity of plane strain shear cracks, Rupture dynamics with energy loss outside the slip zone, Three-dimensional nonplanar simulation of the 1992 Landers earthquake, The 2018 Sulawesi tsunami in Palu city as a result of several landslides and coseismic tsunamis, Source modeling and inversion with near real-time GPS: a GITEWS perspective for Indonesia, Effect of seismogenic depth and background stress on physical limits of earthquake rupture across fault step overs, New computational methods in tsunami science, The GeoClaw software for depth-averaged flows with adaptive refinement, A thermal pressurization model for the spontaneous dynamic rupture propagation on a three-dimensional fault: 1. (2009) use the 3-D, time-dependent displacements from dynamic rupture on a 3-D fault with two planar segments (megathrust and splay) as the source for shallow-water, hydrodynamic tsunami models solved with finite difference methods. Ji et al. This is a far greater (and continuous) range of values than are provided by laboratory measurements on a select number of samples. We end with a look forward. The computational mesh for this structural model has 16 million tetrahedral elements and coarsens gradually off the fault to a maximum mesh size of 100 km. the results show that for the future major earthquakes in the Sunda megathrust, the maximum tsunami wave height in Padang areas can reach 20 m and, therefore, signifi- cant damage and loss may be anticipated in this region. The quake originated in a so-called megathrust … (1). The Sunda megathrust here is advanced in its seismic cycle and may be ready for another great earthquake. This is also seen when comparing the tsunamis from the time-dependent sources from Scenario A versus Scenario B (Fig. However, the peak wave heights from the time-dependent and time-independent sources are similar. Setting the earthquake model initial conditions from a subduction model provides much needed constraint on the earthquake model initial conditions. The structural model in Scenario C is built in GOCAD ( and automatized mesh generation is performed with the software PUMGen (, Rettenberger 2017), which also exports the mesh into the efficient PUML format used by SeisSol. The highest wave height occurs at y = 0 km in both Scenario A and Scenario B. However, the wave heights are asymmetric due to the uni-directional earthquake ruptures. We relax |$\mu _{s}^{\prime }$| and |$\mu _{d}^{\prime }$| at these points to the values of material below, such that |$\mu _{s}^{\prime } = 0.025$| and |$\mu _{d}^{\prime } = 0.0097$|⁠. Jamelot et al. This was pre-European contact. The Sunda megathrust here is advanced in its seismic cycle and may be ready for another great 20 earthquake. As the Indonesian earthquake has shown, megathrust earthquakes can create tsunamis capable of crossing entire oceans. We study REs that reveal fault weakening after a large megathrust earthquake in Costa Rica, followed by fault recovery. Is there a best source model of the Sumatra 2004 earthquake for simulating the consecutive tsunami? Using our virtual laboratory, we simulate tsunamis sourced by two megathrust earthquake scenarios that differ only by their near-trench fault strength and, as a result, slip behaviour. The unstructured output from the earthquake model is bilinearly interpolated to an intermediate uniform Cartesian mesh with a resolution of 1000 m, which is used for the transfer between models. How does megathrust earthquake rupture govern tsunami behaviour? 2000; Day et al. Megathrust earthquakesoccur at subduction zonesat destructive convergent plate boundaries, where one tectonic plateis forced underneath another. These are applied uniformly in the third dimension, along fault strike, in the earthquake model. We see that the highest seafloor displacements over the entire duration of the earthquake do not control the tsunami heights during propagation to first order, but may control the width of the inundation corridor inland from the coast. |\tau _{s}| = c - \mu _{s}^{\prime }\tau _{n}. There is also variability in the static and dynamic friction coefficients with depth. The subduction-initialized earthquake in Section 4 is a Mw 9.0 event. This 2-D subduction model solves for the conservation of mass, momentum and heat in an incompressible viscoelasto-plastic medium (Gerya & Yuen 2007, see Appendix  C). The slip distribution in the Scenario B earthquake is similar to this, though slip in this scenario reaches a maximum of approximately 10 m at the trench, versus the 6 m maximum slip in the South Peru event (Pritchard et al. 2007). Linkage from an earthquake model to a tsunami model requires several considerations. 2015; Bai & Ampuero 2017; van Zelst et al. The process which produces a mega-thrust earthquake would generate a tsunami, not depicted here. Saito et al. This is the value of |$\mu _{s}^{\prime }$| above 40 km depth. For both earthquake scenarios, the maximum and minimum vertical seafloor displacements over the entire earthquake occur at t = 56 s. In the blind rupture, these are 2.6 and −1.0 m, respectively. The resulting 3-D dynamic rupture is linked with the tsunami model through the time-dependent seafloor displacements, following the same methods as in the first two examples. Aochi & Fukuyama 2002; Aagaard et al. We also apply a space–time Fourier filter to remove unwanted signals from the tsunami source that are present in the output from the earthquake model (Section 2.3.1). However, the difference in peak height diminishes during tsunami propagation towards the beach and the inundation patterns are similar in both scenarios; inundation occurs along the same stretch of the beach and has the same run-up. Also, future exploration of the relationship between slip initiation in the earthquake model and strain rate or slip rate in the subduction model may provide insight into the nature of slip nucleation itself. Published by Oxford University Press on behalf of The Royal Astronomical Society. 2005; Kaneko et al. (2019) use a low-pass filter, which does not completely eliminate the seismic waves from the tsunami signal near the source. b(x,y) = \left\lbrace \begin{array}{@{}l@{\quad }l@{}}0.05\, (x-x_0) & \text{for } x \gt x_0 \\ \end{eqnarray}$$. The third column shows temporal differences between inundation in scenarios A and B, with negative values indicating that Scenario B’s tsunami waves arrive later. At 74.7 km depth, high temperatures ensure deformation occurs predominantly through dislocation creep in the subduction model. Megathrust earthquakes and subsequent tsunamis that originate in subduction zones like Cascadia — Vancouver Island, Canada, to northern California — are some of the most severe natural disasters in the world. First, we would like to thank the Volkswagen Foundation (VolkswagenStiftung) for extended funding and excellent support of the ASCETE and ASCETE II projects ( 2012, 2013; Galis et al. SeisSol is the computational model used to simulate 3-D dynamic earthquake rupture and seismic wave propagation (see Appendix A1). In Scenario A, the higher strength near the top of the fault smoothly stops the rupture as it approaches the surface, while in Scenario B, slip continues to the top of the fault and breaks the surface. As shown in (Fig. Maximum run-up is increased in particular. Figs 7(g) and (h) show the differences between tsunamis for each scenario with the change in the source. Most finite fault inversions restricted shallow slip, as for example slip inversions for the 2004 Sumatra–Andaman earthquake (Shearer & Burgmann 2010). Oxford University Press is a department of the University of Oxford. The large modelled slip results from the effect of reverberating seismic waves and the chosen Poisson’s ratio, similar to the 2-D case (van Zelst et al. (a) Structure of the earthquake model for the subduction-initialized earthquake in Scenario C (see Section 4.1.1). The accumulated fault slip reaches maxima of 95.5 m in two locations, as shown in Fig. 2018) or thermal pressurization of pore fluids (e.g., Bizzarri & Cocco 2006; Noda 2008; Gabriel et al. seismological community and general public 2016). (2005) for acceptable errors. The tsunami from the time-independent sources also over predict wave height at y = 0 km. 10e), such constrained Dc varies with depth. (2019) use a series of dislocations derived from dynamic rupture modelling of a potential future earthquake rupture in the Nankai Trough, Japan by Hok et al. The linked initial conditions include a curved, blind fault geometry, heterogeneous fault stresses and strength, and spatially variable material properties. ASAGI automatically replicates or migrates the corresponding data tiles across compute nodes, which greatly simplifies the computing access to material or geographic data at a specific location. This provides sufficient detail for other modellers to run all or parts of these scenarios in their linked model setup and compare their results to these. and related data facilities utilized Kyodo/AP The sudden horizontal and vertical thrusting of the Pacific Plate, which has been slowly advancing under the Eurasian Plate near Japan, displaced the water above and spawned a series of highly destructive tsunami waves. Rupture velocity remains subshear relative to the 4.3 km s–1  S-wave speed in the surrounding material along most of the fault during the blind rupture, but transitions locally to supershear speed up-dip from the nucleation location and along the upper part of the fault in the surface-breaching rupture. Should tsunami simulations include a nonzero initial horizontal velocity? Key earthquake characteristics for the blind (Scenario A) and surface-breaching (Scenario B) ruptures are compared in Table 1. 2008). A4 shows at each depth the shear traction, the static fault strength, and any points at failure, where the absolute magnitude of the shear traction exceeds the fault strength. Future modelling can also strive to quantify differences of the same orders of magnitude attributed to other dynamic earthquake and tsunami characteristics. Similarly, we here restrict the off-fault constitutive behaviour of the earthquake physical models to purely elastic and use a linear slip weakening friction law on-fault. to learn about global and regional seismicity. 2013a,b, 2014). The blind rupture scenario exhibits distinct earthquake characteristics (lower slip, shorter rupture duration, lower stress drop, lower rupture speed), but the tsunami is similar to that from the surface-breaching rupture in run-up and length of impacted coastline. doi: 10.1029/2009GL038295. hydrological, and hydroacoustic data. In the strongly slip rate-dependent friction formulation (van Dinther, $$\begin{eqnarray} Sea surface height (ssh) in the tsunami sourced by the blind (Scenario A) and surface-rupturing (Scenario B) earthquakes (detailed in Section 3) along a cross section at y = 0 km and at (a) the final time the earthquake model, (b) the approximate time of the first inundation and (c) the approximate time of maximum inundation. The beach slope begins at x0 = 500 km, the coast is located at x = 540 km, and the size of the domain extends from x = −600 to 600 km and y = −600 to 600 km (Fig. 2020). This nucleation patch is in the southeast corner of the fault at 26 km depth. The estimated magnitude is approximately up to M9+. This M8.7 Rat Islands earthquake was characterized by roughly 600 km of rupture. 2011) using adaptive mesh refinement. Scenario C’s magnitude and the model fault dimensions are similar to those for the 2011 Mw 9.0 Tohoku megathrust earthquake. IRIS offers a variety of resources for the Davies 2019). 2019; Saito & Kubota 2019). Megathrust fault zone within a generalized subduction zone, highlighting the diverse slip modes observed in the shallow seismogenic, or earthquake-producing, region. Constraints are particularly lacking in locations where observational data is sparse, either because earthquakes have not yet occurred or instrumentation is poor. The archive includes all required configuration files, compilation parameters and input data. 2008). As discussed in Section 2.3, the material properties for the earthquake model are determined by reassigning Poisson’s ratio, here to ν = 0.25. In order to restrict nucleation laterally, we then set |$\mu _{s}^{\prime } = 0.025$| in the regions outside of, but at the same depths as, the nucleation zone. 10). file is included with the download. If the fault location is in a velocity strengthening region of the subduction model, we assign |$\mu _{d}^{\prime }$| to equal the maximum effective friction reached at that location during the entire subduction slip event, which may be locally larger than |$\mu _{s}^{\prime }$|⁠. The rupture speed in this scenario varies along the fault, but averages 2.1 km s–1, somewhat similar to the 2.5 km s–1 mean rupture speed estimated for the Tohoku earthquake by Ammon et al. In both scenarios, the waves reach a maximum runup of 73 m at the centre of the beach (near y = 0). 2009). In general, wave peaks in Scenario B appear delayed relative to peaks in Scenario A, which is at least partially due to the fact that the location of highest seafloor displacement is farther away from the coast in Scenario B. The structural model and mesh for the earthquake physical model used in scenarios A and B are generated with the open-source software Gmsh ( (Geuzaine & Remacle 2009). Citation Wendt, J., Oglesby, D. D., & Geist, E. L. (2009). 2005). 5. In a detailed study of the role of accretionary prisms in 2-D coupled earthquake-tsunami models, Lotto et al. 2020), both of which are at the scale of the modelled differences between tsunamis in Scenarios A and B. United states geological survey, m 8.4 - near the coast of southern peru. We refer to these instabilities as ‘slip events’ to distinguish them from dynamic earthquake ruptures modelled with SeisSol, which capture frictional failure along a pre-existing fault and the accompanying seismic wave emissions. Play now. 9(b). Here, we present methods to harness the potential of complex, 3-D dynamic rupture models as tsunami sources to enable direct studies of how earthquake initial conditions and earthquake dynamics affect tsunami genesis, propagation and inundation. Mai et al. cPropagation speed calculated for wave peak at y = 0 from t = 1000 to t = 1100 s, the time of first inundation. The surface-breaching rupture exhibits 70 per cent larger average fault slip and 40 per cent larger peak fault slip. These include fault boundary rupture, deformation of overlying plate, splay faults and landslides. Ulrich et al. Fig. We here discuss selected aspects of the presented methods for linking subduction, dynamic earthquake rupture and tsunami models. Along the subduction model fault, the average dip is 14.8°, the minimum dip is 2.3°, and the maximum dip is 34.4°. The slip distribution and maximum slip in the Scenario A earthquake are consistent with this event. For example, the assigned fault stress and strength are consistent with the fault geometry and material properties on and surrounding the fault. Slip events occur mainly in the model subduction channel and the accretionary wedge. Tsunami modelling that includes inundation must handle varying spatial scales. VIVA – Ramainya masyarakat membicarakan gempa yang berpotensi menimbulkan tsunami setinggi 20 meter belakangan ini membuat para ahli juga heran. 2019a; Wollherr et al. Fig. Fully coupled simulations of megathrust earthquakes and tsunamis in the Japan Trench, Nankai Trough, and Cascadia Subduction Zone, Fully coupled simulations of megathrust earthquakes and tsunamis in the Japan Trench, Nankai Trough, and Cascadia subduction zone, Dynamic rupture simulations of the m6.4 and m7.1 July 2019 Ridgecrest, California, earthquakes, A self-consistent mechanism for slow dynamic deformation and large tsunami generation for earthquakes in the shallow subduction zone, Dynamic wedge failure and along-arc variations of tsunami genesis in the Japan trench margin, FDM Simulation of seismic waves, ocean acoustic waves, and tsunamis based on tsunami-coupled equations of motion, Significant tsunami observed at ocean-bottom pressure gauges during the 2011 off the Pacific coast of Tohoku Earthquake, Seismic- and Tsunami-wave propagation of the 2011 off the Pacific Coast of Tohoku earthquake as inferred from the Tsunami-coupled finite-difference simulation, Successive estimation of a tsunami wavefield without earthquake source data: a data assimilation approach toward real-time tsunami forecasting, SRCMOD: an online database of finite-fault rupture models, The earthquake-source inversion validation (SIV) project, Clawpack: building an open source ecosystem for solving hyperbolic PDES, Tsunami threat in the Indian Ocean from a future megathrust earthquake west of Sumatra, Parallel, memory efficient adaptive mesh refinement on structured triangular meshes with billions of grid cells, Differences between heterogenous and homogenous slip in regional tsunami hazards modelling, Tsunami Forerunner of the 2011 Tohoku Earthquake Observed in the Sea of Japan, Shallow slip amplification and enhanced tsunami hazard unravelled by dynamic simulations of mega-thrust earthquakes, Tsunamigenic earthquake simulations using experimentally derived friction laws, Frictional constitutive law at intermediate slip rates accounting for flash heating and thermally activated slip process, Cluster design in the earth sciences: Tethys, International Conference on High Performance Computing and Communications, The three-dimensional dynamics of dipping faults, Surface deformation due to shear and tensile faults in a half-space, Mode-wave equivalence and other asymptotic problems in tsunami theory, Three-dimensional dynamic rupture simulation with a high-order discontinuous Galerkin method on unstructured tetrahedral meshes, Verification of an ADER-DG method for complex dynamic rupture problems. Salaree & Okal 2020). Time series of sea surface height (ssh) at 3 measurement points located 10 m from the coast near x = 240 km for tsunamis sourced by the time-dependent and time-independent filtered displacements from (a) the blind rupture in Scenario A and (b) the surface-breaching rupture in Scenario B. Subduction zone earthquakes A megathrust earthquake occurs in subduction zones at convergent boundaries. Stochastic Tsunami Simulation. Elsewhere along the fault, the rupture proceeds at subshear speeds. Therefore, we prevent failure by assigning c = 5 MPa in the sediments above 25 km depth, which is the value of c in the deeper basalt. The events typically nucleate near the downdip limit of the model seismogenic zone in the basalt, after which they progress into the shallow sediments. This megathrust earthquake also triggered a devastating tsunami that caused damage along the Gulf of Alaska, the West Coast of the United States, and in Hawaii. Maeda et al. However, the average time-independent displacements are equivalent at 0.9 m for both scenarios. Pemahaman seperti ini tentu saja kurang tepat,” kata Kepala Bidang Mitigasi Gempa Bumi dan Tsunami BMKG, Daryono dalam keterangan pers, Sabtu, 26 September 2020. Subduction-zone megathrust earthquakes, the most powerful earthquakes in the world, can produce tsunamis through a variety of structures that are missed by simple models including: fault boundary rupture, deformation of overlying plate, splay faults and landslides. 10c). Kame et al. We emphasize that these applications demonstrate the capabilities of the modelling framework; future, more involved and complex applications will certainly result in further knowledge gain. 2017b; Amlani et al. When initializing the earthquake model from a subduction model, we must honor the plane-strain conditions of the 2-D subduction model while mapping the stress field into the 3-D earthquake model. Our mission is to advance awareness and understanding of seismology Alternatively, adjusting the subduction model itself to be region specific, for example to the Japan trench, would provide more direct constraints on fault geometry and other initial conditions. Okada 1985). This isn't likely to happen on the East Coast, but it could. 2016; Uphoff et al. Inundation area is 15 per cent larger for the surface-breaching rupture. The third row shows temporal differences between inundation from the time-dependent and time-independent sources, with negative values indicating that the waves from the time-independent source arrive later. Ocean Bottom Seismograph Instrument Pool (OBSIP), Greenland Ice Sheet Monitoring Network (GLISN), Global Reporting Observatories in Chile (GRO-Chile), INTERNATIONAL DEVELOPMENT SEISMOLOGY (IDS), Recent Earthquake Teachable Moments (RETM), Megathrust Type 1 (excerpt from animation), GIF Megathrust Type 3 (excerpt from animation), GIF Megathrust Type 2 (excerpt from animation), Subduction Zone: Simplified model of elastic rebound. 16(a) shows the history of inundation. The Bengkulu earthquake had slip restricted to below 10 km depth, with most slip occurring at 16–40 km depth and reaching a maximum of 6–7 m (Gusman et al. Use appropriate media player to utilize captioning. $$\begin{eqnarray} However, despite differences in the earthquake slip distributions and maximum slip values, the difference in peak wave height at the coast is small, and coastal inundation extent and run-up are the same in both scenarios. A report, released last week, looked into New Zealand’s worst ever megathrust earthquake, which occurred in the town of Kaikōura on the nation’s South Island in November 2016. Several hydrological tsunami models use a set of 2-D simplifications of the non-linear Navier–Stokes equations, such as the shallow water equations (e.g. 4d), where it is approximately 100 s. The average velocity from t = 1000 s to t = 1100 s of the wave peak from blind rupture source in Scenario A is 157 m s–1, faster than that for the surface-breaching rupture source in Scenario B of 142 m s–1. The failure criterion is met in three locations: within the shallow sediments, at one isolated point at 74.7 km depth, and in the region of 40–43 km depth. Clawpack, Mandli et al. 6(a) shows that the wave peak is initially higher for the surface-breaching rupture source. 2006), the Leibniz Supercomputing Centre (LRZ, projects no. The three-dimensional isotropic case, Earthquake ruptures with strongly rate-weakening friction and off-fault plasticity. Slip then progresses outward in all directions along the fault, producing a Mw 9.0 earthquake with an average dynamic stress drop of 2.2 MPa, which is lower than in Scenarios A and B (Section 3). In this way, earthquake initial conditions are assigned self-consistently and the tsunami source reflects the conditions developed over long-term subduction and seismic cycling. These differences in wave timing are visible in (Fig 8) as well, which shows that the tsunamis from time-independent sources lag behind those from the time-dependent sources at y = 0 km and precede the time-dependent sources at y = 150 km. 7c). \end{eqnarray}$$, To better understand this, we calculate the efficiency of the Scenario A and Scenario B earthquakes in generating the resulting tsunamis as ϵ following Lotto, $$\begin{eqnarray} \mu ^\mathrm{{sc}} = \frac{V_c\mu _s^\mathrm{{sc}}+V\mu _d^\mathrm{{sc}}}{V_c+V}, 2019b) alongside constant frictional parameters and homogeneous material properties. Surprises with devastating consequences in past earthquake-tsunami sequences motivate a better understanding of the physical connections between subduction, earthquake dynamics and tsunami from genesis to inundation.