Site assessment and evaluation of spatial earthquake ground motion of Kyeongju

The earthquake hazard has been evaluated for 10 km×10 km area around Kyeongju. The ground motion potentials were determined based on equivalent linear analysis by using the data obtained from in situ and laboratory tests. In situ tests include 16 boring investigations, 4 crosshole, 12 downhole, 26 spectral analysis of surface waves tests, and in the laboratory, resonant column tests were performed. The peak ground accelerations range between 0.141g and 0.299g on collapse level earthquake and between 0.050g and 0.120g on operation level earthquake, respectively, showing the high potential of amplification in the deep alluvial layer in Kyeongju area. Distribution maps of site amplification for the peak acceleration, amplification factors (F_a and F_v) and dominant site period of Kyeongju are constructed using geographic information system tools. The amplification factor based on the Korean seismic design guide underestimated the motion in short range and overestimated the motion in mid-period range in Kyeongju. The importance of site-specific analysis and the need for the improved site characterization method are introduced.     

The Northridge, California, earthquake of 1994: fire ignition by strong shaking

The Northridge earthquake contributed unprecedented detail and quality of data on strong ground motion and on its effects on man-made structures. About 110 fires have been attributed directly to the effects of this earthquake. Two hypotheses for the principal causative agents leading to fire ignition were examined: differential motion and strains in the soil, and inertial forces. The fire-ignition frequency is described with respect to: (1) simple measures of strain in the soil (via density of water pipe breaks, n), (2) occurrence of severely damaged buildings (via density of red-tagged buildings, N), (3) site intensity of shaking, (I_MM), and (4) inertial forces (via peak horizontal ground velocity, v_m). It is shown that the rate of fires (per unit area) ignited by earthquake shaking can be predicted by several empirical equations of comparable accuracy and in terms of common scaling parameters of strong ground motion.     

Response of the Tarzana strong motion site during the 1994 Northridge earthquake

Exceptionally high ground motions (horizontal peak ground acceleration (PGA) of 1.82g) were recorded at the Tarzana Station during the main shock of the 1994 Northridge earthquake (moment magnitude 6.7 at an epicentral distance of 6 km). At the time of the main shock, the instrument was located near the edge of a 21 m-high ridge with side slopes ranging from 3H:1V to 15H:1V. The ridge is underlain by shallow fill and soft rocks of Medelo Formation.

The objectives of this study were to (1) identify the relative contributions of various factors such as local geology, topography, source mechanism, and travel path on the large ground motions recorded at Tarzana Station and (2) develop an analytical model that could adequately predict observed ground motions at the Tarzana site during the Northridge earthquake and at similar sites during future earthquakes. This study is an integral part of a series of inter-related studies referred to as the ROSRINE research (Resolution of Site Response Issues during Northridge Earthquake) project.

The PGA at the surface of competent bedrock (1 km/s shear wave velocity found about 100 m below ground surface) is estimated by Silva [ROSRINE Study (2000)] at 0.46 gravity (g). To identify the source of ground motion amplification, one-dimensional ( ), two-dimensional (TELDYN and SASSI), and three-dimensional (SASSI) analyses were conducted using both recorded aftershock data and an estimated ground acceleration time histories at a 100 m depth.

The results of the analyses indicate that (1) local geology and topography could only partially account for the observed ground motion amplification, and (2) the PGA and response spectra at a point near the edge of the ridge (the location of the instrument at the time of the main shock) is in good agreement with recorded values when the angle of incident of shear waves (SV waves) at 100 m depth is assumed at 30° from vertical. Considering the local geology and variation of shear wave velocity with depth, the 30° incident angle at 100 m depth corresponds to an 8° incident angle of shear waves at the ground surface. This observation is, in general, consistent with the incident angles of shear waves reported from study of the recorded aftershock data.     

Site effects of the 2002 Molise earthquake, Italy: analysis of strong motion, ambient noise, and synthetic data from 2D modelling in San Giuliano di Puglia

On 31 October and 1 November 2002, the Basso Molise area (Southern Italy) was struck by two earthquakes of moderate magnitude (M _L = 5.4 and 5.3). The epicentral area showed a high level of damage, attributable both to the high vulnerability of existing buildings and to site effects caused by the geological and geomorphological settings. Specifically, the intensity inside the town of San Giuliano di Puglia was two degrees higher than in neighbouring towns. Also, within San Giuliano di Puglia, the damage varied notably. The site response in the city was initially evaluated from horizontal-to-vertical spectral ratios (HVSR) from a limited number of strong motion recordings of the most severe aftershocks. Several microtremor measurements were also available. Both data sets indicated the simultaneous presence of two amplification peaks: one around 6 Hz, attributed in previous studies to the strong, shallow impedance contrast among landfill/clay and calcarenites, and one at 2 Hz related to the first S-wave arrivals and predominantly seen only on one receiver component. Further studies performed on weak-motion recordings also showed strong amplification on the vertical receiver component, thus indicating an underestimation of the amplification by the HVSR technique. Additionally, a 2D-model of the geology of the sub-surface was developed, reproducing the flower-shaped structure generated during the late orogenic transpressive regime. The numerical (finite-difference hybrid) simulation reproduced the two peaks of the observed data at slightly higher frequencies. The model also confirmed that the borders of the flower structure define a boundary between amplification levels, with higher amplification inside.     

Site amplification in lava rock on soft sediments

Seismicity and volcanic activity in Iceland are related to the Mid-Atlantic plate boundary that crosses the island. Due to volcanic activity, different sea levels through the ages and glacial drift, the geology in Iceland is quite complex at many sites. In June 2000, two major earthquakes of magnitude 6.6 (M_w) and 6.5 (M_w) occurred in southern Iceland. Ground motion from these main shocks and a number of aftershocks were recorded at the Icelandic strong motion network operated by the University of Iceland. At one of the instrumented sites considerable amplification was recorded on lava-rock overlying alluvial deposits. This site fits poorly in the soil classification systems of most earthquake codes. In this paper the recorded data is analyzed with different methods and the results are compared with the results of a one-dimensional site response analysis. The different methods produce the same characteristic of soil amplification at the site. The findings of this study have important implications for design criteria for sites with similar geology.     

Two-dimensional nonlinear site response analysis of Adapazarı plain and predictions inferred from aftershocks of the Kocaeli earthquake of 17 August 1999

The objective of this study is to investigate the effects of local site conditions in the Adapazarı plain crossing the severely damaged central part of Sakarya during the 17th August earthquake. A two-dimensional model has been used to obtain the free-field motions in the valley and on surface formations. A preliminary geotechnical profile model developed from available but limited soil data was checked using recordings of some specific locations where a strong ground motion accelerometer array was in operation after the main earthquake. The range of validity of the model was assessed and modifications were made to compare with the actual recorded motions. The purpose of building such a model, which was used to study the aftershocks in Adapazarı, is to provide an improved database to be used for the design of structures in the city. This study also attempts to provide greater insight into the local site response phenomena through the use of a two-dimensional nonlinear analysis. Simulated site responses are in global agreement with the recorded data. In general agreement between the amplification functions of the computed and recorded data is satisfactory for the frequencies of engineering interest. Alluvial sites show amplification factors in the range of 4–6 in the frequencies between 0.6 and 2 Hz. A relatively shallow alluvial station, HASTAHANE, experienced considerable amplification for small aftershocks and deamplification for the strongest aftershock in frequencies between 0.5 and 2 Hz. Another station, TOYOTA, with significant deamplification characteristics and a clearly observed peak response frequency shift at the soil resonant frequency may have experienced soil shear failure during the strongest aftershock. Other alluvial stations, SEKER and GENC, show deamplification characteristics at 0.55 and at around 1 Hz. with the peak frequency shifts during the magnitude 5.8 aftershock. In general, site responses are larger in the alluvial basin for all aftershocks except the magnitude 5.8 event.     

Site effects by generalized inversion technique using strong motion recordings of the 2008 Wenchuan earthquake

The generalized inversion of S-wave amplitude spectra from the free-field strong motion recordings of the China National Strong Motion Observation Network System (NSMONS) are used to evaluate the site effects in the Wenchuan area. In this regard, a total of 602 recordings from 96 aftershocks of the Wenchuan earthquake with magnitudes of M3.7-M6.5 were selected as a dataset. These recordings were obtained from 28 stations at a hypocenter distance ranging from 30 km to 150 km. The inversion results have been verified as reliable by comparing the site response at station 62WUD using the Generalized Inversion Technique (GIT) and the Standard Spectral Ratio method (SSR). For all 28 stations, the site predominant frequency F _p and the average site amplification in different frequency bands of 1.0–5.0 Hz, 5.0–10.0 Hz and 1.0–10.0 Hz have been calculated based on the inversion results. Compared with the results from the horizontal-to-vertical spectral ratio (HVSR) method, it shows that the HVSR method can reasonably estimate the site predominant frequency but underestimates the site amplification. The linear fitting between the average site amplification for each frequency band and the V _s20 (the average uppermost-20 m shear wave velocity) shows good correlation. A distance measurement called the asperity distance D _Aspt is proposed to reasonably characterize the source-to-site distance for large earthquakes. Finally, the inversed site response is used to identify the soil nonlinearity in the main shock and aftershocks of Wenchuan earthquake. In ten of the 28 stations analyzed in the main shock, the soil behaved nonlinearly, where the ground motion level is apparently beyond a threshold of PGA > 300 cm/s² or PGV > 20 cm/s, and only one station coded 51SFB has evidence of soil nonlinear behavior in the aftershocks.     

Site response to multi-directional earthquake loading: A practical procedure

Site response to earthquake loading is one of the fundamental problems in geotechnical earthquake engineering. Most site response analyses assume vertically propagating shear waves in a horizontally layered soil–rock system and simply ignore the effect of site response to vertical earthquake motion, although actual ground motions are comprised of both horizontal and vertical components. In several recent earthquakes very strong vertical ground motions have been recorded, raising great concern over the potential effect of vertical motion on engineering structures. Being a step toward addressing this concern, this paper presents a simple and practical procedure for analysis of site response to both horizontal and vertical earthquake motions. The procedure involves the use of the dynamic stiffness matrix method and equivalent-linear approach, and is built in the modern MATLAB environment to take full advantages of the matrix operations in MATLAB. The input motions can be specified at the soil–bedrock interface or at a rock outcropping. A detailed assessment of the procedure is given, which shows that the procedure is able to produce acceptable predictions of both vertical and horizontal site responses.     

Evidence for fault-related directionality and localized site effects from strong motion recordings of the 2003 Boumerdes (Algeria) earthquake: Consequences on damage distribution and the Algerian seismic code

The Algiers–Boumerdes region has been struck by a destructive magnitude 6.8 (M_w) earthquake on May 21, 2003. The study presented in this paper is based on main shock strong motions from 13 stations of the Algerian accelerograph network. A maximum 0.58g peak ground acceleration (PGA) has been recorded at 20 km from the epicenter, only about 150 m away from a PGA of 0.34g, with both a central frequency around 5 Hz, explained by a strong very localized site effect, confirmed by receiver function technique results showing peaks at 5 Hz with amplitudes changing by a factor of 2. Soil amplifications are also evidenced at stations located in the quaternary Mitidja basin, explaining the higher PGA values recorded at these stations than at stations located on firm soil at similar distances from the epicenter. A fault-related directionality effect observed on the strong motion records and confirmed by the study of the seismic movement anisotropy, in agreement with the N65 fault plan direction, explains the SW–NE orientation of the main damage zone. In the near field, strong motions present a high-frequency content starting at 3 Hz with a central frequency around 8 Hz, while in the far field their central frequency is around 3 Hz, explaining the high level of damage in the 3- to 4-story buildings in the epicentral zone. The design spectra overestimate the recorded mean response spectra, and its high corner frequency is less than the recorded one, leading to a re-examination of the seismic design code that should definitively integrate site-related coefficient, to account for the up to now neglected site amplification, as well as a re-modeling of the actual design spectra. Finally, both the proposed Algerian attenuation law and the worldwide laws usually used in Algeria underestimate the recorded accelerations of the 6.8 (M_w) Boumerdes earthquake, clearly showing that it is not possible to extrapolate the proposed Algerian law to major earthquakes.     

The Beni Haoua,Algeria, Mw 4.9 earthquake: source parameters,engineering, and seismotectonic implications

A moderate Mw 4.9 earthquake struck the Beni Haoua (Algeria) coastal area on April 25, 2012. The mainshock was largely recorded by the accelerograph network of the Centre National de Recherche Appliquée en Génie Parasismique (CGS). The same day the earthquake occurred, eight mobile short period stations were deployed through the epicentral area. In this study, we use accelerogram and seismogram data recorded by these two networks. We combined the focal mechanism built from the first motion of P waves and from waveform inversion, and the distribution of aftershocks to well constrain the source parameters. The mainshock is located with a shallow focal depth, ～9 km, and the focal mechanism shows a nearly pure left lateral strike slip motion, with total seismic moment of 2.8?×?10¹⁶ N.m (M_w?=?4.9). The aftershocks mainly cluster on a narrow NS strip, starting at the coast up to 3–4 km inland. This cluster, almost vertical, is concentrated between 6 and 10 km depth. The second part of this work concerns the damage distribution and estimated intensity in the epicentral area. The damage distribution is discussed in connection with the observed maximum strong motion. The acceleration response spectrum with 5 % damping of the mainshock and aftershocks give the maximum amplitude in high frequency which directly affects the performance of the high-frequency structures. Finally, we tie this earthquake with the seismotectonic of the region, leading to conclude that it occurred on a N–S transform zone between two major compressional fault zones oriented NE–SW.     

An earthquake scenario for the microzonation of Sofia and the vulnerability of structures designed by use of the Eurocodes

The study of the site effects and the microzonation of a part of the metropolitan Sofia, based on the modelling of seismic ground motion along three cross-sections are performed. Realistic synthetic strong motion waveforms are computed for scenario earthquakes (M=7) applying a hybrid modelling method, based on the modal summation technique and finite differences scheme. The synthesized ground motion time histories are source and site specific. The site amplification is determined in terms of response spectra ratio (RSR). A suite of time histories and quantities of earthquake engineering interest are provided. The results of this study constitute a “database” that describes the ground shaking of the urban area. A case study of experiment-based assessment of vulnerability of a cast-in-situ single storey, industrial, reinforced concrete frame, designed according to Eurocodes 2 and 8 is presented. The main characteristics of damage index and storey drift are discussed for the purposes of microzonation.     

Exploring the feasibility of earthquake early warning using records of the 2008 Wenchuan earthquake and its aftershocks

Earthquake early warning system (EEWS) is one of the effective ways to mitigate earthquake damage and can provide few seconds to tens of seconds of advanced warning time of impending ground motions, allowing for mitigation measures to be taken in the short term. After the devastating Ms8.0 Wenchuan earthquake, hundreds of M4-6 earthquakes occurred with depth range of 2–24 km. We explore a practical approach to earthquake early warning in Wenchuan area by determining a ground-motion period parameter τ_c and a high-pass filtered vertical displacement amplitude parameter Pd from the initial 3 s of the P waveforms of these aftershocks with M≥4.0. The empirical relationships both between τ_c and M, and between Pd and peak ground velocity PGV for the Wenchuan area are presented. The τ_c result shows that it is systematically greater for slow earthquakes, leading to a possible false alarm. The moment rate function is used to handle the fact that the Pd parameter alone miss the M=8.0 mainshock. These two relationships can be used to detect the occurrence of a major earthquake and provide onsite warning in the area around the station where onset of strong ground motions is expected within seconds after the arrival of the P wave. The robustness of onsite early warning can be increased by using multistation data when the station density is high or by combing τ_c and Pd as a single indicator.     

Case study of strong ground motion variations across cut slope

We evaluate the influence of topography on motions recorded at the base and crest of an approximate 3H:1V, 20 m single-faced slope. The motions were recorded during the 1983 Coalinga earthquake mainshock and two aftershocks. Mainshock peak accelerations at the crest and base transverse to the slope face were 0.59 and 0.38 g, respectively. The spectral amplification of crest motion occurred across T≈0–2 s. Differences between the crest/base motions are postulated to result principally from soil-structure interaction (base instrument is in a structure), variations in local ground response, and topography. Transfer functions quantifying soil-structure interaction (SSI) effects are evaluated and the base motion is modified at short periods to correct it to an equivalent free-field motion. The different levels of ground response at the crest and base are identified based on location-specific measurements of soil shear wave velocities. Differences between crest/base motions not accounted for by SSI or differential ground response are attributed to topographic effects. By these means, topographic spectral amplification (i.e. amplification relative to level ground conditions) is estimated to be about 1.2 at the crest and about 0.85–0.9 at the base across the period range T≈0.4–1.0 s.     

Variation of earthquake ground motions within a very small distance

At present, dense strong motion observation networks have been established in Japan. One of the important findings based on these networks is that strong ground motions are quite site-dependent. Characteristics of observed ground motions at nearby stations can exhibit a significant variation, even when the stations are within several hundreds of meters. These observations raise one important question; if characteristics of strong ground motions exhibit large variations even for smaller regions, we should be concerned about the application of observed or predicted ground motions for the assessment of structures. In particular, if ground-motion parameters such as PGA, PGV, Spectral Intensity, etc., exhibit large variations for smaller regions, their use for the seismic design and practice will be subject to restriction. In other words, the evaluation of variation of these parameters is an important issue. From such a point of view, the authors investigated the variation of observed ground motions within a very small distance in this study. First of all, fifteen couples of adjacent strong motion stations in Japan, where the distance is within 100 m, were listed up based on our field reconnaissance. Then, microtremor measurements were carried out at each of the station pairs. Next, variation of recorded earthquake ground motions for the station pairs was examined based on various ground motion parameters and response spectra. Moreover, we investigated the key factor which is affecting the variation of observed ground motions.     

Estimation of Site Response in Kachchh,Gujarat, India,Region Using H/V Spectral Ratios of Aftershocks of the 2001 <Emphasis Type="Italic">M</Emphasis><Subscript><Emphasis Type="Italic">w</Emphasis></Subscript> 7.7 Bhuj Earthquake

Site response in the aftershock zone of 2001 Bhuj M_w 7.7 earthquake has been studied using the H/V spectral ratio method using 454 aftershocks (M_w 2.5–4.7) recorded at twelve three-component digital strong motion and eight three-component digital seismograph sites. The mean amplification factor obtained for soft sediment sites (Quaternary/Tertiary) varies from 0.75–6.03 times for 1–3 Hz and 0.49–3.27 times for 3–10 Hz. The mean amplification factors obtained for hard sediment sites (hard Jurassic/Mesozoic sediments) range from 0.32–3.24 times for 1–3 Hz and 0.37–2.18 times for 310 Hz. The upper bounds of the larger mean amplification factors for 1–3 Hz are found to be of the order of 3.13–6.03 at Chopadwa, Vadawa, Kavada, Vondh, Adhoi, Jahwarnagar and Gadhada, whereas, the upper bounds of the higher mean amplification factors at 3–10 Hz are estimated to be of the order of 2.00–3.27° at Tapar, Chopadwa, Adhoi, Jahwarnagar, Gandhidham and Khingarpur. The site response estimated at Bhuj suggests a typical hard-rock site behavior. Preliminary site response maps for 1–3 Hz and 310 Hz frequency ranges have been prepared for the area extending from 23–23.85 °N and 69.65–70.85°E. These frequency ranges are considered on the basis of the fact that the natural frequencies of multi-story buildings (3 to 10 floor) range between 1–3 Hz, while the natural frequencies for 1 to 3 story buildings vary from 3–10 Hz. The 1–3 Hz map delineates two distinct zones of maximum site amplification (>3 times): one lying in the NW quadrant of the study area covering Jahwarnagar, Kavada and Gadadha and the other in the SE quadrant of the study area with a peak of 6.03 at Chopadwa covering an area of 70 km × 50 km. While the 3–10 Hz map shows more than 2 times site amplification value over the entire study area except, NE quadrant, two patches in the southwest corner covering Bhuj and Anjar, and one patch at the center covering Vondh, Manfara and Sikara. The zones for large site amplification values (∼3 times) are found at Tapar, Chopadwa, Adhoi and Chobari. The estimated site response values show a good correlation with the distribution of geological formations as well as observed ground deformation in the epicentral zone.     

Re-evaluation of the large deep earthquake of January 21, 1906

The large deep earthquake of January 21, 1906 is re-evaluated using old seismogram data and updated analysis techniques. From the P and pP-P time data the hypocentre parameters are determined as follows: origin time, 13h 49min 35s; latitude, 33.8°N; longitude, 137.5°E; depth, 340 km. The body-wave magnitude m_B is re-evaluated from the amplitude and periods of P, PP and S waves. The average value of 7.4 is obtained. This value is the smallest among any values assigned previously to this shock, and it is denied that the earthquake is the world's largest deep shock in this century. The focal mechanism is estimated from the P-wave first motions and amplitude distribution of P and S waves. Synthetic body waves are used to constrain the mechanism and to determine the seismic moment. The mechanism solution suggests the down-dip compression typical of this region. A seismic moment of 1.5 × 10²⁷ dyn · cm is obtained. This value and the re-evaluated value of m_B are consistent with the moment-_B relation obtained for other deep earthquakes.     

Site dependence of far-source ground motions during the Wenchuan earthquake

This paper aimed to examine the site dependence and evaluate the methods for site analysis of far-source ground motions. This was achieved through the examination of frequency content estimated by different methods based on strong ground motions recorded at twelve far-source stations in Shandong province during the Wenchuan earthquake. The stations were located in sites with soil profiles ranging from code classes Ⅰ to Ⅲ. Approaches used included the Fourier amplitude spectrum (FAS), the earthquake response spectrum (ERS), the spectral ratio between the horizontal and the vertical components (H/V), the spectral ratio between the spectra at the site and at a reference site (SRRS), and coda wave analysis (CWA). Results showed that major periods of these ground motions obtained by FAS, ERS and H/V ratio methods were all evidently larger than site dominant periods; the periods were also different from each other and mainly reflected the frequency content of long period components. Prominent periods obtained by the SRRS approach neither illuminated the long period aspect nor efficiently determined site features of the motions. The CWA resulted in a period close to site period for stations with good quality recordings. The results obtained in this study will be useful for the evaluation of far-source effect in constructing seismic design spectra and in selecting methods for ground motion site analysis.     

The space and time distribution characteristics of the shear stress field for the sequence of the Wuding earthquake

Follow Chen and Duda's model of spectral fall-off of (3, the dependence of peak parameters of ground motion, peak displacement dm, peak velocity vm and peak acceleration am, upon the environment stress (0-values are studied using near source seismic digital recordings for the sequence of the Wuding, Yunnan, M = 6.5 earthquake, in which, as a new thought, the peak parameters are assumed to be related to the medium Q-value. Three formulae for estimating the environment stress (0-values by the peak parameters of three types of ground motions are derived. Using these formulae, the environment stress (0-values are calculated for the sequence of the Wuding earthquake. The result show that (0-values calculated by the three formulae are constant largely, the averages of (0 are in the range of 5.0～35 MPa for most earthquakes. It belongs to the high-stress earthquakes sequence: the high-stress values are restricted to the relatively small area closely near to the epicenter of the main shock. The fine distribution structure for the contours of the environment stress (0-values is related closely to the strong aftershocks. The analysis of spatial and temporal feature of (0-values suggests that the earthquakes sequence in a rupture process generated at the specific intersection zone of seismo-tectonics under high-stress background.     

VS30, site amplifications and some comparisons: The Adapazari (Turkey) case

The aim of this study was to investigate the role of V_S30 in site amplifications in the Adapazari region, Turkey. To fulfil this aim, amplifications from V_S30 measurements were compared with earthquake data for different soil types in the seismic design codes. The Adapazari area was selected as the study area, and shear-wave velocity distribution was obtained by the multichannel analysis of surface waves (MASWs) method at 100 sites for the top 50 m of soil. Aftershock data following the Mw 7.4 Izmit earthquake of 17 August 1999 gave magnitudes between 4.0 and 5.6 at six stations installed in and around the Adapazari Basin, at Babalı, Şeker, Genç, Hastane, Toyota and Imar. This data was used to estimate site amplifications by the reference-station method. In addition, the fundamental periods of the station sites were estimated by the single station method. Site classifications based on V_S30 in the seismic design codes were compared with the fundamental periods and amplification values. It was found that site amplifications (from earthquake data) and relevant spectra (from V_S30) are not in good agreement for soils in Adapazari (Turkey).     

S-wave velocity structures of sediments estimated from array microtremor records and site responses in the near-fault region of the 1999 Chi-Chi,Taiwan earthquake

The 1999 Chi-Chi, Taiwan earthquake, M_W = 7.6, caused severe damage in the near-fault region of the earthquake. In order to evaluate site effects in the near-field strong motions we estimate S-wave velocity structures of sediments at four sites using array records of microtremors. We also recalculated S-wave velocity structures at other four sites previously reported. To show the validity of the estimated S-wave velocity structures we separate empirical site responses from aftershock records using the generalized inversion method and show the agreement between empirical and theoretical site responses. We also show an observed fact that suggests soil nonlinearity during the Chi-Chi earthquake by comparing horizontal-to-vertical spectral ratios (HVRs) for main shock records with HVRs for aftershock records. Then we calculate one-dimensional equivalent-linear site responses using the estimated S-wave velocity structures and the main shock records observed on the surface. It is found that site amplification due to thick (about 6 km) sediments is one of the important factors for explaining the long-period velocity pulses of about 5 to 10 sec observed at sites in the footwall during the Chi-Chi earthquake. It is also found that the theoretical site responses of shallow soft sediments at sites that sustained severe damage in the hanging wall shows significant amplification around 1 sec. As the amplitude of velocity pulses with period around 1 sec is most critical in causing damage to ordinary buildings of moderate heights, our results suggest that the 1-sec period velocity pulses, amplified by the site response of shallow sediments should contribute to the severe damage during the Chi-Chi earthquake.