In Chapter 4, we analyze and further explore the connections between geodetic observatories of fault blocking and the behavior of seismicity at the base of the seismogenic zone. Given the representative fault structure and geometry in the region, simple elastic modeling suggests that the peak of associated subsurface fault slip coincides with maximum surface uplift and decreases channelward.
Introduction
The assumed fault geometry and elastic structure directly influence the parameterization of the source model and the prediction of surface deformation due to subsurface fault slip. In the following sections, we begin with tsunami observations for the 2011 Tohoku-Oki earthquake, our parameterization of the seabed deformation field, and forward modeling of tsunami excitation and propagation.
Tsunami observations and modeling
All waveforms are calculated to start at zero displacement at the onset time of the earthquake. These surveys form a unique data set in the near field of the source with good azimuthal coverage.
Bayesian inference of seafloor deformation model
In the least-squares case, the conventional least-squares optimization approach is a special case of the Bayesian approach. However, we can always explore the PDF of local averages of the posterior solutions to learn the resolution scales inherent to the problem.
Synthetic scenarios
We use the entire mesh for the quasi-static problem, while adopting a near-source subset of the seabed mesh for the kinematic problem, to reduce the number of free parameters and computational demand. Such a bias is reinforced for the case of a more dispersed tsunami wave generated at the trench (see Fig. 2.20).
Applications to the 2011 Tohoku-Oki earthquake
As a posteriori validation of our models, we evaluate the fit of the posterior data as well as the prediction for the later part of the tsunami waveforms that are not included in the inversion in the figure. For spatially averaging models, the discrepancy between the observations and the mean fit to the data is increased, but still within uncertainty.
Uncertainty
Averaging Scale
Seafloor Uplift
Discussion
Sato et al., 2011) we need to account for the effect of the horizontal displacements in areas with steep bathymetric slopes (Tanioka and Satake, 1996). A cross-sectional profile of the effective seafloor uplift in the model and available seafloor measurements.
A) Trench-normal profiles of fault slip
B) Effect of realistic bathymetry (M1 vs. M2)
D) Considering horizontal displacements
C) Effect of elastic inhomogeneity (M2 vs. M3)
Conclusion
We investigate formulations of Cp based on empirical models and physical considerations that errors in tsunami modeling could be characterized as stochastic deviations in the frequency dispersion relations. Based on analysis of the theory of linear wave (Airy wave), a typical frequency dispersion relation can be derived (e.g. Kundu et al., 2012). Here we perform such waveform corrections in a stochastic sense based on a perturbation approach.
Synthetic waveforms generated from a reference model can be perturbed based on the deviation in dispersion characteristics for each random realization.
Introduction
We adopted a semi-analytical Bayesian approach to the fault slip problem, similar to the approach developed in Chapter 2, which explicitly includes model prediction errors in tsunami modeling and elastic structure. In particular, we examine the importance of tsunami dissipation and elastic structure at inversions and the importance of accounting for these sources of uncertainty. With improved Green's functions and consistent fault models, we aim to constrain the slip distribution of the Maule earthquake, especially the updip profile of the fault slip and the predicted seafloor displacement.
Data and methods
Nodes on other boundaries are assigned with zero slip and not included in the inversion, thereby assuming that they are significantly away from the source region. We use InSAR (interferometric synthetic aperture radar) and GPS (Global Positioning System) measurements based on a composite data set used in previous studies of coseismic slip models (Vigny et al., 2011; Lin et al., 2013). The InSAR measurements are based on image acquisitions from the Japanese L-band Advanced Land Observation Satellite (ALOS), including 6 descending tracks (420 and wideswath 422) and 10 ascending tracks (110 to 119), which were processed and sampled in Lin et al. . .
The surface deformation due to slip on the fault is calculated with a 1D layered structure extracted from 2D models by spatial averaging over the nearby source region (Tassara et al., 2006; Haberland et al., 2009).
Santiago
Depth (km)
Results and discussions
To the south, the shallow slip in the geodetic model is counteracted by including the tsunami data. We note some backward slip to the southernmost end of the geodetic and joint models, mostly within the fault ellipses. Considering Cp actually avoids overfitting the vertical components while still reproducing the large-scale properties well, e.g. the transition from ascent to descent.
Only the first 40 minutes of the waveforms (to the left of the blue lines) are used in the inversions.
Conclusion
Predicted displacements on the surface grid within 30 km of the profile are plotted as circles, with the uncertainties shown by the error bars and their distances from the profile in color. The seismic potential, calculated from our preferred posterior solutions, is shown as a function of distance from the trench. The bump at 120 km to the trench is due to the fault dip variations during the strike in the 3D geometry.
Although the two earthquakes occur under different tectonic conditions, with differences in the distances from the trench to the coast and the locations of the hinge line relative to the coast, such similar uplift patterns near the trench could indicate potentially common failure mechanisms of shallow subduction zones. during large megathrust earthquakes.
Slip (m)
Main Text
The concentrated stress induced at the locked-creeping transition promotes microseismicity at the base of the SZ. 150 to ∼300 years after the previous major seismic events, the locked-creeping transition on these segments is still below the base of the seismogenic zone. The model based on teleseismic waveforms (Robinson et al., 2006) provided similar constraints: fault slip reaches 10 km in most places and about 20 km in the area of highest slip.
The length of the pre-mainshock and postseismic periods are the same as shown in the figures on the right.
Model with SZ-confined rupture (M1)
Model with deeper rupture (M2)
Introduction
For well-instrumented systems of major strike-slip faults, e.g. the San Andreas and San Jac faults in southern California (Figure 5.1), the threshold depths of seismicity (Dseis) determined from a high-resolution earthquake catalog (Nazareth and Hauksson, 2004; Hauksson, 2011) and the locking depth of interseismic faults (Dgeod) obtained from surface geodetic measurements (Smith-Konter et al., 2011; Tong et al., 2014), provide independent estimates of the thickness of the seismogenic zone for different fault segments. With induced earthquake stress perturbations, models that include viscoelasticity and velocity-state friction (e.g., Li and Rice, 1987; Hetland et al., 2010; Takeuchi and Fialko, 2012) have a more physically realistic fault and/or bulk . behavior. Fault geometry is from the Southern California Earthquake Center (SCEC) Fault Model (CFM) (Plesch et al., 2007).
Blue bands represent the range of fault closure depths within 1σ uncertainty determined from surface geodetic measurements (Smith-Konter et al., 2011).
Model setup
The long-term fault slip budget consists of uniform coseismic slip in the seismogenic zone and interseismic fault creep below. Drup during the earthquakes and Dgeod in the interseismic period coincide and define the base of the seismogenic zone. Seismicity is not relevant in such kinematic models, and Dseis is often assumed to delineate the base of the seismogenic zone and must therefore be the same as or close to Dlock.
Building on Chapter 4, here we extend the set of fault models and further investigate and characterize the differences between the models in terms of geodetic and seismic observables arising from long-term fault behavior.
Seismic and aseismic behavior in long-term fault models
Model M2 has earthquake faults confined to SZ, with a sharp transition from the fully locked (ISC equal to 1) to fully creeping (ISC equal to 0) regions in the early interseismic periods. Column (iv) shows an illustrative view of the depth distribution of VW fault spots, with the along-strike dimension compressed on the horizontal axis. We find that this depth can be estimated reasonably well from the depth at which the slip rate just exceeds a threshold value in a relative sense (e.g., V = 0.1 Vmax, where Vmax is the maximum slip rate on the fault).
In addition, we also consider an empirical measure of fault locking depth (D0.5C), where the fault slip rate is equal to half the plate velocity, which, as we show below, provides a rough estimate of Dgeod.
Inferring the geodetic locking depths on major faults
Therefore, Dlock and Dgeod reflect different aspects of fault behavior and may differ, especially in the presence of deeper penetration of coseismic slip and significant afterslip. In the case of the smallest (a−b) (Fig. 5.9F), rapid postseismic fault slip reaches a wider area around the SZ and decreases in amplitude rapidly thereafter, leading to a larger slip deficit zone even in the earlier interseismic period (50 years after the earthquake). , and differences in the derived Dgeod and its time evolution. Dlock, the true locking depth, defined by the depth at which the fault slip rate reaches 0.1Vmax, is shown in red for several time windows in the interseismic period.
Regardless of the depth of the coseismic slip, postseismic slip occurs in the creeping VS region below, expanding in space yet decreasing in amplitude, eventually leading to a slip deficit zone that slips below plate velocity.
Seismicity and fault heterogeneity at the transitional depth
The slip budget is distributed over the co-, post- and inter-seismic fault periods. The long-term fault behavior consists of frequent microseismicity (Mw ∼4) in the interseismic periods of occasional larger events (Mw 6-6.5) (Fig. 5.13). So perhaps Dseis would depend largely on the combined effects of statistical properties of the fault heterogeneity distribution and the fault slip velocity profiles in the LC transition, while Dlock only delineates the upper limit of the seismicity band.
Stacked profiles of the fault-parallel slip rate (A) sampled from the 2D surface velocity field across the fault (B) during the interseismic period of model M5 are used to invert the virtual Dgeod for the segment.
Conclusion
Time evolution of the apparent geodetic closure depth for this segment and the depth of seismicity are shown in (C1), with the magnitudes and overall depth distribution of seismicity in (C2) and (C3), respectively, suggesting that Dseis deeper than Dgeod in this model. This relationship is generally consistent with observations from the Carrizo and Coachella segments on the San Andreas fault, which have hosted large earthquakes in the past, and the discrepancy may be more significant considering the general underestimation of Dgeod becomes due to ignoring elastic heterogeneity. Although we have focused on continental strike-slip faults, the proposed relationship between seismicity, fault closure and large earthquake slip should qualitatively apply to the deeper portion of the subduction zone megathrust.
Chester (1998), Ultracataclasite structure and frictional processes of the Punchbowl fault, San Andreas system, California, Tectonophysics.
