This article was downloaded by:[Andrej, Gosar]On: 21 January 2008Access Details: [subscription number 789742391]Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Earthquake EngineeringPublication details, including instructions for authors and subscription information:http://www.informaworld.com/smpp/title~content=t741771161

Site Effects Study in a Shallow Glaciofluvial Basin UsingH/V Spectral Ratios From Ambient Noise andEarthquake Data: The Case of Bovec Basin (NWSlovenia)Andrej Gosar aa Environmental Agency of Slovenia, Seismology and Geology Office and Universityof Ljubljana, Ljubljana, Slovenia

Online Publication Date: 01 January 2008To cite this Article: Gosar, Andrej (2008) 'Site Effects Study in a Shallow GlaciofluvialBasin Using H/V Spectral Ratios From Ambient Noise and Earthquake Data: TheCase of Bovec Basin (NW Slovenia)', Journal of Earthquake Engineering, 12:1, 17 -

35To link to this article: DOI: 10.1080/13632460701457140URL: http://dx.doi.org/10.1080/13632460701457140

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

This article maybe used for research, teaching and private study purposes. Any substantial or systematic reproduction,re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expresslyforbidden.

The publisher does not give any warranty express or implied or make any representation that the contents will becomplete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should beindependently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings,demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with orarising out of the use of this material.

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

Journal of Earthquake Engineering, 12:17–35, 2008Copyright © A.S. Elnashai & N.N. AmbraseysISSN: 1363-2469 print / 1559-808X onlineDOI: 10.1080/13632460701457140

17

UEQE1363-24691559-808XJournal of Earthquake Engineering, Vol. 00, No. 0, June 2007: pp. 1–20Journal of Earthquake Engineering

Site Effects Study in a Shallow Glaciofluvial Basin Using H/V Spectral Ratios From Ambient Noise and Earthquake Data:

The Case of Bovec Basin (NW Slovenia)

xxxxA. Gosar ANDREJ GOSAR

Environmental Agency of Slovenia, Seismology and Geology Office and University of Ljubljana, Ljubljana, Slovenia

The Bovec basin, which is located in the alpine valley of the So2a river (NW Slovenia), was recentlystruck by two strong earthquakes (1998 and 2004) which caused extensive damage of maximumintensity VII-VIII EMS-98. Macroseismic data for both events showed large variations in damage tobuildings within short distances and indicated strong effects of sediments on ground motion. A siteeffects study was therefore performed using H/V spectral ratios from earthquake data and fromambient noise, as well as standard spectral ratio technique using the reference station located on theedge of the basin. Following the July 12, 2004 (Mw = 5.2) earthquake, six strong motion seismic sta-tions were deployed in a profile across the Bovec basin to record the aftershock sequence. Accelero-grams of eight stronger aftershocks (ML = 2.5–3.6) and additional ambient noise measurements wereused in the study. Spectral ratio analyses showed that ground motion amplification occurs mainly ina frequency range of 5–10 Hz, with corresponding amplitudes in the range of 6–11. The observedrange of amplification cannot be related to the total thickness of Quaternary sediments, which is upto 100 m in the Bovec basin. The variability in the main peak frequencies and in their amplitudes istherefore explained by the complex geological structure of the basin, filled with heterogeneous gla-cial and fluvial sediments. Irregular layers of conglomerate within sand/gravel deposits and layers oftillite result in large impedance contrasts at several interfaces within Quaternary sediments. Spectralratios from earthquake data are therefore quite complex and show a broad range of ground motionamplification. On the other hand, ambient noise data revealed only the first stronger impedance con-trast which is related in the border areas to the flysch bedrock and in the central part of the basin toa shallow layer of conglomerate. Comparison of the two H/V analyses showed that the amplitudeobtained from ambient noise data is always lower than the amplitude from earthquake data. The dif-ference can be as much as a factor of two. These quantitative results apply to weak ground motionand because of nonlinearity cannot be directly extrapolated to damaging earthquake situation. Sinceone and two-story houses prevail in the Bovec basin, with the main building frequency in the range of6–11 Hz, the danger of soil-structure resonance is considerable in the area. It was presumably one ofthe main reasons for the relatively high level of damage observed from both earthquakes.

Keywords Site Effects; H/V Spectral Ratio; Ambient Noise; Ground Motion Amplification

1. Introduction

Several site effect studies based on the H/V spectral ratio method or SSR method haverecently been published, presenting the results from various sedimentary basins. Although

Received 13 November 2006; accepted 27 April 2007.Address correspondence to Andrej Gosar, Environmental Agency of Slovenia, Seismology and Geology

Office and University of Ljubljana, Faculty of Natural Sciences and Engineering, Dunajska 47, Sl-1000, Ljubl-jana, Slovenia; E-mail: andrej.gosar@gov.si

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

18 A. Gosar

their geological setting is different, deep basins filled with unconsolidated Quaternarydeposits, several hundred meters thick, predominate in these studies. The simple 1-D reso-nance frequency of the sediments column is therefore usually low (≤ 2 Hz). Such exam-ples include the Cologne area [Parolai et al., 2004], the Lower Rhine Embayment [Hinzenet al., 2004; Ibs-von Seht and Wohlenberg, 1999] in Germany, the western part of thecity of Basel located in Rhinegraben [Fäh et al., 1997] and the Rhone valley in Valais[Frischknecht et al., 2005] in Switzerland, Thessaloniki [Panou et al., 2004] and theVolvi basin (Euroseistest area) [Theodulidis, 2006] in Greece, and the Venosa area [DiGiacomo et al., 2005] and Molise [Gallipoli et al., 2004] in Italy. Less frequent are exam-ples of shallow deposits (up to several ten meters), which result in a higher resonance fre-quency (> 2 Hz). Such examples are the area of Lisbon [Teves-Costa et al., 1996] inPortugal, the Segura river valley [Delgado et al., 2000] in Spain, basins in Umbria-Marcheregion [Mucciarelli and Monachesi, 1998] in Italy, Victoria basin [Molnar and Cassidy,2006] in British Columbia (Canada), and Parkway basin [Yu and Haines, 2003] in NewZealand. The main reason that studies from deeper basins predominate may be that majorcities for which microzonation studies are normally performed are usually located in deepsedimentary basins. At the same time, high buildings, which are vulnerable at lower fre-quencies are more common in major cities. Therefore, it is important to study the dangerof soil-structure resonance at the lower part of the frequency spectrum.

On the other hand, many towns and settlements are located in smaller and shallowerbasins and valleys. Since family houses (one to two floors high) with a main building fre-quency much above 2 Hz prevail in these regions, it is also very important to study areascovered with thin deposits. Geological situations in which a high heterogeneity of sedi-ments can be expected are also challenging for site effect assessment. Alpine valleys filledwith heterogeneous glacial and fluvial deposits, including lacustrine chalk, till (moraine),clay, silt, sand, and gravel, such as that presented in this study, are one example. More-over, sediments are frequently cemented in conglomerate and tillite. The shear wavevelocity and density of these sediments can therefore vary considerably. This results inlarge variations in the impedance contrast and the depth to the bedrock within short dis-tances and, consequently, in the observed fundamental frequency of sediments and theamplification level.

The Bovec basin is located in the Upper So2a valley in NW Slovenia (Fig. 1), anactive tectonic region undergoing a recent increase in seismic activity. Two strong earth-quakes have recently struck the area, in 1998 (Mw = 5.6) and 2004 (Mw = 5.2). The maxi-mum intensity of the first event was VII-VIII and of the second VI-VII EMS-98. The largeintensity variations within the basin observed during the 1998 earthquake indicated thatsediments had strong effects on seismic motion [Zupan2i2 et al., 2001]. The first siteeffect study which followed was based on the results of sparse geophysical investigations,geotechnical boreholes, and a few ambient noise measurements [Gosar et al., 2001], butno earthquake records within the basin were available at that time. Some areas of probablesoil-structure resonance were identified and large variations in damage to buildings in thetown of Bovec were explained by significant variations in the fundamental frequency ofsediments and ground motion amplification. Later, ambient noise measurements were per-formed in the whole Bovec basin in a 200 m dense grid. Large variations in the fundamen-tal frequency of sediments were obtained, with most of the observed values in the range7–11 Hz [Gosar, 2007]. Additional ambient noise measurements in several one and two-story houses, which prevail in the area, indicated that the danger of soil-structure reso-nance is considerable in the area. The observed frequencies could not be related to theknown thickness (up to 100 m) of Quaternary glaciofluvial sediments, but were explainedby internal heterogeneities, mainly the presence of conglomerate or lithified moraine

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

Site Effects Study in Bovec Glaciofluvial Basin (NW Slovenia) 19

(tillite). Since the lenses of consolidated conglomerate or tillite within unconsolidated sed-iments are highly irregular, the fundamental frequency changes considerably over shortdistances. The study of site effects in such conditions is therefore very complex, becausemore than one interface with significant impedance contrast can be expected. Analysis ofearthquake data recorded within the basin can therefore contribute substantially to the reli-ability of ground motion amplification estimation.

Following the July 12, 2004 (Mw = 5.2) earthquake, several seismographs and strongmotion instruments were deployed in the epicentral area to record aftershocks [Živ2icet al., 2006]. To study ground motion amplification, in addition to a permanent accel-erograph located in Bovec, five strong motion instruments were deployed in the Bovecbasin (Fig. 1) along a profile which extends in a N-S direction. Records of eight strongeraftershocks (ML = 2.5–3.6) were used in this study. Ambient noise measurements wereperformed at the same locations. To assess site effects in terms of fundamental frequencyand corresponding amplitude, horizontal to vertical (H/V) spectral ratios on earthquakedata and ambient noise were computed. In addition, standard spectral ratios (SSR) to thereference station located at the edge of the basin were computed for comparison. Theresults of SSR were considered to be less reliable because: (a) the epicentral distance wasrelatively small with respect to the distance between the recording stations; and (b) the ref-erence station is not completely free of site effects.

2. Seismological and Geological Setting

NW Slovenia marks a kinematic transition between the E-W striking thrust faults of theAlpine system and the NW-SE striking faults of the Dinarides system. Before 1998, rela-tively weak rates of seismicity were characteristic of this area. Nevertheless, in the seismic

FIGURE 1 Shaded relief map of the Bovec basin with epicenters of July 12, 2004 earth-quake in Krn mountains and eight stronger aftershocks. Black triangles are locations ofstrong motion instruments. Boxed area indicates the map shown in Fig. 2.

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

20 A. Gosar

hazard map for a return period of 475 years, the Bovec basin is located in a region ofincreased seismic hazard, with 0.225–0.250 g design ground acceleration [Lapajne et al.,2001]. This is mainly due to the proximity of a high level of seismic activity in the Friuliarea (30–40 km to the W) where the Mw = 6.4 Friuli earthquake occurred in 1976.

The region has undergone a recent increase in seismic activity. The April 12, 1998(Mw = 5.6) and July 12, 2004 (Mw = 5.2) earthquakes occurred on the NW-SE trendingnear-vertical Ravne fault in the Krn mountains at a 7–9 km depth [Zupan2i2 et al., 2001;Živ2ic et al., 2006]. The focal mechanisms of both earthquakes show almost pure dextralstrike-slip. The epicentral distance to the town of Bovec was only 6–7 km. Strong varia-tions in damage to buildings were observed within short distances in the whole Bovecbasin. In general, they cannot be explained by changes in the epicentral distance, bychanges in the radiation of seismic energy from the source or by differences in buildingvulnerability, although these factors also need further attention.

The Bovec basin (6 km long and 2 km wide) developed in the alpine valley of theSo2a river, which is elsewhere very narrow (Fig. 1). The basement consists of Mesozoicplatform carbonates (Fig. 2), overlain by a succession of deep-water flysch or by marlylimestone (scaglia) with intercalated calcarenites, shales, marls, and conglomerates. Qua-ternary sediments are represented from bottom to top by partly lithified glaciofluvial sedi-ments, overlain by lacustrine chalk [Bavec et al., 2004]. During the Holocene, the chalk

FIGURE 2 Simplified geological map of the Bovec basin after Buser [1986], Jurkovšek[1986], and Bavec et al. [2004]. Black triangles are locations of strong motion instru-ments. A-A' indicates cross-section shown in Fig. 3.

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

Site Effects Study in Bovec Glaciofluvial Basin (NW Slovenia) 21

was partly eroded (in some areas totally) and covered by glaciofluvial sand and gravel,weakly cemented in some parts into conglomerate, and by unconsolidated moraine (till).

The lithologic column can be divided into three sections, based on physical proper-ties. The basement consists of platform carbonates. They are overlain by a succession ofdeep water clastics (flysch). Quaternary sediments represent the third section (Fig. 2). Athick sequence of lacustrine chalk covers lithified glacial and fluvial sediments in the cen-tral part of the basin [Bavec, 2002; Kuš2er et al., 1974]. Glacial moraine sediments are inparts cemented (tillite). During the Holocene, the chalk was partly eroded and covered byolder (Bovec terrace) and younger sand and gravel in some parts weakly cemented intoconglomerate, by scree and debris under the steep flanks of the basin and by recent fluvialsediments along the So2a river [Bavec, 2002].

A 2-D cross-section (Fig. 3) was prepared, based on sparse geophysical data becauseno deep boreholes have been drilled in the area. Electrical sounding was used to determinethe thickness of glaciofluvial sediments, while the boundary between flysch and carbon-ates was detectable only in shallower parts on the edge of the basin. Electrical soundingdid not distinguish between flysch and lacustrine chalk [Gosar et al., 2001]. In the vicinityof the cross-section (Fig. 3), according to the geological data, the chalk is found mainly inthe Cezso2a area under younger sand and gravel, while under the Bovec terrace only fly-sch is expected.

3. Methods

Various experimental-empirical methods have been developed during the last decades toanalyze site effects [Pitilakis, 2004]. In the Bovec basin, we applied H/V spectral ratiotechnique using ambient noise data and earthquake records, as well as standard spectralratio technique.

The most widely used technique to characterize site amplification is Standard Spec-tral Ratio (SSR), which is defined as the ratio of the amplitude spectra of a soil-site recordto that of a nearby rocks-site record for the same earthquake and component of motion[Borcherdt, 1970]. The main conditions for the application of this method are: (a) the ref-erence site has to be free of any kind of site effects; (b) the distance between the soil siteand the reference one should be small in comparison to the epicentral distance [Pitilakis,2004]. Because both conditions were only partly fulfilled in the Bovec basin, this methodwas used only in addition to other two methods.

FIGURE 3 Cross-section of Quaternary sediments in the Bovec basin interpreted fromvertical electrical soundings. Black triangles are locations of strong motion instruments.Vertical scale is three times exaggerated.

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

22 A. Gosar

The horizontal to vertical (H/V) spectral ratio technique on earthquake data does notdepend on a reference site. It uses the spectral ratio of the horizontal to the vertical compo-nent of the ground motion recorded at the same station to estimate the transfer function[Lermo and Chavez-Garcia, 1993]. It is based on the assumption that the vertical compo-nent is free of any soil influence at the recording site in cases where the soil stratigraphy isflat and horizontal [Pitilakis, 2004].

The H/V spectral ratio method on ambient noise uses the same approach as the previ-ous method but uses the ambient noise (microtremors). It is now widely used for microzo-nation and site effects studies. However, the theoretical basis of this technique [Lachet andBard, 1994] is still unclear, since two different explanations have been proposed [Bonnefoy-Claudet et al., 2006]. Reviews on the method can be found in Bard [1999] or Mucciarelliand Gallipoli [2001]. In general, it is widely accepted that the method is successful inidentifying the fundamental frequency of the sediments and less certain in estimating thelevel of amplification.

Previous studies (e.g., Bindi et al., 2000; Parolai et al., 2004) have also shown that theamplification derived from H/V spectral ratio obtained from ambient noise is almostalways smaller than that obtained from earthquake data. This observation should be there-fore checked for each area under investigation.

4. Earthquake Data

The aftershock sequence which followed the July 12, 2004 (Mw = 5.2) earthquake wasrecorded with one permanent (BOVC) and five temporary strong motion instruments(Figs. 1, 2) deployed by the Environmental Agency of Slovenia, the University of Trieste –DST, the Istituto Nacionale di Geofisica e Vulcanologia (Italy), and Istituto Nacionale diOceanografia e di Geofisica Sperimentale from Trieste (Italy) in the first three days afterthe main shock [Živ2ic et al., 2006; Pahor et al., 2006]. From the extensive database ofrecords, eight stronger aftershocks (ML = 2.5–3.6), which occurred in the period July 13, –August 18, 2004, were selected for further analysis [Cecic et al., 2006; Kastelic et al.,2006]. Their locations are within 2 km of the epicenrte of the main shock (Fig. 1). The epi-central distance to the closest station (FEJN) was between 3 and 6 km and to the most dis-tant station (RUSC) between 5 and 8 km. The depth of the aftershocks was in the range of7–10 km. Five stations were equipped with Kinemetrics Etna accelerographs and Episen-sor or FBA-23 sensors of different ranges (one station – 1 g, three stations – 2 g, and onestation – 4 g), and one station (CZS6) with a Lennartz 3-D 1s seismometer. The RUSCand FEJN stations were located in stables, the BOVC station in a larger building of thecultural center in Bovec, the BOLE and CZS6 stations in containers, and the CS27 stationin a two-story house in Cezso2a. Analyses have shown that larger buildings, in particular,can have a significant influence on the results compared to ambient noise measurementsperformed in a free field.

Examples of two aftershock records (ML = 2.8–2.9) are shown in Fig. 4. The influ-ence of basin sediments is clearly reflected in the amplitude and duration of the accelero-grams, especially at the BOLE, CZS6, and CS27 stations.

H/V spectral ratio analysis was performed in the following way. After baseline andinstrument response correction, the Fourier spectrum was calculated for tapered window,including S-phase and later arrivals. The amplitude spectra were smoothed and the H/Vspectral ratio computed as the average of the two horizontal component spectra divided bythe vertical one. Spectral ratios for all eight earthquakes were then inspected. In case theresults for one or two events being significantly different than others, they were excludedand the average spectral ratio was computed at the end.

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

Site Effects Study in Bovec Glaciofluvial Basin (NW Slovenia) 23

In addition, SSR was computed using the RUSC station, located on the northern edgeof the Bovec basin as a reference station. Ambient noise measurements at this station (Fig. 5b)showed a nearly flat H/V ratio for frequencies up to 15 Hz, where there is a small peak,probably related to a thin layer of debris. On the other hand, from the accelerogramsshown in Fig. 4 and from H/V spectral ratios from earthquake data (Fig. 5a), it is clear thatthe location is not completely free of site effects. Because the station was located on aslope, some topographic effects may also be present. The results of SSR were thereforeconsidered to be less reliable. As with the H/V spectral ratio, results for all eight earth-quakes used were compared, and possible outlier curves excluded. At the end, the spectralratios for different events were averaged.

5. Ambient Noise Measurements

Measurements of ambient vibrations were performed in a free field in the vicinity of thebuildings in which seismic stations were located, but far enough from them to avoid their

FIGURE 4 Unfiltered accelerograms of two aftershocks recorded on N-S components.(a) ML = 2.9 earthquake on August 18, 2004; (b) ML = 2.8 earthquake on July 23, 2004.The vertical scale for all accelerograms is 0.1 g.

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

24 A. Gosar

influence on recording. A portable Tromino (Micromed) seismograph with three-compo-nent velocity sensor was used. All parts of this seismograph are integrated in a commoncase to avoid electronic and mechanical noise that can be introduced by wiring. Goodground coupling on soft soils was obtained by using long spikes mounted at the base of theseismograph. The recording length was 20 min, which allows spectral analysis down to0.5 Hz. Measurements were performed on a day without wind, because the noise intro-duced by strong wind is known to severely affect the reliability of data. The noisiest con-ditions for measurements were in the Cezso2a area, because of retrofitting constructionactivities and the vicinity of the So2a river.

FIGURE 5 RUSC station used as a reference station for SSR. (a) H/V spectral ratio fromearthquake data; (b) H/V spectral ratio from ambient noise.

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

Site Effects Study in Bovec Glaciofluvial Basin (NW Slovenia) 25

H/V spectral ratio analysis was performed in the following way. Recorded time serieswere visually inspected to identify stronger transient noise. Each record was then split into30 s long nonoverlapping windows, for which amplitude spectra were computed using atriangular window with 5% smoothing and corrected for sensor transfer function. The H/Vspectral ratio was computed as the average of the two-horizontal component spectradivided by the vertical spectrum for each window. From the color-coded plot of H/V spec-tral ratio functions for all 40 windows, those windows including strong transient noisewere identified, in order to be excluded from further computation. At the end, the averageH/V function with a 95% confidence interval was computed.

In addition to earthquake recordings and ambient noise measurements, seismic refrac-tion measurements were performed at the location of the BOVC station and multi-channelanalysis of surface waves (MASW) measurements at the location of the CS27 station, inorder to determine the shallow S-wave velocity structure. Seven vertical electrical sound-ings were performed along a profile (Fig. 3) in order to determine the thickness of theQuaternary sediments [Gosar et al., 2001].

6. Results

The spectral ratios for five seismic stations in the Bovec basin, ordered from north tosouth, are shown in Figs. 6–10. Each figure contains from top to bottom; (a) averaged H/Vspectral ratio from earthquake data; (b) H/V spectral ratio from ambient noise; and (c)SSR to the reference station RUSC, all with 95% confidence interval. In case of ambientnoise the average was computed from 40 windows of subsequent measurements. The stan-dard deviation is therefore usually very small. On the other hand, in the case of earthquakedata average spectra were computed from only 5–7 records what resulted in much higherscattering of values. Standard deviation was similar for H/V spectral ratio from earthquakedata and for SSR data. The fact that we compared results based on earthquake datarecorded with accelerographs located inside buildings and results from microtremor datarecorded in the free field is probably the reason for some discrepancies between the twodata sets described below. However, the differences in the nature of ambient noise andstrong motion wavefields presumably have even greater influence on the results. In gen-eral, the results of ambient noise measurements show only one prominent peak, at the fun-damental frequency, which may be related to the main contrast in acoustic impedance. Onthe other hand, H/V spectral ratios from earthquake data and SSR are more complex andnormally show a broader range of amplification and several peaks within this range.

H/V spectral ratios from earthquake data showed that ground motion amplificationoccurs mainly in a frequency range of 5–10 Hz, with corresponding amplitudes in a rangeof 6–11. The amplitudes of H/V spectral ratio peaks from ambient noise data are in gen-eral much lower than in earthquake data, with the exception of the BOVC station. Asalready mentioned, SSR results are considered as less reliable because of the relativelylarge spacing between compared seismic stations with respect to the epicentral distanceand because the reference station is not completely free of site effects.

At the BOVC station (Fig. 6), both H/V spectral ratios show peaks with amplitudesexceeding 10, but they are at slightly different frequencies. In ambient noise data (freefield) there is a very sharp peak at 9.9 Hz, while in earthquake data (recorded in a build-ing) the main peak is at 7.5 Hz. Additional ambient noise measurements in the buildingshowed that the main building frequency is around 6.8 Hz. The observed shift in the fre-quency is therefore probably caused by the influence of the building on the earthquakerecord. In any case, the relatively high frequencies obtained may be related to the veryshallow velocity structure, also found by the seismic refraction method. Two layers were

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

26 A. Gosar

FIGURE 6 BOVC station. (a) H/V spectral ratio from earthquake data; (b) H/V spectralratio from ambient noise; (c) SSR to reference station RUSC.

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

Site Effects Study in Bovec Glaciofluvial Basin (NW Slovenia) 27

FIGURE 7 BOLE station. (a) H/V spectral ratio from earthquake data; (b) H/V spectralratio from ambient noise; (c) SSR to reference station RUSC.

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

28 A. Gosar

FIGURE 8 CZS6 station. (a) H/V spectral ratio from earthquake data; (b) H/V spectralratio from ambient noise; (c) SSR to reference station RUSC.

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

Site Effects Study in Bovec Glaciofluvial Basin (NW Slovenia) 29

FIGURE 9 CS27 station. (a) H/V spectral ratio from earthquake data; (b) H/V spectralratio from ambient noise; (c) SSR to reference station RUSC.

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

30 A. Gosar

FIGURE 10 FEJN station. (a) H/V spectral ratio from earthquake data; (b) H/V spectralratio from ambient noise; (c) SSR to reference station RUSC.

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

Site Effects Study in Bovec Glaciofluvial Basin (NW Slovenia) 31

identified overlying the 9-m deep flysch bedrock, which has VS≈1000 m/s. The topmostlayer of soil is 2.2 m thick and has VS = 150 m/s. The second layer is 7 m thick and hasVS = 450 m/s. The SSR to the reference station RUSC gave a result that is not comparableto the H/V spectral ratio in terms of amplitude and shows only a small peak around 6 Hz.Since the results were more comparable at all other stations, we have no explanation forthis. The influence of the building on earthquake records was analyzed by inspection ofboth horizontal components, because BOVC station was located in a large building of theBovec cultural center, whereas all other instruments were located in smaller buildings. Ithas shown that influence of the building on the shift of the frequency with respect to ambi-ent noise free-filed measurements is very likely.

At the BOLE station (Fig. 7), there is a good correspondence of frequencies of both H/Vmain peaks at 8.5–8.8 Hz. Previous investigations [Gosar, 2007] have shown that the funda-mental frequency from ambient noise data in this region (Bovec terrace) may be related tothe boundary between unconsolidated sand/gravel and weakly cemented conglomerate at adepth of 5–20 m. The total thickness of Quaternary sediments (up to 100 m), known fromvertical electrical soundings (Fig. 3), is therefore not reflected in ambient noise data, but ispresumably related to the peak near 2.8 Hz in the H/V ratio from earthquake data (Fig. 7a).

At the CZS6 station (Fig. 8), some peaks in the high frequency range (10–15 Hz)were obtained in all three spectral ratios. It is known from outcrops exposed in the nearbycliff of the So2a river that the conglomerate layer here is very shallow (< 10 m). Thisexplains the observed high frequencies. Other peaks at 3.5, 4, 6, and 8 Hz are visible in theH/V from earthquake data and SSR and are related to deeper heterogeneities. Inspection ofother ambient noise measurements [Gosar, 2007] in the central part of the Bovec terrace,where BOLE and CZS6 stations are located, has shown that the amplitude of peaks is ingeneral low (up to 4), probably because of the lower acoustic impedance contrast betweensand/gravel and weakly cemented conglomerate. Earthquake data (Fig. 4) and spectralratios derived from them (Figs. 7 and 8) show higher amplitudes (up to 10), which seemsto be a more reliable assessment of ground motion amplification.

At the CS27 station (Fig. 9) there are clear peaks at 6 Hz in all three spectral ratios.The general shape of the spectral ratio curves is also comparable, with a second range ofhigher values at the high frequency end of the curve. The standard deviation of SSR dataat this location was considerably higher in comparison with other stations, especially atfrequencies lower than 4 Hz. The amplitudes are again much higher in the H/V spectralratio from earthquake data (up to 7.5). The high amplification is in agreement with thehigh damage to buildings in Cezso2a caused by 1998 and 2004 earthquakes. A high levelof noise was characteristic of microtremor data for the whole Cezso2a area, because of ret-rofitting construction activities and the vicinity of the So2a river. This results in a lesssmooth spectral ratio than at other locations. MASW measurements in Cezso2a gave twolayers overlying the bedrock composed of flysch or lacustrine chalk at a depth of 7–11 m,with VS≈900 m/s. The topmost layer of soil is 2.1 m thick and has VS = 130 m/s. The sec-ond layer is 5–9 m thick and has VS = 410 m/s. 7–11 m thick layer of sediments at givenvelocities cannot be directly related to clear peaks obtained at 6 Hz with all three methods.For their modeling deeper velocity structure is therefore needed, which was not resolvedwith shallow geophysical methods used.

At the FEJN station (Fig. 10), which was located near the southern edge of the basin,ambient noise measurements resulted in an almost flat H/V spectral ratio. On the otherhand, H/V spectral ratios from earthquake data and SSR show clear peaks at a frequencyof around 15 Hz, which may be related to the very thin sediment layer (till) expected inthis area. At this station the standard deviation was relatively high at H/V spectral ratiosfrom earthquake data.

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

32 A. Gosar

7. Discussion and Conclusions

The investigations carried out showed that site effects have a considerable influence onearthquake ground motion in the Bovec basin. This observation is consistent with the rela-tively large damage observed during the two earthquakes in 1998 and 2004. Moreover,strong variations in the level and intensity of damage characterize the whole Bovec basin.Analyses have shown that ground motion amplification, derived from spectral ratiosoccurs mainly at relatively high frequencies, in the range of 5–10 Hz, and have corre-sponding amplitudes of 6–10. Analyses and modeling of the July 12, 2004 (Mw = 5.2)main shock record at BOVC station have also shown that spectral acceleration is highestin the frequency range of 2–10 Hz [Vanini et al., 2006]. Since one and two-story housesprevail in the area, the main frequency of which is roughly in the same range (from 6–11Hz) [Gosar, 2007], the danger of soil-structure resonance is considerable. It was presum-ably one of the main reasons for the relatively high level of damage caused by bothearthquakes.

Comparison of the two H/V analyses have shown that the amplitude derived fromambient noise data is almost always lower than the amplitude from earthquake data. Thedifference can be as much as a factor of two. This is in agreement with previous observa-tions that ambient spectral ratios are more reliable in estimating the fundamental fre-quency than the amplitude of amplification, which is almost always greater in earthquakedata. The only exception is the BOVC station. However, the influence of the relativelylarge building in which this accelerograph was located on earthquake recordings is verylikely.

Although earthquake records (Fig. 4) clearly show the highest amplifications in thecentral part of the basin (CZS6, BOLE, and CS27 stations) this is not uniformly reflectedin the amplitude of the highest peak in H/V spectral ratios, which shows higher amplitudesat BOVC (11) and BOLE (10) than at CZS6 (6.5) and CS27 (7.5). Similar observationsapply to the SSR analyses; the highest amplitudes were obtained at CZS6 (11.5) and CS27(13.5) stations and lower at BOLE (6.5) station. Since there are no settlements located inthe central part of the basin (Fig. 2) where the highest amplifications were observed, nomacroseismic data are available in this area for comparison. These quantitative resultsapply to weak ground motion and because of nonlinearity cannot be directly extrapolatedto damaging earthquake situation. The amplifications due to site effects are much smallerin the case of strong ground motion and the site frequencies are normally lower.

The main limitations of the applied methods are the following. The differences in thenature of ambient noise and weak motion wavefields have influence on the results. There-fore, the results of ambient noise measurements show only one prominent peak, whereasspectral ratios from earthquake data are more complex and normally show a broader rangeof amplification and several peaks within this range. The main limitation of SSR methodwas, in our case, that the distance between the recording sites and the reference one wasrelatively large in comparison to the epicentral distance.

The frequency range of amplification cannot be related to the total thickness of Qua-ternary sediments in the basin, which is up to 100 m. The observed variability in terms ofthe main peak frequencies and amplitudes can be explained by the fairly complex geolog-ical structure of the basin, which is not filled with uniform Quaternary sediments. On thecontrary, the combined glacial and fluvial origin of the sediments has resulted in a veryheterogeneous internal structure. Because of irregular layers of conglomerate within thesand/gravel and the presence of tillite in the border areas, large impedance contrasts at var-ious interfaces within the Quaternary deposits can be expected. However, vertical electri-cal soundings (Fig. 3) performed along the investigated profile could not distinguish

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

Site Effects Study in Bovec Glaciofluvial Basin (NW Slovenia) 33

between sand/gravel and weakly cemented conglomerate. It is obvious that ambient noisedata revealed only the first stronger impedance contrast, which, in the border areas isrelated to flysch bedrock or lacustrine chalk and in the central part of the basin to a shal-low layer of conglomerate. On the other hand, spectral ratios from earthquake data aremore complex and show a broad range of ground motion amplification. The several peaksin H/V spectral ratios from earthquake data and SSR analyses, at frequencies below thatobtained by ambient noise measurements, can be explained by additional impedance con-trasts at depth. In addition, 2-D effects, multiple reflections and the secondary generationof surface waves within the basin may have a significant influence.

Acknowledgments

The study was realized with the support of Interreg IIIB Alpine Space project SISMOV-ALP: Seismic hazard and Alpine valley response analysis and NATO SfP project 980857:Assessment of seismic site amplification and seismic building vulnerability in FYR ofMacedonia, Croatia and Slovenia. The aftershock sequence was recorded with strongmotion instruments deployed by the Environmental Agency of Slovenia, the University ofTrieste – DST, Istituto Nacionale di Geofisica e Vulcanologia from Rome (Italy), and Isti-tuto Nacionale di Oceanografia e di Geofisica Sperimentale from Trieste (Italy). We aregrateful to them for providing the data. The author is indebted to Mladen Živ2ic and JurijPahor for their help in data preparation.

References

Bard, P. Y. [1999] “Microtremor measurements: a tool for site effect estimation?,” In: The Effects ofSurface Geology on Seismic Motion, Irikura, K., Kudo, K., Okada, H., and Sasatami, T., eds.(Balkema), pp. 1251–1279.

Bavec, M. [2002] “New temporal, and genetic determination of some late Quaternary sediments inBovec basin and its surroundings (NW Slovenia),” Geologija, 45, 291–298 (in Slovenian, withEnglish summary).

Bavec, M., Tulaczyk, S. M., Mahan, S. M., and Stock, G. M. [2004] “Late Quaternary glaciation ofthe Upper So2a River Region (Southern Julian Alps, NW Slovenia),” Sedimentary Geology, 165,265–283.

Bindi, D., Parolai, D., Spallarosa, D., and Catteno, M. [2000] “Site effects by H/V ratio: comparisonof two different procedures,” Journal of Earthquake Engineering, 4(1), 97–113.

Bonnefoy-Claudet, S., Cotton, F., Bard, P. Y., Cornou, C., Ohrnberger, M., and Wathelet, M. [2006]“Robustness of the H/V ratio peak frequency to estimate 1D resonance frequency,” 3rd Sympo-sium on Effects of Surface Geology on Seismic Motion, Grenoble, pp. 361–370.

Borcherdt, R. D. [1970] “Effects of local geology on ground motion near San Francisco bay,” Bulle-tin of the Seismological Society of America, 60(1), 29–61.

Buser, S. [1986] Basic Geological Map of Yugoslavia 1: 100.000 – Sheet Tolmin and Videm (Geo-logical Survey of Slovenia).

Cecic, I., Živ2ic, M., Jesenko, T., and Kolar, J. [2006] “Earthquakes in Slovenia in 2004,” Potresi vletu 2004, 16–40 (in Slovenian, with English summary).

Delgado, J., López Casado, C., Estévez, A., Giner, J., Cuenca, A., and Molina, S. [2000] “Mappingsoft soils in the Segura river walley (SE Spain): a case study of microtremors as an explorationtool,” Journal of Applied Geophysics, 45, 19–32.

Di Giacomo, D., Gallipoli, M. R., Mucciarelli, M., Parolai, S., and Richwalski, S. M. [2005] “Anal-ysis and modeling of HVSR in the presence of a velocity inversion: The case of Venosa, Italy,”Bulletin of the Seismological Society of America, 95(6), 1–9.

Fäh, D., Rüttener, E., Noack, T., and Kruspan, P. [1997] “Microzonation of the city of Basel,” Jour-nal of Seismology, 1, 87–102.

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

34 A. Gosar

Frischknecht, C., Rosset, P., and Wagner, J. J. [2005] “Towards seismic microzonation-2D model-ing and ambient seismic noise measurements: the case of an embankked, deep Alpine valley,”Earthquake Spectra, 21(3), 635–651.

Gallipoli, M. R., Mucciarelli, M., Eeri, M., Gallicchio, S., Tropeano, M., and Lizza, C. [2004] “Hor-izontal to Vertical Spectral Ratio (HVSR) measurements in the area damaged by the 2002Molise, Italy, earthquake,” Earthquake Spectra, 20(1), 81–93.

Gosar, A., Stopar, R., Car, M., and Mucciarelli, M. [2001] “The earthquake on 12 April, 1998 in Krnmountains (Slovenia): ground motion amplification study using microtremors and modellingbased on geophysical data,” Journal of Applied Geophysics, 47(2), 153–167.

Gosar, A. [2007] “Microtremor HVSR study for assessing site effects in the Bovec basin (NWSlovenia) related to 1998 Mw5.6 and 2004 Mw5.2 earthquakes,” Engineering Geology, 91, 178–193.

Hinzen, K.-G., Scherbaum, F., and Weber, B. [2004] “On the resolution of H/V measurements todetermine sediment thickness, a case study across a normal fault in the lower Rhine embayment,Germany,” Journal of Earthquake Engineering, 8(6), 909–926.

Ibs-von Seht, M., and Wohlenberg, J. [1999] “Microtremor measurements used to map thickness ofsoft sediments,” Bulletin of the Seismological Society of America, 89(1), 250–259.

Jurkovšek, B. [1986] Basic Geological Map of Yugoslavia 1 : 100.000 – Sheet Beljak and Ponteba(Geological Survey of Slovenia).

Kastelic, V., Živ2ic, M., Pahor, J., and Gosar, A. [2006] “Seismotectonic characteristics of the2004 earthquake in Krn mountains,” Potresi v letu 2004, 78–87 (in Slovenian, with Englishsummary).

Kuš2er, D., Grad, K., Nosan, A., and Ogorelec, B. [1974] “Geology of the So2a valley betweenBovec and Kobarid,” Geologija 17, 425–476 (in Slovenian, with English summary).

Lapajne, J., Šket Motnikar, B., and Zupan2i2, P. [2001] “Design ground acceleration map of Slove-nia,” Potresi v letu 1999, 40–49, (in Slovenian, with English summary).

Lachet, C., and Bard, P.-Y. [1994] “Numerical and theoretical investigations on the pos-sibilities and limitations of Nakamura’s technique,” Journal of Physics of the Earth, 42,377–397.

Lermo, J., and Chavez-Garcia, F. J. [1993] “Sitte effect evaluation using spectral ratios with onlyone station,” Bulletin of the Seismological Society of America, 83(5), 1574–1594.

Molnar, S. and Cassidy, J. F. [2006] “A comparison of site response techniques using weak-motionearthquakes and microtremors,” Earthquake Spectra, 22(1), 169–188.

Mucciarelli, M., and Monachesi, G. [1998] “A quick survey of local amplifications and their corre-lation with damage observed during the Umbro-Marchesan (Italy) earthquake of September 26,1997,” Journal of Earthquake Engineering, 2(2), 325–337.

Mucciarelli, M., and Gallipoli, M. R. [2001] “A critical review of 10 years of microtremor HVSRtechnique,” Boll. Geof. Teor. Appl., 42, 255–266.

Pahor, J., Kolar, J., Živ2ic, M., and Cecic, I. [2006] “The earthquake of 12 July 2004 in the Krnmountains and monitoring of the aftershock activity,” Potresi v letu 2004, 88–94 (in Slovenian,with English summary).

Panou, A. A., Theodulidis, N., Hatzidimitriou, P. M., Papazachos, C. B., and Stylianidis, K. [2004]“Ambient noise Horizontal-to-Vertical spectral ratio for assessing site effects in urban environ-ments: The case of Thessaloniki city (N Greece),” Bulletin of the Geological Society of Greece,36, 1467–1476.

Parolai, S., Richwalski, S. M., Milkereit, C., and Bormann, P. [2004] “Assessment of the stability ofH/V spectral ratios from ambient noise and comparison with earthquake data in the Cologne area(Germany),” Tectonophysics 390, 57–73.

Pitilakis, K. [2004] “Site effects,” In Recent Advances in Earthquake Geotechnical Engineering andMicrozonation Ansal A., ed. Kluwer, pp. 139–197.

Teves-Costa, P, Matias, L., and Bard, P. Y. [1996] “Seismic behaviour estimation of thin alluviumlayers using microtremor recordings,” Soil Dynamics and Earthquake Engineering, 15, 201–209.

Theodulidis, N. P. [2006] “Site characterization using strong motion and ambient noise data: Euro-seistest (N Greece),” 3rd Symposium on Effects of Surface Geology on Seismic Motion, Greno-ble, 1, pp. 801–810.

Dow

nloa

ded

By:

[And

rej,

Gos

ar] A

t: 06

:58

21 J

anua

ry 2

008

Site Effects Study in Bovec Glaciofluvial Basin (NW Slovenia) 35

Vanini, M., Villani, M., Faccioli, E., and Gosar, A. [2006] “Modelling of strong ground motion ofthe July 2004, Mw5.2 earthquake in Krn mountains,” 3rd Symposium on Effects of Surface Geol-ogy on Seismic Motion, Grenoble, 1, pp. 1–10.

Yu, J., and Haines, J. [2003] “The choice of reference sites for seismic ground amplification analay-ses: case study at Parkway, New Zealand,” Bulletin of the Seismological Society of America,93(2), 713–723.

Zupan2i2, P., Cecic, I, Gosar, A., Placer, L., Poljak, M. and Živ2ic, M. [2001] “The earthquake of12 April 1998 in the Krn Mountains (Upper So2a valley, Slovenia) and its seismotectonic char-acteristics,” Geologija 44, 169–192.

Živ2ic, M. and Krn-2004 team [2006] “The Krn mountains (Slovenia) MW 5.2 earthquake: dataacquisition and preliminary results,” Geophysical Research Abstracts, 8, 06439.