We developed a 3-D, viscoelastic finite element model of the M9 2011 Tohoku-oki, Japan earthquake capable of predicting postseismic displacements due to viscoelastic relaxation and afterslip. We consider seismically inferred slab geometries associated with the Pacific and Philippine Sea Plate and a wide range of candidate viscoelastic rheologies. For each case, we invert for afterslip based on residual surface displacements (observed GPS minus that predicted due to viscoelastic relaxation) to develop combined viscoelastic relaxation and afterslip models.We are able to find a mechanical model that fully explains all observed geodetic on-land and seafloor horizontal and vertical postseismic displacements. We find that postseismic displacements are in about equal parts due to viscoelastic relaxation and afterslip, but their patterns are spatially distinct. Accurately predicting both horizontal and vertical on-land postseismic displacements requires a mantle wedge viscosity structure that is depth dependent, reflecting the manner in which temperature, pressure, and water content influence viscosity. No lateral heterogeneities within the mantle wedge viscosity structure beneath northern Honshu are required. Westward-directed postseismic seafloor displacements may be due flow via lowtemperature, plastic creep within the lower half of a Pacific lithosphere weakened by plate

Abstract:We investigated the effects of elastic heterogeneity on coseismic deformation associated with the 2011 Tohoku-oki earthquake, Japan, using a 3-D finite element model, incorporating the geometry of regional plate boundaries. Using a forward approach, we computed displacement fields for different elastic models with a given slip distribution. Three main structural models are considered to separate the effects of different kinds of heterogeneity: a homogeneous model, a two-layered model with crust-mantle stratification, and a crust-mantle layered model with a strong subducting slab. We observed two counteracting effects: (1) On large spatial scales, elastic layering with increasing rigidity with depth leads to a decrease in surface displacement. (2) An increase in rigidity from above the slab interface to below causes an increase in surface displacement, because the weaker hanging wall deforms to accommodate coseismic slip. Results for slip inversions associated with the Tohoku-oki earthquake show that slip patterns are modified when comparing homogeneous and heterogeneous models. However, the maximum slip only changes slightly: It increases from 38.5 m in the homogeneous to 39.6 m in the layered case and decreases to 37.3 m when slabs are introduced. Potency, i.e., the product of slip and fault area, changes accordingly. Layering leads to inferred slip distributions that are broader and deeper compared to the homogeneous case, particularly to the south of the overall slip maximum. The introduction of a strong slab leads to a reduction in slip around the slip maximum near the trench. We also find that details of the vertical deformation patterns for heterogeneous models are sensitive to the Poisson's ratio. While elastic heterogeneity does therefore not have a dramatic effect on bulk quantities such as inferred potency, the mechanical response of a layered medium with a slab does lead to a systematically modified slip response, and such effects may bias studies of mega-thrust earthquakes.

Studies of mechanical responses of the Earth crust to large earthquakes can provide us with unique insights into the processes of stress buildup and release. As a complement to geodetic methods that derive crustal strain dynamics from surface observations (e.g., GPS, InSAR), noise‐based seismic velocity monitoring directly probes the mechanical state of the crust, at depth and continuously in time. We investigate the responses of the crust to the Mw 9.0, 2011 Tohoku‐oki earthquake. In addition to the Hi‐net short‐period sensors, we use Hi‐net tiltmeters as long‐period seismometers (8–50 s) to sample the crust below 5 km in depth. The spatial distribution of the strong velocity decreases at short periods appears to be limited to the region of strong ground shaking induced by the 2011 Tohoku‐oki earthquake, while the long‐period velocity changes correlate well with the modeled static strain induced by viscoelastic relaxation and afterslip at depth. Amplitudes of coseismic velocity changes decrease with increasing depth. The temporal evolution of velocity changes in different period bands shows that the maximum drops in the velocity at long periods are delayed in time with respect to the occurrence of the Tohoku‐oki earthquake. The inversion of seismic velocity changes at depth illustrates how S wave velocities evolve down to 40 km at a regional scale after a major earthquake.