GSA DATA REPOSITORY 2010252 He and Peltzer InSAR data analysis: Three pairs of satellite images (Table DR1) are processed into interferograms using the ROI_PAC software (Rosen et al., 2000). The processing follows the standard 2-pass approach using the Shuttle Radar Topography Mission (SRTM) data to remove the topographic phase. To avoid aliasing of the phase in the region of high phase gradient, the data are processed at one range look and referenced to a latitudelongitude grid with a step of 1 arc second. Unwrapping is done using the ROI_PAC default branch-cut algorithm for the pairs on tracks 341 and 348. Higher phase noise required using the SNAPHU unwrapper (Chen et al., 2001) for the pair on track 427. The phase fields of the three interferograms are then masked with a phase correlation threshold of 0.3. The results are shown in Figure 1, b, c, and d. Data inversion: Observed displacement field shows that two distinct faults broke in the sequence (Fig. DR1a). A series of forward models using uniform slip on faults embedded in an elastic half space (Okada, 1992) are performed and compared to displacement profiles perpendicular to the faults to determine an a-priori dip angle for each fault. The main and the second faults are divided into 10 10 and 5 5 rectangular patches, respectively, to resolve slip variations on the two planes. The dip of each fault is later refined using a data-model misfit adjustment. The displacement data of the 3 interferograms are decimated to retain only a sub-set of points (Fig. DR1b). The decimation is based on the local phase gradient in order to keep a denser distribution of points in areas of large deformation (Lasserre et al., 2005). The data subsets are then inverted using a standard least-square procedure to solve for the down-dip and strike-slip components of slip on the faults. Solution smoothing is used in the inversion to reduce high oscillations in the slip distribution on faults and the optimum weight for the smoothing equations is adjusted to minimize the solution roughness without affecting significantly the data-model misfit (Fig. DR1c). Finally a parameter grid search is performed to refine the dip angles of the two faults (Fig. DR1d). Optimum dip angles of 45˚ for the main fault and 58˚ for the second fault are chosen in the region of minimum misfit under the condition that the two faults do not intersect at depth. The optimum fault solution leads to a good agreement between the observed and synthetic interferograms (Fig. DR2). The largest discrepancies occur in the areas where the faults reach the surface and are due to the simplified geometry and discretization of the faults at shallow depth. Pore fluid flow and post-seismic Coulomb Stress change model: The fault geometry and variable slip solution for the main fault are used to calculate the changes of Coulomb stress and pore pressure due to poroelastic deformation of the crust. The Coulomb stress is computed throughout the volume on a direction parallel to the down-dip direction of the second fault. Because we focus on the stress-pressure evolution in a relatively short time period after the Jan. 9, 2008 event (~2 weeks), poroelastic deformation of the crust 1 is assumed to follow the Biot’s theory (Wang, 2000). The components of the stress change ij due to the earthquake-related strain ( ij ) in the solid are given by ij 2G (ij u 1 2 u kk ij ) , (1) where G is the shear modulus, kk the trace of the strain tensor, u the undrained Poisson’s ratio and the Kronecker delta symbol. The change of pore fluid pressure p due to bulk elastic deformation is calculated using (Screaton and Ge, 2007) ij p B G ( 2(1 u ) ) ii , 3(1 2 u ) (2) where B is the Skempton’s coefficient. The evolution of the pore fluid pressure is assumed to be governed by the mass diffusion equation D 2h h t (3) where D is the diffusivity of fluids, h the hydraulic head, and t the time (Screaton and Ge, 2007). We construct a three-dimensional, finite-element model using the Finite Element Program Generator (FEPG) software to compute the coupled solution of elastic deformation and pore fluid flow. The split-node technique is used to impose slip on Fault 1 based on the slip distribution determined by inversion of InSAR data (Fig. 2). Fault 2 is modeled as a planar zone of finite thickness with specific permeability. The thickness of damaged fault zones could range from ~200 m to ~1-2 km (Ben-Zion, et al., 2003; Cochran, et al., 2009). Fault 2 is a relatively small fault and its thickness is assumed to be 80 m (Fig. DR3). To achieve hydrostatic equilibrium in the entire volume before the first seismic event, we run the model over 60 years with 1-year time steps. At t=0, the coseismic displacement of the Jan. 9, 2008 event is imposed on Fault 1 and a poroelastic solution is computed at regular time steps for 10 days. To avoid numerical instabilities immediately after the event, time steps of 1 hour are used for the first 2 days and steps of 12 hours are used between day 2 and day 10 after the first event. 2 Figure DR1. (a) Co-seismic interferogram of Figure 1b draped over shaded topography model of study area. White arrows point at morphologic fault scarps and black, dashed lines indicate approximate location of fault-surface intersection for the two modeled fault planes. (b) Spatial distribution of 966 data points kept in model inversion. Colors refer to 3 interferograms in tracks 341, 348, and 427. See the section on data inversion for details. (c) Trade-off between L2 norm of least-squares solution misfit and model roughness (average slip gradient). Our preferred model has a roughness of ~0.8 cm/km. (d) Model misfit variations represented in the Fault 1 – Fault 2 dip angles parameter space. White dashed line separates regions where faults intersect at depth (upper left), and are disconnected (lower right). White triangle represents the preferred set of dip angles of the two faults. 3 Figure DR2. Comparison between observed and modeled radar line of sight displacement. (a), (b) and (c) are observed interferograms of Track 348, 341 and 427, respectively. (d), (e) and (f) are synthetic interferograms based on fault model for the same tracks, respectively. (g), (h) and (i) are residuals between observed and synthetic displacement on same tracks, respectively. Solid lines show upper edges of modeled fault planes. 4 Figure DR3. (a) 3-dimentional grid of finite elements used in poroelastic model. Earthquake faults are placed in the center of large volume to minimize boundary effects on numerical solution. The volume is divided into 106,200 hexahedral elements with higher node density near the faults. Boundary conditions are: Vertical walls of models are fixed (no movement) in horizontal directions (x,y) and free in vertical (z) direction. Bottom surface is fixed in vertical direction and free in horizontal directions. Upper surface is free in three directions. We assume there is no input and output of pore fluid across the 6 edges of the model, and under the gravitational field the hydraulic head is at the upper surface. Blue rectangle on upper surface is area shown in (b). (b) Details of grid geometry along two modeled faults. Blue box shows map view from top of fault area (see location in (a)). 5 Table DR1. Parameters of Envisat data used in this study. Heading is azimuth clockwise from North of satellite heading direction. Incidence angle is angle between radar line of sight and vertical direction at center of study area. Bp is perpendicular component of baseline between orbits in interferometric pair. T is time interval between two acquisition dates. Dates are written as day/month/year. Track 341 427 348 Date 1 Date 2 09/08/2007 31/01/2008 28/03/2007 06/02/2008 23/11/2007 01/02/2008 Heading (˚) Inc. angle (˚) -13.24 -13.28 -166.5 26.0 42.0 20.4 Bp (m) 131 38 25 T (days) 175 311 70 Additional references: Ben-Zion, Y., Peng, Z., Okaya, D., et al., 2003, A shallow fault-zone structure illuminated by trapped waves in the Karadere-Duzce branch of the North Anatolian Fault, western Turkey: Geophysical Journal International, v. 152, p. 699–717, doi: 10.1046/j.1365–246X.2003.01870.x. Chen, C. W., Zebker, H. A., 2001, Two-dimensional phase unwrapping with use of statistic models for cost functions in nonlinear optimization: Journal of optical Society of America (A), v. 18, p. 338-351. Cochran, E., Li, Y.-G., Shearer, P., et al., 2009, Seismic and geodetic evidence for extensive, long-lived fault damage zones: Geology, v. 37(4), p. 315–318; doi: 10.1130/G25306A.1 Lasserre, C., Peltzer, G., Crampe, F., Klinger, Y., van der Woerd, J., Tapponnoer, P., 2005, Coseismic deformation of the 2001 Mw=7.8 Kokoxili earthquake in Tibet, measured by synthetic aperture radar interferometry: Journal Geophysical Research, v. 110, B12408, doi:10.1029/2004JB003500. 6
© Copyright 2026 Paperzz