Search

In recent years many devastating earthquakes with Mw ≥ 7.5 occurred during 2017-2018 asa result of active plate interactions near the major mega-thrust subduction zones. Theyreveal tectonic complexities and astonishing deformation styles associated with theplate movements. In this study, we determined source mechanism parameters andfinite-fault slip distribution models of four destructive earthquakes (Mw ≥ 7.5) byperforming point-source and finite-fault slip inversions. The shapes and amplitudesof long-period and broad-band P- and SH-waveforms recorded by the Federationof Digital Seismograph Networks (FDSN) and the Global Digital SeismographNetwork (GDSN) stations in the distance range of 30˚ - 90˚ are compared withsynthetic waveforms. Finite-fault slip models of earthquakes are estimated by applyinga hybrid back-projection method that uses teleseismic P-waveforms to integratethe direct P-phase with reflected phases from structural discontinuities near thesource. Overall results indicate various faulting mechanisms at relatively shallowfocal depths (h<70 km). For example, the February 25, 2018 Papua New Guinea(Mw: 7.5) earthquake indicates a thrust faulting mechanism associated with theconvergence of Australia and Pacific Plates. The September 28, 2018 Sulawesi (Indonesia)earthquake (Mw: 7.5), occurred within the Molucca Sea micro-plate at the eastern part ofIndonesia shows a strike-slip faulting mechanism reflecting the complex deformationsrelated to the interactions between Australia, Sunda, Pacific and Philippine Seaplates. Regardless strike-slip mechanism involved, this large earthquake producedunexpected tsunami waves along the coastal planes of Sulawesi. Finite-fault slipmodels reveal several heterogeneous rupture propagations and slip distributions onfault planes of those earthquakes. In addition, distributions of P- wave first motionpolarities recorded at near-field and regional seismic stations are consistent withthe best-fitting minimum misfit source mechanism solutions of earthquakes. Thisstudy is partially supported by the Turkish Academy of Sciences (TÜBA-GEBIP). [ABSTRACT FROM AUTHOR]

Displacement data produced by GNSS observations never saturate for large tsunamigenic earthquakes in contrast to seismometer data that has a limitation of instrument saturation. Geospatial Information Authority of Japan has launched a real-time GNSS analysis system named "REGARD" which estimates finite fault models based on the Japanese nationwide GNSS network GEONET. The REGARD system is designed to provide finite fault models including single rectangular fault and slip distribution models in a few minutes.The system is comprised of three subsystems: (1) Real-time positioning subsystem, (2) Event detection subsystem, and (3) Fault model inversion subsystem. The real-time 1 Hz data from the GEONET stations are analyzed by Real-time positioning subsystem using RTKLIB 2.4.2 software (Takasu, 2013) as the positioning engine. The processed time-series data are monitored by Event detection subsystem to detect earthquake events. EEW messages are also checked for complement of much noisier GNSS data. Once an earthquake event is detected, Fault model inversion subsystem is launched and finite fault models are estimated automatically. The finite fault models include single rectangular fault model, which is applicable to both inland and subduction zone earthquakes, and slip distribution model, which is particularly suitable for large subduction zone earthquakes.The REGARD system has been in operation since April, 2016. The REGARD system is processing ~1,200 stations of GEONET. In this presentation, we introduce an overview of the system design and show the test results for the cases of the 2011 Tohoku earthquake (Mw 9.0) and a simulated Nankai Trough earthquake (Mw 8.7). Furthermore, an operational real-time result for the 2016 Kumamoto earthquake (Mj 7.3) is also shown.The test results showed that reliable finite fault models for the earthquakes could be provided within 3 mintues for rapid tsunami simulations. The result for the 2016 Kumamoto earthquake (Mj 7.3) showed single rectangular fault models with VR of 96.2%, demonstrating the capability for modeling inland earthquakes. These results imply the possibility of improvements in tsunami simulations or hazard information using rapid finite fault models. [ABSTRACT FROM AUTHOR]

The region of Japanese island arc is one of the most seismically active regions of the world due to very high plate convergence rate. On March 11, 2011, the great Mw=9.0 Tohoku earthquake occurred in the central segment of the Japan subduction zone, the region where the previous megathrust earthquake was in 869. The duration of seismic cycle in this part of Japan subduction zone is estimated at about 800-1100 years. To identify the spatiotemporal variations of surface deformations following the 2011 earthquake we analyzed more than 12 years of continuous GPS observations from over 1400 stations of GEONET network that covers the entire territory of Japanese islands. Analysis of GPS data prior to Tohoku earthquake revealed a notable anomaly – a subsidence of coastal region directly opposite the future source zone of 2011 earthquake. This can be an evidence for the termination of elastic stresses accumulation and the transition of corresponding part of the seismogenic zone to the preseismic stage of the seismic cycle. Coseismic displacements captured by GPS stations allowed us to model the slip distribution in the earthquake source and to reveal significant effect of Fossa Magna Graben on the coseismic deformations pattern. Source zone of the Tohoku earthquake affected all possible seismogenic blocks, limited by geological bounds, such as Fossa Magna, which resulted in reestablishing of seismic cycle for the whole northeastern part of Japan subduction zone. The observed surface displacements before, during and after the Tohoku earthquake exhibited good agreement with the expected motions implied by the keyboard model of subduction zones [Lobkovsky et al., 1991] at the appropriate stages of the seismic cycle. At the same time, analysis of the displacement rate variations estimated over 1-year intervals after the Tohoku earthquake showed an intense ongoing postseismic motion of complex mechanism. We tested a hypothesis of significant afterslip in the earthquake source using 1-month cumulative displacements for the first half of the year after the earthquake. Our analysis of the afterslip process showed that its maximal contribution rapidly decreases in the first six months after the earthquake from 1-3 meters to 10-20 centimeters per month and cannot predict the observed long-term displacements. We also assessed the effect of coseismically-induced viscoelastic relaxation in the underlying asthenospheric wedge on observed surface deformations. Such a long-term postseismic process, which can last after megathrust earthquakes up to several decades can significantly affect the seismic cycle in the northeastern part of the Japan subduction zone, providing a probable explanation for seismic cycle prolongation. [ABSTRACT FROM AUTHOR]

The 2011 Tohoku-oki earthquake occurred in the well-studied Japan convergent margin andproduced coseismic and postseismic surface displacements well observed at several GPSsites, including a dense onshore network and offshore GPS-A stations. Such geodeticdata, particularly large landward postseismic displacements in the primary rupturearea, suggest that relocking occurred shortly after the earthquake. This has clearimplications for the slip deficit accummulation and thus the seismic hazard in theregion. However, the landward motion recorded at the sites closest to the trench issignificantly faster than the plate convergence rate, which suggests that it cannot beattributed solely to relocking. Instead, it has been explained through numericalmodelling as due to viscoelastic relaxations of the stresses induced by coseismicdisplacement. Numerical modelling is valuable as it can quantitatively connect the surface displacementfield to the relocking history of the megathrust interface as well as to the postseismicrelaxation history. We thus construct a finite-element model of the Japan subduction zonewith a simplified, spherical-shell 3D geometry. The model is driven to realistic preseismicstresses and strains by imposing the earthquake history of the last several decades. We thenuse it to investigate the geodetic signature of relocking history in the context of specificsubduction zone geometry and physical features. The hypothesis is that a combination ofrapid relocking and significant viscous relaxation is needed to explain the observedpostseismic surface velocities. The lithospheres in the model have linear elastic rheology, neglecting the minor anelasticcomponent of deformation, while the asthenospheres have Maxwell viscoelastic rheologies.The areas of the megathrusts modelled as non-seismogenic are also viscoelastic and thusallow for afterslip in response to stresses induced during coseismic slip. The lithosphericstructure is imported from the Slab2 dataset, derived from seismological data. The effect ofdifferent lithospheric thicknesses of both the overlying and downgoing plates is thenexplored. This accounts for uncertainties, heterogenity and the discrepancy between differentdefinitions of plate thickness. The spatial extent of the coseismically slipping patch in themegathrust is varied in different models, especially its updip limit. This is crucial inattempting to reproduce the surface displacement close to the trench. Furthermore, it reflectsuncertainties and possible variations in locking behaviour of peripheral areas of theinterface. Preliminary results show that rapid relocking is not sufficient to explain the landward anddownward displacements observed offshore on the upper plate. Work is ongoing to fullycharacterize the primary features of the subduction zone responsible for the observed surfacedisplacements. [ABSTRACT FROM AUTHOR]

The east coast of Tasmania is directly exposed to tsunamis originating from the Puysegur subduction zone, and would be affected approximately 90 minutes post-earthquake. Tasmania does not experience tsunamis frequently, and tsunami hazard is poorly understood and planned for in the state. Moreover, the risk to Hobart port operations from a large tsunami has not previously been considered. This study applied numerical modelling to simulate the effects of a maximum credible tsunami (Mw 8.7 earthquake of New Zealand's southwest coast, combined with a highest astronomical tide), with the aim of determining potential coastal inundation levels, and assessing risks to shipping operations in Hobart Port. Inundation modelling was performed using the ANUGA hydrodynamic library, together with a detailed terrain model constructed from LiDAR and bathymetric datasets. In addition, a dune erosion operator was written for ANUGA, to account for tsunami erosion of the sand dunes in front of affected areas. The maritime hazard assessment focused on tsunami impacts for shipping in the navigation channel and ports, including mobilisation and evacuation times required for specific vessel types that frequent the harbour. Modelling results show a maximum incoming tsunami wave of approximately 7 m in exposed coastal areas, which could lead to inundation depths of > 4 m in some places. Significant seiching and wave reflections are evident in Hobart's main channel and embayments, which would pose a considerable risk to marine craft. An important finding of this study is that the feasibility of shipping evacuation from Hobart port is questionable, given the timeframes involved and the nature of simulated water disturbance. This study demonstrates the use of numerical modelling to inform tsunami hazard assessment, and highlights an often overlooked aspect of tsunami hazard. Statewide emergency management plans are currently being updated to make use of this work. [ABSTRACT FROM AUTHOR]

Northeast India is sandwiched between the seismically active plate boundaries of the Eastern Himalaya to the north and the Indo-Burman subduction zone to the east. This plate boundary system is different from the central Nepal Himalaya, due to oblique convergence across two orthogonal plate boundaries, resulting in a zone of distributed deformation both within and away from the plate boundary. Previous studies of spatial distribution and source mechanism of earthquakes have shown that the N20E convergence between India and Tibet is partitioned into N-S underthrusting and E-W subduction of the Indian plate beneath the Himalaya and Burma micro-plate, respectively. However, south-west of the plate boundaries, the middle-to-lower crust deforms by strike-slip faulting in the Kopilli Fault Zone (KFZ); upper-to-mid crustal strike-slip faulting along the western Brahmaputra Valley; and lower crustal strike-slip faulting beneath the Bengal Basin. Strike-slip faulting in the KFZ also extends northwards beneath the Eastern Himalaya and southeastward beneath the Naga fold-thrust belt. In order to understand the role of these strike-slip faults in accommodating the GPS derived convergence across northeast India, we study the source mechanism of recent earthquakes using teleseismic (for earthquakes with Mw>5) and local (4.0

We observed a shallow (H<10 km) seismic activity containing both low frequency events (LFE) and regular earthquakes using a dense local seismic observation network installed recently at northeastern Japan. Significant recurring activities of deep (H>20 km) LFE have been monitored carefully around several subduction zones including Japan. Additionally, activities of shallow and deep LFE are also reported worldwide occurring beneath several potentially damaging active volcanoes. Deep (H>15 km) LFEs have also been observed to occur in inland areas not associated with active volcanoes or volcanism.The shallow seismic activity we observed was located on the Tsugaru Channel, between the Hokkaido and Honshu Islands of Japan. We installed a local seismic network consisting of 36 stations, called AS-net, in 2013 and 2014 and are monitoring shallow and deep seismic activities in the region. AS-net improved both the earthquake detection capability and the accuracy of hypocenter determination around this region significantly. The shallow activity of earthquakes and LFEs was observed at just ~2 km south of an AS-net station. The seismic record at the station revealed that the depth of this seismic activity was shallower than 10 km. From 2014 to 2017, 23 earthquakes and 19 or more LFEs were detected. Both earthquakes and LFEs swarmed individually and intermittently. Magnitudes of all events were less than two. Both types of events seemed to occur in the vicinity of one and other however this is not clear because of the limit of the accuracy of the hypocenter location. At ~10 km eastward of this activity, we observed another activity of deep (15~35 km) LFEs in an area without volcanos. Above this curious shallow seismic activity, Yamagata et al. (1989) pointed out the existence of a volcano based on their research about tephra layers over Hokkaido. They estimated that a tephra layer on tens of thousands of years ago could be from Zenigame Caldera, a depression (1~2 km in diameter) at seafloor of Tsugaru Channel. Because it is under the sea, it is not easy to survey, and no other studies about this potential volcano have been reported.In Japan, opposite to deep LFEs, shallow LFEs are only monitored around potentially damaging volcanoes. Outside of such regions, the presence of shallow LFE is not clear. Shallow LFE activity and its mechanism as well as its relation to volcanoes should be clarified. By doing so, we will be able to locate potential volcanic activity by monitoring shallow LFEs; even underwater. Additionally, we have recently started to install several temporal seismic observation stations in the area surrounding this seismic activity to investigate it in detail. AcknowledgementsThe AS-net has been constructed and is being managed thanks to financial contributions by Tohoku Electric Power Co., Inc., Electric Power Development Co., Ltd., Japan Nuclear Fuel Limited and Recyclable-Fuel Storage Company. [ABSTRACT FROM AUTHOR]

Okal et al. (2014) developed an algorithm, named TS formula, to estimate teleseismic tsunami amplitudes from the seismic moment of the parent earthquake, the azimuth between its fault strike and the receiver, and the epicentral distance; this method was based on least-square regressions over a large data set of numerical simulations (4650), using 10 source regions located in the main subduction zones of the Pacific Basin, and a range of four orders of magnitudes of seismic moment. We apply the inverse methodology to recover the seismic moment of historical earthquakes from their tsunami amplitudes. Inversion of the TS formula to recover the seismic moment was first tested on a set of real tsunami amplitudes, recorded in deep ocean basins by the DART buoy network (see http://www.ndbc.noaa.gov/dart.shtml). The tests were conducted on more than 250 measurements from 20 sources, and reproduced estimates of the seismic moment with an average precision of 0.22 logarithmic units. We point out that this method, which is independent of direct seismic measurements, provides the possibility to highlight non regular earthquakes like 'snappy' or 'slow' ones, and large unusual 'compact' sources like the 2011 Tohoku event. This methodology was then applied to historical events using a data set of 52 earthquakes and tsunamis, between 1854 and 2010, at 164 tide-gauge stations. For this study, the concept of deep sea measurement of tsunami amplitudes has been extended to amplitudes recorded by coastal sea-level stations (tide gauges), using a transfer function based on Green's law, which gives an estimation of the tsunami amplitude near the shore as a function of the deep sea measurements (see Reymond et al. (2011) and Jamelot & Reymond (2015)) and conversely. In this framework, we were able to reassess the seismic moment of some remarkable historical events including Peru-Chile 1868 and 1877, Ecuador 1906, Samoa 1917, Kermadec 1917, Tonga 1919, Chile 1922 and 1960, Alaska 1964..etc. The case of the tsunami earthquakes (Aleutian 1946, Nicaragua 1992, Java 1994 and 2006, Sumatra 2004, etc.) has been also considered.References:Okal, E.A., D. Reymond, and H. Hébert, From earthquake size to far-field tsunami amplitude: Development of a simple formula and application to DART buoy data, Geophys. J. Intl.,196, 340-356, 2014.Reymond, D., E.A. Okal, H. Hébert, and M. Bourdet, Rapid forecast of tsunami wave heights from a database of pre-computed simulations, and application during the 2011 Tohoku tsunami in French Polynesia, Geophys. Res. Letts., 39, (11), L11603, 6 pp., 2012.Jamelot, A., & Reymond, D.. New tsunami forecast tools for the French Polynesia tsunami warning system Part II: numerical modelling and tsunami height estimation. Pure and Applied Geophysics, 172(3-4), 805-819, 2015. [ABSTRACT FROM AUTHOR]

The moment released by postseismic deformation of a large earthquake is typically smaller than its coseismic rupture, which the coseismic rupture should have released the main portion of the accumulated strain. However, it is not uncommon that the accumulative postseismic displacement is larger than or equal to the coseismic displacement based on geodetic observations on some subduction zone environments. The rational cause of such a large postseismic deformation may require a weakly coupled interface near the source rupture to allow prolonged large postseismic slip. Nevertheless, both the coseismic and postseismic deformation of subduction zone earthquakes are mostly derived from inversion of inland geodetic observations, yet the ruptures often occurred at the interface located mostly offshore, which resulted in poor resolution on the slip distribution model. Thus, whether unusually large postseismic deformation may occur on certain subduction zone earthquakes, and if so, what is the main cause of the large postseismic slip require further investigations on earthquakes with such an attribute. The 2016 Mw 6.5 Meinong earthquake occurred beneath the fold-and-thrust belt in SW Taiwan. Based on GPS observations, the two-years accumulated postseismic moment (Mw = 6.60) is larger than the coseismic moment (Mw = 6.43) of the 2016 Meinong earthquake, which renders a paradigm for large postseismic deformation of an earthquake occurred beneath a frontal thrust wedge. We consider that the large postseismic deformation of the 2016 Meinong earthquake is likely activated by slip on the high pore-pressure detachment on the frontal thrust wedge. On the detachment the high pore-pressure fluid overlapped with ~5-km thick sediments, diffused slowly and caused prolonged postseismic deformation observed in SW Taiwan. Some unusually large postseismic deformation are likely caused by activations of localized high pore-pressure mud diapirism, including the one located above the coseismic rupture. [ABSTRACT FROM AUTHOR]

The rupture-zone averaged static stress drop in the 2011 M=9 Tohoku-oki earthquake wasless than 5 MPa according to all published slip models, but it caused a reversal of the state ofstress in most of the offshore forearc, from predominantly compressive to predominantlyextensional. In this work, we demonstrate that this stress reversal has the followingimportant geodynamic implications. (1) The reversal unequivocally indicates avery weak subduction megathrust. In order to reproduce this reversal in a finiteelement model of force balance that quantifies the effects of gravity and megathrustfriction, the effective coefficient of friction of the megathrust has to be about 0.032. Alower value would make the forearc too extensional before the earthquake, and ahigher value would make it too compressive after the earthquake. (2) The very weakmegathrust results in very low differential stresses in the upper plate, consistent withpreviously reported sensitivity of the state of stress to small perturbations. Applying thedynamic Coulomb wedge model, we demonstrate that the inner wedge, and byinference the nearshore-onshore forearc, is in a stable state throughout subductionearthquake cycles, far from failure. (3) The outer wedge is normally in a stable state butmay reach an extensionally or compressively critical state during an earthquake,depending on the behavior of the shallow megathrust. Gravitational collapse of theouter wedge is prevented by a finite strength of the underlying shallow megathrust.Therefore, complete stress drop (i.e., effective coefficient of friction decreases tozero) is unlikely to have happened over the main part of the shallow megathrustexcept for limited local areas. (4) The presence of permanent deformation such asearthquakes and active faulting in the overall stable forearc can be explained bystructural and stress heterogeneities. We propose that for the most part, the upper platenear a subduction zone is much below its yield stress and is an elastic body, butpermanent deformation can locally occur in areas of low strength, high stress, orhigh fluid pressure. (5) The concept of dynamic Coulomb wedge (Wang and Hu,2006, JGR) needs to be expanded to address timescales much beyond subductionearthquake cycles. The strength of the megathrust varies over geological timescales,affecting the stability of the outer wedge and its response to earthquake rupture. [ABSTRACT FROM AUTHOR]

The Nankai Trough Subduction Zone offshore Kii Peninsula, Japan has long been known for presence of great earthquakes (>8 magnitude). It has also been a subject to extensive scientific drilling and geophysical surveys performed by the International Ocean Drilling Programme (IODP) and JAMSTEC. As a result of these, there are currently several hypotheses about the origin, evolution and development of Nankai Trough. The geomechanical modelling is currently primarily reserved for commercial purpose. Well centric 1D, geological and geomechanical 3 and 4D models are regularly utilised by the oil and gas industry not only to optimise production / well placement in the existing fields, but also to avoid costly field/borehole failure. However, there is no research done on computer-based spatial geomechanical modelling of an active subduction zone. We present here preliminary results from Nankai geomechanical modelling. Initial results from constructed 1D wellbore stability models well predicted pore pressure, stress profiles and breakouts orientation in the boreholes. The 3D seismic inversion data from the Nankai Trough area was used to construct preliminary 3D geological model covering in total, area of 12 km x 25 km. Additionally, the 1D geomechanical models served later as calibration points for the subsequent 3D seismic derived geological and geomechanical models. The 3D seismic derived geological model was subsequently translated into the 3D geomechanical model which consists of core area surrounded by the far field stress zone. Physical and mechanical properties of the substrata were derived from seismic data and calibrated where available with measured log data. The geomechanical model covers 50 km x 80 km x 12 km area and consists of 3.3 million elements with 150 m x 150 m horizontal resolution and varying vertical resolution (60 – 150 m). 3D geomechanical model initial results show spatial distribution of regional tectonic stress, pore pressure, mechanical and physical properties such as Young's Moduli, Poisson's ratio, density, and others. Furthermore, the initial model shows distribution of fault criticality and slip tendencies along major fault planes. The geomechanical model of the Nankai Trough will act as a basement of an "early warning system " for the anticipated earthquake in the area. As for the future perspective, the model will be run/updated in Real-Time and with transient (future) time steps in order to predict stress magnitude and orientation, assessing fault criticality and slip tolerance. We hope it will give local population invaluable time to prepare and evacuate from hazardous areas saving lives and property. [ABSTRACT FROM AUTHOR]

Numerical models that capture the complex interplay of the earthquake rupture, the emitted seismic waves, and the resulting displacement at the Earth's surface shed light on the complex processes governing the generation of tsunamis. Physics-based dynamic rupture models can account for complex fault geometry and meets constraints from a variety of data sets. However, parameterization of the complex rheologies and geometries of subduction zones such as on- and off-fault stress and strength, rock cohesion, and friction, is not trivial, yet greatly influences earthquake dynamics and tsunami interaction. For example, the vast majority of seafloor uplift during tsunami genesis occurs within the accretionary wedge, which consists of unconsolidated sediments. Capturing sedimentary constitutive response in order to account for this critical deformation, as well as the non-linear effects of such material on rupture dynamics requires realistically constrained dynamic earthquake models that account for off-fault plastic deformation.We present an innovative coupling approach that constrains the complex initial conditions of elasto-plastic dynamic rupture models for earthquakes in subduction zones by a long-term seismo-thermo-mechanical model (STM, e.g. van Dinther et al., 2014, GRL). The STM model provides geometries for the slab, the accretionary wedge, and the spontaneously developing megathrust, which are all self-consistent with the stress field and rock strength. Such models evolve over millions to hundreds of years according to accretion and thrusting of ocean floor sediments, slab dehydration and fluid percolation, and cycles of megathrust and splay-fault slip episodes. We extract the STM model state prior the onset of slip and use it for the initial conditions in the earthquake model. Dynamic rupture at coseismic timescales coupled to seismic wave propagation is modelled with the software package SeisSol (www.seissol.org) using unstructured tetrahedral computational meshes optimally suited for the complex geometries of subduction zones. Our approach allows to investigate the importance of plasticity and dynamic wave-mediated stresses for tsunami genesis and wedge deformation and helps to evaluate the corresponding tsunami hazard. The across-models ported initial state prior earthquake rupture indicates that the uppermost part of the accretionary wedge is close to plastic failure. Energy dissipation due to the occurrence of co-seismic, off-fault plastic deformation halts rupture propagation at the transition to a shallow velocity-strengthening region while the comparable, but purely elastic simulation ruptures through this region. The corresponding vertical coseismic seafloor uplift is considerably increased by the occurrence of plastic strain within the accretionary wedge and its effect on the fault dynamics.The coupled, complex initial stress conditions and the heterogeneous material properties along the megathrust lead to pronounced rupture pulses that are smeared out when accounting for off-fault plasticity. The distribution of plastic strain after one dynamic rupture simulation resembles steeply dipping splay faults evolving over many seismic cycles within the wedge in STM models, suggesting high dynamic stress concentrations during co-seismic rupture. [ABSTRACT FROM AUTHOR]

After the 2011 Mw 9.0 Tohoku earthquake ruptured the subduction megathrust offshore Japan, many normal faulting earthquakes were observed in regions that previously hosted dominantly thrust faulting events. The rate of normal faulting seismicity has decreased since 2011, but is still significantly higher than before the Tohoku earthquake. Similar normal faulting earthquake sequences have been observed after other Mw 8.5+ megathrust events and in some cases appear to last for decades. In this study, we investigate whether the spatial distribution and temporal evolution of normal faulting seismicity in subduction zones can be explained by earthquake cycle deformation processes, particularly co-seismic and post-seismic stress changes. We develop a three-dimensional finite element model of the subduction zone offshore Honshu with an appropriate plate interface geometry and a rheology consisting of an elastic upper plate above a linear visco-elastic mantle wedge. We spin up the model using internally consistent earthquake cycles to generate pre-stresses; stresses build across the locked plate interface during the inter-seismic stage, driving co-seismic slip. Afterwards, the interface is re-locked as afterslip and bulk viscous relaxation occur, leading back into the subsequent inter-seismic stage. We incorporate the co-seismic slip from known megathrust earthquakes occurring before 2011 to better represent the slip deficit and state of stress at the time of the Tohoku earthquake. The model results show that the immediate co-seismic effect of a great megathrust earthquake is to cause approximately trench-normal extension of the upper plate (relative to the pre-earthquake stress field). This extension occurs above and landward of the rupture zone in the regions that currently host significant normal faulting earthquakes. Early afterslip occurring on the subduction interface down-dip of the rupture results in further extension of both the upper and subducting plates, promoting normal faulting earthquakes in these regions. Although the upper plate continues to move trenchward as a consequence of the bulk viscous relaxation of the deeper visco-elastic regions, this does not have a significant effect on the stress field in the shallower, elastic regions. The dominant process responsible for reestablishing the pre-earthquake state of stress is the plate convergence after the re-locking of the subduction megathrust. We find good agreement between the predicted spatial and temporal pattern of tensile stresses and the post-seismic normal faulting earthquakes in Japan. [ABSTRACT FROM AUTHOR]