ects of Negative Stress Drop on Fault Rupture,**. mid-niigata (Chuetsu), Japan, earthquake

Size: px

Start display at page:

Download "ects of Negative Stress Drop on Fault Rupture,**. mid-niigata (Chuetsu), Japan, earthquake"

Ashlyn Fletcher
5 years ago
Views:

1 Bull Earthq Res Inst Univ Tokyo Vol 2-,**2 pp,1-,13 The E# ects of Negative Stress Drop on Fault Rupture The,** mid-niigata (Chuetsu), Japan, earthquake Takashi Miyatake*, Aitaro Kato, Kazuhito Hikima and Takeshi Kimura Earthquake Research Institute, University of Tokyo Abstract We investigate spatiotemporal distribution in a kinematic slip model of the,** mid-niigata (Chuetsu) earthquake, focusing on the relationship between rupture time and stress drop Stress drop distribution is calculated from the final slip distribution inferred from waveform inversion There are negative stress drop regions near the edges of inferred asperities Rupture delays were detected in the mean slip rate time functions Combining the above distribution, we found delayed rupture within regions of negative stress drop In these negative stress drop regions, rupture propagation is expected to arrest because a positive stress drop promotes rupture It should be emphasized that our results were obtained without reconstructing the dynamic faulting process Key words: Rupture Process, The,** mid-niigata earthquake, Stress drop + Introduction teri and Spudich (,***) demonstrated that two key The faulting process, ie, space-time distributions dynamic parameters, Dc and peak stress, cannot be of fault slips of past earthquakes, have been esti- uniquely estimated from band-limited waveforms mated by many seismologists ( eg, Liu, et al,,**0, In this manuscript, we study the details of the Liu and Archuleta,,**; Wald and Heaton, +33 ; kinematic model of the,** mid-niigata (Chuetsu) Yagi,,**) In addition to these kinematic fault earthquake, which occurred within a dense network models, dynamic models of earthquakes have also of seismographs (Fig + ), and obtained important inbeen constructed with the constraints of the kine- formation on physical faulting process of the event matic parameters ( eg, Quin, +33* ; Miyatake, +33, ab; without reconstructing a sophisticated dynamic Mikumo and Miyatake, +332 ; Miyatake et al,,**; model Olsen et al +331 ; Peyrat et al,,**+) Knowledge of dynamic parameters is important not only to under-, The,** mid-niigata earthquake stand earthquake source physics, but also strong The,** mid-niigata earthquake of M 0 0 ocground motion simulations, which require realistic curred in the back-arc area of the main Japanese source models for probabilistic modeling (Irikura et island as a shallow inland dip-slip earthquake The al,,**) However, the dynamic model depends on source region was located in the eastern margin of a the kinematic model selected Several di# erent kine- thick sedimentary basin (Kato et al,**0) It caused matic models often exist for a single earthquake, serious damage in and around the source region The each obtained with a di# erent inversion method, main shock was followed by significant aftershock waveform data set, or modeling of kinematic source activity, and the number of large aftershocks was Depending on numerous additional factors includ- significantly greater than those associated with ing computational method and computational pa- other recent inland earthquakes in Japan rameters adopted, dynamic models based on kinematic Kinematic model of the,** mid-niigata earthquake models are also not unique As one example of Several inversion studies have been performed on the additional complexity of dynamic models, Guat- the,** mid-niigata earthquake, each of which has * miyatake@eriu-tokyoacjp ( + + +, Yayoi, Bunkyo-ku, Tokyo ++- **-,, Japan) 273 w

T Miyatake, A Kato, K Hikima and T Kimura Fig, Slip distribution inferred from waveform inversion results obtained by Hikima and Koketsu (,**/) Fig + Hypocenter of the,** Chuetsu earthquake is

Honda et al,,**/; Asano and Iwata,,**3) In general, the waveform inversion results di# er because of specific choices made regarding, for instance, the inclusion of geodetic data or teleseismic data,

teleseismic using Okada s ( +33, ) Green s functions data, but used near-field ground motion records, although number of waveforms are di# erent On the other hand, the Green s functions or the

2 T Miyatake, A Kato, K Hikima and T Kimura Fig, Slip distribution inferred from waveform inversion results obtained by Hikima and Koketsu (,**/) Fig + Hypocenter of the,** Chuetsu earthquake is plotted as a black star Red triangles represent KiKnet stations that recorded data compared to synthetic seismograms (see also Fig /) yielded slightly di# erent results ( eg, Hikima and Koketsu,,**/; Honda et al,,**/; Asano and Iwata,,**3) In general, the waveform inversion results di# er because of specific choices made regarding, for instance, the inclusion of geodetic data or teleseismic data, the inversion algorithm itself, and di# erent Green s functions ( eg + D, hybrid + D, or -D) The Fig - Static stress change distribution calculated above three studies did not use geodetic nor teleseismic using Okada s ( +33, ) Green s functions data, but used near-field ground motion records, although number of waveforms are di# erent On the other hand, the Green s functions or the structure the time windows used in their inversion is,3 km/s, models for the functions di# ered This earthquake which is fast enough coupled with the relatively long occurred along a structural boundary between low time window ( 2 sec) to estimate rupture delay times seismic velocity sediments in a hanging wall and In their approach, rupture time is not an estimated basement rocks in the footwall (Kato et al,**0) The parameter Instead, the amplitudes of subfault slip source region is structurally heterogeneous, and rate time functions at successive time steps, which many of the di# erences among the results of the propagate at a constant speed from a starting point, three groups listed above may stem from di# erences are obtained If the velocity of the moving window is in the Green s functions used for inversion There- greater than rupture velocity, and if the total durafore, a well-calibrated Green s function is indispensa- tion of the time windows is su$ ciently long, it is ble for building a precise kinematic model (Graves possible to detect a delayed rupture The waveform and Wald,,**+) inversion results obtained by Hikima and Koketsu We chose to use the results obtained by Hikima (,**/) are amenable for detecting rupture delay Aland Koketsu (,**/) because: + ) they estimated hybrid though Asano and Iwata (,**1) also computed a hy- + D structures for each station via inversion of the brid + D crustal structure when obtaining wellwaveforms of aftershocks to calculate accurate calibrated Green s functions, the time window propa- Green s functions, and,) the propagation speed of gation speed of +3 km/s is too slow to detect any 274

The E# ects of Negative Stress Drop on Fault Rupture Fig Stress change distribution and mean slip rate time functions on the fault plane Arrows indicate rupture time delays The yellow star indicates

3 The E# ects of Negative Stress Drop on Fault Rupture Fig Stress change distribution and mean slip rate time functions on the fault plane Arrows indicate rupture time delays The yellow star indicates the initial point of fault rupture Solid orange lines demarcate negative stress drop zones Dashed orange lines encompass zones of negative stress drop and shadow regions See text for further details rupture delay It is worth noting, however, that the and red color indicates stress rises, or negative stress slip distributions of Hikima and Koketsu (,**/) and drops A maximum stress drop of about 3MPa oc- Asano and Iwata (,**1) are very similar Honda et al curred at the deeper asperity of maximum slip, (,**/) used a similar inversion algorithm to that of which is comparable with those for other inland Asano and Iwata (,**1), but di# erent crustal models Japanese earthquakes (Miyatake, +33, b), but much Honda et al (,**/) used only three separate crustal lower than the +** MPa maximum stress drops obmodels, ie, two velocity structure models for sta- tained by Bouchon ( +331) for several earthquakes in tions on the hanging wall and one model for footwall California side stations In contrast, Asano and Iwata (,**1) and Negative stress drop regions are evident in the Hikima and Koketsu (,**/) inferred the structure for shallower half of the fault and also on the eastern each station separately using a genetic algorithm part of the fault (around - km along strike and km and waveform inversion, respectively from the bottom in Fig -, see also B- in Fig ) At Slip and stress change distribution shallow depths, a negative stress drop (Fig -, see also The kinematic slip distribution obtained by Hikima B+ in Fig ) separates the positive stress drop reand Koketsu (,**/) is illustrated in Fig, It pos- gions A+ from A * (Fig ) This negative stress drop sesses a maximum slip value of +1 m at deeper por- (B + ) corresponds to a slip deficit (Fig,) There is also tions of the fault The shallower ( +3 km along the a corresponding slip deficit in the kinematic model strike and +* km up-dip from the bottom) of the two estimated by Asano and Iwata (,**1) and Honda et al main asperities is narrow and has a slip of + m con- (,**) Therefore, the inference that negative stress necting the deeper asperity Based on this slip distri- drops correspond to locations of slip defect seems bution, we calculated the stress change due to the reasonable earthquake (Fig -) using Green s functions derived Delay of rupture propagation by Okada ( +33, ) Blue color indicates stress drops The moment rate (or mean slip rate) time functions 275

4 T Miyatake, A Kato, K Hikima and T Kimura of the subfaults estimated by Hikima and Koketsu namic modeling Despite the procedure adopted, we (,**/) are superimposed on the stress change distri- successfully obtained physically meaningful results bution in Fig Each plot starts at the arrival time In general, negative stress drops are caused by one of the first time window used in the inversion or more of the following factors: + ) frictional properprocess, which propagates at a speed of,3 km/s as ties of the fault,,) inappropriate or inaccurate prenoted above Arrows indicate the delayed onset of sumed fault geometry, or -) poor inversion resoluslip at each location, ie, rupture delay, which we tion In the first explanation, rate-dependent fricmeasured There are delays at many subfaults, par- tional properties possibly account for negative stress ticularly in the upper half part of the fault Regard- drops (Cochard and Madariaga, +33 ) According to ing the relation to stress drop, delayed ruptures are the rate-and state friction law, if A-B *, slip hardenin both positive ( eg A+ and A,; Fig ) and negative ing, and therefore negative stress drop occur where stress drop regions (ie, B +, B,, and B -) A and B are parameters of the rate-and state friction Accounting for rupture delays at negative stress law (see, eg, Scholz,,**,) Likewise, a discrepancy drop regions is straightforward Because a positive between the fault s true geometry and that used for stress drop promotes rupture, a negative stress drop modeling may generate negative stress drops (Ando, (such as in regions B+ and B,) would be expected to,**/), particularly if fault kinks or o# sets decrease arrest or decelerate rupture propagation On the slip locally If we calculate stress distribution on the other hand, rupture onset delays in positive stress assumption of a planar fault, a negative stress drop drop regions, for example, A+ and A,, can be ex- will be obtained, and in this case the region surplained as follows: a negative stress drop at B+ will rounding a fault kink or step will act as a rupture cause delays not only there but also within the as- barrier (Harris and Archuleta, +33+ ) The third explaperities, A+ and A,, because the onset of rupture at nation for negative stress drops is poor model resoluthose two locations is a# ected by any delays at B + tion in the source inversion process Poor resolution In other words, regions A+ and A, are shaded by causes fault slip to leak towards the periphery of the B+ with respect to the hypocenter (yellow star in Fig fault This smoothing around the fault edge pro- ), and the delays measured there on the region de- duces the negative stress drop In the case of the,** pend on the delay at B + On the analogy of travel mid-niigata earthquake, the event occurred within a time calculation, the delays at A+ and A, are the dense seismic network with several seismographs integration of (positive or negative) delay times located very close to the fault, so inversion resolution along the propagation paths is not considered likely to be a significant factor A similar argument may be made for A- lying Kato, Miyatake, and Hirata (,**1) observed that the within the shadow of the negative stress drop re- negative stress drop region is located near an abrupt gion B, Another important factor controlling rup- transition in sediment thickness (see Fig - and Fig ture delay may be peak stress (or fracture energy) 1 of Kato et al,,**0), above the mainshock hypocen- This could account for the delays at A+ and A, if ter They also suggested that the ductile flow of peak stresses were high enough locally However, sediments above the mainshock hypocenter might given that delays within positive stress drop regions act as a soft barrier against coseismic rupture beare consistently observed to occur in the shadows cause the low-velocity sediments would not sustain cast by negative stress drops, we consider that nega- high stresses This argument implies that stresses tive stress drops play a key role in determining the prior to the earthquake were low Low stress or rupture onset of the earthquake ductile flows themselves produce a low or zero stress - Discussion drop, and would easily raise the residual stress level, so cause negative stress drop when combined with We carefully studied the stress drop and the delay the above + ) fault friction property or the above,): of moment rate time functions on the fault of the kink or bend (or small bump on the fault plane),** mid-niigata earthquake, Japan, and found that corresponding to the macroscopic assumption rupture delay is related to the negative stress drop We investigated rupture delay However, to what on the fault In our approach, we did not use dy- extent are di# erences in waveform produced by de- 276

5 The E#ects of Negative Stress Drop on Fault Rupture Fig / Comparison between synthetic velocity waveforms from the kinematic model (Hikima and Koketsu,,**/) and those from same kinematic model without rupture delays detected in Fig - Red and green lines indicate kinematic model waveforms and those without indicate rupture delays, respectively The observed waveforms are also plotted as black lines Numerals are peak amplitudes in cm/s 277

6 T Miyatake, A Kato, K Hikima and T Kimura lay time? If such a waveform change is small enough, it is impossible to detect rupture delay Fig / shows a comparison between synthetic waveforms from the original kinematic model of Hikima and Koketsu (,**/) and those from the same kinematic model without rupture delays detected in Fig The kinematic model waveforms are clearly di# erent from those without rupture delays Waveform inversion is a complex system It depends on many factors including di# erences in source modeling, data processing (low-pass or band-pass filter), and inversion technique Besides, in the inversion using the multitime window method, the rupture delay is not an inversion parameter Therefore, it is di$ cult to calculate the resolution of the rupture delay Instead, we demonstrated that the delays of the rupture times provide clear di# erences in the seismic waveforms from those without delays Conclusion We carried out a detailed investigation of a kinematic model of the,** mid-niigata earthquake and found that negative stress drops play an important role in controlling rupture propagation, particularly rupture onset delays Because a positive stress drop promotes rupture, a negative stress drop is expected to consume elastic energy and impede rupture propagation It is important to emphasize that these results were obtained without reconstructing the dynamic faulting process Acknowledgments This research was supported in part by Special Project for Earthquake Disaster Mitigation in Urban Areas and the Grand-in-Aid for Scientific Research funding from the Ministry of Education, Culture, Sports, Science and Technology of Japan We thank Prof T Yamashita for critically reading the manuscript and for his helpful suggestions References Ando, R,,**/, Development of e$ cient spatio-temporal boundary integral equation method and theoretical study on dynamics of fault zone formation and earthquake rupture, (in Japanese, English Abstract), PhD Thesis, University of Tokyo Asano, K and T Iwata,,**3, Source rupture process of the,** Chuetsu, mid Niigata prefecture, Japan, earthquake inferred from waveform inversion with dense strong motion data, Bull Seism Soc Amer, 33, +,- +*, doi: +* +12/ /*+,**2*,/1 Bouchon, M, +331, The state of stress on some faults of the San Andreas system as inferred from near-field strong motion data, J Geophys Res, +*,, ++, , 1 Cochard, A and R Madariaga, +33, Dynamic faulting under rate-dependent friction, PAGEOPH, +,, +3 / Graves, RW and D J Wald,,**+, Resolution analysis of finite fault source inversion using one-and threedimensional Green s functions: + Strong motions, J Geophys Res, +*0 (B/), 21/ 2100, +* +*,3 /,*** JB 3**-0 Guatteri, M and P Spudich,,***, What can strong-motion data tell us about slip-weakening fault friction laws? Bull Seis Soc Am 3*, Haris, RA and R J Archuleta, +33+, Fault steps and the dynamic rupture porcess :, D numerical simulations of a spontaneously propagating shear fracture, Geophys Res Lett, +2, Hikima, K and K Koketsu,,**/, Rupture processes of the,** Chuetsu (mid-niigata prefecture) earthquake, Japan : A series of events in a complex fault system, Geophys Res Lett, Doi : +* +*,3 /,**/ GRL *,-/22 Honda, R, S Aoi, N Morikawa, H Sekiguchi, T Kunugi and H Fujiwara,,**/, Ground motion and rupture process of the,** Mid Niigata Prefecture earthquake obtained from strong motion data of K-Net and KiK-net, Earth Planets Space, /1, /,1 /-, Irikura, K, H Miyake, T Iwata, K Kamae, H Kawabe and L A Dalguer,,**, Recipe for predicting strong ground motions from future large earthquakes, +-th World Conference on Earthquake Engineering, Paper No +-1+ Kato, A, S Sakai, N Hirata, E Kurashimo, T Iidaka, T Iwasaki and T Kanazawa,,**0, Imaging the seismic structure and stress field in the source region of the,** mid-niigata prefecture earthquake : Structural zones of weakness and seismogenic stress concentration by ductile flow, J Geophys Res, +++, B *2-*2, doi: +* +*,3 /,**/ JB ***+0 Kato, A, T Miyatake and N Hirata,,**1, E# ects of structural heterogeneities of fault zone on source process of the,** mid-niigata Prefecture Earthquake, S +- **2, Japan Geoscience Union Meeting Liu, P and R J Archuleta,,**, A new nonlinear finite fault inversion with three-dimensional Green s functions: Application to the +323 Loma Prieta, California, earthquake, J Geophys Res, +*3, B *,-+2, doi: +* +*,3 /,**-JB **,0,/ Liu, P, S Custódio and R J Archuleta,,**0, Kinematic Inversion of the,** M 0 * Parkfield Earthquake Including an Approximation to Site E# ects, Bull Seismol Soc Am, 30,S+- S +/2 Mikumo, T and T Miyatake, +33/, Heterogeneous distribution of dynamic stress drop and relative fault strength recovered from the results of waveform inversion: The +32 Morgan Hill California earthquake, Bull Seism Soc Amer, 2/, Miyatake, T, +33, a, Dynamic rupture processes of inland earthquakes in Japan Weak and Strong asperities, Geophysical Research Letters, +3, +*+ +* Miyatake, T, +33, b, Reconstruction of dynamic rupture process of an earthquake with constraints of kinematic parameters, Geophysical Research Letters, +3, -3-/, 278

7 The E# ects of Negative Stress Drop on Fault Rupture Miyatake, T, Y Yagi and T Yasuda,,**, The dynamic the October +/, +313 Imperial Valley, California, earthrupture process of the,**+ Geiyo, Japan, earthquake, quake, Tectonophysics, +1/, Geophys Res Lett, -+, L +,0+,, doi: +* +*,3 /,** GL Scholz, C H,,**,, The mechanics of earthquakes and fault- *+31,+ ing, second edition, Cambridge University Press Okada, Y, +33,, Internal deformation due to shear and Wald, D and T H Heaton, +33, Spatial and Temporal tensile faults in a half-space, Bull Seism Soc Amer, 2,, Distribution of Slip for the +33, Landers, California, +*+2 +** Earthquake, Bull Seism Soc Am, 2, Olsen, KB, R Madariaga and R J Archuleta, +331, Three- Yagi, Y,,**, Source rupture process of the,**- Tokachidimensional dynamic simulation of the +33, Landers oki earthquake determined by joint inversion of teleearthquake, Science,,12, seismic body wave and strong ground motion data, Peyrat, S, K B Olsen and R Madariaga,,**+, Dynamic Earth Planet and Space, /0, ,,** modeling of the +33, Landers earthquake, J Geophys (Received June,/,,**2) Res +*0,,0,01,0,2, (Accepted March,-,,**3) Quin, H, +33*, Dynamic stress drop and rupture dynamics of 279

SOURCE MODELING OF RECENT LARGE INLAND CRUSTAL EARTHQUAKES IN JAPAN AND SOURCE CHARACTERIZATION FOR STRONG MOTION PREDICTION

SOURCE MODELING OF RECENT LARGE INLAND CRUSTAL EARTHQUAKES IN JAPAN AND SOURCE CHARACTERIZATION FOR STRONG MOTION PREDICTION Kimiyuki Asano 1 and Tomotaka Iwata 2 1 Assistant Professor, Disaster Prevention