Tomography of the source zone of the great 2011 Tohoku earthquake

The 2011 Mw 9.0 Tohoku-oki megathrust earthquake occurred along the subduction interface between the Pacific and Okhotsk plates and produced devastating tsunami and damage in NE Japan. A key unresolved issue is why extraordinarily large shallow slip occurred near the Japan Trench and why high-frequency radiation was weak in that region while being strong down-dip. Competing explanations invoke weak clay-rich materials and/or dynamic fault weakening, as well as upper-plate structural controls and along-strike variations in pore fluids and lithology. Prior tomography suggested a high-velocity asperity near the hypocenter, but resolution near the trench was limited due to lack of near-field observations. With the new S-net ocean-bottom seismic network, the study aims to image short-wavelength structural heterogeneity along the megathrust, test whether rupture nucleated at a boundary between contrasting velocity anomalies, and clarify how along-dip structural variations relate to slip and frequency-dependent radiation.

Multiple studies after the 2011 event documented large near-trench slip and complex rupture dynamics, including high-frequency radiation localized down-dip and smoother shallow slip. Drilling (JFAST) recovered weak, clay-rich (smectite) materials and low coseismic shear stress at shallow depths, supporting fault weakness near the trench. Alternative perspectives emphasize upper-plate structure controlling interplate coupling and seismogenic behavior, and along-strike differences in pore fluids and lithology influencing slip modes. Prior regional tomography imaged a high-velocity anomaly at the hypocenter, interpreted as an asperity, but lacked sufficient near-trench resolution to fully resolve shallow structures. Depth-dependent frictional and thermal models have been used to explain high-frequency radiation patterns and rupture dynamics, but require detailed structural constraints that were previously unavailable offshore.

Data: P-wave arrival times from 3757 local earthquakes (2446 Hi-net events beneath the onshore network and 1311 S-net events offshore) spanning June 2002–December 2018, recorded by 480 Hi-net land stations and 120 S-net OBS stations. From high-quality vertical-component S-net seismograms, P-wave picks were made with estimated accuracy of 0.05–0.15 s. Events with hypocentral uncertainty <3 km were retained, yielding 109,890 P-wave times. Tomographic inversion: The Zhao et al. seismic tomography method was employed to determine a detailed 3-D P-wave velocity (Vp) model, incorporating known geometries of the Conrad, Moho, and the upper boundary of the subducting Pacific slab (UBP). A 3-D grid with 0.33° lateral spacing and 5 km vertical spacing near the UBP was used. Vp perturbations at grid nodes relative to a starting model (constructed with reference to active-source surveys of the Tohoku forearc and an initial +4% Vp perturbation within the slab) were solved by damped, smoothed least squares. A 3-D ray tracer computed theoretical travel times and ray paths, accounting for station elevations and topography. Earthquakes were relocated during inversion. Derivatives and residual images: After inversion, 3-D Vp perturbations were referenced to an average 1-D velocity model. A 2-D Vp image of the overriding Okhotsk plate above the UBP was derived by comparing vertical travel times through the 1-D and 3-D models (ΔVp/Vp = (f1 − f2)/f1). Residual Vp images were obtained via spectral averaging along UBP contours, analogous to residual topography/gravity calculations, and subtracting the averaged component to emphasize lateral heterogeneity along the megathrust and in the overriding plate. Sensitivity and robustness: Inversions were repeated with altered depths of discontinuities (Conrad, Moho, UBP), with different 1-D reference models, and with a starting model including a low-V oceanic crust atop the slab; all yielded similar patterns. A joint inversion including previous arrival-time data produced a consistent model. Resolution tests: Checkerboard resolution tests with Gaussian noise (σ = 0.1 s; range −0.2 to +0.2 s) indicated lateral resolution of ~33 km and vertical resolution of ~5 km around the UBP. Additional synthetic tests confirmed recovery of key features. Interpretation overlays: Results were compared with residual topography and gravity maps, coseismic slip distributions, high-frequency radiation and strong motion areas, tremor/VLFE catalogs, slow-slip events, and postseismic GPS-A displacement fields to interpret structural controls on rupture and radiation.

• The 2011 mainshock hypocenter lies at the boundary between a down-dip high-V anomaly and an up-dip low-V anomaly along the megathrust.
• A pronounced shallow (<20 km) low-V anomaly extends from the mainshock area to the Japan Trench and coincides with the region of largest coseismic slip and with weak high-frequency radiation.
• Down-dip (20–50 km) high-V anomalies beneath the Pacific coast correlate with strong ground motions and strong high-frequency radiation, while coseismic slip there was relatively modest.
• Along-strike variations: From the forearc segment boundary (FSB) to ~39°N, high Vp aligns with positive residual topography and gravity; north of ~39°N, Vp, residual topography, and gravity are negative, matching the Sanriku-oki low-seismicity region.
• The shallow low-V anomaly likely reflects low-rigidity materials (e.g., clay-rich sediments, elevated dynamic pore fluid pressure), consistent with JFAST findings, enabling large shallow slip and suppressing high-frequency radiation.
• Prior large M≥7.0 megathrust earthquakes ruptured deeper high-V/high-Q regions interpreted as strong asperities or granite batholiths; their ruptures likely did not propagate up-dip into the shallow low-V zone. In 2011, strong inertial rupture from the high-V region overcame this barrier, producing extreme shallow slip.
• The shallow low-V zone may also relate to a preceding slow slip event (M ~7.0) in the source region.
• Postseismic GPS-A shows landward motions in the large-slip area predominantly within low-V regions; exceptions (stations G03, G05) in high-V patches show small trench-ward motions, indicating spatial variations in frictional/elastic properties and lithology.
• Compared with previous studies, inclusion of S-net data resolves along-dip Vp variations near the trench, supporting depth-dependent frictional behavior inferred from source spectra and radiation patterns.
• Tomographic resolution is ~33 km laterally and ~5 km vertically near the UBP; robustness checks (alternative discontinuity depths, 1-D models, low-V oceanic crust) do not alter the main patterns.

The high-resolution Vp images directly link structural heterogeneity along the Tohoku megathrust to rupture mechanics of the 2011 event. The nucleation at the high-V/low-V boundary indicates that strong, down-dip asperities can drive rupture upward into weak, shallow materials. In the 2011 earthquake, the shallow low-V zone lacked sufficient strength to arrest the dynamically propagating rupture, yielding extraordinary near-trench slip and a deficiency of high-frequency radiation. Conversely, strong high-frequency energy and damaging ground motions originated within deeper high-V domains, consistent with higher rigidity and rapid moment release. Along-strike variations in Vp align with residual gravity/topography and seismicity patterns (e.g., Sanriku-oki low-seismicity region), reinforcing the role of both upper- and lower-plate structure in controlling interplate coupling, slip behavior (including slow slip and VLFEs/tremor), and radiation. The integration of S-net enables near-trench imaging that reconciles large shallow slip with weak high-frequency radiation, resolving prior controversies stemming from limited offshore resolution.

This study delivers a high-resolution 3-D P-wave tomography of the Tohoku forearc megathrust leveraging S-net and Hi-net data. It demonstrates that: (1) the 2011 Mw 9.0 Tohoku-oki rupture initiated at a boundary between down-dip high-V and up-dip low-V anomalies; (2) large near-trench slip and weak high-frequency radiation are associated with a shallow low-V, low-rigidity zone; and (3) strong high-frequency radiation originates from deeper high-V regions. These results unify observations of slip distribution, radiation, and postseismic deformation under a structural framework, improving understanding of megathrust seismogenesis and informing seismic hazard assessment. Future work could integrate multiparameter imaging (e.g., S-wave/attenuation tomography, anisotropy), pore-fluid and temperature constraints, and time-dependent (4-D) monitoring to capture transient changes in fault properties, as well as extend similar near-field imaging to other subduction margins.

• The S-net was deployed after the 2011 event; images reflect post-earthquake structure. Although frictional heating and coseismic changes are thought to be minor, any small rupture-induced velocity changes below the model’s resolution may not be captured.
• Spatial resolution (~33 km lateral, ~5 km vertical near the UBP) may miss finer-scale heterogeneities and thin fault-zone structures.
• The model uses P-wave data only; complementary S-wave and attenuation (Q) constraints could better characterize rigidity and fluids.
• Residual Vp along the megathrust can be influenced by heterogeneities in the overriding plate despite efforts to isolate along-fault variations.
• Assumed geometries of major discontinuities (Conrad, Moho, UBP) carry uncertainties, though sensitivity tests suggest limited impact on main features.