FEM model of coseismic deformation measured by DInSAR and GPS: Wenchuan (China) 2008 and L’Aquila (Italy) 2009 Earthquakes IMPLEMENTING REALISTIC TOPOGRAPHY IN FEM IMPLEMENTING SEISMIC TOMOGRAPHY IN FEM C.Kyriakopoulos, E.Trasatti, S. Atzori, C.Bignami, M.Chini, S.Stramondo, and C.Tolomei Istituto Nazionale di Geofisica e Vulcanologia, Rome, Italy The very steep topographic relief of LongmenShan Range (figure 10) is implemented in our FE model (figure 9,11). The topographic database used is the 90 m Digital Elevation Model (DEM) from the Shuttle Radar Topographic Mission (SRTM). A large number of DEM layers corresponding to East China is used to achieve the complete coverage of the FEM model. In order to incorporate the 90 m topography we perform a low-pass filtering and subsampling of DEM data (ENVI 4.3 ITT software) obtaining a smoothed version of the original topography without loosing the significant topographic relief features. These operations are necessary to avoid large files that are very difficult to be handled by the mesh generation algorithm. The 3D seismic tomographic images (Lei et al., 2009) are computed using a large number of P and S wave arrival times from the aftershocks of the 2008 Wenchuan earthquake and other local events. The results reveals significant structural heterogeneities within the LongmenShan fault zone in the crust and upper mantle. The importance of implementing seismic tomography in the calculation of Green Functions is to evaluate the effects of both vertical and lateral heterogeneities of the medium on slip distribution. 3D seismic tomography by Lei et al., 2009 The FE model (figure 8) is composed by 145’000 3D-brick elements. The whole FE domain extends 2400 km x 2400 km x 500 km. The large extension is imposed by the lenght of the fault reaching almost 300 km and is necessary to avoid border effects that can affect the evaluation of static deformation. CLOSE UP OF CENTRAL DOMAIN The finite element mesh is generated by CUBIT 12.0 (SANDIA National Laboratory). The mesh is build by a large number of structured volumes. Mesh regularity is ensured by the Map/Submap scheme. LomgmenShan (Pengguan Massif) Eastern Tibet ~4500 m G13A km 8 km 18 km Flat Model Topographic model Chengdu Anxian Beichuan Maoxian Wenchuan Guanxian Yingxiu Qingchuan L ’ AQUILA (ITALY) 2009 EARTHQUAKE SLIP DISTRIBUTION preliminary results We have performed two experiments. In the first one we have used for the linear inversion only the YBQ main fault (figure 14). In the second experiment both the YBQ and GA fault (figure 17) are implemented. The reduced chi-square is close to 4.0 for all the models. We develop a procedure to perform linear inversions of spatially dense geodetic data (DInSAR displacements maps) of earthquakes. The technique is based on the Finite Element Method (FEM) and accounts for a more realistic description of the earth crust. and the computation of the fault Green functions. The FEM based method is used to explore two earthquakes with significantly different magnitude: the L’Aquila 2009 earthquake, Italy and the Wenchuan 2008 earthquake, China. The dataset available for these earthquakes were different. Indeed the L’Aquila earthquake is imaged by ascending and descending orbits of COSMO SkyMed and ENVISAT satellite missions while for the Wenchuan earthquake only the ALOS-PALSAR ascending orbit is available. The DInSAR data are integrated with 8 near field GPS data for the L’Aquila case and with 158 GPS data for the Wenchuan earthquake. Finally we retrieve slip distribution for both the L’Aquila and Wenchuan earthquakes in a heterogeneous and topographic medium and we perform for each one the comparison with the respective homogeneous and flat model. Results highlight the influence of the medium complexities in both Wenchuan and L’Aquila cases, the homogeneous and heterogeneous models reveals differences in slip distribution values up to 20%. Accounting for the topographic relief in the model produces more appreciable effects for the Wenchuan earthquake, where height variations are more significant with respect to the L’Aquila region. The FE model of the L’Aquila earthquake is composed of about 100’000 3D-brick elements contained in a cylindrical domain having radius of 100 km and height of 80 km. The fault plane, initially constrained by non-linear inversion, is extended up to 20 km of length and 15 km of width from the dimensions HOF=Homogeneous and Flat HOT=Homogeneous and Topography HEF=Heterogeneous and Flat DIFF=Differences YBQ=Yingxiu-Beichuan-Qingchuan GA=Guanxian CONCLUSIONS L’AQUILA: The homogeneous and heterogeneous models show discrepancies in the fault slip distribution values up to 20%. A new maximum of slip is found above the hypocenter, in accordance with seismological studies. The topography doesn’t seem to affect significantly the solution. WENCHUAN: The homogeneous (HOF) and heterogeneous (HEF) models show discrepancies in the fault slip distribution values up to 20%. The results are influenced by the low resolution tomography. The topographic (HOT) and flat (HOF) models show discrepancies in fault slip distribution values up to 25%. Topography affect the final solution. References: Chini et al., 2010, IEEE GEOSCIENCE AND REMOTE SENSING LETTERS, VOL 7, No 2, APRIL 2010 Atzori et al., 2008, J. Geophys. Res., 113, B9, doi: /2007JB005504, 2008 Zhang et al., 2008, Sci. China, Set. D, Earth Sci., vol. 38, no. 10, pp. 1195–1206, in Chinese Shen et al., 2009, Nature Geoscience, doi: /NGEO636 Lin et al., 2009, Tectonophysics, 471, , doi: /j.tecto Lin et al., 2009, Tectonophysics, article in press Lei et al., 2009, Geochemistry, Geophysics, Geosystems, vol 10, No, 10, doi:1029/2009GC Maximum displacement 16 cm24 cm27 cm Atzori et al., 2009 The Mw 6.3 L’Aquila earthquake occurred on the 6th of April The mainshock, located at 9.5 km depth, was followed in the next week by 7 aftershocks M > 5 (figure 1 ). The moment tensor solution of the mainshock and most of the aftershocks indicatesa normal faulting mechanism ( in agreement with the extensional tectonics of the Central Appennines. The activated fault is a NW-SE structure dipping SW (Atzori et al., 2009). Fault Central Topography Gran Sasso L’Aquila Fossa The dataset used in this work is composed by data from COSMO-SkyMed and Envisat satellite missions. The maximum displacement is located between the city of L’Aquila and the town of Fossa, where a subsidence of 16 cm (COSMO-Skymed), 24 cm (Envisat ascending) and 27 cm (Envisat descending) is detected (figure 2). In our analysis we also consider 8 GPS data by Anzidei et al. [2009] in addition to the DInSAR data. The 3D tomography (figure 3) is computed from the aftershocks. It evidences vertical and lateral vp heterogeneities, with a potential effect on displacement and stress fields. In particular, high vp values are found belowL’Aquila city between 4-10 km. vp and vp/vs velocities are transformed into elastic parameters (e.g., Poisson coeff., rigidity). Tomography from Di Stefano et al., submit In the heterogeneous models HEF and HET the slip is elongated horizontally and it is concentrated between 3 and 7 km depth, since the slip is inhibited at greater depths by the high rigidity values. A new slip concentration appears above the hypocenter, according to seismological inversions [Cirella et al., 2009]. The rake (only a few arrows are shown) is characterized by a low variability in the range -96°±4°, especially for the most slipping areas. To better appreciate the changes introduced by the 3D elastic structure, Figure 5c and Figure 5f show the differences between HEF and HOF, and between HET and HOT, respectively. The differences between models amount up to 20% of the maximum slip in both cases. A destructive (Mw 7.9) earthquake struck the Sichuan province (China) on May 12, 2008 (figure 6). The seismic event ruptured approximately 280 km of the Yingxiu-Beichuan fault and about 70 km of the Guanxian-Anxian fault. Surface effects were sufered over a wide epicentral area (about 300 km E-W and 250 km N- S). The huge earthquake took place within the context of long term uplift of the Longmen Shan range in eastern Tibet. The coseismic deformation is imaged by a set of ALOS-PALSAR L- band SAR interferograms (figure 7). We use an unprecedented high number of data (25 frames from 6 adjacent tracks) to encompass the entire coseismic area. THE WENCHUAN (CHINA) 2008 EARTHQUAKE We emphasize that flat models, frequently adopted, are inappropriate to represent earth surface topographic features and especialy in the case of Wenchuan earthquake. Futhermore in these specific case the calculation of the elastic solutions (regarding surface deformation due to a unite dislocation on a fault patch) have to take into account variations in height of the same order of the patches dimension. (figure 12) TOMOGRAPHY FE MODEL SPACE DATA SLIP DISTRIBUTION Slip distribution for the HOF model. The displacement pattern reveals high slip near the hypocenter area (Yingxiu town) and Beichuan county Slip distribution for the HOT model. Differences between HOF and HOT model Figure 2 Figure 4 Figure 1 Figure 3 Figure 5 Figure 10 Figure 6Figure 7 Figure 8 Figure 16 Figure 12 Figure 13 Figure 9 Figure 15 Work partially funded by the Italian Space Agency (ASI) in the framework of the SIGRIS project. ERS, ENVISAT and ALOS data courtesy of ESA and JAXA. homogeneous (HOF) and heterogeneous (HEF) models show discrepancies in the fault slip distribution values up to 20%. The differences between the two models (topographic and flat) are modified due to the presence of the GA fault. YBQ fault GA fault The front thrust GA fault is implemented into the FE model Figure 17 Figure 14 The resulting fault plane dips shallowly (~40°) at SW and became steeper at NE (~90°). Figure 11 NE SW NE SWNE SW NE Trasatti et al., ready to submit