INVESTIGATION OF EARTHQUAKE-CYCLE DEFORMATION IN TIBET FROM ALOS PALSAR DATA PI 168 Roland Bürgmann 1, Mong-Han Huang 1, Isabelle Ryder 2, and Eric Fielding 3 1 Berkeley Seismological Laboratory, University of California, Berkeley, California, USA. 2 School of Environmental Sciences, University of Liverpool, Liverpool, UK. 3 Jet Propulsion Laboratory, California Institute of Technology, Pasadena, California, USA. 1. INTRODUCTION There is much debate about the most fundamental aspects of the style of continental deformation in the India- Eurasia collision zone. One school of thought envisions Tibet to be a thickened, weak and fluid-like zone, while others consider the active tectonics in the region as that of rigid microplates bounded by major lithospheric faults. An alternative view suggests a primary role of lowviscosity channel flow in the lower crust. To resolve this debate, better knowledge of the rheology of the deep interior of the Tibetan plateau is needed. Current estimates of average effective viscosity in the lower Tibetan crust span five orders of magnitude. Quantitative information about the viscous strength of the deep Tibetan crust will allow for improved models of the tectonic evolution and current dynamics of the collision zone. The aim of this project is to rigorously determine the rheological properties of Tibet s interior from geodetically measured postseismic deformation. Such measurements effectively probe the constitutive properties of viscous flow at depth resulting from earthquakeinduced stress changes. Fig. 1: Earthquakes with M > 6 on the Tibet Plateau during recent years shown with examples of coseismic interferograms.
Given the lack of ground-based GPS measurements near many of the events we consider, much of this work relies on InSAR data collected spanning a number of recent major earthquakes and several years of postseismic deformation. We use ALOS PALSAR data and data from other SAR satellites to investigate a number of normal faulting earthquakes that have occurred on the Tibetan plateau, and the large thrust and strike-slip earthquake on the eastern edge of Tibet during 28. These include the M 6.4 Nima-Gaize earthquake in January 29 [1][2], the M 7.2 Yutian earthquake in March 28, the M 6.6 Zhongba County earthquake in August 28, the M 6.3 Damxung earthquake in October 28, and the M 7.9 Wenchuan earthquake in May 28 [3][4]. Collectively, these earthquakes cover a large area of the Plateau (Figure 1), and so have the potential to reveal lateral heterogeneity in crustal rheological structure. Both coseismic and postseismic interferograms are constructed to constrain models of the coseismic rupture and the postseismic relaxation processes. Ultimately, the goal of this investigation is to rigorously constrain the viscous properties of the deep interior of Tibet from the distribution and rate of transient postseismic deformation, including how these properties vary across the Plateau. We need to consider the relative importance of factors such as lithospheric layering, lateral crustal heterogeneity, coseismic source model details, afterslip vs. flow, and linear vs. non-linear viscosity in constructing numerical models to explain the observations. 2. ALOS PALSAR DATA AND PROCESSING The ALOS PALSAR platform provides data that are playing an increasingly important role in our investigations of earthquake cycle deformation in Tibet. The SAR data are used in this project to create interferograms of co- and postseismic surface deformation following the various earthquakes we target. In addition to the quota of data obtained through our PI project, we also benefited from a large number of PALSAR scenes obtained through the US Government Research Consortium (USGRC) data pool at the Alaska Satellite Facility. Coherence of the L-band data is excellent; however, limitations are presented by (1) relatively sparse sampling in time (usually greater than the 46-day repeat interval), (2) lack of descending orbit acquisitions as other instruments are switched on during the daytime flyovers and (3) substantial atmospheric and ionospheric artifacts. We thus complement the ALOS data with acquisitions from other spacecraft such as the ESA ENVISAT satellite (Figure 1) and, for a few events, with GPS data made available by our collaborators from China. All SAR data from the ALOS and Envisat satellites were processed from the raw signal data (Level 1. for PALSAR and Level for ASAR). The high-resolution fine beam (FB) PALSAR and image mode (IM) ASAR images, acquired in stripmap modes, were processed with the JPL/Caltech ROI_pac interferometric SAR (InSAR) package [5]. We used the ALOS PALSAR preprocessor that is part of GMTSAR for the stripmap FB data. We also used it to convert the ScanSAR WB1 data to faux stripmap data using the method described in [6] and completed the WB1-WB1 InSAR processing with ROI_pac. Pixel tracking or sub-pixel correlation calculations were also performed using ROI_pac. Corrections for topography rely on a version of the Shuttle Radar Topography Mission (SRTM) 3-arcsecond (9 m) spacing digital elevation model (Farr and Kobrick, 2) that has the voids filled with other data sources. 3. RESULTS 28 Nima-Gaize Earthquake: For the Nima earthquake in the central plateau, interferograms with excellent coherence (Fig. 2) reveal coseismic surface displacements due to slip on two faults, the mainshock (M w 6.4) fault plane and the fault plane of the largest aftershock (M w 5.9). The postseismic deformation field is closely correlated spatially with the coseismic field, suggesting that afterslip on the same fault planes occurred during the first few months following the initial ruptures. There is no evidence yet of postseismic viscoelastic relaxation, which enables a lower bound of ~3x1 17 Pa s to be placed on the viscosity of the lower crust [1]. ALOST59A: 7116!816 5 A A 4 32 3' 3 2 1!1 2 4 distance(km) 2 1!1 ALOST59A: 816!832 A A 3 2 4 distance(km) 32 15' 32 3' 32 15' 5 T59 7116 8116 (coseismic) 85 ' 85 15' 85 3' Fig. 2: Interferograms and line-of-sight displacement profiles for the 28 Nima earthquake. The coseismic and postseismic PALSAR interferograms show a similar deformation pattern. The double lobe pattern of the both interferograms results from slip occurring on both mainshock and aftershock planes (fault traces are shown by black lines). 5.5 T59 7116 8116 (postseismic)..5 postseismic coseismic 85 ' 85 15' 85 3'
Postseismic Envisat and ALOS interferograms show several interesting features, which are interpreted as shallow afterslip. Fig. 4 shows a time series of postseismic deformation, which has been used to invert for afterslip (lower panel). In the northern part of the fault zone, the afterslip appears to occur in the top few km of the upper crust where coseismic slip tapers off, while in the southern part of the rupture zone, the part that slipped coseismically simply re-slipped subsequently. Our interpretation of this unusual postseismic signature is ongoing. As for the Nima-Gaize case, there is no clear signal of viscoelastic relaxation, which will enable us to place a lower bound on lower crustal viscosity in the so far unprobed northwestern part of the Tibetan plateau. Fig. 3: Coseismic interferograms for the March 28 Yutian earthquake (M w 7.2). Curved black line shows inferred fault trace. The Guliya ice cap causes incoherence near the fault trace, but L-band images are able to capture high gradients of displacement in the hanging wall (west of the fault trace). Inferred slip distribution on a curved fault interface is shown at the bottom of the figure. New ALOS PALSAR scenes: We have recently ordered and received 27 new PALSAR scenes for the Nima and Yutian earthquakes, and also for another earthquake that occurred in the southeastern part of the plateau in 28, in Damxung County. Our research in the near future will incorporate these additional scenes into our analysis. 28 Yutian earthquake: The Yutian earthquake was the largest normal fault earthquake in Tibet in the last decade. InSAR coherence for the Yutian earthquake in NW Tibet is compromised by the permanent Guliya ice cap in the study area; however, in part of the ice-free hanging wall graben, we obtain a clear coseismic signal with L-band radar, which can be inverted for slip on a curved fault plane (Fig. 3). 28.43 28.52 28.71 28.8 4 4 28.9 29.1 29.1 29.39 Envisat T155 8222 852 (coseismic) 81 ' 81 3' 82 ' Fig. 4: Postseismic Envisat time series for the 28 Yutian earthquake and afterslip model. Dates are given above each image. For comparison, a coseismic interferogram on the same track is shown on the right. White ellipses highlight areas of similarity/difference between the coseismic and postseismic deformation patterns. 36 ' 35 3' 35 ' 2 15 1 5 5 1 coseismic Fig. 5: Envisat interferograms showing coseismic deformation for the 25 Zhongba earthquake and the 28 Damxung earthquake. Our new ALOS scenes, once processed, should contribute towards our analysis of these events. 28 Wenchuan Earthquake: A devastating Mw 7.9 earthquake occurred on 12 May 28 beneath the Longmen Shan mountain front in the Sichuan province of China. The earthquake epicenter was located in Wenchuan County and the rupture continued into Beichuan County, where damage to the city of Beichuan was extreme. The earthquake caused great damage extending at least 3 km along the Longmen Shan and adjacent areas. The coseismic deformation field was well captured by PALSAR interferograms (Fig. 6), which when combined with GPS measurements constrained detailed models of the geometry and slip distribution of the earthquake [3][4]. For the Wenchuan earthquake, we processed both ascending fine beam (8 paths, 47 477, with 8 to 13 frames on each path) and descending wideswath interferograms to completely cover the 3-
km rupture. A total of more than 15 PALSAR frames were used in the coseismic analysis. Joint inversion of the InSAR, GPS and teleseismic waveforms allow a detailed model of the slip evolution during this earthquake [4]. Ground deformation in the ascending radar line-of-sight (LOS) is complicated due to the combination of horizontal motion to the east combining with vertical motion upward on the hanging wall or northwest side of the fault. In addition, several bends in the fault cause localized deformation as the rupture proceeded towards the northeast. One pair of WB1 scenes was specially acquired with burst synchronization by JAXA for the Wenchuan earthquake on descending path 124 (Fig. 7). This pair has a longer spatial baseline in addition to being in the ScanSAR mode, so the coherence in areas of steep slopes is very low (masked out in the unwrapped interferogram). It provides valuable information on the earthquake deformation at moderate distances from the fault rupture that helps to constrain the fault slip at depth. Fig. 7: Zoom of ALOS PALSAR interferograms in area of main ruptures (same data as Fig. 6). Note that LOS displacements change sign close to much of the northwest side of the fault because the vertical motion towards the satellite is larger than the horizontal motion to the east away from satellite. From [4]. Fig. 6:. Mosaic of ALOS PALSAR unwrapped interferograms from 8 ascending tracks. Wave-like features elongated in the NW-SE direction are likely due to ionospheric radar propagation delays. Black lines show surface ruptures and green star shows epicenter. Positive LOS motion is away from satellite (i.e., east or down). From [4]. Fig. 8: Wide-beam ALOS PALSAR interferogram from descending path 124. Phase is unwrapped and converted to range change. Five ScanSAR subswaths, with incidence angles varying from 16 on the east side to 43 on the west side, were combined so LOS vectors change across scene. From [4].
We also mapped the surface ruptures by measuring the distortions of SAR images after the earthquake with pixel tracking or sub-pixel correlation, a technique that measures large surface displacements at coarse spatial scales of 1 2 to 1 3 m (Figure 9). Our map of the 28 surface rupture locations is very similar to the field maps and shows that the main surface rupture occurred on the Beichuan Fault with significant slip on a subparallel thrust fault system called the Pengguan Fault. Both of these faults were identified in geologic studies as active before the earthquake. In general, the Wenchuan postseismic interferograms are highly perturbed by different sources of noise distributed in different wavelengths. Three main sources of the perturbation are: topographic perturbation, the atmospheric perturbation, and ionospheric perturbation. The high relief of topography in Longmenshen area also largely decreases the coherence between the two SAR scenes for pairs with long baselines. In addition, the signal is generally perturbed by topography or ionosphere (Figures 1a and 1b). We are focusing on assembling all acquisitions along track 474 and process all of possible pairs of interferograms in order to identify and mitigate the perturbations. As a result, even though the postseismic InSAR result is not yet able to resolve an unambiguous relaxation signal, we will be able to further reduce the errors if we can increase the number of post-earthquake scenes in each track, and by applying more sophisticated filtering or time series analysis tools. A B Fig. 9: Surface displacements in the radar line of sight (LOS) from sub-pixel correlation of ALOS PALSAR images. Green line shows surface ruptures mapped from discontinuities in these LOS offsets and other sub-pixel correlation measurements from ALOS and Envisat SAR. Positive deformation is away from the satellite that means east or down. From [4]. Fig. 1 (a) ALOS PALSAR interferograms of Wenchuan postseismic deformation (track: 471(879-9224), 473(874-9219), 475(8622-9625)). (b) Postseismic interferograms (track: 472(8617-985), 474(8721-998), 476(879-9224)). The Wenchuan earthquake surface rupture is shown in red lines. Note that the signals are highly perturbed by the ionospheric noise
4. ACKNOWLEDGMENTS We acknowledge the JAXA PI project No. 169 award, without which much of this work would not have been possible. We thank the staff at the JAXA AUIG helpdesk and the AADN and ASF User Services Office for their help with providing and processing the data products. We thank Craig Dobson at NASA, Eva Zanzkerkia at NSF and Dan Dzurisin at USGS for their creation and support of the US Government Research Consortium data pool at the ASF DAAC. Part of this research was performed at the Jet Propulsion Laboratory, California Institute of Technology under contract with the National Aeronautics and Space Administration. The research presented in this report is supported by NSF grants EAR 73829 and EAR 11488. 5. REFERENCES [1] Ryder I., R. Bürgmann, and J. Sun, Tandem afterslip on connected fault planes following the 28 Nima-Gaize (Tibet) earthquake, J. Geophys. Res., 115, 21. [2] Sun J., Shen Z., Xu X., and Bürgmann R., Synthetic normal faulting of the 9 January 28 Nima (Tibet) earthquake from conventional and along-track SAR interferometry, Geophys. Res. Lett., 35, L2238, 28. [3] Shen, Z.-K., Sun, J., Zhang, P., Wan, Y., Wang, M., Bürgmann, R., Zeng, Y., Gan, W., Liao, H., and Wang, Q., Slip maxima at fault junctions and rupturing of bariers during the 28 Wenchuan earthquake, Nature Geosci., 2, 718-724, 29. [4] Fielding, E. J., Sladen, A., Li, Z., Avouac, J.-P., Bürgmann, R., and Ryder, I., Kinematic fault slip evolution model for the 28 M7.9 Wenchuan earthquake in China from SAR interferometry, GPS and teleseismic analysis: J.Geophys. Res., in preparation. [5] Rosen, P.A., Hensley, S., Peltzer, G. & Simons, M., Updated repeat orbit interferometry package released, EOS, Trans. AGU,85, p. 47 24. [6] Tong, X., Sandwell, D.T. & Fialko, Y., 21. Coseismic slip model of the 28 Wenchuan earthquake derived from joint inversion of interferometric synthetic aperture radar, GPS, and field data, J. Geophys. Res., 115, B4314.