Content uploaded by Kou-Cheng Chen
Author content
All content in this area was uploaded by Kou-Cheng Chen on Jun 04, 2018
Content may be subject to copyright.
A preview of the PDF is not available
An Observation of Rupture Pulses of the 20 September 1999 Chi-Chi, Taiwan, Earthquake from Near-Field Seismograms
Abstract
The ground-velocity recordings of the 20 September 1999, Chi-Chi, Taiwan earthquake recorded at stations near the ruptured fault trace show a simple, large-amplitude, and long-period pulse following the S wave, which is closely associated with the surface faulting and the rupture process of thrust faulting. The conspicuous pulse on the ground-velocity seismogram following the S -wave arrival, called the S 1 phase, is interpreted as the superposition of the rupture pulses that nucleate at an asperity near and underneath the station and propagate up-dip and laterally along the fault plane toward the surface stations. The arrival times of the S 1 phase and the onsets of the permanent displacement at stations near and along the ruptured fault trace increase with hypocentral distance, suggesting that the rupture of the Chi-Chi earthquake might have initiated at the hypocenter of the mainshock and propagated both upward and laterally from south to north. On the basis of the travel-time differences between the S 1 phase and the direct S wave at the stations near and along the ruptured fault trace, the rupture velocities varied from 2.28 to 2.69 km/sec, with an average rupture velocity of about 2.49 km/sec. The rupture velocities decreased from south to north.
... The near-fault forward directivity (NFFD) ground motions have been recorded in many earthquakes during the past 50 years (e.g., Port Hueneme earthquake in 1957 [42]; Parkfield earthquake in 1966 [43]; San Fernando earthquake in 1971 [19,56]; Landers earthquake in 1992 [27]; Northridge earthquake in 1994 [69]; Kobe earthquake in 1995 [39]; Marmara earthquakes in 1999 [3]; Chi-Chi earthquake in 1999 [32]; L′Aquila earthquake in 2009 [33]; Christchurch earthquake in 2011 [23]). Their distinct features in terms of dynamic source characteristics as well as large-amplitude impulsive horizontal and vertical waveforms that increase the damage potential on structures have appealed many seismological and engineering studies. ...
We investigate the role of major seismological (magnitude, pulse period, fault length, seismic activity, orientation of incident seismic wave with respect to fault-strike) and geometrical (fault-site geometry) parameters to understand the variations in ground-motion demands due to near-fault directivity (NFD) effects. To this end, we used a suite of probabilistic strike-slip earthquake scenarios and established the elastic spectral amplitude distributions conditioned on the above investigated parameters. The probabilistic earthquake scenarios also provided information on the sensitivity of directivity dominant near-fault (NF) ground motions to mean annual exceedance rates. We implemented different narrow-band directivity models to observe the significance of seismological modeling in the directivity dominant NF ground-motion amplitudes. The observations from these case studies suggest that each one of the above parameters have implications on the amplitude and spatial variation of directivity dominating NF ground-motion demands. The influence of each investigated parameter on NFD spectral amplitudes is dependent of the implemented directivity model. We also establish some rules to map the spatial extent of directivity dominant ground motions considering the variations in the investigated seismological parameters. The outcomes of the paper can be used to incorporate the NFD effects into design spectra representing different annual exceedance rates.
... The near-fault forward directivity (NFFD) ground motions have been recorded in many earthquakes during the past 50 years (e.g., Port Hueneme earthquake in 1957 [42]; Parkfield earthquake in 1966 [43]; San Fernando earthquake in 1971 [19,56]; Landers earthquake in 1992 [27]; Northridge earthquake in 1994 [69]; Kobe earthquake in 1995 [39]; Marmara earthquakes in 1999 [3]; Chi-Chi earthquake in 1999 [32]; L′Aquila earthquake in 2009 [33]; Christchurch earthquake in 2011 [23]). Their distinct features in terms of dynamic source characteristics as well as large-amplitude impulsive horizontal and vertical waveforms that increase the damage potential on structures have appealed many seismological and engineering studies. ...
We collected a data set of mainshocks and their respective largest foreshocks of 38 earthquake sequences in Taiwan. The plot of local magnitude, M L , of a mainshock (denoted by M Lm ) versus M L of its largest foreshock (denoted by M Lf ) shows an increase in M Lm with M Lf . This indicates that for Taiwan’s earthquakes the bigger the largest foreshock is, the larger the mainshock is. The plot of the epicentral distance, Δ (in km), from the largest foreshock to the mainshock versus M L of the mainshock exhibits a weak increase in Δ with M Lm as Δ<10 km. The plot of the focal depth of the largest foreshock and that of the mainshock shows a linear increase in the former along with the latter for most event-pairs. Let T be the interval between the occurrence time of the largest foreshock and the mainshock. The plot of T versus M Lm exhibits that the mainshock will occur within 5 days, with the highest probability of 1 day, after the occurrence of the largest foreshock. Let H be the hypocentral distance between the largest foreshock and the mainshock. The plot of T versus H reveals a slight increase in T with H when T>1 day.
We collected a data set of mainshocks and their respective largest foreshocks of 38 earthquake sequences in Taiwan. The plot of local magnitude, M L , of a mainshock (denoted by M Lm ) versus M L of its largest foreshock (denoted by M Lf ) shows an increase in M Lm with M Lf . This indicates that for Taiwan’s earthquakes the bigger the largest foreshock is, the larger the mainshock is. The plot of the epicentral distance, Δ (in km), from the largest foreshock to the mainshock versus M L of the mainshock exhibits a weak increase in Δ with M Lm as Δ <10 km. The plot of the focal depth of the largest foreshock and that of the mainshock shows a linear increase in the former along with the latter for most event-pairs. Let T be the interval between the occurrence time of the largest foreshock and the mainshock. The plot of T versus M Lm exhibits that the mainshock will occur within 5 days, with the highest probability of 1 day, after the occurrence of the largest foreshock. Let H be the hypocentral distance between the largest foreshock and the mainshock. The plot of T versus H reveals a slight increase in T with H when T >1 day.
A new automatic baseline correction method based on the Hilbert spectral analysis (HSA) is proposed to recover a relatively reasonable time history of baseline offset, permanent displacement, and stable peak ground motion displacement (PGD) from a raw near-fault ground motion record. This method can get rid of the selection of the start time of strong motion (corresponding to t1 in traditional methods) by an adaptive extraction procedure, then a stable PGD as well as an accurate permanent displacement can be obtained. For baseline offsets, there is no two-stage-segment assumption in this procedure, and the extracted time histories of these offsets include complex frequency changes, which is more consistent with the explanations of causes for baseline offsets. All causes of baseline offsets are treated as contaminants that exist in all frequency ranges in a record. Uncontaminated components can be extracted from the original record by the HSA, then information about strong motions can be retained as much as possible. The residual part is regarded as a heavily contaminated component of which the energy of the noise rivals that of the real ground motion. It is corrected based on only one time point. Finally, the corrected record is contributed by both of the uncontaminated and adjusted contaminated components. The HSA method, which depends on the energy distribution of original record in different frequency component, can obtain the baseline offset, PGD, and permanent displacement reasonably, stably, and automatically.
Multi-pulse characteristics of near-fault ground motions, such as the number of inherent pulses, pulse periods, and amplitudes, have notable influences on the response of structures. To investigate these important parameters, an automatic detection procedure, which is conducted on the rough pulse signal that is extracted by the HHT method, is proposed in this work. This procedure can localize all inherent pulses in the time domain independently and discontinuously. Important parameters can be automatically obtained at the same time. Then, statistical relationships between these multi-pulse parameters and earthquake parameters, including moment magnitudes, site conditions (Vs30), rupture distances and types of faults, are investigated comprehensively. With an increasing number of pulses, the multi-pulse ground motions are more likely to be recorded in relatively limited areas. All pulse periods in a velocity record are similar to each other and can be represented by the period of the pulse with the largest energy (TPE). TPEs are almost identical to periods of the first pulse in the time domain (TP1). They are related not only to magnitudes but also to fault-types and site conditions. New empirical models are proposed in this work according to fault-types that can predict most TPEs across all magnitudes (from 5.7 to 7.6). For amplitudes, all PGVs of inherent pulses can be expressed by the PGV of the pulse with the largest energy (PGVE) in a linear attenuation relationship. PGVEs are related to rupture distances, site conditions, and fault-types. New empirical models are also developed in this work.
A series of large-scale shaking table tests on a typical subway station structure were carried out for pulse-like earthquake motions. The propagation of earthquake motions in soils is explored. The effects of velocity pulse parameters (i.e., pulse shape, pulse frequency (fp), and pulse duration (Tpd)) on dynamic responses of the structure are analyzed from the viewpoint of the dynamic characteristics of soils. It is found that the site amplifies the frequency components of dynamic excitation in specific frequency bands (i.e., the amplification frequency band, AFB). With an increase in earthquake motion intensity, the AFB gradually moves towards low frequencies. Additionally, the bandwidth gradually narrows. The velocity pulse increases the amplitude of the first-mode shear deformation of soils and thus the seismic internal force response of underground structures. In addition, the effect of a one-sided pulse in terms of increasing the first-order shear deformation amplitude of soil is more pronounced than the effect of a two-sided pulse. Pulse-like earthquake motions with long pulse duration and multiple pulse cycles may trigger remarkable resonance-like phenomena of soils and further aggravate soil and structure responses when the main pulse frequencies are located in the AFB of soils.
This article investigates the spatial distribution, predominant direction, and variations in the intensity measures (IMs) with orientation for classified pulse‐like and nonpulse motions during Chi‐Chi Mw 7.6 earthquake. The results show evidence of high polarization for long‐period spectral accelerations at relatively large source‐to‐site distances (50–100 km) north of the Chelungpu fault. The polarization of long‐period motions shows a clear correlation with the directivity parameters’ isochrone directivity predictor and ξ, indicating a connection between directionality and rupture directivity. The variation in strong‐motion directionality with the period is also studied. The discrepancy in directionality caused by strong directivity increases with the period from 1 to 10 s, which confirms a clear correlation of period‐dependent directionality with directivity effects. This study finds stronger directionality of pulse‐like motions than nonpulse motions for long periods over 3 s with higher maximum‐to‐median and maximum‐to‐minimum IM ratios. For periods over 3 s, the maximum‐to‐median ratios of pulse‐like motions are higher than the mean prediction of the Shahi and Baker (2014a) model, whereas those of nonpulse motions are lower than the prediction. However, this study does not find simple and clear results for the directions of the maximum component at different periods for pulse‐like and nonpulse motions. Despite clear differences between the unidirectional fling‐step and bidirectional forward directivity pulses, the effects of fling‐step and forward directivity are actually coupled in the waveforms.
We investigated 22 broadband teleseismic records of the 1999 Chi-Chi earthquake to determine its temporal and spatial slip distribution. The results show an anomalous large slip region centered about 40 to 50 km north of the hypocenter at a shallow depth. The largest amount of slip was about 6 - 10 m. The slip near the vicinity of the hypocenter had a relatively smaller amount of slip. The spatial slip distribution pattern coincides well with the observed strong-motion displacement and surface break. In the largest dislocation region, the slip was dominated by dip-slip. Some strike-slip component in the middle of the fault was found during the rupture. The Southern portion of the fault showed relatively constant rupture velocity with an average slip of about 1 m, whereas the northern portion of the fault showed significant variations in rupture velocity and produced a large amount of slip.
The strong motion accelerograph recordings of the 24 January 1980 main shock and the 27 January 1980 aftershock of the Livermore Valley earthquake sequence are analyzed for systematic variations with azimuth or station location. The variation of the peak accelerations with epicentral azimuth is apparently reversed for the two events: the main shock accelerations are larger to the south, and the aftershock accelerations are larger to the northwest. We eliminate the site effects by forming the ratio of the peak accelerations recorded at the same station, after correcting for the epicentral distance. This analysis indicates that source direcUvity caused a total variation of a factor of 10 in the peak accelerations. Comparison of this variation with the spatia ! extent of the after-shock sequences suggests that the strong directivity in the radiated accelera-tions is the result of unilateral ruptures in both events. The accelerograms recorded at 10 stations within 35 km of the events were digitized to analyze the azimuthal variation of the rms acceleration, the peak velocity, and the radiated energy flux. The variation of rms acceleration corre-lates almost exactly with the variation of the peak accelerations. This correlation is analyzed using both deterministic and stochastic models for the acceleration waveforms. The peak velocities, corrected for epicentral distance, vary with azimuth by a factor of 5 for both events, while the radiated energy flux varies by a factor of 30 for the main shock and 15 for the aftershock. The peak velocities are strongly correlated with the radiated energy flux. The radiated seismic energies are estimated to be 2.6 ± 0.9 x 102° dyne-cm for the main shock and 1.5 ± 0.3 x 1020 dyne-cm for the aftershock.
Teleseismic body waves recorded at IRIS Global Seismographic Network are analyzed to investigate the source rupture process of the 1999 Chi-Chi earthquake. The azimuthal coverage of seismograph stations is good enough to resolve some details of heterogeneous moment-release. The source parameters obtained for the total source are: (strike, dip, slip)=(22°, 25°, 67°); the seismic moment=2.8×1020 [Nm] (Mw=7.6); the source duration=30 [s]; the fault area=75×40 [km2]; the average dislocation=3.1 [m]; the stress drop=4.2 [MPa]. These values reveal typical features of large subduction zone earthquakes. A notable aspect of this earthquake lies in the highly heterogeneous manner of the moment-release. In the southern source area including the initial break, several small subevents were derived, which may be interpreted as ruptures of fault patches with a smaller length-scale. On the other hand, two large asperities with a length-scale of 20km were obtained in the northern part of the source area: one in a shallow western part and the other in a deeper eastern part. The local stress drop on the former asperity was as high as 20MPa and the fault slip exceeded 8m, which is comparable to the ground displacement as inferred from both strong motion and GPS data. A high dip-angle nature of branch-faults in accretionary prism resulted in a considerable uplift of the hanging wall. グローバル広帯域地震計観測網の遠地実体波記録を用いて, 1999年台湾中部地震の震源過程を調べた.観測点の方位分布は良好であり,大まかな不均一断層すべり分布が抽出された.主な震源パラメーターは次の通り.断層メカニズムは東傾斜面の衝上断層で(走向,傾斜,すべり角)=(22°, 25°, 67°);地震モーメント=2.8×1020Nm (Mw=7.6);破壊継続時間=30s;断層面積=75×40km2;平均すべり量=3.1m;平均応力降下=4.2MPa.これらの震源パラメーターは沈み込み帯におけるプレート境界大地震の標準的な値である.しかし今回の地震の際だった特徴は断層すべりの不均一性にある.まず,破壊開始点を含む断層面の南側の領域では,比較的小さい断層パッチが次々と動いたことが示唆された.一方,断層面の北側では,さしわたし約20kmの2つの大きなアスペリティ(大きい地震時すべりを起こす領域)が動いたとの結果が得られた. 1つは西側の浅いところ,もう1つは東側の深めに位置する.浅い方のアスペリティでは応力降下が26MPa,すべり量は8m強である.このすべり量は, GPSデータや強震記録から推定された断層上盤側の最大変位量と同程度である.付加帯内部への分岐断層の特徴として断層面の傾斜が比較的高角であるため,プレート上面の地震の場合と比べて,上盤側の隆起量が大きい.仮に震源が海底にあったとすると,地震の規模から経験的に予測されるより大きい津波を引き起こしたはずである.
The 21 September 1999 Chi-Chi earthquake (ML=7.3) was the largest inland quake occurred in Taiwan in the past 100 years. According to direct field observations, a significant thrust rupture was occurred along the active Shuangtung fault and Chelungpu fault. From the integrated displacement derived from free-field strong motions around the source area, horizontal displacements over 9 m and vertical offsets reaching 3.4 m are obtained at the station TCU068 located nearby the northern section of Chelungpu fault. Based on the field observations of slip distribution over the entire 77-km surface break and the integrated displacement waveforms, we have arrived at the following findings: An obvious small offset at about 3 sec prior to the principal offset is observed at two sites close to the southern part of Shuangtung fault. This small offset is not shown on the other stations. We propose that a relatively small event, probably with a magnitude of about 6.0, occurred about 3 sec before the major rupture, that is located at the Shuangtung fault about 8 km east of the Chelungpu fault. The induced northward propagating rupture along the Chelungpu fault has reached a maximum permanent offset over 9 m at the northern section of fault. This northern section of the Chelungpu fault was identified as one of the heaviest damaged areas.
Particle motions in a foam rubber model of shallow-angle thrust faulting show many features different from those commonly assumed in dislocation models of subduction thrusts (Brune, 1996). As a complement to a physical foam rubber experiment, we have carried out dynamic simulation using a 2D lattice numerical model. The model is constructed as a triangular block sliding over a rectangular elastic block. It consists of a 2D set of particles interacting with each other by non-linear Hooke's forces and obeying Newton's equations of motion. Rough surfaces are introduced on the contact plane of the two blocks to simulate realistic friction. The numerical simulations demonstrate several key features of thrust faulting, including stick slip, fault opening, and strong breakout and shaking at the hanging-wall toe (the wedge-shaped tip of the outcropping hanging-wall block) of the thrust fault, all consistent with the foam rubber experiments. The stick-slip motion shows an approximate time and slip repeatable behavior. Each slip event is characterized by a self-healing pulse associated with fault opening and frictional locking. The rupture pulse propagates steadily along the fault. When rupture reaches the toe of the fault outcrop, the hanging wall breaks away from the foot wall and creates a large opening vibration of the hanging wall. The peak acceleration observed at the hanging-wall toe of the fault outcrop is about 3 to 4 times the peak acceleration in the center part of the fault, and about 2 to 3 times the motion of the foot wall. Such a large increase in peak acceleration at the toe is caused mainly by multiply reflecting stress waves trapped in the wedge-shaped hanging wall of the fault. The strong asymmetry of the particle velocity between the hanging wall and foot wall is an important feature of the results consistent with the foam rubber model but different from standard dislocation models. This dynamic result illustrates the important effects of thrust faulting geometry on fault slip and ground motion.
Model studies have demonstrated that the stopping phase, an event which originates at the termination of rupture, may be identified on long-period strain recordings. A tentative identification of the stopping phase has been made on ultra-long-period seismograms of three major earthquakes (Montana, 1959; Chile, 1960; and Alaska, 1964). The events chosen lead to reasonable estimates of the length of rupture associated with the earthquake. A second type of stopping event, the breakout phase, must occur when a rupture intersects a free surface. A two-dimensional model study indicates that the breakout phase should be a prominent seismic event, particularly if the first motion is emergent. A review of studies of seismograms of earthquakes which produced surface faulting indicates that a prominent second event is often observed. However, there does not appear to be an adequate criterion to distinguish the breakout phase from the pP phase. Thus no certain identification can be made.
A simple, uncontrived model of the rupture during the San Feruando earth- quake can explain the main features of the particle velocity traces derived from the accelerograms recorded at Pacoima Dam. This result, combined with the probable small effect of surface topography on the velocity traces, strengthens the case for acceptance of the peak particle velocity at Pacoima Dam (115 em/sec) as a valid ground-motion parameter for design purposes ill earthquake engineering. Most of the conspicuous motion on the velocity traces during the first 4 sec after triggering seems to result from thrust faulting, starting at a focal depth within several kilometers of 14 km, on a fault surface dipping at least 50 ° and extending only part way to the surface at a velocity near 2.5 km/sec. The data also indicate that this faulting continued to the surface at a slower rupture velocity (less than 2 km/sec) along a less steeply dipping surface. The amount of relative offset across the fault surface is difficult to determine, both because of inherent limitations in the two-dimensional model and because of nouuniqueness in the fitting of the data. The estimates of this dislocation, however, are consistent with the wide range of values reported by other authors in studies using various types of data. The data are also consistent with a model suggested by Alewine and Jordan (1973) and Trifunac (1974) in which the total dislocation has a minimum near the center of the fault surface, with approximately equal amounts of total offset on the fault near the hypocenter and near the Earth's surface.
Near fault tip strong motion records from the northern part of the major earthquake (Mw=7.6), namely the Chi-Chi earthquake on September 21, 1999 in central Taiwan demonstrated systematic differences on the hanging wall and footwall, and simulated by the finite element method. The extraordinary ground motion differences on either side of the northern fault tip can be explained by a 2-dimension kinematic source model with fault rupture breaking to surface. In this study, the earthquake faulting was considered as bilateral from the center of a low angle thrust fault which is 30 km in length with a dip angle of 31°. Based on waveform modeling, the source rupture velocity, rise-time and dislocation of 2.0 km/sec, 5 sec and 6 meters, respectively are suggested. The results of this study show that on the northern part of the Chi-Chi earthquake fault there was lower rupture velocity and longer rise-time of the fault slip than that previously reported. Furthermore, the effects of surface breaking from the fault movement contributed large ground deformations near the fault tip and, consequently, induced extensive damage.
'The large displacement pulse occurring at Pacoima dam approximately 2.5 s after the triggering of the accelerogram is identified as the shear radiation emanating from the point of initial rupture, this interpretation suggesting that the San Fernando, California, earthquake was initiated with massive but localized rupture in the hypocentral region. On the basis of the resulting S-P time at Pacoima dam and teleseismic observations of the reflected phase pP, together with the estimated uncertainties in the local hypocentral determination of Allen et al. (1973), the point of initial rupture is located at 34ø27.0'N, 118ø24.(rW and h -- 13 kin. The breakout phases resulting from the rupture of the earth's surface are tentatively identified on the Pacoima dam accelerograms and teleseismic World-Wide Standard Seismograph Network seismograms. For the San Fernando earthquake the breakout phases do not appear to be particularly energetic, but this may only reflect the especially energetic nature of the initial rupture phases. The inferred length of faulting and the estimated origin times of the initial rupture and the breakout phases yield an average rupture velocity. of at least 2.8 km/s. Although they are subject to the uncertainties of the suggested effects of source propagation and of single-station observations, the source parameters of the initial rupture are estimated as follows: source dimension, 6-3 km; stress drop, 350-1400 bars; average slip, 4.6-9.2 m; and seismic moment, 1.7-0.85 X 10 dynes cm. A 0.8-s time interval after the arrival of Sx contributes 40% of the radiated energy flux at Pacoima dam; the initial rupture event may have generated as much as 80% of the total energy radiated by the San Fernando earthquake. The stress difference accompanying the initial rupture is comparable to an estimate of the north-south compresslye stress that is assumed to maintain the continued uplift of the San Gabriel Mountains; the stress drop obtained for the entire faulting process, however, is smaller than either of these by 1-2 orders of magnitude. That the San Fernando earthquake was apparently initiated with massive but localized failure not representative statically or dynamically of the full faulting process most likely reflects a highly nonuniform distribution of strain energy density in the incipient source region. This possibility adds a new dimension to laboratory studies, numerical models, and conventional seismic investigations of the earthquake mechanism.