G12A-01 INVITED 10:20h
Postseismic Deformation: Different mechanisms in different times and places.
Improved understanding of postseismic deformation may elucidate time dependent stress transfer and triggered seismicity following large earthquakes. Afterslip, distributed viscoelastic flow, and poroelastic relaxation alter crustal stress and pore pressure distributions and in many cases lead to distinctive surface deformation patterns. Delayed triggering, due to rate and state dependent friction, on the other hand need not lead to detectable surface deformation. Postseismic deformation recorded following the 1999 ChiChi, Taiwan, 2003 Tokachi-Oki, Japan, and 2000 south Iceland earthquakes can be used to test for the effects of these processes. Horizontal displacements of 10 cm accumulated in the first year following the Chi-Chi quake. These are best explained with continued slip on the Chelungpu fault (Hsu et al, G.R.L. 2002). Inversions indicate the afterslip was roughly localized in a ring around the locus of maximum coseismic slip. The observed displacement pattern is inconsistent with predictions from viscoelastic and poroelastic models. Viscoelastic relaxation of the lower crust produces shortening of the hanging wall instead of the observed extension. The fully drained poroelastic response predicts deformation concentrated near the fault ends, which was not observed. Fully time dependent calculations, however, are still required because poroelastic displacements need not be monotonic. Afterslip following the M 8 Tokachi Oki earthquake is also localized around the area of high mainshock slip (Miyazaki et al, GRL, 2004). Surprisingly, the slip is not located downdip of the mainshock, but along strike of the source region. This indicates that the transient deformation is not caused by deceleration of the earthquake instability, but rather by stress increases due to the mainshock. A major question is whether intermediate depth afterslip following the Tokachi Oki and ChiChi earthquakes occurs in stable (steady state velocity strengthening) areas which will never initiate fast earthquake slip, or on unstable (velocity weakening) parts of the fault that slipped in a stable fashion following the earthquakes due to the pre-earthquake fault state and stress. Resolution of this question has important implications for future earthquakes in these areas. Postseismic deformation in the month following the South Iceland earthquakes was clearly detected by InSAR. The spatial and temporal patterns are inconsistent with both afterslip and viscoelastic deformation. The InSAR data are, however, well explained by a rapid poroelastic response. This was confirmed by water level changes with the same spatial and temporal scales as the deformation (Jansson et al, Nature, 2003). The decay of aftershocks is substantially longer that the poroelastic relaxation, suggesting that poroelastic effects do not control the timing of triggered earthquakes. The InsAR data, however, are insensitive to pore pressure changes at the depths of most aftershocks. At longer time scales, other processes dominate the observed deformation (Arnadottir, this meeting). An inescapable conclusion of these studies is that different physical processes dominate postseismic deformation in different geologic environments at different time scales.
G12A-02 INVITED 10:35h
Transient Rheology of the Upper Mantle: Evidence From two Case Studies
Geodetic observations indicate that both the M7.1 1999 Hector Mine and M7.9 Denali earthquakes were followed by a brief period ($\sim 0.1$ year) of rapid postseismic movements and a longer period ($\geq \sim 2$ years) of slower postseismic movements still elevated well above pre-earthquake velocities. The involvement of GPS sites more than 50 km and 100 km from the Hector Mine and Denali ruptures, respectively, suggests a deep source for the observed postseismic deformation. Viscoelastic relaxation of the regional lower crust and/or mantle governed by either a nonlinear or transient rheology (represented by a Burghers body) would be consistent with the observed time dependence. Detailed examination of 3-component GPS time series covering the first $\sim 2$ years after each earthquake indicates that relaxation of the regional upper mantle governed by a Burghers body rheology is consistent with the temporal and spatial patterns of relaxation observed after both earthquakes. A relaxation model involving a linear Maxwell rheology in the upper mantle fails to explain the same patterns. These results suggest that a combination of anelastic and viscous deformation mechanisms in the uppermost mantle is generally important in shaping the Earth's response to large earthquakes.
G12A-03 INVITED 10:50h
Deep Lithospheric Mantle and Heterogeneous Crustal Flow Following the 2002 Denali, Alaska Earthquake
A large earthquake can be utilized as a rock deformation experiment in which sudden stress changes trigger an observable postseismic response that can be used to infer rheological properties of the lithosphere. Such experiments seek to understand the relative contributions of postseismic viscous flow, poroelastic rebound, and afterslip, the depth at which such mechanisms are most prominently active, and the nature of the strain rate-to-stress relationship (linear versus non-linear). Here we study the response of the Alaskan lithosphere to the 2002 M7.9 Denali, Alaska earthquake. We utilize 2 years of campaign and continuous GPS observations of surface deformation to constrain finite element models of the various candidate postseismic mechanisms. We find that observed far-field deformation (100 km or more from the rupture surface) can only be explained by processes extending deep into the lithosphere. This broadly distributed deformation can be explained by viscous flow at a depth of 50 to 100 km or by deep afterslip on a downdip extension of the Denali fault, dominantly at a depth of 45 to 60 km. The problem with such deep afterslip, other than the fact that it results from a kinematic optimization with no constraining physics, is that it implies a large separation between the depth of coseismic slip (0 to 20 km depth) and the depth of afterslip, which may be difficult to physically explain. Deep flow that satisfies long wavelength surface deformation cannot, however, fully explain observed surface deformation within 50 km of the fault. Near-field displacements require additional flow near the base of the crust (24 to 30 km deep) or an afterslip distribution that includes significant (more than 1 m) shallow slip. Poroelastic rebound appears to play a minor role in the postseismic deformations. Viscous models that only vary flow parameters with depth cannot explain significant deformation observed to occur to the south of the Denali fault near its junction with the Totschunda fault. We find that this deformation can, however, be explained by localized mid-crustal flow in a region which coincides with anomalously high seismic velocities. This region could correspond to a mafic (high velocity) pluton that may be viscously weak because of partial melts or a high water content, though such an interpretation is only conjecture at this stage in our analysis. Time series data from continuous GPS stations indicate that the relationship between postseismic strain rate and stress is not linear (i.e. not Newtonian). We find that the decay of observed surface deformations with time can be explained by viscous relaxation involving a power-law in which strain rate is proportional to stress raised to a power between 2 and 3. This is below the experimentally derived powerlaw exponent of about 3.5 for dislocation creep of olivine, the mineral most likely controlling deformation within the mantle. This may indicate that viscous flow in the mantle beneath the Denali region may be comprised of a combination of dislocation and diffusion creep or contributions from other shallow mechanisms.
G12A-04 INVITED 11:05h
Spatio-temporal Signatures of Post-seismic Relaxation due to the Mojave Desert (S. California) Earthquakes from InSAR and GPS Data, With Implications for the Driving Mechanisms
The 1992 $M_w7.3$ Landers and 1999 $M_w7.3$ Hector Mine earthquakes in the Mojave desert (southern California) produced some of the best ever documented post-seismic deformation transients. I use a well-populated catalog of the Synthetic Aperture Radar (SAR) data from the ERS 1 and 2 satellites that includes more than 200 interferable acquisitions, and time series from several tens of Global Positioning System (GPS) stations to investigate the spatial and temporal characteristics of the post-seismic deformation following the two earthquakes. The post-Landers geodetic data reveal a deformation transient with a characteristic relaxation time of several years. The stacked InSAR data show prominent lobes of post-seismic relaxation with a characteristic wavelength of 30-40 km, and amplitude of $\sim$5 cm along the satellite line of sight (LOS). The LOS displacements exhibit high gradients across the surface trace of the Landers rupture, implying a shallow origin of the relaxation process. The spatio-temporal signatures of the post-Landers InSAR data are consistent with percolation of pore fluids in the upper crust in response to the co-seismic stress changes imposed by the Landers earthquake. The observed time dependence of post-seismic response implies effective hydraulic diffusivity of the order of $0.1-1$ m$^2$/s, consistent with laboratory, borehole, and field measuremets. This conclusion is confirmed by data from several boreholes in the near field of the Landers rupture that show gradual changes in the water level over a time period of 2-3 years following the Landers earthquake. Pore fluid flow, however, underpredicts horizontal displacements in the far field, suggesting a contribution of another deformation mechanism, such as the deep afterslip or visco-elasto-plastic relaxation. Models assuming pure visco-elastic behavior of the lower crust and upper mantle imply that the deformation is still essentially transient several years following the earthquakes, and predict that significant post-seismic deformation is yet to occur in the future.
http://sioviz.ucsd.edu/~fialko/res_land.html
G12A-05 11:20h
Kinematic and dynamic models of deep afterslip following the 1999 Izmit-Duzce earthquake sequence
Kinematic inversions of GPS velocity data from the first four years after the Izmit-Düzce earthquake sequence indicate that most postseismic slip has occured below two high coseismic slip patches, separated by a central coseismic slip patch that extends to a depth of at least 22 km. Integrated over four years, up to 2.5 meters of afterslip has occurred in the mid- to lower crust. Deep afterslip does not completely fill the gaps between the high coseismic slip patches, suggesting that interseismic slip in the mid-crust may occur at a significant fraction of the relative plate rate. This is consistent with localized interseismic strain around the North Anatolian Fault Zone (e.g., McClusky et al., 2000) and with earthquake-cycle models incorporating frictional fault zone rheology (Lapusta et al., 2000). By four years after the Izmit earthquake, the maximum afterslip velocity (beyond the interseismic rate) had decreased to about 30 mm/year. Most features of the deep afterslip are reproduced by dynamic finite element (FE) models of fault zone creep driven by coseismic stresses, but only if coseismic slip gradients in the mid-crust are large (that is, the high-slip patches have sharp bottom edges). The FE models assume velocity-strengthening friction in the fault zone, so we have not ruled out other rheologies that could yield nonlinearly stress-dependent slip rates. Some kinematic afterslip features cannot be explained by the dynamic FE models; for example, a patch of non-decaying afterslip (~20 mm/yr) which persists below the Yalova segment of the NAFZ. This may represent either dynamically triggered slip or a local error in the secular deformation model used to correct the postseismic velocities.
G12A-06 INVITED 11:35h
Post-Seismic Deformation Following the June 2000 Earthquake Sequence in Southwest Iceland
Two Mw6.5 earthquakes occurred in the South Iceland seismic zone (SISZ) in June 2000, the first one triggering at least three Mw$>$5 events on Reykjanes Peninsula. Modeling of co-seismic deformation, and aftershock locations indicate the two main shocks ruptured N-S, right-lateral strike slip faults, spaced about 17 km apart. We use campaign GPS observations from 1992-2000 to estimate the steady state plate motion in the area, and subtract from the station displacements during 2000-2004 to obtain the post -seismic deformation field. We observe significant changes in the velocity field in the SISZ, as well as on Reykjanes Peninsula due to the June 2000 earthquake sequence. The perturbation of the velocity field is most profound during the first year (2000-2001) in the main shock epicentral area. In the SISZ we observe post-seismic deformation over two spatio-temporal scales. On the first scale, we see a rapidly decaying deformation transient within 5 km of the two main shock ruptures, lasting no more than 2 months. This local month-scale transient is captured by several radar interferograms (InSAR) and is also observed at a few campaign GPS sites located near the main shock faults. The deformation pattern has been explained by poro-elastic rebound due to pore-fluid flow in response to the main shock induced pore-pressure changes [Jansson et al., Nature, 2003]. The second scale has a characteristic time of the order of a year and is detected by GPS measurements alone. Different models are tested to explain the year-scale deformation. Models of visco-elastic relaxation of the lower crust and upper mantle in response to the coseismic stress changes suggest that the viscosity of the lower crust and upper mantle must be of the order 10$^{18}$ Pa s to fit the observations. Simple models of afterslip to fit the post-seismic deformation data are also tested. We use these models to calculate changes in Coulomb failure stress in the main shock epicentral area from 2000-2004.
G12A-07 11:50h
Postseismic Deformations Following The 2003 Tokachi-oki Earthquake Detected By Dense GPS Observation
A Mjma=8.0 earthquake hit Hokkaido, northern Japan, on September 26, 2003 (JST). This is the first great interplate earthquake since the deployment of the nation-wide continuous GPS observation network in Japan. Up to 1m coseismic displacements were detected in the southeastern Hokkaido by this network. Since there may be a seismic gap or unbroken asperity on the plate interface east of the source region of this earthquake, how released stress is redistributed is crucial. It is necessary to monitor the postseismic deformations as well as seismicity. Postseismic deformations can provide us with frictional characteristics of plate interface, which is critically important to the physics of earthquake. Postseismic movements have been observed by the nation-wide network as well. However this network is a bit sparser in Hokkaido than in other areas of Japan. We can not expect enough resolving power of slip distribution on plate interface, especially its deeper extension. Therefore, we newly established GPS sites just after the mainshock in the eastern part of Hokkaido to fill gaps of the network as a collaborative work of researchers from Japanese universities. Observation at some sites started on September 28. Thirty GPS sites with dual frequency receivers were established by the end of October in the region of about 250 km x 150 km in southeastern Hokkaido. Some sites are telemetrically monitored. Data are stored in receiver or PC and collected periodically at the rest of them. All the data are archived in the server of Hokkaido University. We have analyzed the data till March, 2004, by using Bernese 4.2 software, and obtained time series of coordinates of our temporary sites and GEONET sites. It is confirmed that postseismic deformation still continues 6 months after the main shock. Most sites moved SE-ward and their magnitude decay according to the distance from the epicenter. We found rate changes in mid-Oct. at some sites and reverse of motion in eastern Hokkaido around the New Year day. These characteristics in temporal variation in coordinates can be explained well by a logarithmic decaying function based on the rate and state-dependent friction law. Some sites suffered from snow cover and showed anomalous motion in winter. We tried inversion of the observed displacements to estimate spatio-temporal variation in afterslip on the interface between Pacific and North American plates. We divide whole observation period into three according to temporal changes. We can find large afterslip, which amounts to ~50cm for the first month, in the source region during all three period, but their magnitude is decaying according to time. So far we have not resolved separation of area of afterslip and the main shock. Afterslip in deeper part of plate interface seems to have died out a month after the main shock. In the eastern part, interplate coupling dominates during the whole period and afterslip may not have proceeded into this region.
G12A-08 12:05h
Toward an Improved Understanding of Subduction Zone Mechanics: Finite Element Modeling Results From the Jalisco GPS Project, 1993-2004
We use continuous and periodic measurements from a 30-station GPS network along the Pacific coast of western Mexico, adjacent to the subducting Rivera plate, to derive new constraints on subduction mechanics. Two megathrust earthquakes have ruptured the Rivera plate subduction interface immediately offshore from our network since we began measurements in the mid-1990s, the 9 October 1995 M=8.0 Colima-Jalisco earthquake and the 21 January 2003 M=7.6 Tecom\'{a}n earthquake. Continuous GPS measurements after the 1995 earthquake reveal a strong postseismic response in which both the rate and direction evolve through time. Models that attribute the response solely to fault afterslip or postseismic viscoelastic flow fit the observations poorly, independent of the viscosities employed to model the latter response. Motion is instead well fit by a model that superimposes the effects of fault afterslip, viscoelastic flow, and relocking of shallow areas of the subduction interface. Possible triggering of the 2003 earthquake via coseismic and postseismic loading associated with the 1995 earthquake is examined via Coulomb stress analysis. We further present modeling of the unusual GPS displacement field associated with the 2003 earthquake, including the possibility that the earthquake triggered slip, possibly aseismic, along a fault in the upper crust.