Interrogation of the Megathrust Zone in the Tohoku-Oki Seismic Region by Waveform Complexity: Intraslab Earthquake Rupture and Reactivation of Subducted Normal Faults

Abstract

Results from the 2011 Mw 9.1 Tohoku-Oki megathrust earthquake display a complex rupture pattern, with most of the high-frequency energy radiated from the downdip edge of the seismogenic zone and very little from the large shallow rupture. Current seismic results of smaller earthquakes in this region are confusing due to disagreements among event catalogs on both the event locations (>30 km horizontally) and mechanisms. Here we present an in-depth study of a series of intraslab earthquakes that occurred in a localized region near the downdip edge of the main shock. We explore the validity of 1D velocity model and refine earthquake source parameters for selected key events by performing broadband waveform modeling combining regional networks. These refined source parameters are then used to calibrate paths and further simulate secondary source properties, such as rupture directivity and fault dimension. Calculation of stress changes caused by the main event indicate that the region where these intraslab events occurred are prone to thrust events. This group of intraslab earthquakes suggest the reactivation of a subducted normal fault, and are potentially useful in enhancing our understanding on the downdip shear zone and large outer-rise events.

Appendix: Generating Empirical Green’s Functions

To generate empirical Green’s Functions using E2, based on the generalized ray theory (Helmberger 1983), the characteristic travel time of a generalized ray in a layered half-space is given by:

$$t_{0} = p_{0} r + \mathop \sum \limits_{i} \eta_{i} d_{i}$$

(1)

where r is the source-receiver distance, η _i the vertical slowness of the ray in each layer, and d _i the vertical distance of the ray segment in each layer. For two very close sources, the paths to the receiver will be highly similar in shape and differ only by a small time shift dt ₀ . This time variance (dt ₀ ) can be approximated by using Taylor series expansion for t ₀ around the position of the point source (r, h).

$$\partial t_{0} = \frac{{\partial t_{0} }}{\partial r}{\text{d}}r + \frac{{\partial t_{0} }}{\partial h}{\text{d}}h$$

(2)

∂t ₀ /∂r is essentially p ₀ , which is treated as a constant here. ∂t ₀ /∂h = −εη _s , where ε = 1 for down-going rays and ε = −1 for up-going rays. η _s , which equals [(1/ν ² _s ) − p ² ₀ ]^1/2, is the vertical slowness of the ray p ₀ in the source region. The velocity in the source region is represented by ν _s . p ₀ and η _s in this study are numerical estimation from synthetics generated at different depths based on the 1D velocity model used for CAP inversion.

Here we assume E1 to be a finite-fault earthquake which is 45 times larger than E2 in moment magnitude. Thus, we discretize the rupture region into a line of 45 elements, each represented as an E2 point source. The total response (R (t)) of E1 at the receiver can then be represented by a summation of the 45 rays, each properly lagged in time according to the relative position from the reference point source.

$$R\left( t \right) = \mathop \sum \limits_{i = 1}^{45} E2_{i} (t - {\text{d}}t_{0i} )$$

(3)

Since we assume four rupture scenarios, diagonally northward and southward on the two auxiliary focal planes, there is a set of four empirical Green’s Functions R (t) generated, which is then compared with the data obtained to determine rupture directivity.