Analysis of Ground-Motion Amplification Characteristics in Kobe City Considering a Deep Irregular Underground Structure Interpretation of Heavily Damaged Belt Zone During the 1995

J. Phys. Earth, 44, 577-590, 1996

Analysis of Ground-Motion Amplification Characteristics in Kobe City Considering a Deep Irregular Underground Structure Interpretation of Heavily Damaged Belt Zone - during the 1995. Hyogo-ken Nanbu Earthquake-—

Masato Motosaka 1,* and Masayuki Nagano 2

1 Disaster Control Research Center, Faculty of Engineering, Tohoku University, Aramaki-Aoba, Aoba-ku, Sendai 980-77, Japan 2 Kobori Research Complex , Kajima Corporation, KI-Building, Minato-ku, Tokyo 107, Japan

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.

unknown seismic fault beneath the belt zone (e.g. 1, Introduction Watanabe et al., 1995). Some reports suggest the The characteristic damage distribution around "Nagisa phenomena" (e.g. Suzuki et al., 1995), Kobe City during the 1995 Hyogo-ken Nanbu which means the amplification of ground motion earthquake is reported for a "heavily damaged belt due to shallow surface geology near the basin edge. zone" (e.g. Takemura and Tsuji, 1995) 1-2 km wide Is the shallow surface geology the only reason for (shown in Fig. 1), where the highest death toll and the belt zone? At a very high amplitude level, most of the injured were concentrated and where maximum acceleration exceeded 800 cm/s2 and seismic intensity, as measured by JMA, was VII. maximum velocity 100 cm/s, basically causing the Seismic observations recorded very large velocity soil to become a high damping material. Therefore, and acceleration amplitudes perpendicular to the the "Nagisa phenomena" is unlikely to occur, al- Rokko faults. The maximum recorded acceleration though one-dimensional amplification could be at Fukiai (FKI), in the belt zone, exceeded 800 cm/s2 expected. Interested in the deep irregular under- and the maximum acceleration and velocity at ground structure of the northwest edge of the Osaka Takatori (TKT) exceeded 600 cm/s2 and 130 cm/s, basin, the authors performed an analytical inves- respectively, in two horizontal directions, although tigation (Motosaka and Nagano, 1995) from the the maximum acceleration observed at Kobe Uni- standpoint that the damage in the belt zone is versity (KBU), a rock-site observation station, was attributed to the amplification of ground motion about 300 cm/s2 or less and the maximum velocity due to focusing effects in the deep irregular under- was 55 cm/s (BRI, 1995; Toki et al., 1995). ground structure as well as in the shallow surface Various interpretations of the damaged belt zone layers. In this study, the focusing effect is used in a have been reported. Some reports suggest an wide meaning, including amplification due not only

Received July 15, 1995; Accepted November 20, 1995 * To whom correspondence should be addressed .

577 578 M. Motosaka and M. Nagano

Fig. 1. Heavily damaged zone in Kobe City during the 1995 Hyogo-ken Nanbu earthquake and schematic figure diagram of an underground structure near the Rokko fault zone. The JMA seismic intensity was VII in the hatched heavily damaged zone. Line a-a' corresponds to the analyzed section. The ratios of structural damage along line b-b' are referred to in Fig. 19. The locations of the Hankyu line (HK), JR line (JR), and Hanshin line (HS) are indicated. Route 2 (R2) is located between JR and HS. Locations of major strong-motion observation sites: Kobe University (KBU), Motoyama (KOB), Fukiai (FKI), Takatori (TKT), Shin Kobe (SKB), and the Kobe Harbor Office (KBH) are indicated. to focusing of diffracted body waves at the irregular (R43) is analytically investigated. Section 2 describes boundary and vertically propagated body waves but the underground structure model investigated. The also to the superposition of basin-induced surface analytical method is briefly described in Sec. 3. In wave and vertically propagated body waves. The Sec. 4, the ground motion at the surface of the Osaka focusing effect is also suggested by such authors as group layers for incident planar waves assumed from Akamatsu et al. (1995) and Nakagawa (1995), but the wavefield due to the source mechanism are ad- analytical investigations have not been performed. dressed. A planar 2-D analysis for vertical incident In the meantime, for investigating the structural S-waves is performed, as well as a 3-D (commonly damage to various structures and for reconstruction called 2.5-D) analysis for the oblique incident of the damaged area, it is important to evaluate the SH-wave propagating along the fault. The latter features of ground motion in the source area with incident wavefield is assumed for Kobe in the strong geological variety through simulation anal- direction of SH-wave dominance in the source ysis. In this case, it is necessary to estimate the mechanism of the earthquake; namely, the strike- strength of ground motion at the earth's surface as slip of the vertical fault. The observation records well as at the upper surface of the Osaka group obtained at Kobe University are used as control layers. data in these analyses. Section 5 discusses ground This paper describes the wave propagation in two- motion at the soil surface affected by amplification dimensional (2-D) irregular underground structure resulting from the shallow surface layers, indicating models of an orthogonal cross section of the fault strong nonlinearity. Section 6 discusses the sensitiv- through Kobe University in Nada ward (refer to ity of the adopted Q-values to the maximum ground line a-a' in Fig. 1). The amplification characteristic acceleration. The maximum values obtained analyt- of ground motion in the heavily damaged zone from ically are compared to the observation records and the Hankyu line (HK) through the JR line (JR), the estimated distribution of the maximum accelera- Route 2 (R2), and Hanshin line (HS) to Route 43 tions are compared to the distribution of structural

J. Phys. Earth Effects of Deep Irregular Underground Structures on Ground Motion in Kobe City 579 damage. Section 7 makes concluding remarks. clear geological discontinuity is seen due to the fault between the mountain area and the sediment. Figure 2. Underground Structure and Analytical Model 2 shows the bedrock depth contours of the Osaka With regard to the deep underground structure basin based on a geological survey and seismic of the Hanshin area, it is suggested that the bedrock refraction data (Iwasaki et al., 1995). In particular, comprised of Rokko granite has a very steep gra- the boring data indicating 724 m at Ninomiya-cho, dient at the foot of the Rokko mountain area, Chuo-ku, Kobe (Iwami, 1987) is added in the figure. corresponding to the edge of the Osaka basin. A The gravity anomaly around the Osaka basin (e.g.

Fig. 2. Contours of depth to the bedrock in the Osaka basin based on refraction and boring data (after Iwasaki et al., 1995). The bedrock depth of-0.72 kin at Ninomiya-cho in Kobe, based on boring data, is added (after Iwami, 1987).

Fig. 3. Seismic reflection profile along the Ishiyagawa line by the Committee of Earthquake Observation and Research in the Kansai Area (after Iwasaki et al., 1995).

Vol. 44, No. 5, 1996 580 M. Motosaka and M. Nagano

Kobayashi et al., 1995) also suggests a vertically reflection survey (refer to Fig. 3) made by the discontinuous deep underground structure along the Committee of Earthquake Observation and Re- south foot of the Rokko mountain range. This search in the Kansai Area (CEORKA) suggest that suggests that there are thick sediment layers con- there is a major vertical discontinuity almost sisting of: from the top, alluvium, diluvium, Osaka orthogonal to the Rokko fault, beneath the HK and Kobe group layers and weathered granite on along the Ishiyagawa measuring line including the the granite bedrock south of the vertical discon- Kobe University site (Iwasaki et al., 1995). The tinuity. velocity structure, up to a depth of GL-70 in below In numerical modeling of a 2-D section orthog- the strong-motion site at Kobe University, has also onal to the Rokko fault, the following geological been investigated by CEORKA (Iwasaki et al., information is taken into account. The results of a 1995).

Fig. 4. Two-dimensional models of an underground structure orthogonal t M o the Rokko fault plane. (a) odel-1 and (b) Model-2. HK, JR , R2, HS, and R43 indicate Hankyu line, JR line line , Route 2, Hanshin , and Route 43 (Harbor Expressway), respectively . The 2.5-D wave propagation characteristics are investigated for the incident SH-wave propagating along the unde i rground structure with a vertical ncident angle of ƒÆ. The incident wave is assumed from the wavefield from th e source mechanism of nearly strike-slip on a vertical fault .

J. Phys. Earth Effects of Deep Irregular Underground Structures on Ground Motion in Kobe City 581

Table 1. Soil profiles of rock site in Model-1 and Model-2.

Table 2. Soil profiles of transit zone in Model-1. Fig. 5. Dispersion curves of fundamental modes of Love and Rayleigh waves. (a) Rock site; (b) sediment site.

the Q-value of each layer is assumed to be Vs (m/s)/15. The sensitivity for the lower Q-value for the layer with a shear-wave velocity of 500 m/s is investigated in Sec. 6 taking into account the high strain level during the earthquake. The dispersion Table 3. Soil profiles of sediment site in curves of the surface waves for the velocity structure Model-1 and Model-2. are shown in Fig. 5(a) and (b) for the rock site and sediment site, respectively, for consideration of wave propagation of the secondary generated surface wave at the rock-sediment boundary. Two analytical models, Model-1 and Model-2, are used in the wave propagation analyses to investigate the sensitivity of the vertically discontinuous part in the underground structure. Model-1, with 3 hyperelements, has two vertical discon- In this study, the two 2-D models shown in Fig. tinuities. The transit zone is between the two vertical 4 are used. In these models, the shallow surface boundaries. Model-2 is a simplified version having layers indicating strong nonlinearity during the only two hyperelements. earthquake are excluded. The effects of these layers are evaluated by one-dimensional nonlinear analysis 3. Analytical Method (equivalent linear analysis) as described in Sec. 5. The hyperelement method is applied to the wave The sediment below the Osaka group is modeled as propagation analysis of the 2-D models described a two-layer structure. The soil properties of the above. This method was originally proposed by layered model for the rock site, the transit zone Kausel and Roesset (1977) for 2-D planar strain and between the rock and sediment sites and the sediment axisymmetric cases. The authors have extended it site are listed in Tables 1-3. The assumed velocity to a 3-D response problem for a 2-D structure for structure for the rock site is determined from the the case of planar incident waves with an arbitrary geological structure by CEORKA (Iwasaki et al., azimuth and incident angle (Nagano and Motosaka, 1995) and Iwata et al. (1995). The velocity structure 1995). In this method, an objective structure is of the sediment site is also based on Iwata et al. divided into plural regions with horizontal layers. (1995), being determined from the delay of the Each region is modeled using a hyperelement. The arrival time due to sediment layers for the SP- nodes of the hyperelement are placed at the inter- converted waves of an aftershock. It is noted that faces of the thin-layer elements on both sides of a

Vol. 44, No. 5, 1996 582 M. Motosaka and M. Nagano

Fig. 6. Rotated waveforms and corresponding Fig. 7. Distribution of maximum velocity at the pseudo velocity response spectra. (a) Rotated surface of upper part of Osaka group layers (for velocity waveforms. The particle orbits are Models-1 and -2 subjected to vertical incident calculated in the two indicated time sections; i.e. S-waves). (a) Model-1 and (b) Model-2. The section-1 (7.5-9.0s) and section-2 (10.0-11.5 s) solid line indicates the distribution of maximum (refer to Fig. 10). (b) Pseudo velocity response velocities in the OR direction. The broken line spectra (h =5%). The thick lines are the spectra indicates maximum velocities in the UD direc- of rotated waveforms and the thin lines are those tion. The thin broken line indicates the maxi- of original observed waves in the NS and EW mum velocity from 1-D analysis. directions. into account the source's location. It is noted that region. The displacement at an arbitrary internal the SH-wave is dominant in the location in Kobe point in the hyperelement is calculated using an City for seismic waves from sources with a source interpolation function specified as the wave function mechanism of nearly strike-slip in a vertical fault. once nodal displacements are obtained. Each node The waveforms are calculated by a commonly used has three degrees of freedom. The method enables frequency response analysis using FFT.

3-D response analysis not only for incident body The velocity ground motion in the NS and EW and surface waves but also for a source moving at directions at Kobe University as observed by a constant velocity. CEORKA (Toki et al., 1995) are rotated to com- The objective frequency range of the analyses is ponents orthogonal (OR, N 160•‹E) and parallel (PA, 0-10 Hz. For the two models (Model-1 and N250•‹E) to the fault axis; that is, 20•‹ retrograde to Model-2), planar 2-D response analyses are per- the EW direction. The rotated waveforms and the formed first. For Model-2, the 3-D response anal- corresponding pseudo velocity response spectra are ysis is performed for an oblique incident SH-wave shown in Fig. 6. It is noted that the peak near and propagating along the irregular 2-D structure. The at 2.5 s of the response spectrum in the original vertical incident angle is assumed to be 45•‹ taking EW component disappears in the PA component ,

J. Phys. Earth Effects of Deep Irregular Underground Structures on Ground Motion in Kobe City 583 as shown in Fig. 6. In the simulation analysis, the amplification characteristics is small. The maximum velocity waveform in the OR component is used as value of horizontal velocity increases with distance the control data for the Kobe University location from the rock-sediment boundary. The maximum after eliminating the effects of the soft, weathered values are large 0.5-2.5 km from the boundary. The granite layers with shear-wave velocities of less maximum amplification factors for sediment to rock than 850 m/s using a 1-D equivalent linear analysis. are about 2.5 in the two models, and occur at the That is, the incident waveform is determined such 0.8 km point located between JR and R2. that the calculated waveform becomes the control In the region corresponding to reclaimed land, motion at Kobe University. It is noted that the x, the maximum velocity response decreases as the y, and directions defined in Fig. 4 correspond to points approach the coast. It is recognized that the the OR, PA, and UD directions, respectively. maximum velocities of the 2-D analyses are about 75% larger than those of the 1-D analysis near the 4. Ground Motion Characteristics at Osaka Group 0.8 km point and 60% larger at a point 1.8 km from Layer Surfaces the rock-sediment boundary. It can be seen from Firstly, the amplification characteristics in the this figure that the amplitude of vertical velocity sediments of the two models are compared for the induced by the irregular underground structure is planar 2-D analysis due to vertical incident S-waves. large in the sediments near the rock and decreases The maximum velocities of the two models are with distance from the boundary. The ratios of compared in Fig. 7. The maximum velocities of the vertical and horizontal motion are less than 0.5, soil site using the 1-D analysis for sediment are except in the region near the boundary (0-0.5 km). indicated by the broken line. It is found from this To investigate the amplification in the 2-D figure that the overall shapes of the maximum analyses, the calculated waveforms for Model-2 are distributions for Model-1 and Model-2 are almost shown in Fig. 8(a), (b), and (c) for the total wavefield the same. The sensitivity of the transit zone to the in the OR direction, the total wavefield in the UD

Fig. 8. Calculated velocity waveforms at the surface of Model-2 subjected to vertical incident S-waves. (a) OR direction; (b) UD direction; (c) scattered wave contents in the OR direction. The arrows in (b) and (c) at the rock-sediment boundary correspond to the wave propagation at a group velocity of the fundamental mode of Rayleigh waves in the Airy phase.

Vol. 44, No. 5, 1996 584 M. Motosaka and M. Nagano

Fig. 9. Calculated velocity waveforms at the surface of Model-2 subjected to oblique incident SH-waves. (a) OR direction; (b) PA direction; (c) UD direction. The arrows in (b) and (c) at the rock-sediment boundary correspond to the wave propagation at group velocities of the fundamental modes of Love and Rayleigh waves in each Airy phase. direction and the scattered wavefield in the OR propagating body waves and the scatted waves direction. The scattered wavefield is defined as the comprising various diffracted waves. The amplifica- subtraction of the incident wavefield (i.e. the wave- tion of ground motion due to the superposition of field obtained by 1-D analysis using the velocity waves passing through different paths seems to structure of the sediment site) from the total produce a kind of focusing effect. Dispersive wave wavefield. The arrows in Fig. 8(b) and (c) at the propagation can also be seen in the induced vertical rock-sediment boundary correspond to wave prop- motions. It is recognized that the waves propagate agation at a group velocity (0.24 km/s) of the at a higher velocity than 0.24 km/s, with the excep- fundamental mode of the Rayleigh wave in the Airy tion of the latter part of the waveforms. The duration phase. It is recognized that the amplitudes of the increases with distance from the vertical discon- scattered waveforms in the OR direction do not tinuity; that is, the duration at a point 3 km from monotonously decrease but fluctuate in space , and the rock-sediment boundary become 15 s, compared only the waves in the latter part of the waveforms to 6 s at Kobe University. propagate at a velocity of 0.24 km/s. This means Next, 3-D response characteristics caused by that the scattered waveforms consist mainly of the oblique incident SH-waves are investigated using body waves caused by refraction at the rock- Model-2. The calculated velocities in the OR , PA, sediment boundary and their transmitted and and UD directions are shown in Fig. 9. The reflected waves at the layer boundaries of the maximum velocity in the OR direction becomes sediment site rather than of the basin-induced 126 cm/s at a point 1 km from the rock-sediment surface waves, which contribute to the dispersive boundary, compared to 55 cm/s at Kobe University . waveforms of the total wavefield. The amplification The waveforms in the PA direction are the induced of ground motion between the JR and R2 seems horizontal motions at the rock-sediment boundary , to be caused by the superposition of the vertically consisting mostly of anti-planar motions including

J. Phys. Earth Effects of Deep Irregular Underground Structures on Ground Motion in Kobe City 585

Fig. 10. Calculated and observed particle orbits of velocity at Kobe University for Model-2 subjected to oblique incident SH-waves. (a) Section-1 (7.5-9.0 s); (b) section-2 (10.0-11.5 s). The two time sections are indicated in Fig. 6.

Love waves. The arrow in Fig. 9 (b) corresponds to the wave propagation at the group velocity

(0.27 km/s) of the fundamental mode of Love waves in the Airy phase. Wave propagation characteristics in the OR and UD directions are similar to the results of the vertical incidence case.

Figure 10 shows the particle velocity orbits for the two major wave groups at Kobe University. The prograde orbits indicated in the report by Iwasaki

(1995) are expressed analytically. The calculated Fig. 11. Distribution of maximum response values particle orbits closely coincide with the observed at the surface of Model-2 subjected to oblique orbit in the second wave group. The calculated and incident SH-waves. The solid line indicates the observed particle orbits show discrepancy in the first distribution of maximum velocities in the OR wave group. This may be because of the assumption direction, the chain line in the PA direction, and that the strike of the seismic fault is N70•‹E, and the thick broken line in the UD direction. The that only the SH-waves are taken into account for thin broken line indicates the maximum velocity from 1-D analysis. the incident wave. Actually, the azimuth angle of the incident SH-waves may be different for the first wave group. The effect of other incident waves may In Fig. 12, the velocity waveforms at points P-1 also lead to this discrepancy. to P-4 in the sediment sites are compared to those

Figure 11(a) and (b) show the maximum velocity of the 1-D analysis. It is recognized that the and maximum acceleration for Model-2 subjected interference from scattered waves caused by the to the oblique incident SH-waves. The maximum rock-sediment boundary occurs at a later time and velocities in the OR and UD directions are slightly with increasing distance from the boundary, as larger than those in the vertical incidence case, but marked in this figure. almost the same. The maximum velocity in the PA The acceleration response spectra at the same direction is less than half of that in the OR direction. points are compared to those of the 1-D analysis, The shape of the maximum acceleration distribu- as shown in Fig. 13. The shapes of the spectra are tion is similar to that of the velocity distribution, almost the same and the spectral values are pro- indicating that the maximum values become large portional to the maximum accelerations at points near a point 0.8 km from the rock-sediment P-1 and P-2. At points P-3 and P-4, the spectra boundary, where maximum acceleration becomes shapes and values are almost the same as those of

800 cm/s2 against 300 cm/s' at Kobe University (the the 1-D analysis. The spectral values at P-2 are rock site). amplified in the wide period range shorter than 4 s.

Vol. 44, No. 5, 1996 556 M. Motosaka and M. Nagano

The above investigation confirms that the ground propagation analysis is performed using waveforms motion between JR and R2 was amplified by su- calculated at equally spaced points (100 m) for perposition of vertically propagating body waves Model-2 subjected to oblique incident SH-waves. and scattered waves, including body and surface The conventional equivalent linear analysis is used waves, from the rock-sediment boundary, which can as a practical method for estimating the distribu- be interpreted as a focusing effect. tions of maximum velocities and accelerations taking into account the relatively small effect of liquefaction 5. Surface Ground Motion Characteristics Taking in the objective region in this study. The diluvium into Account Shallow Surface Layers and alluvium layers on the Osaka group layers are To estimate nonlinear soil amplification due to taken into account. Figure 14 shows the configura- shallow surface layers, the 1-D nonlinear wave tion of these shallow surface layers, which was determined based on boring data (e.g. Japan Society of Soil Mechanics and Foundation Engineering, Kansai Branch, 1992). The thickness of these layers

Fig. 12. Comparison of calculated velocity waveforms for Model-2 and the 1-D soil model subjected to oblique incident SH-waves. The solid lines indicate the waveforms from Model-2 and the broken lines indicate those from 1-D Fig. 13. Acceleration response spectra at the soil model. The interference seems to occur surface of Model-2 subjected to oblique incident roughly from the marked time. SH-waves. (a) P-1; (b) P-2; (c) P-3; (d) P-4.

Fig. 14. Configuration of shallow surface layers . A 1-D equivalent linear analysis is performed at equally spaced points with 100m separation using the calculated waveforms at the surface of Model -2 at the corresponding locations .

J. Phys. Earth Effects of Deep Irregular Underground Structures on Ground Motion in Kobe City 587

Fig. 15. Nonlinear characteristics of soils, G/G0-y, h-y. G/G0, h, and y are shear modulus ratio , damping coefficient and shear strain , respectively.

increases as the points approach the coast . The assumed initial shear wave velocities of the

alluvium and diluvium layers are 200 and 300 m/s , respectively. Their corresponding densities are 1.6 •~

103 and 1.8 •~ 103 kg/m3 respectively. The G/G0-y

and h-y curves of the two layers are shown in Fig . 15. These curves are based on the average nonlinear

characteristics reported by Imazu and Fukutake

(1986). Figure 16(a) and (b) show the distribution of Fig. 16. Distribution of maximum values at ground maximum velocity and maximum acceleration, re- surface (for Model-2 subjected to oblique spectively, taking into account amplification result- incident SH-waves). (a) Maximum response ing from the shallow surface layers indicated by velocity; (b) maximum response acceleration.

thick solid lines. To compare the obtained results SUOG stands for "surface of upper part of

with those of the 1-D analysis for the deep Osaka group." underground structure, the latter are indicated by thick broken lines. In this figure, the distributions maximum accelerations are almost the same as those of the maximum values on the free surface of the of the 1-D analysis. It is recognized that the Osaka group are also shown by thin lines for maximum velocities at the ground surface are larger comparison. "SUOG" in this figure stands for the over a wider area than the maximum accelerations, surface of the upper part of Osaka group. It is found because that the maximum velocities in the re- from this figure that the maximum velocities and claimed land are amplified by the surface layers, but accelerations at the ground surface become large the maximum accelerations are not amplified there between JR and R2, where the ratio of structural resulting from the soil nonlinearity. damage is very high, as reported in Sec. 6. The Figure 17 shows the acceleration response spectra maximum velocities and accelerations in the of four major points. It is found that the spectral damaged zone are amplified 3-4 times those amplification at a point 0.8 km from the rock- recorded at the rock site. The maximum accelera- sediment boundary is significant at the periods of tion, 0.5-1.5 km, is much larger (50%) than that of 0.6-0.7 s. With increasing distance from the rock the entire 1-D analysis. It is emphasized that the site, spectral values of the long-period range are amplification of ground motion between JR and amplified. R2 cannot be explained without considering the It is noted that maximum strain is about 1.2% at focusing effect of the deep irregular structure. a point 0.9 km from the rock-sediment boundary.

However, as the points approach the coast, the The strain is large but may not be beyond the

Vol. 44, No. 5, 1996 588 M. Motosaka and M. Nagano

Fig. 17. Acceleration response spectra at ground surface (for Model-2 subjected to oblique incident SH-waves). (a) P-5; (b) P-6; (c) P-7; (d) P-8.

applicable level of the equivalent linear analysis for estimating maximum accelerations and velocities.

Effective stress analysis taking into account the

effects of liquefaction may be desirable and necessary Fig. 18. Distribution of maximum accelerations for two cases of Q-values at the first layer of in the investigation of nonlinear soil amplification Model-2 with a shear-wave velocity of 500 m/s, of specific sites, where detailed soil parameters of surface layers are known. (a) At the surface of Osaka Group layers; (b) at ground surface.

6. Discussion

6.1 Sensitivity analysis for Q-values of the Osaka reduced 25% for Q= 10, compared to Q = 33 at the

group layers and incident angles top of the Osaka layers, and are reduced 30% at To investigate the sensitivity of Q-values at the the ground surface. But the distribution shape,

first layer of the deep underground model with a indicating that the maximum acceleration values

shear-wave velocity of 500 m/s to maximum ac- become large between JR and R2, does not change.

celerations, the amplification characteristics of the The sensitivity for the two incident angles is

ground motion for a Q-value of 10 (damping factor; relatively small. The shaded zones of the figures h=5%) and Vs (m/s)/15 (h= 1.5%) are compared show estimated accelerations through this sensitivity for two incident angles: ƒÆ =45 and 0•‹. The Q-value analysis.

of 10 is based on the maximum strain of this layer

in the preliminary analysis. The results for the two 6.2 Comparison of the maximum observed values

Q-values are shown in Fig. 18(a) and (b) for free and the analytical results surface motion in the deep structure model re- The maximum horizontal acceleration observed

presented at the top of the Osaka group layers, at Fukiai (FKI) in the heavily damaged zone was and for ground surface taking into account the 833 cm/s2. The maximum horizontal acceleration in surface geology. It is found from this figure that the the principal direction observed at Kobe Harbor maximum accelerations in the sediment part are Office (KBH) over reclaimed land was 538 cm/s2.

J. Phys. Earth Effects of Deep Irregular Underground Structures on Ground Motio n in Kobe City 589 tions coincide closely with the distribution of structural damage. It is also recognized by comparing Fig. 18(b) with Fig.19 that the maximum accelerations in the region where the damage ratio exceeds 50% are 800-1,200 cm/s' for the Q-value of Vs/15 and 600-800 cm/s2 for the Q-value of 10. 7. Concluding Remarks In this study, the amplification characteristics of ground motion in Kobe City during the 1995 Hyogo-ken Nanbu earthquake were analytically investigated based on a 2-D deep irregular underground structure model with vertical discontinuity. Three-dimensional wave propagation analyses of the Fig. 19. Structural damage ratio in Nada ward 2-D model were performed using the hyperelement (after Ono et al., 1995). Damage survey is method for planar incident waves assumed from performed along line b-b' in Fig. 1. The investigated structures are of wood and the wavefields due to the source mechanism; that is, for the incident SH-wave propagating along the reinforced concrete (RC) . irregular underground structure impinging at an oblique incident angle. The observation records at The corresponding velocity was 122 cm/s. These Kobe University (a rock site) were used as control observation values are consistent with the analytical data. The ground motion at the surface of the Osaka results, although the locations of these two points layers and at the earth's surface were calculated. The cannot be plotted in the analyzed section . It is also effects of shallow surface layers on ground motion noted that relatively large vertical motion was amplification were estimated by the conventional observed near the rock-sediment boundary . The 1-D nonlinear wave propagation analysis. The con- maximum vertical velocity observed at the third clusions derived from this report are as follows. basement floor of a high-rise building near Shin 1) It has been shown analytically that ground Kobe Station (SKB) was 47 cm/s as compared to 31 motion in the heavily damaged belt zone is amplified and 25 cm/s for the two horizontal components due to a focusing effect caused by a deep irregular (Building Research Institute, 1995). The maximum underground structure as well as shallow surface vertical velocity at Motoyama (KOB) was 49 cm/s layers. It is not necessary to incorporate a seismic as compared to 55 and 77 cm/s for the two horizontal fault beneath the belt zone to interpret the damaged components, although these are the maximum values zone. of the restored velocity waveforms from saturated 2) The calculated maximum acceleration dis- records (Kagawa et al., 1995). These observations tributions coincide closely with the structural seem to be consistent with the analytically estimated damage distribution. It may be concluded that a tendency for the vertical motion to be relatively large true characteristic of the heavily damaged belt zone near the rock-sediment boundary (refer to Fig. 10). is a focusing effect in the deep irregular underground structure at the edge of the Osaka basin as well as 6.3. Comparison of structural damage distribution the shallow surface layers. and maximum accelerations 3) It is suggested that vertical motion was in- Figure 19 shows the damage ratio of structures duced by the deep irregular underground structure (wooden structures and reinforced concrete struc- even though the incident wavefield consisted of only tures) along line b-b' in Fig. 1, which is closed by horizontal motion. The ratio of vertical motion to the analyzed section (line a-a' in Fig. 1). Damage horizontal motion is less than half, with the ratio means the ratio of heavily damaged and exception of the sediment site in the vicinity of the collapsed structures to total structures. This figure vertical discontinuity, where relatively large vertical is plotted based on the data by Ono et al. (1995). motion was induced. Comparing this figure with Fig. 18(b), it is found In this study, the wave propagation analyses were that the calculated maximum acceleration distribu- performed for the planar incident SH-waves. The

Vol. 44, No. 5, 1996 590 M. Motosaka and M. Nagano

amplification characteristics of the ground motion Japan Society of Soil Mechanics and Foundation in the analyzed irregular structure shoud also be Engineering, Kansai Branch, Soil in Kansai Area, 1992 investigated for the wavefield, taking into account (in Japanese). a finite moving source with inhomogeneous slip Kagawa, T., K. Irikura, and I. Yokoi, Saturation distribution. characteristics of servo velocity seismograph-applica- It is noted that research for determining geological tion for the records obtained under the Hyogoken structures has become important. In determining the Nanbu Earthquake-, Programme and Abstracts , parameters of geological structures, the sensitivity Seismological Society of Japan, No. 2, A91, 1995 (in of each parameter should be investigated from the Japanese). viewpoint of its effects on ground motion character- Kausel, E. and J.M. Roesset, Semianalytical hyperelement istics. for layered strata, J. Engineer. Mechan., ASCE, 103, 569-588, 1977. We are grateful to the Committee of Earthquake Kobayashi, S., S. Yoshida, S. Okubo, R. Shichi, T. Observation and Research in the Kansai Area (CEORKA) Shimamoto, and T. Kato, 2.5-dimensional analysis of , the gravity anomaly across the Rokko fault system who provided valuable earthquake observation data , , valuable information on the velocity structure at the Kobe Programme and Abstracts, Seismological Society of University site and the results along the Ishiyagawa Japan, No. 2, A61, 1995 (in Japanese). measuring line obtained from the reflection survey referred Motosaka, M. and M. Nagano, Analytical study on to in this study. We also thank Dr. Takuji Kobori, amplification characteristics of ground motions taking Professor Emeritus of Kyoto University and Chief Adviser account of irregular bedrock structure in Kobe City, of Kajima Corporation, who provided valuable advice in Soil Found., 43 (7), 15-20, 1995 (in Japanese). this study. Comments by two anonymous reviewers were Nagano, M. and M. Motosaka, Response analysis of 2-D helpful in revising this manuscript . structure subjected to obliquely incident waves with arbitrary horizontal angles, J. Struct. Construct , REFERENCES Engineer. (Trans. AIJ), 474, 67-76, 1995 (in Japanese with English abstract). Akamatsu, J., H. Morikawa, H. Saito, and M. Jido, Nakagawa, K., Relation between earthquake ground Relation between the distribution of damage caused by failure and subsurface structure , Proceedings of the 1995 Hyogoken-Nambu earthquake and vibration Symposium on the Great Hanshin-Awaji Earthquake and Its Geo-Environments characteristics inferred from microseisms, J. Nat. Disas. , 233-238, 1995 (in Japanese). Set., 16(2), 63-70, 1995. Ono, S., K. Ishikawa, and S. Mizoguchi, Earthquake Building Reserach Institute, Interim Report of Damage damage survey of buildings and houses during the 1995 Hyogo-ken Nanbu earthquake Survey during the 1995 Hyogo-ken Nanbu Earthquake , , Proceedings of the 50th pp. 123-125, Japan Association for Building Research Annual Conference of the Japan Society of Civil Promotion, Tokyo, 1995 (in Japanese) . Engineers, 1-(B), 948-949 , 1995 (in Japanese). Imazu, M. and K. Fukutake, Dynamic shear modulus and Suzuki, T., M. Hakuno, and S. Igarashi , Numerical damping of gravel materials , The 21st Japan NationaI simulation on ground motion amplification on dipping Conference on Soil Mechanics and Foundation soft soil layers, Proceedings of the 23rd JSCE Eathquake Engineering, 509-512, 1986 (in Japanese) . Engineering Symposium-1995 , 73-76, 1995 (in Japa- Iwami, Y., Town and Soil in Kobe, Kobe Shinbun Press nese). Center, Kobe, p.14, 1987 (in Japanese) . Takemura, M. and Y. Tsuji, Strong motion distribution Iwasaki, Y., Earthquake environment in Hanshin area and in Kobe area due to the 1995 Southern Hyogo earth - strong ground motions during the Hyogo-ken Nanbu quake (M=7.2) in Japan as inferred from the topple Earthquake, Soil Found., 43 (3), 2-6, 1995 (in Japanese). rate of tombstones, J. Phys. Earth, 43, 747-753, 1995. Iwasaki, Y., T. Hongo, H. Yokota, and S. Ito, Ground Toki, K., K. Irikura, and T. Kagawa, Strong motion re- characteristics of the Rokko-dai (Kobe University) cords in the source area of the Hyogoken-Nambu Ground Motion Monitoring Site, Programme and Earthquake, January 17, 1995, Japan, J. Nat. Disas, Sci., 16(2), 23-30 Abstracts, Seismological Society of Japan , No. 2, P77, , 1995. 1995 (in Japanese). Watanabe, M. and Y. Suzuki , Late Quaternary activity of' Iwata, T., K. Hatayama, H. Kawase, K. Irikura, and K. active faults in Kobe area , Programme and Abstracts, Matsunami, Aftershock observations at Higashinada Seismological Society of Japan , No. 2, A85, 1995 (in ward, Kobe City, J. Nat. Disas. Sci., 16(2), 41-48, Japanese). 1995.

J. Phys. Earth