1998 Kluwer Academic Publishers. Printed in the Netherlands. 223. Site response study in Abruzzo (Central Italy): underground array versus surface stations.
Journal of Seismology 2: 223–236, 1998. © 1998 Kluwer Academic Publishers. Printed in the Netherlands.
223
Site response study in Abruzzo (Central Italy): underground array versus surface stations G. De Luca1,5 , E. Del Pezzo2 , F. Di Luccio3 , L. Margheriti3 , G. Milana4 & R. Scarpa1,5 1
Universit`a dell’Aquila, Dipartimento di Fisica, Via Vetoio, 67010 Coppito (L’Aquila), Italy Universit`a di Salerno, Dipartimento di Fisica, Via S. Allende, 84081 Baronissi (Salerno), Italy 3 Istituto Nazionale di Geofisica, Via di Vigna Murata 605, 00143 Rome, Italy 4 Servizio Sismico Nazionale, Via Curtatone 3, 00185 Rome, Italy 5 Laboratori Nazionali del Gran Sasso (LNGS-INFN), 67010 Assergi (L’Aquila), Italy 2
Received 27 May 1997; accepted in revised form 23 April 1998
Key words: site response, Abruzzo, underground array
Abstract In this site response study we examined local earthquakes recorded at surface stations of a local seismic network and at a temporary underground seismic array installed in a tunnel underneath the Gran Sasso Massif in Abruzzo (central Italy). This allowed us to compare the seismic site response beneath the mountain and on the surface in similar geological environment (soft rock sites). We applied spectral ratios method on different segments of the seismograms and used different reference spectra in the 1–20 Hz frequency band. We found little or no amplification effects at most of the surface stations whereas site transfer functions evaluated with respect to underground sites show an amplification factor up to 6 in the 1–8 Hz frequency range. Coda spectral ratios estimated at soft rock sites are confirmed as good estimates of shear wave transfer function.
Introduction It is well known that significant peak amplification of ground motion can occur in sites with loose sedimentary deposits or other non-compact rock conditions (e.g., Aki, 1988; Field and Jacob, 1995; and references there in), but it is also quite well observed that even in apparently uniform geological conditions, stations of local networks show significant different site transfer functions in the short period range (Tucker et al., 1984; Blakeslee and Malin, 1991; Steidl et al., 1996, 1997; Theodulidis et al., 1996). This effect needs to be estimated for seismic risk assessment as well as for source and attenuation studies at a local scale. In central Italy, a large underground laboratory (Laboratori Nazionali del Gran Sasso – LNGS – of Istituto Nazionale di Fisica Nucleare – INFN) for the study of astroparticle physics was built in the Gran Sasso Mountain area. The installation of the physics laboratories in a tunnel excavated through the Gran
Sasso Massif, allowed to study seismic ground motion in an underground environment characterized by an average coverage of 1 km thick compact limestone. Moreover, due to the presence of the physics laboratories, a seismic site response study was soon considered by the scientific community as one of the results to achieve. The aim of the present work is to measure the site response in this region. In the Gran Sasso region background seismicity suitable for the site response study exists and the geological setting offers a large variety of limestones and sedimentary rocks strongly deformed during the orogenesis of Apennines, but no real hard rock sites are present. Site response can be measured using spectral ratio between seismic signals recorded at particular sites and a reference spectrum. In this study we focused our attention in the short period frequency range (1– 20 Hz). Among the different techniques that have been developed to (partly) overcome the difficulty of not
GSB/prepr: Pipsnr. 168465; (josekap:bio2fam) v.1.1 jose52.tex; 17/08/1998; 19:32; p.1
224 having a suitable reference site, we applied the approach that uses the spectrum log-averaged over the stations as a reference (Su et al., 1996). Moreover we used the single station horizontal over vertical (H/V) spectral ratio technique (Lermo and Chávez-García, 1993). We used surface and underground stations to compare the seismic site response beneath the mountain with that calculated on the surface, obtaining interesting indirect information about the complexity of the wave propagation in such underground conditions. As a result of our analysis we found two cases of amplification factors that can be interpreted as topographic effects. For underground stations we found uniformity in the seismic site response. Underground array and external seismic network: different settings in a similar geological condition In the present paper we examined data from a local seismic network and an underground array of temporary stations, operating in the period January–March 1993, located underneath the Gran Sasso Massif. A comparison between seismograms recorded at a surface station located just outside the tunnel entrance and the other station installed in the underground laboratories in 1992 (Figure 1) showed a relevant modification of waveforms from local earthquakes. This effect can be summarized in a different frequency content and the almost absence of Coda-waves in underground condition, as theoretically expected (Aster and Shearer, 1991a, 1991b; Carter et al., 1991; Malin et al., 1988; Shearer and Orcutt, 1987). The results from the experiment reported in this paper have the aim to present systematically the differences of ground motion observed underground and at surface stations of the seismic network. The local seismic network was composed by 10 seismic stations covering an area of approximately 100 × 100 km2, centered in the town of L’Aquila, Abruzzo region, central Italy. The network was set up with the purpose of monitoring seismicity in an active region with present small to moderate seismicity level, which is however located in a seismic region of the Apennines with the occurrence of large historical events (magnitudes up to 7.0–7.5) and the presence of many active normal faults (Scarpa and Zollo, 1985; Westaway et al., 1989; Capuano et al., 1992; Michetti et al., 1996; Scarpa et al., 1996). A simplified map of the geological and structural setting of the Abruzzo region is given in Figure 2.
The 10 stations of the network operating in early 1993 were equipped with Mark L4C (1 Hz), three component sensors, with local digital recording (trigger system based on the STA/LTA ratio) and high (120 dB) dynamical range (Lennartz MARS-88). The stations are time synchronized with radio frequency signal DCF (77.5 kHz) and permit a local data storage up to 330 Mbytes with a sampling rate of 125 Hz per channel. A map of the station’s locations is reported in Figure 3 and the location and technical characteristics of the seismic network are summarized in Table 1. The stations are installed mainly on limestone rocks, with exception for the stations ACCI and CMPF that are located on slope debris, and the station BARI which is located on an old alluvial conglomerates terrace. Considering the geological conditions at the sites we do not expect strong amplification phenomena even if the rough topography of the Gran Sasso area can have some influence on the site transfer function. The temporary array (Figure 4) has been composed by 13 underground and one surface digital three-component seismic stations, whose position is also reported in the map of Figure 3. Coordinates and technical characteristics of the array stations are reported in Table 2. These stations were installed on the compact limestone which is present underneath the Gran Sasso Massif, in a gallery excavated through the mountain. During the temporary experiment several local and regional earthquakes were recorded both at the surface and at the underground stations. Some examples of records are shown in Figure 5. Spectral ratios have been computed for shear waves, in a time window of about 4 sec (512 points), for all events in the distance range of 30–40 km from the tunnel (i.e. S–P ∼ 4– 5 sec). Data were corrected for geometrical spreading and attenuation, assuming an average Q value of 200 (Del Pezzo and Zollo, 1984; Del Pezzo and Scarcella, 1986). The ratios of the log-average spectra are illustrated in Figure 6. These figures show an amplitude enrichment in the frequency band 1–8 Hz, by a factor up to 6 for the stations located at surface, as compared to the underground ones. In particular Figure 6d has been obtained from log-average spectral ratios of 10 earthquakes recorded at 9 surface and 12 underground stations. This observation can be explained with the absence both of free-surface and the resonance effects due to shallow low velocity layers in the underground sta-
jose52.tex; 17/08/1998; 19:32; p.2
225
Figure 1. Three component seismograms of a microearthquake recorded at two MARS-88 stations located just outside the tunnel (lower traces) and in the underground laboratories (upper traces). The seismic event (ML ∼ 1) is located at the same distance from the two stations.
tions, as theoretically expected from simple models of wave propagation (see Shearer and Orcutt, 1987). Local seismicity We examined 52 local earthquakes, selected on the basis of the best signal-to-noise ratio, recorded at the local network and 44 local and regional earthquakes recorded at the 4 selected stations of the underground array. Most of the events were located with
HYPO71PC computer code (Lee and Valdes, 1985) by using the velocity model (Nicolich, 1981) reported in Table 3. For the events with a location at the border or outside the network, we used the location of the National Seismic Network (ING seismic bulletins). In Figure 3 the locations of the events recorded at the seismic network and the locations of some of the events recorded also at the undeground array are respectively reported. Among the 44 events recorded at the underground array we selected 15 events for
jose52.tex; 17/08/1998; 19:32; p.3
226 Table 1. Technical characteristics of the seismic network Station name
Data logger
Latitude north
Longitude east
Altitude (mt.)
Sampling rate (Hz)
Channels
ACCI – Acciano AIEL–Aielli BARI–Barisciano CMPF–Campo Felice FILE–Filetto FONT–Fontecchio LUCO–Lucoli OCRE–Ocre SANG–San Giuliano TAGL–Tagliacozzo
MARS-88 (Lennartz) MARS-88 (Lennartz) MARS-88 (Lennartz) MARS-88 (Lennartz) MARS-88 (Lennartz) MARS-88 (Lennartz) MARS-88 (Lennartz) MARS-88 (Lennartz) MARS-88 (Lennartz) MARS-88 (Lennartz)
42◦ 42◦ 42◦ 42◦ 42◦ 42◦ 42◦ 42◦ 42◦ 42◦
13◦ 13◦ 13◦ 13◦ 13◦ 13◦ 13◦ 13◦ 13◦ 13◦
594 945 853 1642 1065 680 1012 745 771 1010
125 125 125 125 125 125 125 125 125 125
V–N–E V–N–E V–N–E V–N–E V–N–E V–N–E V–N–E V–N–E V–N–E V–N–E
100 040 180 120 220 130 180 180 220 030
39.7" 26.9" 05.2" 32.8" 43.1" 52.8" 03.6" 01.6" 25.3" 55.5"
430 330 350 280 310 360 200 280 230 140
08.7" 47.7" 35.9" 11.8" 15.8" 27.2" 28.7" 47.3" 33.3" 49.5"
42 ° 5 5 '
(1)
LA T IT U D E N O R T H
(2) (3) (4) (5) (6) (7)
(a)
Km 0
10
20
30
(b) (c) (d)
41 ° 2 7' 12 ° 5 1'
L O N G IT U D E E A S T
14 ° 1 8'
Figure 2. Geological map of the region: (1) Pleistocene and Holocenic continental deposits of Fucino and Sora; (2) Middle Pliocene and lower Pleistocene units; (3) Alanno-Maiella unit; (4) Flakes region; (5) Marchigiano units (e.g., Montagnone unit); (6) Laziale-Abruzzese units (e.g., Mt Morrone unit and Gran Sasso unit); (7) Umbrian unit; (a) Thrust; (b) Anticline folds; (c) Transcurrent shear faults; (d) Normal faults.
the site response study, while all the 44 events were utilized for the evaluation of the quality factor. The magnitude (ML ) of the earthquakes utilized in the present study ranges from 0.9 to 3.9.
function is reported in Field and Jacob (1995). Here we give only a brief discussion of the S-wave spectral ratio, the Coda-wave spectral ratio and the H/V ratio method, that are utilized in the present paper.
Spectral ratios technique using different phases and reference spectra: methods and data analysis
The spectral ratio method
A wide discussion about the different techniques that can be utilized in the evaluation of the site transfer
The spectral ratio method (Borcherdt, 1970) assumes that the Fourier spectrum of a seismic signal, Oij , can be expressed as a function of the source function (Sij ),
jose52.tex; 17/08/1998; 19:32; p.4
227
43 .5 A dria tic S ea
LA T IT U D E N O R T H
43 .0
G ra n S a sso (2 9 1 2 m t)
42 .5
P e sca ra
L in e a r A rra y L 'A q u ilaSANG
FILE
LUCO OCRE BARI CM PF
FO NT ACCI
AIEL
TAG L
42 .0
A ve zza n o
S o ra
41 .5 12 .5
13 .0
13 .5
14 .0
14 .5
LO N G IT U D E E A S T Figure 3. Map of the station locations (closed triangles), linear array in the tunnel of Gran Sasso Massif (closed stars) and epicenters (small crosses for events recorded at the seismic network and open circles for events recorded also at the underground array).
L IN E A R A R R A Y (JA N U A R Y - M A R C H 1 9 9 3 ) SO U TH - W EST
NO RTH - EAST M t. A q u ila (2 3 7 0 m )
L 3 (E xt. S ta tion ) R1
E1
E2
R3
E3
E4
R5
E5
M1 L .N .G .S .
M2
M3
L1
L2
T u n n e l o f G ra n S a sso 0
1 km
Figure 4. Map of the underground array composed by 13 temporary three-component digital seismic stations located in the tunnel of Gran Sasso Massif and one just outside (L3). The tunnel extends for 10.2 km and its average altitude is 960 mt a.s.l.. The location of the underground laboratories (LNGS) is indicated by a large closed star.
jose52.tex; 17/08/1998; 19:32; p.5
228 Table 2. Technical characteristics of the seismic array in the tunnel of Gran Sasso Station name
Data logger
Distance from South-West tunnel entrance (mt.)
Latitude north
Longitude east
Altitude (mt.)
Sampling rate
Channels (Hz)
R1 E1 E2 R3 E3 E4 R5 E5 M1 M2 M3 L1 L2 L3
REFTEK EDA (Scintrex) EDA (Scintrex) REFTEK EDA (Scintrex) EDA (Scintrex) REFTEK EDA (Scintrex) MARS-88 (Lennartz) MARS-88 (Lennartz) MARS-88 (Lennartz) MARS-5800 (Lennartz) MARS-5800 (Lennartz) MARS-5800 (Lennartz)
750 1450 2530 3020 3540 4100 5060 5650 7000 7680 8320 9050 9560 external
42◦ 42◦ 42◦ 42◦ 42◦ 42◦ 42◦ 42◦ 42◦ 42◦ 42◦ 42◦ 42◦ 42◦
13◦ 13◦ 13◦ 13◦ 13◦ 13◦ 13◦ 13◦ 13◦ 13◦ 13◦ 13◦ 13◦ 13◦
960 960 960 960 960 960 960 960 960 960 960 960 960 990
125 120 120 125 120 120 125 120 125 125 125 125 125 125
V–N–E V V–N–E V–N–E V V–N–E V–N–E V V–N–E V–N–E V–N–E V–N–E V–N–E V–N–E
250 250 250 250 250 260 260 260 270 270 270 280 280 250
04.8" 12.0" 30.1" 38.8" 52.3" 07.7" 31.8" 47.2" 21.8" 39.1" 55.2" 01.2" 14.4" 48.0"
320 320 320 330 330 330 340 340 340 350 350 350 360 320
24.0" 38.4" 55.4" 04.7" 19.1" 35.0" 01.8" 18.5" 55.5" 14.6" 32.1" 49.2" 03.3" 42.0"
21 F E B R U A R Y 19 93 02 :56 (U T ) - V E R T IC A L (a) M1
21 F E B R U A R Y 19 93 02 :56 (U T ) - V E R T IC A L (b)
G R O U N D V E LO C IT Y (m /s) A M P L IT U D E = +/- 1 .5 4E -5 m /s
SANG G R O U N D V E L O C IT Y (m /s) A M P LIT U D E = + /- 2 .64 E -5 m /s
M2
M3
L1
0
4
8 T IM E (s)
12
16
BARI
CM PF
0
4
8
12
16
T IM E (s)
Figure 5. Example of vertical component seismograms recorded: (a) at the underground seismic array; and (b) at the stations of the regional seismic network. The y-axis scales of (b) are double with respect to (a).
jose52.tex; 17/08/1998; 19:32; p.6
229
10
10 (a )
(b ) A ve ra g e S A N G / IN T
A ve ra g e B A R I/ IN T 8
V S P E C T R A L R A T IO (S -W A V E )
S P E C T R A L R A T IO (S -W A V E )
8
N -S E -W 6
4
N -S E -W 6
4
2
2
0
0 0
5
10
15
V
20
0
5
F R E Q U E N C Y (H z)
10
10
20
10 (d )
(c) A ve ra g e M 1 / IN T 8
A ve ra g e E X T / IN T 8
V S P E C T R A L R A T IO (S -W A V E )
S P E C T R A L R A T IO (S -W A V E )
15
F R E Q U E N C Y (H z)
N -S E -W 6
4
N -S E -W 6
4
2
2
0
0 0
5
10 F R E Q U E N C Y (H z)
15
20
V
0
5
10
15
20
F R E Q U E N C Y (H z)
Figure 6. Ratios of the log-average spectra of regional events at distance lower than 50 km from the underground array, for: (a) BARI over underground stations (INT); (b) SANG over underground stations (INT); (c) M1 over underground stations (INT); (d) all available surface stations (EXT) over all available underground stations (INT).
jose52.tex; 17/08/1998; 19:32; p.7
230 800
Table 3. Velocity model used to localize the events (Nicolich, 1981)
FREQUENCY 1 8 .0 H z 1 2 .0 H z
600
6 .0 H z
4.0 5.5 6.5 7.0 8.1
Q -C o d a
0–4 4–12 12–20 20–30 Half space
3 .0 H z 1 .5 H z
400
200
0 20
the path (Pij ), the instrument transfer function (Ii ) and the site transfer function (Rij ): (1)
where f is the frequency, j indicates the earthquake and i the station. If we divide Oij for the spectrum measured at a reference station, ORj we can isolate the site transfer function, correcting for the path and the instrumental response, if needed, simply as Rij (f ) =
Oij (f ) ORj (f )
e−πf Dij /vS Q Dij
50
600 L A P S E T IM E 50 s 40 s
400
30 s 20 s
200
(2)
The most of the site transfer function measurements are carried out on the horizontal components of the S-wave time window, which is most energetic and important for engineering purposes. In this case Pij can be estimated by Pij (f ) =
40
L A P S E T IM E (s)
Q -C o d a
Oij (f ) = Sij (f ) · Pij (f ) · Ii (f ) · Rij (f )
30
0 0
4
8
12
16
20
F R E Q U E N C Y (H z)
Figure 7. Qcoda curves versus the frequency and lapse time (the time elapsed from the origin along the Coda).
(3)
where Dij is the hypocentral distance, vS is the shearwave velocity and Q is the quality factor, 1/Dij accounts for the geometrical spreading for body waves. It is important to note that the ratio of equation (2) cannot be independent on some effects due to the radiation pattern of the source. If many earthquakes widely distributed in azimuth are examined together, the effect of radiation pattern on the log-average spectra from two components of motion can be considered as negligible. Coda-wave method Considering the Coda-waves in a time window centered at a given lapse time measured from the origin time, which is at least twice the travel time of the S-waves, it is possible to disregard the effects of source radiation pattern (Rautian and Kalturin, 1978; Kato et al., 1995) and to estimate the site transfer function for shear waves. The method is the same of that
described above for S-waves, with the substitution of the S-wave window with Coda-wave window. H/V ratio method The so called H/V ratio method consists in the evaluation of the spectral ratio between the horizontal and vertical component of motion. It was developed for seismic noise by Nakamura (1989) and is based on pure empirical observations, even if it has been partially justified with theoretical considerations by Field et al. (1992) and Lachet and Bard (1995). Lermo and Chávez-García (1993) applied this technique on direct S-waves giving some theoretical explanations. A wide and sound description of this method is reported in Field and Jacob (1995), and we applied this technique on both S- and Coda-waves. We first estimated the quality factor for Codawaves using the standard IASPEI procedure (Lee et al., 1986). We find a clear dependence of Qcoda from the frequency and lapse time (the time elapsed from
jose52.tex; 17/08/1998; 19:32; p.8
231
(1) the horizontal component spectra (RMSH) and the log-average of RMSH; (2) the horizontal component spectra (RMSH) and the RMSH of the reference station, which was chosen on the basis of the largest number of recorded events (SANG); (3) the horizontal component spectrum (RMSH) and the vertical component spectrum (H/V ratio method). We evaluated for each site average spectral ratios with respect to the reference spectrum and its standard deviation (we used standard deviation of the means (sd/N) because it takes into account the number of earthquake (N) included in the evaluation). Standard deviations of ratios relative to the log-average of RMSH are gener-
10
ACCI
A IE L
BARI
CMPF
F ILE
FO NT
LUCO
OCRE
SANG
TAG L
5
0 10
5
S -W A V E - S P E C T R A L R A T IO
the origin along the Coda). Results are reported in the plots of Figure 7. We used the Qcoda to correct spectra for attenuation, assuming the hypothesis that Qcoda shares the same frequency dependence as QS , the quality factor for S-waves. Even if this is not a strictly correct assumption , because Qcoda is closer to the intrinsic S-wave Q than to the total Q (Hoshiba, 1993), the bias in the estimate of Q poorly affects the value of the spectral ratio estimates due to the short distance range. S-wave spectra are also corrected for a geometrical spreading term equal to the inverse of the hypocentral distance and for the instrument transfer function. Then we examined the data from the seismic network both for S- and Coda-waves. S-wave analysis was performed for each seismogram evaluating the FFT on a time window of 8 sec long, starting from the beginning of the S-wave train: Coda-wave analysis was performed applying the same technique on a 8 sec long time window starting at time t"; t" was taken as twice the travel time of S-waves at the farthest station. In the hypothesis of a single scattering model, the Coda-wave spectrum at lapse times greater than twice the travel time for S-waves is practically independent on epicentral distance when source is close to station. This means that t" > 21/vS where 1 is the hypocentral distance and vS is the S-wave velocity. The values of t" used in this work fulfill always this condition. RMS average of the two horizontal components of spectra (RMSH) were evaluated both for S-waves and for Coda-waves and for each earthquake the logaverage (over the stations) of the RMSH was performed. For each earthquake and both windows we performed spectral ratios between:
0 10
5
0 10
5
0 10
5
0 0
5
10
0
5
10
F R E Q U E N C Y (H z)
Figure 8. Results for S-wave spectral ratio for the seismic network. The three different curves and symbols refer to the different reference spectra: ratio with respect to the log-average spectra (bold lines), ratio with respect to the log-average spectra from reference station SANG (curves with small open squares) and ratio between horizontal and vertical spectra (curves with small closed dots).
ally smaller than the others (which are generally on the order of a 20% of the mean). Results are reported in Figure 8 for S-waves and in Figure 9 for Coda-waves. Then we examined the data from the array under the Gran Sasso Massif. For this set of data the reference station was changed, because the station which recorded the largest number of events was the station M1. We examined the data recorded simultaneously at four selected stations of the array (M1, M2, M3, and L1) and at 4 stations of the seismic network (BARI, CMPF, FILE and SANG), in order to compare the site transfer function obtained for the stations inside the Massif located in the tunnel and a group of stations at the surface. For each earthquake we evaluated the log-average of the spectra of each component of the ground motion; we used 4 sec time windows of S-waves and we performed the spectral ratios between:
jose52.tex; 17/08/1998; 19:32; p.9
232 10
ACCI
A IE L
BARI
CMPF
F ILE
FO NT
LUCO
OCRE
SANG
TAG L
5
0
C O D A -W A V E - S P E C T R A L R A T IO
10
5
0 10
5
0 10
5
0 10
5
0 0
5
10
0
5
Figure 10. Underground array results for S-waves with respect to reference M1 station. The three curves and symbols refer to the vertical (bold lines), North-South (curves with small closed dots) and East-West (curves with small open squares) components. For a partial default of sensor the vertical component of Filetto station (FILE) is not shown.
10 10
Figure 9. Results for Coda-waves spectral ratio for the seismic network. The three different curves and symbols refer to the different reference spectra: ratio with respect to the log-average spectra (bold lines), ratio with respect to the log-average spectra from reference station SANG (curves with small open squares) and ratio between horizontal and vertical spectra (curves with small closed dots).
(1) the North-South, East-West and Vertical component spectra and the corresponding spectra of the reference station (2) the North-South, East-West and Vertical component spectra and the corresponding log-averages. Then we calculated the spectral ratios, standard deviations resulted of the order of 20% of the mean values. Results are reported in the Figure 10 and in Figure 11.
S -W A V E - S P E C T R A L R A T IO O N T H E A V E R A G E
F R E Q U E N C Y (H z)
BARI
CMPF
F ILE
SANG
L1
M1
M2
M3
5
0 10
5
0 10
5
0 10
5
0
Discussion Results for the surface seismic network Results obtained for the seimic network show that the spectral ratios of S-waves (Figure 8) and Coda-waves (Figure 9) are generally flat and share the same pattern, indicating that in the examined geological conditions
0
5
10
0
5
10
F R E Q U E N C Y (H z)
Figure 11. Underground array results for S-waves with respect to log-average spectra. The three curves and symbols refer to the vertical (bold lines), North-South (curves with small closed dots) and East-West (curves with small open squares) components. For a partial default of sensor the vertical component of Filetto station (FILE) is not shown.
jose52.tex; 17/08/1998; 19:32; p.10
233 the site transfer function obtained using the Codawaves is a good estimate of the site transfer function for S-waves. The comparison of Coda-wave site amplification and S-wave site amplification has been investigated by several researchers: Tsujiura (1978), Tucker and King (1984), Su et al. (1993, 1996), Kato et al. (1995), Margheriti et al. (1994), Steidl et al. (1995), finding in general good agreement. In the last two studies mentioned above the authors found that at sites in alluvial basins, Coda-wave site amplification overestimates that of S-wave, owing to the presence of trapped waves. Our results confirm the consistency between the two site amplification estimates in the case of soft rock or stiff soil sites. The spectral ratios with respect to the reference station (curves with small open squares) and with respect to the log average (bold lines) share the same pattern in the examined frequency range (1–10 Hz) except for the station LUCO, which shows a peak around 8 Hz in the S-wave spectral ratios with respect to the reference station. The same station, LUCO, does not show this peak in the Coda spectral ratio. The difference of site amplifications between the ratio with respect to the log-average and the ratio with respect to SANG shown in Figures 8 and 9 is very small, which is quite consistent with the fact that the SANG amplification relative to the log-average are almost unity for all the frequencies of interest here. Large differences are present at the other stations, especially ACCI for S and Coda and TAGL for Coda. Even the differences at CMPF, FILE, and FONT are large, if we take the ratios between the amplifications by log-average and those by SANG. This fact is explained by the low hit rate (small number of earthquakes) observed at the stations ACCI and TAGL, lower than 30% of the total amount observed in the remaining stations. In spite of the low number of available seismograms, two sites show clear evidence of site amplification effects. ACCI shows an amplification peak around 3 Hz with amplifications higher than 5 for both Sand Coda-waves, consistently among the three different reference spectra. TAGL station shows significant peaks between 2.5 and 5 Hz for the spectral ratios evaluated with the H/V ratio technique (curves with small closed dots) while the ratios respect to the logaverage and respect to the reference station are much lower (from 2.5 up to 5 in the Coda-waves window). It is notewhorthy that H/V ratio method (curves with small closed dots) is inconsistent with the spectral ratio method for AIEL site showing an amplification of about 5 in the frequency band centered at 5 Hz.
However, in terms of the peak frequencies FONT and OCRE stations also show big differencies. These peaks in H/V ratios are nonexistent in the H/H spectral ratios and so they may not reflect the true amplification of these sites. This is a significant fact since Lachet and Bard (1995) theoretically proved that at least predominant frequencies can be estimated by the H/V ratio method. This method in the other cases shows a good agreement in the trend, but overestimate the other ratios of a factor of 2. Results obtained for the linear underground array For this set of data we studied separately the three components of the ground motion and we used also the H/V ratio technique. Results are reported in Figure 10 (spectral ratios with respect to the reference station, M1) and in Figure 11 (spectral ratios with respect to the log-average). The obtained results confirm that the general trend of the spectral ratios for S-waves is quite uniform at different stations: almost flat for the entire band of frequency considered. As compared to the external stations there is still an amplification factor, in the frequency band 1–10 Hz, confirming the result shown in Figure 6. The different data set used in this case includes also more distant earthquakes and does not show generally a strong amplification factor of surface stations as compared to the underground ones. The underground spectra are contaminated with the reflected waves from the surface even if it is observed in the genuine rock formation. For the very low frequency range both the surface station and the underground station have the same amplitude (∼ 2), but for the higher frequency range the underground station has deamplification at certain frequencies at which destructive interferene between the incident and reflected waves are taking place. This does not hold if we take a very short time-window at which the reflected phase will be outside the timewindow, but is not the case here. The fundamental frequency of this through corresponds to vS /4h where vS is the shear wave velocity and h is the depth of the station. If vS in the Gran Sasso Massif is 2 km/s and the depth of the station is 1 km, then this frequency will be as low as 0.5 Hz. From that frequency we will have nodal frequencies at every odd multiples, that is, 1.5 Hz, 2.5 Hz, etc. If we take a ratio for a surface station relative to the borehole station with such deamplification, the site factor of the surface station will be very large at these frequencies. Since this is not the
jose52.tex; 17/08/1998; 19:32; p.11
234 4
from this figure the noise and S-wave analysis does not illustrate any influence from the geometry of the tunnel on the spectral ratios. The homogeneity of site response at underground stations is a result of great relevance for the installation of a small aperture underground seismic array with the aim of high-sensitive monitoring microearthquake activity in this region of the Apennines (Scarpa et al., 1996; Amoruso et al., 1997).
(a) 3 2 1
S P E C T R A L R A T IO
0 4 (b) 3
Conclusions
2
We find three main results from the present analysis: (a) consistency between S-wave and Coda-wave site response at at rock sites; (b) similarity among rock sites both on the surface (the seismic network) and in the rock (the temporary seismic array installed in the tunnel through the Gran Sasso Massif); (c) strong amplification factors and larger complexity of waveforms of surface stations as compared to the underground ones. The consistency between Coda-waves and direct S-wave spectral ratios indicates that at sites classified as soft rock or stiff soil sites it is possible to use the Coda portion of the seismograms to achieve the transfer function for shear waves. Moreover since the Coda-waves are expected to be more independent from source than the S-waves, their use in the site transfer function studies avoids the uncertainties due to radiation pattern effects and their results are more stable. We compared different types of ratios applied both to body waves and to Coda-waves. The results show generally that the spectral ratios computed with respect to a reference station share almost the same pattern of those carried out with respect to a logaverage spectrum. This results shows that it may be possible to estimate the differences between the site factors to the log-average and those to a single reference site, although not as small as expected. This is so here since our dataset contains a lot of rock site stations. If there is no hard site among the stations used, then the log-averaged site factor itself contains strong effects of site amplification (i.e. the average site amplification) and the ratios relative to the logaverage will be relatively small compared to the ratios relative to the true rock site spectrum. Even with 50% of rock sites among all the site, log-average site factors are highly affected by the presence of soft-soil sites, at which site factors can reach easily one order of magnitude. As long as the final spectra are the only concern, site factors with respect to anything is still
1 0 4 (c) 3 2 1 0 0
5
10
15
F R E Q U E N C Y (H z) Figure 12. Underground array results for S-waves (bold lines) and noise (dashed lines): (a) spectral ratios between horizontal and vertical components; (b) spectral ratios between parallel (to the tunnel) and vertical components; (c) spectral ratios between transversal (to the tunnel) and vertical components.
case observed here, the underground stations probably do not have a coherent reflection from the surface. The cause of such incoherency of reflected waves can be due to the topographic feature of the Massif, strong scattering inside the Massif, or shallow surface layers with high absorption, or a combination of all these causes. We evaluated also the possible influence of the tunnel effects using the H/V technique. Figure 12 shows the spectral ratios between horizontal and vertical components for noise and S-waves. We extracted 36 windows of 256 points for noise signals and 7 windows of 256 points for S-waves for each component at different underground stations and at different dates. We rotated the horizontal components to obtain the parallel and transversal components with respect to the tunnel and then we computed the ratios between log-averages of spectra (Figure 12). As it can be seen
20
jose52.tex; 17/08/1998; 19:32; p.12
235 T A G L - 2 JU N E 1 994 - 1 7:3 8 (U T )
G R O U N D V E L O C IT Y (m /s) A M P LIT U D E = + /- 1 .65 E -4 m /s
V ertica l
N o rth-S outh
E ast-W e st
0
10
20
30
40
T IM E (s)
Figure 13. Example of a typical three-component seismogram recorded at the station Tagliacozzo (TAGL).
correct, but then the physical meaning of the other factors will be changed. For example, if we use the log-average as a reference, then the source term Sij (f) is no longer the source term since it includes the average site amplification, the exact amount of which is never known. Two sites presented anomalous amplification peaks: ACCI and TAGL. ACCI is located on slope debris on a side of a quite steep valley (Valle Subequana) which probably produces the observed anomaly. TAGL is located directly on limestone near a sub-vertical scarp East-West elongated, and the amplification peak is clearly visible for the H/V ratio method. This site effect could be produced mainly by the peculiar topography; in fact TAGL site is characterized by a scarp elongated for about 1000 m in the East-West direction and for ∼ 500 m in the North-South direction, with an average height of 250 m above the underlaying valley. The observation of several seismograms indicates that the North-South component of motion is larger than East-West component, independently on source location and mechanism (Figure 13). The observed peak of the H/V ratio close to 3 Hz is compatible with the fundamental mode resonant frequency of an infinite long, 500 m wide hill having vS velocity around 2 km/s. This result is compatible with findings by Ashford and Sitar (1997) and Ashford et al. (1997) on the values of peak amplification caused by steep slopes and by Chávez-
García et al. (1996) on the potential of H/V method for revealing topographic site effects but a more detailed numerical 3D-simulation should be needed to confirm this hypothesis. It is notewhorthy that there is no full agreement between the H/V ratios and the H/H ratios. As pointed out in the previous chapter AIEL shows very strong difference as well as FONT and OCRE. The H/V ratios are always larger than the H/H ratios even at the sites where the frequency distribution is quite similar enlightning that the H/V ratio method does not provide correct absolute values of amplication site factors. The comparison of the signals observed at the surface stations and the underground array show that freesurface effects and mostly the resonance due to surface layers influence significantly the spectral shape, particularly for events at distances smaller than about 50 km. It would be very interesting to compare the Coda spectra between surface and underground station, in order to study the scattering processes inside the earth crust, but the data did not allow us to obtain stable results. A future permanent underground array, planned to be located inside the Gran Sasso Massif, could answer this open question. Acknowledgements We acknowledge the Director of LNGS (INFN), Prof. P. Monacelli, and Ing. R. Adinolfi for providing facilities and assistence during the field experiment. Prof. E. Bellotti, President of Consorzio di Ricerca del Gran Sasso, is also aknowledged for the continuous support and encouragements. We are grateful to Hiroshi Kawase for the comments and suggestions which greatly contributed to the presentation of results. We thanks also A. Cirella, G. De Sanctis, A. Deschamps, L. Filippi and M. Martini for technical assistance and A. Gorini for the realization of Figure 2. References Aki, K., 1988, Local site effects on strong ground motion, Proc. of Earthq. Eng. and Soil Dynamics II, 103–155. Amoruso, A., Crescentini, L., De Luca, G., Scarpa, R., Abril, M. and Cirella A., 1997, Underground earth strain and seismic radiation measurements with a laser interferometer and a dense small-aperture seismic array, Annali di Geofisica XL, 995–1005. Ashford, S. A. and Sitar, N., 1997, Analysis of topographic amplification of inclined shear waves in a steep coastal bluff, Bull. Seism. Soc. Am. 87, 692–700. Ashford, S. A., Sitar, N., Lysmer, J. and Deng, N., 1997, Topographic effects on the seismic response of steep slopes, Bull. Seism. Soc. Am. 87, 701–709.
jose52.tex; 17/08/1998; 19:32; p.13
236 Aster, R. C. and Shearer, P. M., 1991a, High-frequency borehole seismograms recorded in the San Jacinto fault zone, Southern California. Part 1: Polarizations, Bull. Seism. Soc. Am. 81, 1057– 1080. Aster, R. C. and Shearer, P. M., 1991b, High-frequency borehole seismograms recorded in the San Jacinto fault zone, Southern California. Part 2: Attenuation and site effects, Bull. Seism. Soc. Am. 81, 1081–1100. Blakeskee, S. and Malin, P., 1991, High-Frequency site effects at two Parkfield downhole and surface stations, Bull. Seism. Soc. Am. 81, 332–345. Borcherdt, R. D., 1970, Effects of local geology on ground motion near San Francisco Bay, Bull. Seism. Soc. Am. 60, 29–61. Capuano, P., Gasparini, P., Peronaci, M. and Scarpa, R., 1992, Strong ground motion and source parameters for earthquakes in the Apennines, Italy, Earthquake Spectra 8, 529–554. Chávez-García, F. J., Sánchez, L. R. and Hatzfeld, D., 1996, Topographic site effects and HVSR. A comparison between observations and theory, Bull. Seism. Soc. Am. 86, 1559–1573. Carter, J. A., Barstow, N., Pomeroy, P. W., Chael, E. P. and Leahy, P. J., 1991, High-frequency seismic noise as a function of depth, Bull. Seism. Soc. Am. 81, 1101–1114. Del Pezzo, E. and Scarcella, G., 1986, Three-component coda Q in the Abruzzi-Molise region, central Apennines, Ann. Geophys. 4, 589–592. Del Pezzo, E. and Zollo, A., 1984, Attenuation of coda waves and turbidity coefficient in central Italy, Bull. Seism. Soc. Am. 74, 2655–2659. Field, E. H. and Jacobs, K. H., 1995, A comparison and test of various site response estimation techniques, including three that are not reference site dependent, Bull. Seism. Soc. Am. 85, 1127– 1143. Field, E. H., Jacobs, K. H. and Hough, S. H., 1992, Earthquake site response estimation: a weak-motion case study, Bull. Seism. Soc. Am. 82, 2283–2307. Hoshiba, M., 1993, Separation of scattering attenuation and intrinsic absorption in Japan using the multiple lapse time window analysis of full seismogram envelope, J. Geophys. Res. 98, 15809–15824. Kato, K., Aki, K. and Takemura, M., 1995, Site amplification from Coda waves: validation and application to S-wave site response, Bull. Seism. Soc. Am. 85, 467–477. Lachet, C. and Bard, P. Y., 1995, Numerical and theoretical investigations on the possibilities and limitations of the ‘Nakamura’s’ technique, J. Phys. Earth 42, 377–397. Lee, W. H. K. and Valdes, C. M., 1985, HYPO71PC: a personal computer version of the HYPO71 earthquake location program, U. S. Geol. Surv. Open-File Report 85–749, 43 pp. Lee, W. H. K., Aki, K., Chouet, B., Johnson, P., Marks, S., Newberry, J. T., Ryall, A. S., Stewart, S. W. and Tottingham, D. M., 1986, A preliminary study of Coda Q in California and Nevada, Bull. Seism. Soc. Am. 76, 1143–1150. Lermo, J. and Chávez-García, F. J., 1993, Site effects evaluation using spectral ratios with only one station, Bull. Seism. Soc. Am. 83, 1574–1594. Malin, P. E., Waller, J. A., Borcherdts, R. D., Cranswick, E., Jensen, E. G. and Van Schaack, J., 1988, Vertical seismic profiling of Oroville microearthquakes: velocity spectra and particle motion as a function of depth, Bull. Seism. Soc. Am. 78, 401–420.
Margheriti, L., Wennerberg, L. and Boatwright, J., 1994, A comparison of Coda and S wave spectral ratios as estimates of site response in the Southern San Francisco bay area, Bull. Seism. Soc. Am. 84, 1815–1830. Michetti, A. M., Brunamonte, F., Serva, L. and Vittori, E., 1996, Trench investigations of the 1915 Fucino earthquake fault scarps (Abruzzo, central Italy): Geological evidence of large historical events, J. Geophys. Res. 101, 5921–5936. Nakamura, Y., 1989, A method for dynamic characteristics estimation of subsurface using microtremor on the ground surface, QR. Railway Tech. Res. Inst. 30, 1. Nicolich, R., 1981, Il profilo Latina-Pescara e le registrazioni mediante OBS nel Mar Tirreno, Atti del 1◦ Convegno annuale Gruppo Nazionale di Geofisica della Terra Solida (CNR) II, 621–637. Rautian, T. G. and Khalturin, V. I., 1978, The use of the Coda for the determination of the earthquake source spectrum, Bull. Seism. Soc. Am. 68, 949–971. Scarpa, R. and Zollo, A., 1985, Seismotectonics and seismic quiescence in the Southern Apennines, Italy, Earthq. Predict. Res. 1, 81–93. Scarpa, R., Cirella, A. and De Luca, G., 1996, Borehole and underground geophysical experiments in central Apennines, Italy, Proc. Workshop ‘Monitoring and research for mitigating seismic and volcanic risk’, Rejkjavik, EUR 16696 EN, Brussels, pp. 87–98. Shearer, P. M. and Orcutt, J. A., 1987, Surface and near-surface effects on seismic waves – Theory and borehole seismometer results, Bull. Seism. Soc. Am. 77, 1168–1196. Steidl, J. H., Bonilla, F. and Tumarkin, A. G., 1995, Seismic Hazard in the San Fernando Basin, Los Angeles, CA: A site effect study using weak motion and strong motion data, Proc. of the Fifth Intern. Conf. on Seis. Zonation. Nice, France, Oct 17–19, pp. 1149–1156. Steidl, J. H., Tumarkin, A. G. and Archuleta, R., 1996, What is a reference site? Bull. Seism. Soc. Am. 86, 1733–1748. Steidl, J. H., Tumarkin, A. G. and Archuleta, R., 1997, Erratum, Bull. Seism. Soc. Am. 87, 516. Su, Feng, Anderson, J. G., Brune, J. N. and Zeng, Y., 1996, A comparison of S wave and Coda wave site amplification determined from aftershocks of the Little Skull Mountain earthquake, Bull. Seism. Soc. Am. 84, 1006–1018. Su, Feng, Anderson, J. G., Brune, J.N., Zeng, Y., Sheehan, A., Smith, K., Savage, M. and Anooshehpoor, R., 1993, Site response in the Yucca mountain area determined from aftershocks of the Little Skull Mountain earthquake, EOS 74, p. 422. Theodulidis, N., Bard, P. Y., Achuleta, R. and Bouchon, P., 1996, Horizontal to Vertical spectral ratio and geological conditions: the case of Garner Valley downhole array in Southern California, Bull. Seism. Soc. Am. 86, 306–319. Tsujiura, M., 1978, Spectral analysis of the Coda waves from local earthquakes, Bull. Earthquake Res. Inst. Tokio Univ. 53, 1–48. Tucker, B. E. and King, J. L., 1984, Dependence of sediment filled valley response on the input amplitude and the valley properties, Bull. Seism. Soc. Am. 74, 153–165. Tucker, B. E., King, J. L., Hatzfeld, D. and Nersesov, I. L., 1984, Observations of hard rock site effects, Bull. Seism. Soc. Am. 74, 121–136. Westaway, R., Gawthorpe, M. and Tozzi, M., 1989, Seismological observations of 1984 Lazio-Abruzzi earthquakes: implications for the active tectonics of Italy, Geophys. J. 98, 489–514.
jose52.tex; 17/08/1998; 19:32; p.14
