2009 AGU Fall Meeting abstracts

B. Enescu, T. Takeda, K. Obara, S. Sekine, W. Suzuki, Y. Asano, Y. Yukutake,
The 2008 Iwate-Miyagi Inland Earthquake (Mw6.9): The Distribution and Focal Mechanism Solution of Aftershocks in relation to the Crustal Structure, S44B-06

Abstract. Several large earthquakes, including the 2008 Iwate-Miyagi Inland earthquake (Mw6.9), occurred recently in the High Strain Rate Zone of Japan. In the framework of a large national project of observations and research in this region, we investigate the earthquake distribution, crustal structure and aftershock focal mechanisms in the area of the 2008 Iwate-Miyagi sequence. The detailed aftershock distribution, spanning roughly from south to north, shows a conjugate fault plane structure (i.e., planes dipping towards NW and SE, respectively) that is particularly clear in the northern part of the aftershock area. However, such a conjugate fault structure is also apparent south of the mainshock. To clarify the relation between seismicity and faults, we further investigate the earthquake focal mechanisms. While most aftershocks have a focal mechanism consistent with that of the mainshock, some others are strike slip events but with the same orientation (NW-SE) of the compression axis as that of the mainshock. Moreover, there is a distinct group of thrust events located south from the mainshock that have fault planes dipping to north or south. Our results indicate that at least in the south there is one fault structure dipping to NW (likely the mainshock fault) and several clusters of earthquakes with different focal mechanisms that cannot be easily associated with a fault. Taking advantage of the NIED dense linear seismic array located south of the mainshock, we investigate the detailed crustal structure on a profile positioned along the linear array (and perpendicular to the fault). We use the P- and S-wave arrival times recorded by the NIED array stations, as well as other stations in the area, in a tomographic inversion using the Lotos software (Koulakov, 2009).  The velocity structure under the NIED array shows two relatively low-velocity, shallow structures, separated by a higher velocity body that continues in the deeper part. The presence of such a higher velocity structure is also highlighted by the analysis of seismograms recorded at array stations. The main fault (dipping towards NW), which can be delineated from the distribution of aftershocks, is located at the transition between lower and higher velocities. To explain the distribution and the various focal mechanism solutions of the aftershocks, we estimated the Coulomb stress changes caused by the mainshock in the area. We used the slip distribution of Suzuki et al. (2009) and computed the static stress changes on optimum-oriented faults. The stress changes can explain well the most important characteristics of the aftershock distribution. We are presently computing the stress changes on the two conjugate fault planes of aftershocks for which focal mechanisms are available. To eliminate possible artifacts in the deltaCFF distribution close to the fault, we remove from the analysis the aftershocks that are closer than 2km from the fault. Preliminary results indicate that most of the aftershocks occurred in areas of positive Coulomb stress changes.

S. Toda, B. Enescu, R. Stein, CoulombExpress: Automated near-realtime forecasts of earthquake-induced stress transfer and expected seismicity rate changes, S22C-02 (INVITED)

Abstract. A principal tenet of the Coulomb hypothesis is that stress increases promote, and decreases inhibit, fault failure. In support of such a simple hypothesis, a growing number of studies have found that seismicity rates climb where the stress increases and fall where the stress drops. However, they are all evaluated retrospectively, which may permit unintentionial bias to enter into data interpretation. Further, retrospective tests do not contribute to earthquake disaster mitigation. In particular, the probability rate for triggered seismicity is highest immediately after a mainshock, as suggested by rate/state friction. Thus, to make the stress-based earthquake forecasting rapidly available and to permit objective prospective testing, we have developed CoulombExpress, an automatic stress calculation system, which uses near real-time information, such as earthquake magnitude, location, depth, and its rapid moment tensor solution. Two versions of CoulombExpress, “CoulombExpress Global” and “CoulombExpress Regional” are in progress.
1) CoulombExpress Global: The system quickly computes the Coulomb stress change caused by an M≥6 around the globe. It automatically accesses the USGS National Earthquake Information Center parameters. In order of time, the system uses the NEIC W-phase, central, and body-wave moment tensor solutions, when available, to make the two-nodal-plane source fault models using the empirical scaling relations of Wells and Coppersmith [1994]. The stress changes are resolved on receiver faults parallel to the sources and also on both nodal planes of the nearby 1967-2005 Global CMT earthquakes, as stand-in's for active faults. The color-coded displays allow the viewer to grasp where and by how much the off-fault aftershocks might become active. Keeping long-term records of our results in an official archive will allow evaluators to examine rigorously the forecasting skills of our model.
2) CoulombExpress Regional: This is a forecasting system that computes not only Coulomb stresses but also the space-time seismicity rate evolution (Toda et al., JGR, 2005) for a particular region or country where high-quality local data are available. It has been implemented for M≥4 in Japan first (with California to follow). Moderate earthquakes are implemented as point sources to generate local stress perturbations. Since the local catalog is available, we analyze the background rate of seismicity, aftershock durations, and regional G-R parameters needed to translate the calculated stress changes into expected seismicity rate changes. We have built matrices of assumed receiver fault planes (strike, dip, and rake) based on focal mechanism data and structural controls. Since stress increase should immediately increase the seismicity rate, we select and use for further calculations the maximum stress changes throughout the various receiver planes and calculation depths. We will submit this regional model to CSEP (Collabratory for the Study of Earthquake Predictability), Japan, starting at the end of 2009.

Z. Peng, B. Enescu, P. Zhao, S. Hainzl, Detecting early aftershocks in California and Japan based on a matched filter technique, S54A-06

Abstract. A large shallow earthquake is immediately followed by numerous aftershocks with a significant portion missing in existing earthquake catalogs, mainly due to masking of the mainshock coda and overlapping arrivals. Recovering these missing early aftershocks is important for understanding the physical mechanisms of earthquake triggering, and tracking post-seismic deformation around the mainshock rupture zone. An effective way of detecting and locating those missing early aftershocks is the matched filter technique. It utilizes waveforms or travel time information of existing events as a template, or “matched filter”, to search for similar patterns in the continuous recordings as a suggestive of an event. Here we apply this technique to systematically detect early aftershocks of moderate-size events in California and Japan. In the first study, we use waveforms of relocated events along the Parkfield section of the San Andreas Fault (SAF) as templates, and scan through continuous waveforms for 3 days around the 2004 Mw6.0 Parkfield earthquake to detect missing aftershocks. We identify 11 times more aftershocks than reported in the standard Northern California Seismic Network (NCSN) catalog. The newly detected aftershocks show clear migration in both along-strike and down-dip directions with logarithmic time since the mainshock, consistent with the numerical simulations on expansions of aftershocks caused by propagating afterslip. The cumulative number of early aftershocks increases linearly with postseismic deformation in the first 2 days, suggesting that early aftershocks could be driven by significant afterslip along the SAF induced by the Parkfield mainshock. In the second study, we apply the same technique to detect and locate early aftershocks of the 2008 Mw6.9 Iwate-Miyagi Nairiku earthquake in northeastern Honshu, Japan. The existing aftershock locations show complex spatial distributions with many occurring away from the mainshock fault plane. Well-resolved focal mechanism solutions of aftershocks show considerable variations compared to the mainshock focal mechanism. However, most of the off-fault events occur in areas of positive Coulomb stress changes. A small coefficient of friction (~0.2) provides in general a better consistency between stress changes and aftershock locations. After detecting the missing events, our next step is to examine the spatio-temporal migration patterns to quantify whether early aftershocks were driven by coseismic stress changes, afterslip, fluid migration, or a combination of these factors.

T. Takeda, H. Sato, K. Obara, B. Enescu,
NW-SE Trending Fault-Segmentation Boundaries in the High-Strain-Rate Zone of Japan, T53B-1578

Abstract. In the high-strain-rate zone of Japan, located in the eastern margin of the Sea of Japan, large earthquakes of magnitudes up to 7.5 have often took place. Recently, two M6.8 earthquakes occurred in 2004 and 2007, showing reverse fault mechanisms, with a NW-SE compression. The aftershock distributions of both mainshocks are oriented in a NE-SW direction and seem to have a sharp cut-off boundary at the northeastern edges by a common NW-SE line. This implies the existence of a structural boundary trending in the NW-SE direction. Since revealing structural constraints on fault segmentation is important for strong motion prediction and earthquake hazard assessment, we investigate here the boundary that controls the fault segmentation, using accurately located hypocenters and their focal mechanisms. We investigate earthquakes which occurred in the target area from 2001 to 2004 since later data is dominated by aftershocks that can obscure subtle structural features. The hypocentral distribution shows three major linear alignments with a strike of a NW-SE direction. Two of them are located near an estimated epicenter of a large earthquake in 1828, and the other is located in the northern part of the aftershock area of the 2004 event. To assess the accurate spatial distribution of hypocenters, we relocated these earthquake alignments by using the double-difference method (Waldhauser and Ellsworth, 2000) and classified the focal mechanism using the criterion of Frohlich (1992). As a result, two alignments near the 1828 earthquake appear more concentrated and show subvertical planar distributions, 2.5-3 km wide x 3-5 km deep. The average focal mechanism has a P-axis E-W oriented. These observations indicate that the both alignments have seismic activity along NW-SE trending strike-slip faults. On the other hand, the alignment near the 2004 earthquake separates into two main groups and does not show a clear NW-SE trending strike-slip fault since it includes not only strike-slip type but also reverse type of focal mechanisms. However the existence of strike-slip mechanisms with N-S compression is supporting the existence of a NW-SE trending strike-slip fault. These results indicate the existence of NW-SE trending strike-slip faults, that is, structural boundaries having the same orientation. Sato (1994) suggests that in this area a rift structure, formed as a result of normal faulting when the Sea of Japan opened in the Miocene, is presently reactivated as a reverse fault by stress field inversion. Therefore the seismic activity would be controlled by the rift structure and be bounded by a transfer fault with a NW-SE strike. For example, there is a NW-SE trending lateral ramp between the Shitada hill and the Niitsu anticline, which is bounding the source fault of the 1828 earthquake. It is worth noting that the alignment is located at this boundary and distributes along a steep gravity anomaly gradient. There are many linear alignments other than those previously mentioned. Linear alignment analysis using various geophysical methods could result in a more accurate detection of the boundaries that constrain fault segmentation in the high-strain-rate zone of Japan.

M. Cocco, S. Hainzl, J. Woessner, B. Enescu, F. Catalli, A. Lombardi, Sensitivity study of forecasted aftershock seismicity based on Coulomb stress calculation and rate- and state-dependent frictional response, S22C-05 (INVITED)

It is nowadays well established that both Coulomb stress perturbations and the rate- and state-dependent frictional response of fault populations are needed to model the spatial and temporal evolution of seismicity. This represents the most popular physics-based approach to forecast the rate of earthquake production and its performances have to be verified with respect to alternative statistical methods. Despite the numerous applications of Coulomb stress interactions, a rigorous validation of the forecasting capabilities is still missing. In this work, we use the Dieterich (1994) physics-based approach to simulate the spatio-temporal evolution of seismicity caused by stress changes applied to an infinite population of nucleating patches modelled through a rate- and state-dependent friction law. According to this model, seismicity rate changes depend on the amplitude of stress perturbation, the physical constitutive properties of faults (represented by the parameter Aσ), the stressing rate and the background seismicity rate of the study area. In order to apply this model in a predictive manner, we need to understand the variability of input physical model parameters and their correlations. We first discuss the impact of uncertainties in model parameters and, in particular, in computed coseismic stress perturbations on the seismicity rate changes forecasted through the frictional model. We aim to understand how the variability of Coulomb stress changes affects the correlation between predicted and observed changes in the rate of earthquake production. We use the aftershock activity following the 1992 M 7.3 Landers (California) earthquake as one of our case studies. We analyze the variability of stress changes resulting from the use of different published slip distributions. We find that the standard deviation of the uncertainty is of the same size as the absolute stress change and that their ratio, the coefficient of variation (CV), is approximately constant in space. Second, we demonstrate that all model parameters are strongly correlated for physical and statistical reasons. We discuss this correlation emphasizing that the estimations of the background seismicity rate, stressing rate and Aσ parameter are strongly correlated to reproduce the observed aftershock productivity. Our results demonstrate the impact of these model parameters on the Omori-like aftershock decay (the c-value and the productivity of the Omori law), implying a p-value smaller or equal to 1. Finally, we discuss an optimal strategy to constrain model parameters for near-real time forecasts. Our case studies demonstrate that accounting for realistic uncertainties in stress changes as well as for the correlation among model parameters strongly improves the forecasting performances, although the original deterministic approach is converted into a statistical method.