SUMMARY
During the last decades, many numerical models have been developed to investigate the conditions for seismic and aseismic slip. Those models explore the behaviour of frictional faults, ...embedded in either elastic or inelastic media and submitted to a far field loading (seismic cycle models), or initial stresses (single dynamic rupture models). Those initial conditions impact both on-fault and off-fault dynamics. Because of the sparsity of direct measurements of fault stresses, modellers have to make assumptions about these initial conditions. To this day, Anderson’s theory is the only framework that can be used to link fault generation and reactivation to the 3-D stress field. In this work, we look at the role of the 3-D stress field in modelling a 2-D strike-slip fault under plane-strain conditions. We show that setting up an incorrect initial stress field, based on Anderson’s theory, can lead to underestimation of the damage zone width by up to a factor of six, for the studied cases. Moreover, because of the interactions between fault slip and off-fault deformation, initial stress field influences the rupture propagation. Our study emphasizes the need to set up the correct initial 3-D stress field, even in 2-D numerical simulations.
The Longitudinal Valley Fault (LVF) in Taiwan is creeping at shallow depth along its southern half, where it is bounded by the Lichi Mélange. By contrast, the northern segment of the LVF is locked ...where it is bounded by forearc sedimentary and volcanoclastic formations. Structural and petrographic investigations show that the Lichi Mélange most probably formed as a result of internal deformation of the forearc when the continental shelf of South China collided with the Luzon arc as a result of the subduction of the South China Sea beneath the Philippine Sea Plate. The forearc formations constitute the protolith of the Lichi Mélange. It seems improbable that the mechanical properties of the minerals of the matrix (illite, chorite, kaolinite) in themselves explain the aseismic behavior of the LVF. Microstructural investigations show that deformation within the fault zone must have resulted from a combination of frictional sliding at grain boundaries, cataclasis (responsible for grain size comminution) and pressure solution creep (responsible for the development of the scaly foliation and favored by the mixing of soluble and insoluble minerals). The microstructure of the gouge formed in the Lichi Mélange favors effective pressure solution creep, which inhibits strain-weakening brittle mechanisms and is probably responsible for the dominantly aseismic mode of fault slip. Since the Lichi Mélange is analogous to any unlithified subduction mélanges, this study sheds light on the mechanisms which favor aseismic creep on subduction megathrust.
•Lithology controls the mode of slip on the Longitudinal Valley Fault (LVF).•The collision Lichi Melange clearly promotes aseismic creep.•The Lichi Melange likely formed as a result of internal deformation of the forearc.•Fault zone fabric, rather than minerals content, control the creep on the LVF.•Frictional sliding and pressure-solution creep accommodate the creep on the LVF.
The Longitudinal Valley Fault (LVF, Taiwan) is a fast‐slipping fault (∼5 cm/yr), which exhibits both seismic and aseismic slip. Geodetic and seismological observations (1992–2010) were used to infer ...the temporal evolution of fault slip. This kinematic model is used here to estimate spatial variations of steady state velocity dependence of fault friction and to develop a simplified fully dynamic rate‐and‐state model of the LVF. Based on the postseismic slip, we estimate that the rate‐and‐state parameter
(a−b)σ̄ decreases from ∼1.2 MPa near the surface to near velocity neutral at 19 km depth. The inferred (a − b) values are consistent with the laboratory measurements on clay‐rich fault gouges comparable to the Lichi Mélange, which borders the LVF. The dynamic model that incorporates the obtained
(a−b)σ̄ estimates as well as a velocity‐weakening patch with tuned rate‐and‐state properties produces a sequence of earthquakes with some realistic diversity and a spatiotemporal pattern of seismic and aseismic slip similar to that inferred from the kinematic modeling. The larger events have moment magnitude (Mw ∼6.7) similar to the 2003 Chenkung earthquake, with a range of smaller events. The model parameterization allows reproducing partial overlap of seismic and aseismic slip before the earthquake but cannot reproduce the significant postseismic slip observed in the previously locked patch. We discuss factors that can improve the dynamic model in that regard, including the possibility of temporal variations in (a − b) due to shear heating. Such calibrated dynamic models can be used to reconcile field observations, kinematic analysis, and laboratory experiments and assess fault behavior.
Key Points
Time series of afterslip and dynamic modeling are used to estimate friction properties
Inferred frictional properties are consistent with laboratory values for the effective normal stress distribution with fluid overpressure
Fully dynamic modeling is used to explore how field observations, geodetic analysis, and laboratory experiments can be reconciled
Delineating seismic and aseismic slip on faults allows the exploration of the complex relationship between these different modes of slip. Further, quantifying them helps in the assessment of ...seismogenic potential. We present a distributed slip model of the rate of aseismic slip along the Leyte island section of the Philippine Fault and make comparisons with the extent of seismic slip from the latest significant earthquakes in July 2017. We derived both coseismic and aseismic slip distributions from kinematic inversions of synthetic aperture radar interferometric (InSAR) observations using a probabilistic (Bayesian) framework. Velocity maps from stacking and time series analysis of ALOS interferograms spanning 4 years (2007–2011) show a step change in ground deformation right on the fault. Inverting for slip at depth reveals along‐strike variations of aseismic slip rate over ∼100 km. Aseismic slip on the surface reaches rates of more than 3 cm/yr, equivalent to the long‐term slip rate (3.3 ± 0.2 cm/yr). Over the same period, a 20‐km segment in Tongonan appears to be locked. This segment ruptured in Mw 6.5 and Mw 5.8 earthquakes on 6 and 10 July 2017, respectively, as constrained by Sentinel‐1 and ALOS‐2 InSAR data. Seismic slip appears to be restricted within the Tongonan segment, with up to 152 ± 21 cm of left‐lateral displacement. The slip budget and complementarity between the extents of interseismic and coseismic slip suggest that a seismogenic asperity exists in Tongonan. The presence of active hydrothermal systems and rate‐strengthening materials provide physical conditions that can promote aseismic slip.
Plain Language Summary
In July 2017, the Philippine Fault on Leyte island produced a damaging earthquake that was unexpected. Previous studies that tracked the movement of the Earth's crust have suggested that the fault in Leyte does not accumulate strain, and hence stress, as the fault primarily slides stably and slowly. This is in contrast to earthquake‐generating faults that go on a cycle of building up stress for a long period of time, due to the crust staying stuck (or locked) on the fault, and then releasing the stored stress as an earthquake. With analysis of high‐resolution satellite radar, we now show that while 100 km of the fault indeed appears to be moving slowly and gradually, a 20‐km‐long portion in the middle of the fault appears to be locked. Our observations and models of the July 2017 earthquake confirm that this same section of the fault alone moved then. The geometry of the fault, the abundance of fluids, and the presence of specific types of minerals may all play a role on how the fault behaves. This work is the first of its kind on the Philippine Fault, highlighting the importance of exploring relatively less well studied faults and their physical properties.
Key Points
We present the first probabilistic model of interseismic aseismic slip throughout seismogenic depth along 100 km of the Philippine Fault
Joint inversion for anthropogenic subsidence and fault deformation reveals downdip locking from the surface on a 20‐km section in Tongonan
Extent of seismic rupture of the July 2017 magnitude 6.5 and 5.8 earthquakes correlates with geometric complexity and lack of slip deficit
Physics‐based numerical simulations of earthquakes and slow slip, coupled with field observations and laboratory experiments, can, in principle, be used to determine fault properties and potential ...fault behaviors. Because of the computational cost of simulating inertial wave‐mediated effects, their representation is often simplified. The quasi‐dynamic (QD) approach approximately accounts for inertial effects through a radiation damping term. We compare QD and fully dynamic (FD) simulations by exploring the long‐term behavior of rate‐and‐state fault models with and without additional weakening during seismic slip. The models incorporate a velocity‐strengthening (VS) patch in a velocity‐weakening (VW) zone, to consider rupture interaction with a slip‐inhibiting heterogeneity. Without additional weakening, the QD and FD approaches generate qualitatively similar slip patterns with quantitative differences, such as slower slip velocities and rupture speeds during earthquakes and more propensity for rupture arrest at the VS patch in the QD cases. Simulations with additional coseismic weakening produce qualitatively different patterns of earthquakes, with near‐periodic pulse‐like events in the FD simulations and much larger crack‐like events accompanied by smaller events in the QD simulations. This is because the FD simulations with additional weakening allow earthquake rupture to propagate at a much lower level of prestress than the QD simulations. The resulting much larger ruptures in the QD simulations are more likely to propagate through the VS patch, unlike for the cases with no additional weakening. Overall, the QD approach should be used with caution, as the QD simulation results could drastically differ from the true response of the physical model considered.
Key Points
With enhanced weakening, fully dynamic runs produce pulse‐like ruptures
In the same cases, quasi‐dynamic runs yield much larger crack‐like ruptures
The rupture‐mode differences dramatically affect long‐term fault behavior
The Longitudinal Valley Fault (LVF) in eastern Taiwan is a high slip rate fault (about 5 cm/yr), which exhibits both seismic and aseismic slip. Deformation of anthropogenic features shows that ...aseismic creep accounts for a significant fraction of fault slip near the surface, whereas a fraction of the slip is also seismic, since this fault has produced large earthquakes with five Mw>6.8 events in 1951 and 2003. In this study, we analyze a dense set of geodetic and seismological data around the LVF, including campaign mode Global Positioning System(GPS) measurements, time series of daily solutions for continuous GPS stations (cGPS), leveling data, and accelerometric records of the 2003 Chenkung earthquake. To enhance the spatial resolution provided by these data, we complement them with interferometric synthetic aperture radar (InSAR) measurements produced from a series of Advanced Land Observing Satellite images processed using a persistent scatterer technique. The combined data set covers the entire LVF and spans the period from 1992 to 2010. We invert this data to infer the temporal evolution of fault slip at depth using the Principal Component Analysis‐based Inversion Method. This technique allows the joint inversion of diverse data, taking the advantage of the spatial resolution given by the InSAR measurements and the temporal resolution afforded by the cGPS data. We find that (1) seismic slip during the 2003 Chengkung earthquake occurred on a fault patch which had remained partially locked in the interseismic period, (2) the seismic rupture propagated partially into a zone of shallow aseismic interseismic creep but failed to reach the surface, and (3) that aseismic afterslip occurred around the area that ruptured seismically. We find consistency between geodetic and seismological constraints on the partitioning between seismic and aseismic creep. About 80–90% of slip on the southern section of LVF in the 0–26 km, seismogenic depth range, is actually aseismic. We infer that the clay‐rich Lichi Mélange is the key factor promoting aseismic creep at shallow depth.
Key Points
About 80% of slip on the LVF in the seismogenic depth range is aseismic
Seismic slip during the 2003 earthquake occurred on a previously locked patch
Aseismic afterslip occurred around the area that ruptured seismically
Using borehole strainmeters, we detected a 13‐day long slow slip event in 2011 with Mw = 5.45 on the Longitudinal Valley Fault, Taiwan. It has been likely promoted by the significant Coulomb stress ...changes (∼0.5–1 MPa) imparted by a combination of coseismic and postseismic slip of the 2003 Chengkung earthquake. Using kinematic slip models, we infer that the slow event has accommodated about 15%–35% of the slip deficit accumulated from 1997 to 2011 in its source region. Further, we find that inherited state of stress may have controlled the slow event southward propagation. We also highlight a spatiotemporal correlation between the slow event and a cluster of repeating microearthquakes, suggesting a complex interplay between seismic and aseismic processes on the fault, where transient aseismic regions are adjacent to highly coupled zones.
Plain Language Summary
Tectonic faults display a large range of slip behaviors, ranging from fast slip (earthquakes) to episodic or continuous aseismic slow slip. Episodic aseismic slip events (typically called 'slow slip events') are now widely observed in active regions and play an important role in stress redistribution in the Earth's crust. The Longitudinal Valley Fault is the most active fault in Taiwan, hosting large to destructive earthquakes. However, very little is known about the presence and the role of slow slip events on the fault deformation. We document a 13‐day long slow slip event with a magnitude of 5.45, which represe'nts the largest event detected on the fault to date. We demonstrate that the slow event was likely encouraged by stress accumulation due to the 2003 Chengkung earthquake. Monitoring and characterizing the sources of aseismic slip is fundamental to identify areas with high seismic hazard on the fault and to shed some light on the interactions between seismic and aseismic processes.
Key Points
We document a slow slip event on the Longitudinal Valley Fault (Taiwan) with magnitude of 5.45, which represents the largest detection to date
Slow slip has likely been induced by a combination of coseismic and postseismic stress changes (>0.5 MPa) from the 2003 Chengkung mainshock
Inherited state of stress on the fault has likely controlled the slip distribution and slip mode