G21D-01 INVITED
Multi-year Postseismic Deformation in South Iceland Dominated by Viscoelastic Relaxation, but not Afterslip
Several time-dependent processes follow large earthquakes, such as pore-fluid flow, postseismic slip (afterslip), and viscous relaxation of the lower crust and upper mantle. These processes cause additional stress changes in the surrounding crust and therefore influence the occurrence of aftershocks and other neighboring earthquakes. However, it has proven difficult to distinguish between these processes based on field observations, as models of the different processes can predict similar ground displacements and more than one process may be acting at the same time. The postseismic deformation observed following two South Iceland earthquakes (magnitude 6.5) in June 2000 has provided a unique opportunity to work on this problem, as different postseismic processes appear to have operated at different timescales. The early phase (a few months) of the postseismic deformation was dominated by localized displacements (within 5-10 km from the faults), showing quadrants of uplift and subsidence around the two faults. This pattern of deformation was caused by postseismic groundwater pressure changes during the 2-3 months when coseismic pore-pressure changes were dissipating. Annual GPS measurements revealed another pattern of post-seismic deformation with a much slower decay than the poroelastic relaxation. However, both afterslip and viscoelastic relaxation models are able to predict the GPS observations equally well, making it impossible to reject one of these two potential explanations based on these data alone. Additional information about the multi-year deformation was derived from times-series analysis of multiple satellite radar interferograms (InSAR) using radar data from 2000-2005. The InSAR images were multi-looked and spatially unwrapped, and then the line-of-sight displacement for each multi-looked pixel was estimated at each acquisition date. The deformation pattern revealed by the time-series analysis is consistent with models of viscoelastic relaxation, but clearly inconsistent with afterslip model predictions. The results also show decaying deformation rate with a rather poorly constrained exponential decay time-constant of 500-1000 days, suggesting a low upper-mantle viscosity of circa 1×1018 Pa s. The results provide a stark contrast to results focusing on some other moderate-size earthquakes, e.g.\ the 2004 Parkfield earthquake, where afterslip was found to be the dominating postseismic mechanism.
G21D-02 INVITED
Long-lasting afterslip due to a magnitude-5 earthquake at the Chaman fault, Afghanistan
The Chaman fault system forms a prominent ~900-km-long left-lateral transform plate boundary between the Indian and Eurasian plates in Afghanistan and Pakistan. Whereas geological investigations on the fault indicate a long term slip rate of 20-40 mm/yr, historical records show an absence of large earthquakes at a 300- 400 km segment of the Chaman Fault, suggesting that this segment of the fault is either locked or slipping aseismically. Here we show satellite radar interferometry (InSAR) data that revealed that afterslip signal following an earthquake of magnitude 5.0 lasted for more than a year and at least quadrupled its moment release. We detected no contiguous surface slip before the earthquake during the 1.5 years sampled by our data. This finding of long-lasting afterslip suggests that the plate motion along the Chaman Fault is at least accommodated by slow slip episodes following moderate earthquakes in the fault.
G21D-03
Evolution of Deformation, Pore Pressure, and Coulomb Stress Following the M9 Sumatra- Andaman Earthquake.
The M9 Sumatra-Andaman Earthquake of 2004 ruptured the interface of the subducting Indo-Australian plate and overriding Burma microplate. Near-field GPS measurements of the coseismic deformation are on the order of several meters. This deformation induced a devastating tsunami and generated transient stress and pore pressure changes that triggered numerous aftershocks, including the M8.7 Nias earthquake. Finite element models (FEMs) are uniquely capable of simulating the coseismic load and induced evolution of postseismic deformation, pore pressure, and Coulomb stress; while simultaneously honoring the known geologic complexity of the subduction zone. We construct FEMs that simulate deformation of the earthquake for a three-dimensional problem domain partitioned to account for the distribution of material properties of the subducting slab, mantle wedge, forearc, volcanic arc, and backarc. The coseismic slip distribution is estimated from the near-field GPS data via standard inverse methods and FEM-generated Green's functions. Forward models, driven by this slip distribution, predict the evolution of poroelastic and viscoelastic deformation, stress, and pore pressure following the earthquake. Preliminary results suggest poroelastic deformation may be up to several tens-of-centimeters in offshore regions, although predicted poroelastic displacements for near-field GPS sites are generally a few centimeters. Initial pore pressure magnitudes, due to the load of the coseismic slip, exceed 1 MPa in the near- field region. Predicted postseismic pore pressure recovery correlates to the observed spatial and temporal distribution of aftershock swarms, in accord with the poroelastic formulation of Coulomb failure theory. Although the predicted poroelastic displacements are resolvable by GPS measurements in the near-field region, more than a meter of viscoelastic deformation is expected for near-field GPS sites over the next decade.
G21D-04
Far-Reaching Transient Motions After Mojave Earthquakes Require Broad Mantle Flow Beneath a Strong Crust
The variation of strength with depth of continental lithosphere continues to be much debated: is a weak viscously deforming lower crustal layer sandwiched between strong upper-crustal and upper-mantle layers or is the upper mantle simply weaker than the lower crust? It also continues to be debated whether continental deformation at depth occurs along discrete, strain-weakened shear zones or is broadly distributed in viscously deforming layers. Because of the difficulty to directly determine viscoelastic strength and the degree of localization of deformation, there is no consensus on either issue. A useful approach for inferring the strength of the lithosphere is to utilize earthquakes as large rock deformation experiments where coseismic stress changes induce a variety of postseismic responses, including afterslip, poroelastic rebound, and viscoelastic relaxation. However, given the limited spatial and temporal resolution of postseismic observations, it has proven difficult to sort out the relative postseismic contributions of each mechanism, let alone to determine how viscosity varies as a function of depth. Here we utilize 7 years of observed postseismic transient surface displacements following the 1999 M7.1 Hector Mine quake in southern California as recorded at an extensive array of continuous GPS stations that demonstrate a previously unrecognized broad pattern of transient deformation throughout southern California and into Nevada, more than 200 km from the epicenter. This first of a kind far-field view (over 4 rupture lengths) of a postseismic deformation field following a strike-slip earthquake, allows us to much more uniquely determine the mechanism responsible for this broad deformation pattern. We use a 3-D viscoelastic finite element model of the region that incorporates both the Hector Mine rupture and the nearby 1992 M7.3 Landers earthquake that is still influencing transient displacements. We considered a wide range of possible viscoelastic structures ranging from viscosity being uniform with depth below 20 km depth, to structures where viscosity decreases rapidly with depth (as might be expected due to increasing temperatures). The best models are those where the viscosity below 40 km depth is an order of magnitude or more lower than the viscosity of mantle above and two orders of magnitude less than the viscosity of the lower crust. We are able to rule out significant contributions of localized afterslip below the seismogenic zone to far-field postseismic deformation, as well as a significant contributions from poroelastic rebound. Thus, we are able to determine that the pattern of broad postseismic transients following these Mojave Desert earthquakes can only be explained by viscoelastic flow in a region of the mantle 100s of km wide and below a depth of 40 km. This result enables two robust conclusions regarding the nature of lithospheric strength in this region to be reached: the mantle is weaker than the lower crust, and flow occurs over a wide region of mantle as opposed to within a narrow shear zone beneath the fault. Considering the broad region of mantle sampled by this study, it is possible that such a model may be appropriate for much of western North America.
G21D-05 INVITED
Influence of velocity strengthening regions on the seismic cycle: Implications for postseismic slip and the generation of slow earthquakes.
Postseismic slip is thought to be located in velocity strengthening (VS) regions which promotes stable slip. Deep afterslip is observed following large earthquakes, and occurs in such a VS zone, the fragile-ductile transition zone also named brittle creep fault zone (BCFZ). Superficial afterslip is sometimes observed and may represent the relaxation of VS patches embedded in the seismogenic fault zone (SFZ). Slow earthquakes are now widely reported and appear to be located in the upper part of the BCFZ. However, the link between those events and postseismic slip is not obvious and we present here a model that may reconcile those observations. We study the response of a rate and state strengthening fault to static Coulomb stress changes. If the size of the fault is larger than a characteristic length Lb, a slip transient could be generated, Lb being analogous to the nucleation length of a velocity weakening fault. A positive Coulomb stress step induces a slip transient followed by a relaxation similar to the case of a pure velocity strengthening region. This slip transient may escape from detection because it occurs right after the onset of perturbation and could be misinterpreted as part of the main perturbation (e.g., rupture of the mainshock). Unlike what one could expect, a negative perturbation in Coulomb stress also generates a slip transient. The main difference with the positive Coulomb stress step is that the transients corresponding to unloading steps are of smaller amplitude and occurs way after the onset of perturbations, and could be misinterpreted as a ‘spontaneous' creep burst. Postseismic slip corresponds to the response of a VS zone to a large scale positive Coulomb stress change while spontaneous slow earthquakes may correspond to the response of a VS zone to a localized negative Coulomb stress change. The implications of our results in terms of geodetic slip inversions and numerical modelling of the seismic cycle will also be briefly discussed. Influence of velocity strengthening regions on the seismic cycle: Implications for postseismic slip and the generation of slow earthquakes.
G21D-06 INVITED
3-D FEM derived elastic Green's functions for the coseismic and postseismic deformation of the 2005 Mw 8.7 Nias-Simeulue, Sumatra earthquake
We adopt a finite element method to investigate the effect of 3D variation of material properties in the subduction zone using the coseismic and postseismic deformation of the 2005 Mw 8.7 Nias-Simeulue, Sumatra earthquake. In this study, we construct a simple subduction model using the mesh generation software, Cubit, developed by Sandia National Lab., USA. The fine element code, PyLith, is used to compute Green's function responses due to unit dislocation. To validate the FEM results, we compare simple modeling results between FEM and Okada analytic solutions. Preliminary analysis shows the difference of surface displacement calculated from homogeneous and 3-D heterogeneous material models can be as large as 20%. Ignoring the spatial variation of material properties leads to systematic misfits in surface horizontal and vertical displacements. Inverting fault slip distributions with assumption of a homogeneous, isotropic earth, results in biased fault slip distributions and fault geometries in our synthetic tests. For the coseismic and postseismic deformation of the Nias-Simeulu earthquake, we infer a model with less up-dip slip when using a more realistic 3-D elastic structure. We find the spatial variation of coseismic and postseismic slip distribution in various models remains similar, while integrated potency along depth in 3-D elastic models shift along the down-dip direction comparing with that in an elastic half-space model. The down-dip shift of maximum integrated potency along the depth depends on the material contrast in the fault zone. In addition, the impact of 3-D fault geometry seems to play a more important role comparing to the effect of heterogeneity.
G21D-07
Models of Postseismic Deformation Following the 2003 Tokachi-oki Earthquake and Interseismic Deformation in Northern Japan
Transient deformation is frequently observed following large earthquakes on subduction interfaces, and is commonly interpreted to result from aseismic slip on the megathrust fault-zone (MFZ). Kinematic models typically reveal that postseismic afterslip is spatially anti-correlated with coseismic slip, indicating that the majority of afterslip occurs in non-seismogenic portions of the MFZ. Since the spatio-temporal distribution of afterslip is dictated by the MFZ rheology, geodetic observations of postseismic deformation have the potential to place constraints on the MFZ rheology. In general, postseismic deformation can rarely be considered independent of all previous ruptures, as transient postseismic deformation affects the interseismic deformation late in a seismic cycle. Hence, simultaneously modeling postseismic deformation with the steady deformation observed in the years prior to a megathrust earthquake, may further constrain the rheology. We recently developed a 3D model consisting of a finite fault, embedded in a half-space, with imposed ruptures on regions of the fault, and stress- dependent postseismic and interseismic slip on areas that are not locked during the interseismic periods (either stable or conditionally stable regions of the MFZ). This model is based on the formulation of Rice (1993); however, we only solve for aseismic slip, and impose coseismic slip in prescribed regions. The rheology of the MFZ can be linear viscous, stress-dependent viscous, or rate- and state-dependent friction. Our model is applicable to any fault geometry, but in this presentation we only consider the 2003 M8 Tokachi-oki earthquake on the southern Kuril trench megathrust in northern Japan.
G21D-08
Nucleation of afterslip following the 2003 Tokachi-oki earthquake
We model early postseismic deformation of the 2003 Tokachi-oki earthquake using high-rate GPS posititions calculated every twelve minutes for the first 36 hours following the main shock. The data show little or no motion at GPS sites immediately after the earthquake with sudden acceleration at about 1.5 hours after the mainshock followed by gradual deceleration. The sudden acceleration is nearly coincident in time with the largest aftershock (M = 7.1). We interpret the GPS signal during this first 36 hours as a response to afterslip on the subduction interface. The coincidence of the largest aftershock and the onset of afterslip may indicate that afterslip and aftershocks are governed by the same physical process. We model the time series of the postseismic deformation using a spring-slider model that is assumed to obey a rate- and state-dependent friction law. A nonlinear inversion scheme is devised to estimate rate-state friction parameters with the spring-slider model and the GPS time series. We find that the 1.5 hour delay in afterslip places tight constraints on the critical slip distance, L, which we estimate to be on the order of 10-4 m. This value is at the very upper end of laboratory measurements of L which are of order 10-6 to 10-4 m. The estimated value of a-b, an indicator of frictional stability, is on the order of 10-3, consistent with other estimates using postseismic data and rate-dependent friction laws.