Estimation of the 2010 Mentawai tsunami earthquake rupture process from joint inversion of teleseismic and strong ground motion data

Transcription

1 geodesy and geodynamics 2015, vol 6 no 3, 180e186 Available online at ScienceDirect journal homepage: Estimation of the 2010 Mentawai tsunami earthquake rupture process from joint inversion of teleseismic and strong ground motion data Zhang Lifen a,b,c,*, Liao Wulin a,c, Li Jinggang a,b,c, Wang Qiuliang a,c a Key Laboratory of Earthquake Geodesy, Institute of Seismology, China Earthquake Administration, Wuhan , China b Institute of Geophysics, China Earthquake Administration, Beijing , China c Wuhan Base of Institute of Crustal Dynamics, China Earthquake Administration, Wuhan , China article info abstract Article history: Received 13 January 2015 Accepted 20 February 2015 Available online 23 May 2015 Keywords: 2010 Mentawai earthquake Rupture process Tsunami earthquake Joint inversion Teleseismic recording Strong ground motion Sunda megathrust fault Joint inversion of teleseismic body-wave data and strong ground motion waveforms was applied to determine the rupture process of the 2010 Mentawai earthquake. To obtain stable solutions, smoothing and non-negative constraints were introduced. A total of 33 teleseismic stations and 5 strong ground motion stations supplied data. The teleseismic and strong ground motion data were separately windowed for 150 s and 250 s and bandpass filtered with frequencies of 0.001e1.0 Hz and 0.005e0.5 Hz, respectively. The finitefault model was established with length and width of 190 km and 70 km, and the initial seismic source parameters were set by referring to centroid moment tensor (CMT) solutions. Joint inversion results indicate that the focal mechanism of this earthquake is thrust fault type, and the strike, dip, and rake angles are generally in accordance with CMT results. The seismic moment was determined as Nm (Mw7.8) and source duration was about 102 s, which is greater than those of other earthquakes of similar magnitude. The rupture nucleated near the hypocenter and then propagated along the strike direction to the northwest, with a maximum slip of 3.9 m. Large uncertainties regarding the amount of slip retrieved using different inversion methods still exist; however, the conclusion that the majority of slip occurred far from the islands at very shallow depths was found to be robust. The 2010 Mentawai earthquake was categorized as a tsunami earthquake because of the long rupture duration and the generation of a tsunami much larger than was expected for an earthquake of its magnitude. 2015, Institute of Seismology, China Earthquake Administration, etc. Production and hosting by Elsevier B.V. on behalf of KeAi Communications Co., Ltd. This is an open access article under the CC BY-NC-ND license ( * Corresponding author. address: zhanglf112@163.com (Zhang L.). Peer review under responsibility of Institute of Seismology, China Earthquake Administration. Production and Hosting by Elsevier on behalf of KeAi / 2015, Institute of Seismology, China Earthquake Administration, etc. Production and hosting by Elsevier B.V. on behalf of KeAi Communications Co., Ltd. This is an open access article under the CC BY-NC-ND license (

2 geodesy and geodynamics 2015, vol 6 no 3, 180e Background Indonesia is located at the triple junction of the Australian Plate, Eurasian Plate, and Pacific Plate. The Australian Plate is converging with the southeastern segment of the Eurasian Plate, called the Sunda Plate, at a relative plate motion rate of approximately 60 mm/a [1,2]. This plate motion is oblique to the Sumatran Subduction Zone. The boundary-parallel shear motion of the convergence transpires mainly along the Sumatran Fault (Fig. 1), while the boundary-perpendicular convergent motion is accommodated by underthrusting along the Sunda Megathrust at a rate of approximately 45 mm/a [2]. Because of these complex tectonics, Indonesia is an earthquake-prone country. The seafloor off the west coast of the island of Sumatra has produced several great interplate earthquakes not only from subduction zone activity but also from the movement of the Sunda Fault and the active fault systems along the Sumatra Island. During the past decades, much of the Sunda Megathrust has slipped, resulting in large earthquakes, including the December 26, 2004 SumatraeAndaman (Mw9.1, from USGS) and March 28, 2005 Nias (Mw8.6) events, thereby rupturing the zone from 0 to 14 N [3,4]. The SumatraeAndaman Mw9.1 earthquake was the third largest earthquake to occur since 1900 and caused in excess of 286,000 casualties in the 14 countries surrounding the Indian Ocean [3]. A subduction interface from 2 S to 5 S, near the Pagai Islands (southeastern Mentawai Islands), ruptured and resulted in three large events: the September 12, 2007 Southern Sumatera Mw8.5 earthquake, the Kepulauan Mw7.9 earthquake, and the October 25, 2010 Mentawai Mw7.8 earthquake [5]. At 9:42 PM local time, the Mentawai Mw7.8 earthquake, with its epicenter located near the trench where the Australian Plate is subducted beneath the Sunda Plate, hit Indonesia and produced ground shaking on Pagai Island. Although tsunami warnings were broadcast on local television, many of the isolated coastal areas were unprepared for the large tsunami waves that swept onto Pagai Island's southwestern coast. According to the casualty information from Pusdalops PB Sumbar, the Disaster Management Operational Control Center for West Sumatra Province, this earthquake caused substantial damage and 509 human causalities. Some researchers identified this earthquake as a tsunami earthquake [6,7]. The so-called tsunami earthquake is a special class of events that can create a tsunami that is much larger than is expected. To gain an understanding of the mechanism of this kind of tsunamigenic earthquake, it is important to investigate the kinematic rupture process. 2. Methodology and data processing 2.1. Method Source process inversion theory follows the standard waveform inversion scheme [8] and numerical method developed by Yagi et al. [9]. Generally, the source can be modeled using point source and finite-fault source models. The point source model is suitable to represent a small earthquake because the size of its epicenter can be approximated as point, the duration is very short, and the slip can be assumed to be uniform. However, for a large earthquake, the finite-fault model is needed to describe its slip variation in space and time. The fault plane is divided into m n subfaults and each point source is set at the center of every subfault. Slip rate functions are described by a series of triangle functions with rise time t, and the fault slip vector, denoted by K basis slip vectors. The observed seismic waveform at the station j is stated by: W obs j ðt i Þ¼ X X mnlk G mnkj ðt i ðl 1Þt mnlk qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (1) ðmdxþ 2 þðndyþ 2 =V r Þþe j Fig. 1 e Tectonic map of the 2010 Mentawai earthquake region. An expanded view of the study region is shown in the upright corner. where X mnlk is the kth component of slip at the mnth subfault with the lth time step; G mnkj (t) is the Green's function at the mn-subfault; Vr is the rupture velocity; and e j is the Gaussian error with variance of j. In general, high frequency radiation is one of the most important causes of significant earthquake damage. After a large earthquake has occurred, teleseismic data can be quickly retrieved from some websites; this data mainly contains the long period (low frequency) event components. Therefore, the overall event characteristics, such as the overall moment releasing rate and the depth range of the rupture area, can be roughly described. However, sometimes detailed information can be unobtainable. Strong ground motion data with high frequency information can be utilized to fill these gaps. Thus, if we want to obtain more accurate kinematic rupture information, it would be better to combine both types of data for the source process inversion.

3 182 geodesy and geodynamics 2015, vol 6 no 3, 180e186 In this paper, teleseismic recordings and strong ground motion data were jointly used to estimate the source rupture process, as opposed to previously published studies on this earthquake Data processing Teleseismic recordings were retrieved from IRIS-DMC and near source data was provided by BMKG (Fig. 2). A total of 33 teleseismic stations' waveforms were selected because of their good azimuthal coverage and high signal-noise ratio. In order to avoid the upper mantle contamination to the crustal structure, all of the stations ranged from 30 to 90 [10]. Near-source data was retrieved from five strong ground motion stations in BMKG, with the distance between the epicenter and the stations ranging from 120 km to 300 km. Starting 10 s before P-wave arrival, the teleseismic waveforms were windowed for 150 s, band pass filtered between Hz and 1.0 Hz, and then integrated into displacement with a sampling time of 0.25 s. Starting 5 s before P-wave arrival, the near source data was windowed for 250 s, band pass filtered between Hz and 0.5 Hz, and then integrated into displacement with a sampling time of 0.25 s. In the joint inversion, same weight was assumed to the two kinds of datasets. Green's Function for teleseismic data was calculated using the method of Kikuchi and Kanamori [11], in which the multilayer structure is used to compute response of source, station, and PP bounce point structures. The discrete wave number method developed by Kohketsu [12] was utilized to obtain the Green's Function for the near source data. The 1- D velocity structure model from Kopp et al. [13] was adopted. To reduce the instability caused by increasing the model parameters, smoothing constraints to the slip distribution, with respect to time and space, were applied. We used a finite difference Laplacian smoothing matrix to impose smoothness constraints on the model, to regularize the inversion. We also employed a non-negativity constraint [14], with the assumption that no backslip should be accommodated in Fig. 2 e Distribution map of teleseismic stations and strong motion stations. the earthquake, and constrained the slip to zero along the sides and bottom of the fault patch. 3. Results 3.1. Inversion parameters To estimate co-seismic slip distribution, joint inversion techniques of teleseismic and strong ground motion data for the assumed planar fault geometry were employed. According to Sunda Fault geometry and the aftershock distribution over the week after the mainshock, a planar fault model with an area of km was constructed. The fault plane was optimally discretized into 133 subfaults with a subfaultdimensionof10km.thestrikewasfixedtoaconstant value of 324, the dip was set to 10, and the rake was set to 96; these parameters are slightly different from those estimated in the Global Centroid Moment Tensor solution. The epicenter location, at a focal depth of 13.5 km, was determined by BMKG. The slip rate function of the subfault was expanded into a series of 15 triangle functions with a rise time of 2 s. The optimal rupture velocity, Vr, determined from the range of 0.5 km/s to 3.5 km/s, was 2.0 km/s, for maximum velocity with a minimum variance. The rigidity, m, was assumed to be GPa from the structural model Inversion results The inversion results are shown in Fig. 3. Figs. 3a-e show the fault mechanism, the source duration, the slip distribution, the teleseismic waveform fitting, and the strong ground motion waveform fitting, respectively. The waveform fitting results testify to the robustness of our results, Figs. 3d and e. The focal mechanism of the 2010 Mentawai earthquake was determined as the thrust fault type. A total seismic moment, M 0 of Nm (Mw7.8) was released during a period of 102 s, Fig. 3b, which is in agreement with the value estimated by the global CMT moment solution. The determined rupture duration is anomalously long, in comparison with conventional earthquakes of similar magnitude, such as the 1996 Peru earthquake (Mw7.6), whose rupture duration was around 45e50 s [15]. Due to the excessively long rupture duration and the generation of a tsunami that was much larger than was expected, the 2010 Mentawai earthquake is categorized as a tsunami earthquake. The final dislocation showed that the maximum slip for this earthquake was around 3.9 m near the hypocenter. The rupture mainly extended along the dip direction, Fig. 3c, and propagated from the hypocenter to the shallower subfaults. Fig. 4 shows the map view of co-seismic slip and aftershock distribution for the week after the mainshock. The aftershock data was obtained from BMKG, while the focal mechanisms were taken from global CMT solution results. The rupture propagates mainly along the dip direction and the rupture area appears to extend parallel to the subduction zone between the Australia and Eurasian Plates. Most of the

4 Fig. 3 e Joint inversion result of the Mentawai earthquake.

5 184 geodesy and geodynamics 2015, vol 6 no 3, 180e186 Fig. 4 e Slip distribution map view for the week after the mainshock. aftershocks were also distributed near the trench. The focal mechanism of the aftershocks, as shown in Fig. 4, predominantly was normal faulting near the trench. The slip distribution probably describes the rupture of a locked subducted seamount on an otherwise decoupled zone, resulting in extension of the outer rise causing the normal faulting aftershocks. 4. Discussion and conclusion 4.1. Comparison with other studies Lay et al. [7] developed finite-source models for the rupture process of the Mentawai earthquake using teleseismic recordings of P, SH, and short-arc Rayleigh wave recordings from global seismic network stations. Inversion indicates a slight preference for Vr of approximately 1.5 km/s, and the corresponding slip distribution covers a 100 km long region of 2e3.5 m or 3.5e4.3 m slip, concentrated seaward and beneath the Pagai Islands. It is suggested that low seismic moment, but large slip, also occurred in low rigidity material extending out to the trench. Slip models were generally consistent with a tsunami recording from DART buoy 56001, although the amplitude was 10%e30% underestimated. Newman et al. [6] calculated a finite-fault solution from teleseismic data and scaled the source upward by an average of 5.6 times to get seafloor displacements great enough to reproduce the observed tsunami effects. It seems that their source model produced unrealistic static geodetic displacements, far greater than those observed. Hill et al. [17] examined this event using a combination of high-rate GPS data, from instruments located on the nearby islands, and a tsunami field survey. The results show that any large patch of slip, >4 m, must be accommodated in a very shallow and narrow strip of the megathrust, at depths <6 km and no further than 50 km from the trench. Maximum slip for their preferred model is 9.7 m, and maximum seafloor uplift is 1.9 m. This model comes much closer to predicting the measured tsunami. Satake et al. [18] estimated the slip distribution on the fault plane through the inversion of tsunami waveforms. Their results show that slip on the shallow subfault ranged from 1 to 6 m, while slip on the deep subfaults was smaller. The largest slip, 6.1 m, was estimated on the shallowest subfault near the epicenter. The average slip on shallower subfaults was approximately 2 m. Yue et al. [19] estimated the rupture process using finitefault rupture models obtained by joint inversion of the highrate-gps time series and numerous teleseismic broadband P and S wave seismograms together with iterative forward modeling of the tsunami recordings. The models indicated rupture propagated about 50 km up dip and about 100 km northwest along strike from the hypocenter, with a rupture velocity of approximately 1.8 km/s. Subregions with large slip extend from 7 to 10 km depth nearly 80 km northwest from the hypocenter, with a maximum slip of 8 m and from roughly a 5 km depth to beneath thin horizontal sedimentary layers beyond the prism deformation front for about 100 km along strike, with a very small localized region having >15 m of slip. This rupture model indicates that local heterogeneities in the shallow megathrust can accumulate strain that allows some regions near the toe of accretionary prisms to fail in tsunami earthquakes. This study obtained the rupture process by joint inversion of teleseismic and strong ground motion waveform data. By comparison, the overall slip distribution is similar with others results. Our maximum slip was determined to be 3.9 m, which is in accordance with Lay et al. [7], however, it is somewhat smaller than the slip estimated from GPS and tsunami

6 geodesy and geodynamics 2015, vol 6 no 3, 180e modeling (9.7 m, Hill et al. [17]; 6.1 m, Satake et al. [18]; >15 m, Han Yue et al. [19]). references 4.2. Conclusions The 2010 Mentawai earthquake generated a locally devastating tsunami much larger than expected based on the seismic magnitude. Source process inversion results indicate a shallow dip, consistent with an origin on the Sunda Megathrust. The rupture nucleated around the hypocenter and propagated to the southwest and broke the first asperity centering at 14 km from the epicenter with maximum slip amounting to 3.9 m, then propagated along the strike direction to the northwest where the second asperity was broken, which was centered about 78 km from the epicenter. We identify this earthquake as a tsunami earthquake because of its excessively long rupture duration and its generation of a greater than expected tsunami. However, there are large differences in slip distributions from the different modeling methods and datasets. Although we cannot reconcile these complexities and large uncertainties on the amount of slip still exist, we found the conclusion that the majority of slip occurred far from the islands at very shallow depths to be robust. As mentioned above, many large earthquakes have occurred in this region. After comprehensive analysis, the 2010 Mentawai Mw7.8 earthquake ruptured immediately updip of and was probably triggered by stress changes following the September 2007 Mw8.5 Sumatran earthquake [6,20]. This area may have last ruptured as part of the 1797 Mw8.6 and 1833 Mw8.9 events, described by Natawidjaja et al. [21] as having about 18 m of megathrust slip to explain the co-seismic uplift. Further north, the 2005 Mw8.6 Nias earthquake ruptured the same approximate area [22]. Available high resolution bathymetry along the trench suggests that significant faultingintheregionmaybeduetorupturethroughthe prism toe during the 2004 Sumatran Giant earthquake and previous earthquakes [23]. The large slip estimated in the shallow trench and the considerable faulting near the trench toe further north support the hypothesis that the subduction zone off western Indonesia is capable of supporting shallow megathrust slip. It challenges the conventional wisdom that the shallow tips of subduction megathrusts are aseismic and, therefore, raises important questions both about the mechanical properties of the shallow fault zone and the potential seismic and tsunami hazards of this shallow region. Acknowledgements We would like to express our gratitude to Professor Yuji Yagi and two anonymous reviewers for their great help and creative suggestion. Also great thanks are given to Iman Fatchurchoman for providing seismic data. This work was supported by National Natural Science Foundation of China ( ). [1] Bock Y, Prawirodirjo L, Genrich JF, Stevens CW, McCaffrey R, Subarya C, et al. Crustal motion in Indonesia from global positioning system 21measurements. J Geophys Res 2003;108(B8):2367. [2] Chlieh M, Avouac JP, Sieh K, Natawidjaja DH, Galetzka J. Heterogeneous coupling of the Sumatran megathrust constrained by geodetic and paleogeodetic measurements. J Geophys Res 2008;113:B JB [3] Lay T, Kanamori H, Ammon CJ, Nettles M, Ward SN, Aster RC, et al. The great Sumatra-Andaman earthquake of 26 December Science 2005;308:1127e32. [4] Ammon JC, Ji C, Thio H, Robinson D, Ni S, Hjorleifsdottir V, et al. Rupture process of the 2004 Sumatra-Andaman earthquake. Science 2005;308:1133e9. [5] [6] Newman AV, Hayes G, Wei Y, Convers J. The 25 October 2010 Mentawai tsunami earthquake, from real-time discriminants, finite-fault rupture, and tsunami excitation. Geophys Res Lett 2011;38:L [7] Lay T, Ammon CJ, Kanamori H, Yamazaki Y, Cheung KF, Hutko AR. The 25 October 2010 Mentawai tsunami earthquake (Mw7.8) and the tsunami hazard presented by shallow megathrust ruptures. Geophys Res Lett 2011;38:L [8] Hartzell SH, Heaton TH. Inversion of strong ground motion and teleseismic waveform data for the fault rupture history of the 1979 Imperial Valley, California earthquake. Bull Seism Soc Am 1983;73:1553e83. [9] Yagi Y, Mikumo T, Pacheco J, Reyes G. Source rupture of the tecoman, Colima, Mexico earthquake of 22 January 2003, determined by joint inversion of teleseismic body-wave and near-source data. Bull Seism Soc Am 2004;94:1795e807. [10] Lay T, Wallace TC. Modern global seismology. San Diego: Acad. Press; [11] Kikuchi M, Kanamori H. Inversion of complex body wave-iii. Bull Seism Soc Am 1991;81:2335e50. [12] Kohketsu K. The extended reflectivity method for synthetic near-field seismograms. J Phys Earth 1985;33:121e31. [13] Kopp H, Flueh ER, Klaeschen D, Bialas J, Reichert C. Crustal structure of the central Sunda margin at the onset of oblique subduction. Geophys J Int 2001;147:449e74. [14] Lawson CL, Hanson RJ. Solving least squares problems. New Jersey: Prentice-Hall, Inc; [15] Jenifer LS, Susan LB. Source characteristics of the 12 November 1996 Mw7.7 Peru subduction zone earthquake. Pure Appl Geophys 1999;154:731e51. [16] [17] Hill EM, Borrero JC, Huang Z, Qiu Q, Banerjee P, Natawidjaja DH, et al. The 2010 Mw7.8 Mentawai earthquake: very shallow source of a rare tsunami earthquake determined from tsunami field survey and near-field GPS data. J Geophys Res 2012;117:B [18] Satake K, Nishimura Y, Putra PS, Gusman AR, Sunendar H, Fujii Y, et al. Tsunami source of the 2010 Mentawai, Indonesia earthquake inferred from tsunami field survey and waveform modeling. Pure Appl Geophys 2013;170:1567e82. [19] Yue H, Lay T, Rivera L, Bai Y, Yamazaki Y, Cheung KF, et al. Rupture process of the 2010 Mw7.8 Mentawai tsunami earthquake from joint inversion of near-field hr-gps and teleseismic body wave recordings constrained by tsunami observations. J Geophys Res Solid Earth 2014;119:5574e93. [20] Stein R. The role of stress transfer in earthquake occurrence. Nature 1999;402:605e9.

7 186 geodesy and geodynamics 2015, vol 6 no 3, 180e186 [21] Natawidjaja DH, Sieh K, Chlieh M, Galetzka J, Suwargadi BW, Cheng H, et al. Source parameters of the great Sumatran megathrust earthquakes of 1797 and 1833 inferred from coral microatolls. J Geophys Res 2006;111:B /2005JB [22] Briggs R, Sieh K, Meltzer AS, Natawidjaja D, Galetzka J, Suwargadi B, et al. Deformation and slip along the Sunda megathrust in the Great 2005 Nias-Simeulue earthquake. Science 2006;311:1897e901. [23] HenstockT J, McNeill L, Tappin D. Seafloor morphology of the Sumatran subduction zone: surface rupture during megathrust earthquakes. Geology 2006;34(6):485e8. dx.doi.org/ / Zhang Lifen, Associate Professor, now is a PhD student in the Institute of Geophysics, China Earthquake Administration. She is mainly engaged in research on focal mechanisms, kinematic and dynamic earthquake rupture processes, and reservoir-induced seismicity in the Three Gorges area. Since 2006, she has published more than 30 research papers.