The source process of the 1995 Kobe earthquake is determined using teleseismic body waves. The source parameters obtained for the total source are: focal mechanism [strike, dip, rake] equal to [233°, 86°, 167°], nearly pure strike-slip with an EW compression axis; the seismic moment, Mo= 2.5×1019 Nm (Mw= 6.9); and source duration T=11s. The rupture process consists mainly of three subevents with source durations of 5-6 s: two are nearly pure strike-slip with slightly different fault strikes, and the other is dip-slip. Combining the source location with the aftershock distribution, we infer that the first subevent is a bilateral rupture and the later subevents are unilateral propagating NE and reaching the Kobe area. The total fault length is 40 km, the averaged dislocation is 2.1 m, and the averaged stress drop is 8 MPa. These source parameters indicate that the Kobe earthquake was a typical shallow inland earthquake which occurred on a previously mapped Quaternary fault zone. The directivity toward Kobe determined from teleseismic data was probably one of the main factors responsible for the heavy damage. The combined use of regional networks and teleseismic networks as demonstrated in this study will continue to be important for rapidly assessing the overall social impact of an earthquake in the very early stage of sequence.
The geodetic data, strong-motion waveforms, and far-field waveforms from the 1995 Kobe earthquake (MJMA 7.2) were inverted to determine the rupture process. The geodetic data, including the horizontal displacements measured at 43 stations and the vertical displacements measured at 61 stations, were provided by the Geographical Survey Institute of Japan and the Japanese University Consortium for GPS Research. We assumed two fault planes, one each Awaji Island and Kobe, with different strike directions and dip angles, and which were consistent with the active fault map and the aftershock distribution. We first inverted the geodetic data alone to determine the heterogeneous distribution of slip on the assumed faults. Second, the strong-motion records provided by the Japan Meteorological Agency, the Committee of Earthquake Observation and Research in the Kansai Area, and the Kobe City government were inverted to obtain the temporal variation of slip. Third, the displacement waveforms of far-field P- and SH-waves were analyzed. Finally, we performed a triple joint inversion to determine the source model most suitable for explaining all the data sets. The solution of the joint inversion expressed very strong ground motions, which well simulated the ones observed in Kobe at frequencies lower than 1 Hz. Concentrated slips are recovered in the shallow northeastern portion of the Nojima fault on Awaji Island and beneath the Strait of Akashi between Awaji and Kobe. Using the geodetic data, we also examined the possibility of another hidden fault beneath the extremely damaged area in Kobe, but matching of the data was not improved by the introduction of this fault.
We analyze strong ground motions and geodetic data simultaneously with the multiple time-window method in order to infer the source process and the fault geometry of the Hyogo-ken Nanbu earthquake . Considering aftershock distribution and geodetic data, we assume two faults with a step-over near the hypocenter and different strike angles: a northeastern fault (Fault #1) and a southwestern fault (Fault #2) . We divide Fault #1 further into two parts with different dip angles. Large slip occurred at the deeper part (below 8 km) of Fault #1. The rupture propagated smoothly along this area, but decelerated near the boundary of the above-mentioned two parts of Fault #1. We therefore believe that the boundary is a geometric barrier. Large slip occurred over the entire area of Fault #2. The rupture at the shallower part of the fault was of long duration . The seismic moment is estimated to be 2.9×1019Nm. The geodetic data clearly suggest that Faults #1 and #2 have a step-over at a shallower part beneath northern Awaji Island. We also consider other fault configurations . When not dividing Fault #1, some of the geodetic data cannot be explained. We invert the data, assuming the one-fault model whose shallower part bends to the northwest at the southwestern side, and this model explains the data as well as the previous one. In this sense, we cannot deny that the two faults connect at a deeper part . Comparing our results with previous inversion results, we found that two characters of main-shock faulting relate to the magnitude of the largest aftershocks: fault segmentation and slip distribution along the dip direction . This suggests that a detailed investigation of the source process of a main shock is useful for predicting the magnitude of the largest aftershock.
We estimated the source process of the 1995 Hyogo-ken Nanbu (Kobe), Japan, earthquake based on: 1) locating buried fault planes in the Kobe area by examining particle motion at observation stations in the near-source area and 2) multi-time window linear waveform inversion of strong ground motion seismograms. S-wave particle motion diagrams created theoretically show reverse rotation in the horizontal plane at two stations located on opposite sides of the intersection of the earth's surface and the extension of the buried fault plane. We simulated ground motion at various locations surrounding the buried fault plane and compared their particle motions with observed records to obtain constraints on the location of the fault plane. At least two planes are needed for the rupture area northeast of the epicenter if the causative fault is assumed to be made up of a few perfectly planar structures, and the rupture extended at least 26 km northeast from the epicenter. Using a fault model consisting of three planes, two planes on the Kobe side and on the Awaji side, and assuming a step-over at the Akashi strait based on the aftershock distribution, we performed a waveform inversion. The main rupture extended about 45 km horizontally. Three regions had relatively large moment releases: 1) around the rupture starting point; 2) in the shallow (less than 10 km) part of the Nojima fault, which extends along the northwest shore of Awaji Island: and 3) deep (about 10 km) under Kobe City. Even though the second subevent on the Nojima fault had a large moment release, it did not generate pulsive waves, because its rise time was long. The two remarkable pulses seen in the seismograms in the Kobe area came from the first and third subevents.
Near-source ground motions, teleseismic body waveforms, and geodetic displacements produced by the 1995 Kobe, Japan, earthquake have been used to determine the spatial and temporal dislocation pattern on the faulting surfaces. A linear, least-squares approach was used to invert the data sets both independently and in unison in order to investigate the resolving power of each data set and to determine a model most consistent with all the available data. A two-fault model was used, with a single rupture plane representing faulting beneath Kobe and a second plane representing slip underneath Awaji Island. The total seismic moment is estimated to be 2.4×1019Nm (MW 6.9), with rupture partitioned such that about 40% of the slip was relatively deep (5-20 km) and northeast of the epicenter toward Kobe, and about 60% was toward the southwest and shallower (mostly 0-10 km) beneath Awaji Island. Analysis of the slip model indicates that the ground motions recorded within the severely damaged region of Kobe originated from the region of relatively low slip (about 1 m) deep beneath Kobe and not from the shallow, higher slip regions (about 3 m) beneath Awaji Island. Although the slip was relatively low beneath Kobe, the combined effects of source rupture directivity, a short slip duration, and site amplification conspired to generate very damaging ground motions within the city.
We estimated the high-frequency wave radiation process on the fault plane of the 1995 Hyogo-ken Nanbu earthquake (MJMA=7.2), which was a strike slip event along active inland faults, by the envelope inversion of strong-motion acceleration seismograms. The horizontal extent of the region that radiated high-frequency (2-10 Hz) waves is about 45 km, which matches the source region estimated from the inversion of strong-motion displacement waveforms. High-frequency waves were mainly radiated at the periphery of this source region, while a clear gap of radiation was seen in the center. Along the Nojima fault, both high- and low-frequency wave radiation were large. This is explained by the fact that the rupture broke the ground surface here. Highfrequency waves were also radiated at the step-over of faults between Awaji Island and the Kobe area. The step-over is interpreted to have behaved as a geometrical barrier and generated high-frequency waves when ruptured. The radiation of high-frequency waves in the northeastern part from the hypocenter to Kobe was small in the large moment release areas estimated from the displacement waveform inversion, and large at their peripheries. This suggests these high-frequency waves were stopping phases from the subevents corresponding to the large moment release areas. The extent of the source region of high-frequency waves roughly matches that of the heavily damaged areas. We inferred that the strong high-frequency waves that came directly from the source was one of the causes of the heavy damage to low-rise structures.
A clear aftershock gap was found in the central portion of the belt shaped aftershock area of the 1995 Hyogo-ken Nanbu earthquake. We term this gap an "earthquake bright spot." In this paper, the rupture growth process of the main shock is discussed in the framework of the earthquake bright spot model. The resultant process is as follows: The rupture initiated at the depth of 14 km, which was the bottom of the seismogenic zone, and it propagated in a semicircular direction. The early stage of rupture was very slow with small dislocation. After about 4 s, it accelerated by successive fault step-over and some subfaults were produced in the confined region, the "bright spot." The preliminary duration of 4s satisfies Umeda's empirical formula, which expresses the relationship between the nucleation time and earthquake magnitude. Powerful high-frequency waves of 0.5-1.5 Hz were concentrically radiated from the bright spot within a short time. After about 9s, just beneath Kobe City, predominant long-period waves of 5-6 s were produced by moment release from smooth dislocation of a fault.
It is important for disaster prevention and public safety design to know whether subsurface faults exist, not only in Kobe, which was heavily damaged by the 1995 Hyogo-ken Nanbu earthquake, but also in any urban area which is tectonically active. However, there is still little information on subsurface faults. In this paper, we propose a new approach for sensing subsurface faults using teleseismic-wave array records. Our new method employs an imaging technique similar to the reflection method, but is much less expensive and can be used in urban areas. The target of our method is not the velocity structure itself, but rather the position and size or a subsurface fault. The imaging algorithm is based on the steepest descent method, which is also used in waveform inversion of reflection data. However, since iteration is not required by our approach, only one forward computation together with reverse-time computation of the wavefleld is required. The method proposed in this paper consists of the following steps: 1) pre-processing observed records, 2) forward calculation, 3) calculation of the steepest descent direction of the model parameters, and 4) image analysis and identification of faults. The feasibility of the method is demonstrated using SH synthetic data.
We estimated the basement depth in Kobe City from array records of aftershocks of the 1995 Hyogo-ken Nanbu earthquake. The array measurements were conducted along the Sumiyoshi River in Kobe City using 12 stations from north to south within a distance of about 2 km. Travel time analyses of initial P- and S-waves were made to obtain travel time residuals due to the sediments. We also obtained the travel time difference between S-wave and SP-wave (i.e., P-wave converted from S-wave) at the basement-sediment boundary. By compiling three travel time data sets, we estimated the distribution of the basement depth. The basement depth suddenly increases from 0 to 1 km within a distance of 1 km, and then the depth gently increases to the south.
Site amplifications during aftershocks of the 1995 Hyogo-ken Nanbu earthquake were examined in and around the disaster area of Higashinada ward, Kobe City, Japan. Five temporary stations together with the KOB station (CEORKA) were installed just after the main shock to arrange an array for strong ground motion observations across a severe damage band about 1.5 km wide. Several aftershock records were obtained from these stations. The peak values of the observed ground accelerations and velocities at the soil sites were 3-20 times larger than those at the rock site. Spectral ratios between the soil sites and the rock site have peak frequencies characteristic to each station pair. We also estimated a thickness of about 1 km for the sedimentary basin just under the soil sites using SP-converted waves recorded at all soil sites. Waveform simulations using a dislocation point source in a horizontally stratified medium were performed to explain the observed records at both the rock and soil stations. We found that large amplifications at the soil sites were generated by a thick sedimentary layer including low-velocity surface layers. The heavy damage caused by the main shock is therefore strongly connected with the sedimentary structure.
Ground motion records from aftershocks of the January 17, 1995, Hyogo-ken Nanbu earthquake provide a good opportunity for evaluating the Osaka basin edge effects on wave propagation in the Kobe area. Aftershock ground motion recorded in a small array deployed at sites within and outside the heavily damaged zone, east of Kobe City, show very large amplification at the sedimentary sites located close to the basin edge. The peak velocities at sediment sites were up to 15 times larger than those at the rock site. To model the effect of the basin edge structure and the source radiation in the frequency range 0.1-2.5 Hz, we calculate double-couple point source SH seismograms using two-dimensional finite difference techniques. By comparing the synthetic seismograms with the transverse component of the observed waveforms from two selected aftershocks, we find that the northern edge of the Osaka basin has a fault-like structure, dipping north. The rather complex basin edge structure causes seismic energy to focus at sites close to the edge. The constructive interference of different basin generated waves amplifies the ground motion. The focal area and the amplification due to the basin structure depend on the source location, i.e. the arrival direction of seismic waves. Our results indicate that the basin edge effect is one of the reasons why the zone heavily damaged during the main shock lies along a narrow belt. They also indicate that the thin surface sediments in the Kobe region amplify the ground motion at frequencies around 2 Hz considerably.
To estimate the amplification characteristics of ground motion in the heavily damaged belt zone in Kobe City during the 1995 Hyogo-ken Nanbu earthquake, three-dimensional wave propagation analyses of a two-dimensional, deep irregular underground structure model with vertical discontinuity were performed using the hyperelement method for incident planar waves expected from the wavefields due to the source mechanism. The observation records from Kobe University, a rock site, are used as control data. The ground motion at the surface of the Osaka group layers and at ground surface are calculated, The effects of the deep irregular underground structure and shallow surface layers on ground-motion amplification are discussed. The analytical results show that ground motions in the heavily damaged belt zone was amplified due to a focusing effect in the deep irregular underground structure as well as the shallow surface layers, and that the calculated maximum acceleration distributions coincide closely with the distribution of structural damage.
The 1995 Hyogo-ken Nanbu earthquake of JMA magnitude 7.2 caused serious damage widely throughout the Hanshin and Awaji areas in the Kinki district of southwest Japan . Damage to artificial structures was observed widely, even in the Osaka area 30-50 km east of the epicenter. Most of the damage was accompanied by ground failures such as liquefaction, lateral spreading and landslides of reclaimed lands along the coastal zone and river channels and ponds in the inland areas . Damage to artificial structures not accompanied by ground failure was also observed in several areas. Most of this type of damage was concentrated in narrow zones along the Butsunenji-yama fault and the Uemachi fault trending in the N-S direction . As the concentration of damage suggests that seismic waves were strongly amplified in limited zones around faults, we examined the focusing effect of seismic waves as one of possible interpretation . A velocity structure model composed of four homogeneous layers with curved interfaces referring to the result of a seismic reflection survey was utilized for this purpose. Two-dimensional calculations of wave paths and surface amplitudes support the possibility that vertical displacement of a subsurface geologic structure along the fault caused seismic waves to focus at the margin of the subsided block, resulting in the narrow zone of damage.
A later phase propagating horizontally southeast was detected to the south of Japan Railway line in the lowlands of Sumiyoshi, Kobe, by the tripartite installed at the Nada junior high and high school station of the Sumiyoshi Linear Seismic Observation Array, for aftershocks of the 1995 Hyogo-ken Nanbu earthquake. A frequency wave-number spectral analysis revealed a phase velocity of 300-400 m/s and energy concentration in the frequency range 2.0-4.0 Hz. According to the delay after S-wave arrival, its phase velocity and its propagating direction, it is believed this phase was a secondary wave converted from incident S-wave at the geological fault along the base of the Rokko mountain range.
After the Hyogo-ken Nanbu earthquake of 1995, an array observation of aftershocks was done from north to south along the Sumiyoshi River in Kobe City. The degree of damage varied drastically following the river from north to south. The spectral ratio shapes for different sites within the heavily damaged area were almost the same, suggesting that the variation in soil condition among the sites is small in this area. A clear difference in spectral ratio was observed between the observation points in the heavily damaged area and other areas. The amplification of spectral ratio within the frequency range 2-3 Hz, which possibly affected the wooden houses in the area, showed drastic change around the Hankyu Kobe line, with large values between SH3 and SME. This trend of variation in spectral ratio might be similar to the damage distribution of wooden houses. The above observation indicates that the site might have a relation to the damage distribution of wooden houses, even though aftershocks are a low strain level phenomena.
We obtained an attenuation relation of response spectra to magnitude, source-to-site distance and soil classification by performing regression analysis of strong-motion data recorded in California. We applied the attenuation relation to the 1995 Hyogo-ken Nanbu earthquake and found that the estimates at pre-Quaternary sites agree well with the observed values. Peak ground accelerations on pre-Quaternary stratum in Kobe are consistent with the source models, and they are estimated to be about 350 cm/s2, while the estimates for Awaji Island vary with the source models from 600 to 1, 200 cm/s2. We calculated the amplification factors of Quaternary stratum during the Hyogo-ken Nanbu earthquake by taking ratios between the observed records at Quaternary sites and estimates on pre-Quaternary stratum. As the amplifications in the heavily damaged area are large near the natural period of wooden houses (04-0.5 s), these short-period amplifications might be one of the reasons Kobe suffered extensive damage. The estimated amplification factors generally agree with those estimated from other earthquake records as well as with one-dimensional wave propagation theory, while the amplifications at alluvial sites were larger at periods of more than 1 s and might be affected by surface waves.