Natural Hazards

0 downloads 0 Views 4MB Size Report
May 26, 1991 - Professor, Technical University of Kaiserslautern ... Arnaldo Barchiesi. Universidad Nacional de ..... by the Z924 (Molise-Gargano) and the Z925 (Ofanto) areas (Figure 4). Earthquakes in ...... varicoloured clay shales (AV). Fig.

Natural Hazards Lessons learned from two case-histories of seismic microzonation in Italy --Manuscript Draft-Manuscript Number: Full Title:

Lessons learned from two case-histories of seismic microzonation in Italy

Article Type:

Manuscript

Keywords:

Seismic zonation, soil characterization, reference input motion, seismic response analysis, Benevento, San Giuliano di Puglia

Corresponding Author:

Filippo Santucci de Magistris, Ph.D. University of Molise Termoli CB, ITALY

Corresponding Author Secondary Information: Corresponding Author's Institution:

University of Molise

Corresponding Author's Secondary Institution: First Author:

Filippo Santucci de Magistris, Ph.D.

First Author Secondary Information: Order of Authors:

Filippo Santucci de Magistris, Ph.D. Anna d'Onofrio, Associate Professor Augusto Penna, PhD Rodolfo Puglia, PhD Francesco Silvestri, Professor

Order of Authors Secondary Information: Abstract:

The prediction of seismic motion variability in a given urban area is considered an effective tools to plan appropriate urban development, to undertake actions on seismic risk mitigation, to understand the damage pattern caused by a strong motion event. Even though the procedures for studying the seismic response and the seismic zonation of an urban area are well-established, some controversial points still exists and are discussed here. In this paper, the selection of a reference input motion, the construction of a subsoil model, and the site response analysis procedures are specifically dealt with. These points are analyzed partially based on literature reports and mainly on the Authors' experience in two Italian case-histories that are the seismic zonation of the city of Benevento, which is a predictive study, and the analysis of seismic response and damage distribution in the village of San Giuliano di Puglia after a strong-motion earthquake, which is a retrospective analysis.

Suggested Reviewers:

Christos Vrettos Professor, Technical University of Kaiserslautern [email protected] Well known expert in Earthquake Geotechnical Engineering Anastasiadis Anastasios Aristotile University of Thessaloniki [email protected] Expert in seismic zonation. Secretary of TC203 of ISSMGE Arnaldo Barchiesi Universidad Nacional de Cuyo [email protected] member of the TC203 “Earthquake Geotechnical Engineering and Associated Problems” of the ISSMGE

Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation

Susumu Yasuda Professor, Tokyo Denki University [email protected] Well known expert in seismic zonation and earthquake engineering Carlos Alberto Ferreira de Sousa Oliveira Professor, IST Lisboa [email protected] Well known expert in seismic zonation and earthquake geotechnical engineering Opposed Reviewers:

Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation

*Manuscript Click here to view linked References

Lessons learned from two case-histories of seismic microzonation in Italy Filippo Santucci de Magistrisa, Anna d’Onofriob, Augusto Pennab, Rodolfo Pugliac, Francesco Silvestrib a

University of Molise, Structural and Geotechnical Dynamics Laboratory StreGa, Di.B.T. Department, via Duca degli Abruzzi 86039 Termoli CB, Italy b

University of Naples Federico II, DICEA Department, via Claudio 21, 80125, Naples, Italy c

Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Milano-Pavia, via Bassini 15, 20133 Milan, Italy

Corresponding Author: Filippo Santucci de Magistris, Ph.D. University of Molise Structural and Geotechnical Dynamics Lab. StreGa via Duca degli Abruzzi 86039 Termoli CB - Italy tel/fax: +39 0874 404952 Di.B.T. Department Contrada Fonte Lappone, 86090 Pesche (IS) email: [email protected] web: http://docenti.unimol.it/index.php?u=filippo.santucci mob: +39 338 1303406 +39 320 4794782 A PAPER SUBMITTED TO NATURAL HAZARD OCTOBER 2013

Abstract The prediction of seismic motion variability in a given urban area is considered an effective tools to plan appropriate urban development, to undertake actions on seismic risk mitigation, to understand the damage pattern caused by a strong motion event. Even though the procedures for studying the seismic response and the seismic zonation of an urban area are well-established, some controversial points still exists and are discussed here. In this paper, the selection of a reference input motion, the construction of a subsoil model, and the site response analysis procedures are specifically dealt with. These points are analyzed partially based on literature reports and mainly on the Authors’ experience in two Italian case-histories that are the seismic zonation of the city of Benevento, which is a predictive study, and the analysis of seismic response and damage distribution in the village of San Giuliano di Puglia after a strong-motion earthquake, which is a retrospective analysis.

Keywords: Seismic zonation, soil characterization, reference input motion, seismic response analysis, Benevento, San Giuliano di Puglia

1. INTRODUCTION In the last decades, several case histories showed that the effects of an earthquake can significantly differ in a limited part of a territory due to intrinsic heterogeneities in geometric, physical, hydraulic and mechanical characteristics of the subsoil. The empirical understanding of the importance of the soil conditions on the effects of earthquakes date back at least at the XIX century. Mline (1891) stated that “sometimes the harder ground provided better foundation, sometime the softer. The superiority of the one over the other depends on local circumstances”. Studies on seismic site response got a strong burst because of the heavy damages in Mexico City after the 1985 Michoacán earthquake, widely documented by the two special issues of Earthquake Spectra (1985). Well-supported evidences of correlations between damages and soil conditions during earthquakes were collected by scientific journals and bulletins about the 1989 Loma Prieta earthquake (BSSA 1991; NRC 1994), the 1994 Northridge earthquake (BSSA 1994) and the 1995 Kobe earthquake (Soils and Foundations, 1996; 1998). Representative pictures of urban areas damaged by earthquakes can macroscopically document the role of site effects on building damage at a large scale. For instance, Figure 1a shows an aerial view of downtown of Managua, capital city of Nicaragua, hit by a 6.2 magnitude earthquake on Dec. 23, 1972, while Figure 1b refers to the city of Armenia (Colombia) hit by 6.1 magnitude Quindio earthquake, on Jan. 25, 1999. In both cases, clear differential damages in limited areas were easily noticed and could be ascribed to different local soil conditions. The modifications induced by local soil conditions on the seismic response of an urban area can, therefore, produce strongly variable damage patterns after a strong-motion earthquake. Such modifications can be the subject of either a retrospective analysis of an observed damage after a recent strong-motion earthquake, or a predictive study aiming at mapping the likely effects of ‘expected’ scenario earthquakes. In the latter case, being the ground motion primarily dependent on the characteristics of the earthquake source, it is possible to define a seismic macrozonation as the mapping at large scale (prefecture, region, and nation) of areas having the same level of shakeability, represented by a given ground motion parameter defined at a conventional bedrock. On the other hand, the seismic microzonation can be seen as the analysis and representation, for a given urban territory, of the spatial distribution of a ground motion parameter including or representing soil amplification in free field conditions, i.e. without considering buildings and other constructions. Therefore, a seismic microzonation map can be considered as an effective tool supporting land use planning at the urban scale. Hereafter, if not explicitly specified, the seismic microzonation at urban scale will be denoted as ‘seismic zonation’ (or briefly SM), while aspects related to the seismic macrozonation will not be dealt with. The first historical microzonation studies were purely empirical, and based on retrospective analyses of recently occurred earthquakes. A microzonation map of Tokyo, drawn using the damage pattern of the 1854 earthquake was reported in Imamura (1913) as indicated in ISSMGE-TC4 (1999). In more recent years, in a very large number of journal and conference papers (see for instance Singh et al. 1998; Finn 1991; Suetomi and Yoshida 1998; Seed et al. 1991; Scott et al. 1995; Jimenez et al. 2000, and their related references) and several guidelines (see for instance CGS 1997; CGS 2004; NESC 2000; RISC 1994; WIDRM 2004; GdL 2008) rational approaches for seismic microzonation have been proposed. From the above referenced literature, it appears that the international scientific and technical community agrees on a common key points of the zonation process: the dimension of the area and the scale of the maps should be related to the degree of accuracy, rapidity and complexity of the zonation method (see for instance CNR 1986; WIDRM 2004; Vinale et al. 2008).

Nowadays, a well-established approach is the 3-level methodology currently shared by international (e.g. ISSMGE-TC4 1999) and national (e.g. GdL 2008) guidelines, although the use of empirical, simplified and advanced methods (respectively for grade I, II and III mapping) is characterized by variable specifications from case to case. As previously stated, in a wide sense, seismic zonation can be seen as a tool for seismic risk reduction and land management. Therefore, the results of a SM study should not be confused or mixed with the specifications of building codes. Crespellani and Martelli (2008), for instance, state that “certainly, the fact that many studies of post-earthquake seismic zonation, beginning with that of Tarcento (Brambati et al. 1980), were also used for the structural design, may have created the idea that studies of seismic zonation are essentially design-oriented. It is also clear that there are cases in which a thorough study of zonation, for example in an area intended for the construction of a strategic structure or infrastructure like a dam, providing that the site response analysis is carried out with advanced algorithms supported by adequate geotechnical on-site and laboratory testing, can be directly used for the structural design. But it is clear that this is a condition in which the issue is perhaps more nominal than substantial”. Again in WIDRM (2004) it is stated that zone-specific building regulations should reflect the degree of hazard, controlling also the degree of investigations aimed at providing more reliable design criteria. In general, structures can be built in all zones, provided that appropriate measures are taken to encounter the different earthquake hazards. Zone-associated building regulations are only one among several criteria which will lead to a sustainable land use management in a municipality. The final product of a seismic microzonation process is a map in which areas of similar features in terms of seismic hazard and subsoil vulnerability are grouped together, accounting for ground motion amplification, liquefaction and slope stability. Therefore, a combined hazard map for a given area derives from the superposition of a ground shaking map to those of liquefaction susceptibility and of slope stability (WIDRM 2004). Whatever the purpose, three main steps can be recognized in an advanced (grade III) SM study: (1) the selection of a reference input motion, (2) the construction of subsoil model, and (3) the site response analysis. In the following, the discussion on these steps will be developed referring to two Italian case-histories widely studied by the Authors in the last years: the first is the seismic zonation of the city of Benevento (Campania Region), which is a predictive study which adopted a historical earthquake as a seismic scenario; the second is the retrospective analysis of the seismic response and the damage distribution in the village of San Giuliano di Puglia (Molise Region) after a strongmotion earthquake.

2. THE ANALYZED CASE-HISTORIES Benevento is a small town of about 60.000 inhabitants, located close to the Southern Apennine chain, around 200 km SE of Rome. The urban area belongs to the so-called Sannio region, which is considered one of the most seismically active in Italy. The town is characterized by a valuable monumental heritage and the presence of many buildings (reinforced concrete, masonry and mixed type) constructed between the fifties and seventies of the last century, when the country lacked of specific regulations on a seismic design of structures. The above factors suggested considering Benevento as a pilot case study in projects addressed to seismic risk mitigation developed in last years (Marcellini et al. 1995a,b; GNDT 2004; MIUR 2004). In the whole urban area, seismic site effects are expected, i.e., local variations of earthquake ground motion are highly probable, since the subsoil is characterized by lithological heterogeneity and significant widespread seismic impedance contrasts (Improta et al. 2005; Di Giulio et al. 2008). San Giuliano di Puglia is a small village of about 1.000 inhabitants, located in the Molise region, around 250 km E of Rome. It lies on the ridge crest of a narrow hill and shows the typical characteristics of the Southern Apennines villages: buildings are mostly located on outstanding

positions, and frequently connected in heterogeneous structural aggregates. The historical residential houses have typically rubble masonry walls, while in the newer part, built after the 1940s, rubble and brick masonry are mostly present, with sometimes recent additions of reinforced concrete buildings (Mucciarelli et al. 2003). Overall, both historical and newer buildings of San Giuliano di Puglia are usually 2-4 storeys high, resulting in fundamental periods of 0.1 to 0.5s (Baranello et al. 2003). As a consequence of the historical development of the town, the older part was edified on a soft rock formation (Faeto flysch) outcropping in the southern part of the town; in the progressive extension towards North, the newer buildings were raised up along the ridge, on a thick fine-grained soil (Toppo Capuana marly clays). Worldwide attention focused on San Giuliano di Puglia because of the October 31st 2002 Molise earthquake (moment magnitude, MW=5.7) that caused the collapse of a primary school and the subsequent death of 27 pupils and a teacher. The strongly non-uniform damage distribution observed in the town suggested that site amplification significantly affected the seismic response of the area, as confirmed by different studies and research projects developed in the last years (Earthquake Spectra 2004, Cara et al. 2005; Puglia et al. 2007; Lanzo and Pagliaroli 2009; RIG 2009; Puglia et al. 2013).

2.1 The case-history of Benevento 2.1.1 Geological setting

The town of Benevento rises on a hill dominating the confluence of the Calore and Sabato rivers. The geological setting of Benevento is described in detail in Improta et al. (2005), following the groundwork of Improta (1998) and Pescatore et al. (2005). Figure 2 shows a sketch of the geological map of the city of Benevento together with a representative stratigraphic section (Costanzo et al. 2007). The subsoil is essentially made of a Pliocenic clay formation covered by coarse alluvial mixed to fluvial clayey deposits (the so-called Ariano unit). The top of the Pliocenic clay formation ranges from tens of meters (Sabato river valley) to hundreds of meters below the ground level (Calore river valley and the Benevento hill). The historic center lies on a hill made of a Pleistocenic (Rissian) conglomeratic formation, overlying the Pliocenic clay. Towards the S-E side of the hill, fluviolacustrine deposits (the so-called Cretarossa unit) overlay the conglomerate formation. The Cretarossa succession consists of silty and clayey layers with polygenic and heterometric clasts in a sandy matrix. Debris and colluvial deposits deriving from the disintegration of the Rissian conglomerates and/or remoulding of Phlegrean and Vesuvius pyroclastic materials usually overlay the hill slopes. Terraced alluvia are present on the banks of Sabato and Calore rivers; recent alluvial materials are found along their beds as well as along the marginal areas of the hill, where they overlay the Rissian conglomerates. Throughout the whole urban territory, man-made deposits, including large masonry blocks and archaeological ruins, are also present. 2.1.2 Seismic hazard

Benevento has a long history of disastrous earthquakes. According to intensity database of damaging earthquakes in the Italian area covering the time interval 1000-2006 aC (Locati et al. 2011), the city was hit by at least ten earthquakes with site intensity, IMCS, equal or larger than grade VII of the MCS scale (Figure 3). Two earthquakes have macroseismic intensity equal to IX including the 1702 Irpinia-Benevento and the 1688 Sannio earthquake. The latter event was characterized by an epicentral maximum intensity, I0, equal to XI, an estimated moment magnitude MW=6.98 and a source located around 40 km N of Benevento (Rovida et al. 2011). This earthquake (Baratta 1979) had a huge impact in the whole Centre-South of Italy, destroying entire towns such as Cerreto Sannita and most of the built environment of Benevento. Therein and in the surrounding area the death toll was limited to 1577, since the earthquake occurred in daytime, when most of the

people were working as farmers. In the last century, the earthquakes that struck the area had lower intensity, reaching grade VIII of the MCS scale for the 1930 Irpinia earthquake. Benevento is located inside the seismogenic zone Z927 Sannio - Irpinia – Basilicata (Meletti and Valensise 2004) that represents the main earthquake source for the city. Minor contribution is given by the Z924 (Molise-Gargano) and the Z925 (Ofanto) areas (Figure 4). Earthquakes in zone Z927 have a mode depth of 10 km and normal faulting mechanisms are most likely to occur. The epicenter alignment of past earthquakes approximately corresponds to that of the Apennine chain, i.e. to that of the main faults feeding the seismicity of Southern Italy. The recurrence law, obtained for the same seismogenic area using the earthquake database (Rovida et al. 2011), showed that the 1688 Sannio earthquake is characterized by a returning period of about 400 years, while the largest events for the whole seismogenic area were the 1456 Molise and 1854 Basilicata earthquakes, both having an estimated magnitude MW7.0. The recent probabilistic hazard mapping for the whole Italy (Gruppo di Lavoro 2004) for the town of Benevento show that peak ground acceleration on outcropping rock of 0.257 g and a peak spectral acceleration of 0.607 g are expected for earthquakes with a probability of exceedance of 10% over 50 years, i.e. a return period of 475 years.

2.2 The case history of San Giuliano di Puglia 2.2.1 Geological setting

On the basis of surface surveys and of the analysis of 35 borehole data (15 of which cored onpurpose after the earthquake), Melidoro (2004) and Guerricchio (2005) proposed an interpretation of the geological setting, summarized in Puglia (2008), according to which the main formations present at San Giuliano di Puglia (see Figure 5) are: the Faeto flysch, that is a sedimentary succession of mainly calcareous soils, either coarse or fine-grained; and, a deep layer of Toppo Capuana marly clays, whose maximum thickness is of the order of few hundred meters. This latter formation consists of three principal units: a “debris cover”, of less than five meters thickness, of disturbed soil and landslide debris; a layer, of two to ten meters thickness, of “tawny clays”, characterized by medium to intense fissuring, resulting from weathering and disturbance; a deep layer, called “grey clays”, less intensely fissured than the weathered tawny clays, being the discontinuities mainly induced by tectonic stresses. The marly clay formation is in lateral contact with the flysch formation, which emerges in the Southern part of the ridge, where the flysch appears to be less fractured and constitutes the foundation soil of the historical part of San Giuliano di Puglia. In the Northern part of the town, the Faeto flysch is heavily tectonized and broken up. 2.2.2 Seismic hazard

Until 2002, the town was not classified as hazardous by the Italian Seismic Code. As a matter of fact, the area of San Giuliano di Puglia seems to be characterized by low local seismicity. From 1853 to 2001, only three earthquakes with magnitudes between 3.5 and 4 occurred in a 22 km radius (Decanini et al. 2004). The above mentioned intensity database of damaging earthquakes in the Italian area (Locati et al. 2011) indicates that significant damages in the village of San Giuliano di Puglia are categorized only for earthquakes occurring in the last decades (Table 1).

However, Galli and Molin (2004) underlined that the low damage macroseismic intensity of old earthquakes was due to the lack of historical chronicles, being San Giuliano di Puglia always a poorly developed village. Assigning a local seismic intensity Is value consistent with those observed in the neighboring localities and adopting an appropriate isotropic attenuation law, they concluded that the maximum intensity historically felt can be estimated as high as Is= VIII÷IX. This value was probably reached during the December 5, 1456 Bojano earthquake (MW=7.2, epicental distance of about 45 km), one of the strongest events that interested Italy. It is worth noting that the intensity assigned for the 1456 event is the same as that of the 2002 earthquake (MW=5.7, epicental distance of about 6 km), and that this latter resulted higher than the epicentral value essentially because of the unusually high damage suffered by the modern part of the village (IS= IX÷X MCS). From the Italian seismic hazard map (Gruppo di Lavoro 2004) at the village of San Giuliano di Puglia a reference peak ground acceleration of 0.209 g and a peak spectral acceleration of 0.566 g are estimated for earthquakes with a returning period of 475 years. The reference ground motion appears therefore slightly lower than that predicted for the city of Benevento.

3. SELECTION OF THE REFERENCE INPUT MOTION Some questions might be formulated regarding the role of the reference input motion, thereafter abbreviated as RIM, in seismic zonation studies. It is well known that, for a given soil layering, the amplification function that relates the bedrock RIM to the ground motion at surface is strongly dependent on the characteristics of the seismic motion itself, due to the soil non-linear behaviour (see for instance Kramer 1996). Therefore, a key point in carrying out a seismic microzonation might be the definition of most suitable RIM. Seismic zonation approaches classified as Grade-1 or Grade-2 according to ISSMGE-TC4 (1999) guidelines are RIM-independent, being based on geological mapping and extrapolation of empirical low-energy soil amplification or related parameters. Although this constitute one of the main limitations of these methods, it yields great operative advantages, providing that some approximations are accepted. When quantitative Grade-3 zonation procedures are adopted to support actions for urban planning and land development (i.e., a predictive study), different hypotheses on the reference seismic scenario need to be accurately evaluated. Instead, the description of a specific earthquake is required in retrospective studies (i.e., to analyze the damage pattern produced by a single strong motion event). In order to reproduce a reference accelerogram at the bedrock formation, several methods may be adopted (see, for instance, Marcellini and Pagani 2004). Deterministic, stochastic and probabilistic approaches were adopted by Franceschina et al. (2006) in an attempt to simulate the 1997 UmbriaMarche earthquake at the city of Fabriano. It appears that the response spectra computed at the surface of given layering are sensitively dependent on the approach adopted to define the RIM. Ansal and Tönük (2007) stated that, when using acceleration records, the scaling rule of the selected time histories becomes an important decision point. Referring to the seismic microzonation of the area of Zeytinburnu in Istanbul, they concluded that for two sets of scaled records microzonation maps were different, even though the differences were not as significant as those from a map obtained with a synthetic RIM. Similar considerations were expressed by Santucci de Magistris et al. (2004) in a preliminary Grade-3 seismic microzonation for the town of Benevento. On the other side, synthetic RIMs are becoming much more trustworthy in the last years, since the availability of digital recordings allows to accurately validate the modeling (e.g. Graves and Pitarka, 2010; Mena et al. 2010). For instance, the broadband synthetics of the April, 2009, MW 6.3, L’Aquila earthquake has been validated through 14 digital strong motion stations within 50 km from the epicenter whereof 5 within the surface projection of the fault (Ameri et al. 2012). Another key point about the selection of the RIM for seismic zonation could be if and how the selected acceleration time histories must be related to the design specifications of relevant seismic

codes. Eurocode 8 (2004) and the New Italian Building Code for Constructions (2008) give specific indications on the evaluation of seismic actions for buildings according to the limit states design, which are therefore related to given earthquake return periods. Following the performance-based design philosophy, for different types of buildings, different life cycles, hence limit states and return periods can be considered. Then, a unique RIM selected for the seismic microzonation of an urban area, typically characterized by variable building types and re-construction purposes cannot be representative as the basis for all design applications. Hence, in the Turkish regional microzonation manual (WIDRM, 2004) it is specified that for seismic zonation it is essential to conduct a regional seismic hazard study with reference to a return period of 100 years. Such value does not correspond to the return period of the design actions for ordinary buildings (Tr=475 years) specified by the Turkish Building Code at the Life Safety limit state. Building codes specify seismic actions through conventional response spectra, which cannot be directly used as a RIM for site response analysis, hence for Grade-3 zonation methods, which require an appropriate selection of acceleration time histories on outcropping bedrock. Moreover, most dynamic analysis procedures for geotechnical design require the selection of natural accelerograms with reliable frequency content and duration. NTC (2008), for instance, states that recorded rather than synthetic accelerograms need to be used for seismic geotechnical design. Natural accelerograms might be obtained from either regional or international databases, preferably selecting records of strong motion earthquakes recorded in the same seismogenic zone and as close as possible to the studied area. Alternatively, recorded signals belonging to different areas might be employed; providing that focal mechanism, magnitude, depth and distance are comparable to the expected earthquake. In both cases, a set of compatible recorded signals need to be selected so that at least the PGA level and, possibly, the average spectrum match with the hazard studies for the given returning period. In order to search accelerograms compatible with a target spectrum, several codes were developed (e.g. REXEL, Iervolino et al. 2010; ASCONA, Corigliano et al. 2012) and implemented in some databases, such as ITACA (http://itaca.mi.ingv.it/) or the PEER Ground Motion Database (http://peer.berkeley.edu/peer_ground_motion_database/). Since no standardized procedures are currently specified nor adopted by most codes and guidelines, the gathering of worldwide experience is of primary utility to answer to this question. In the following, a brief overview on the procedures adopted for the definition of the RIM for Benevento and San Giuliano di Puglia are presented.

3.1 Reference Input Motion for Benevento In order to reproduce the seismic scenario of the historical Sannio earthquake occurred on June 5, 1688, macroseismic, geological, geomorphological, paleoseismic and seismological data were preliminarily analyzed to recognize the position and the dimension of the seismogenic fault (GNDT 2004). The radiation pattern for the 1688 Sannio earthquake was simulated using a hybrid statisticaldeterministic approach. This method is based on the simulation of a large number of possible rupture processes along the same fault, parameterized by different rupture nucleation points and final slip distributions on the fault plane. Complete wave-field seismograms in a flat layered velocity structure were computed using the discrete wavenumber method. The simulation process produced three sets of 150 synthetic seismograms, i.e. one for each component of the motion (NS, EW and UD). Table 2 summarizes the average and the standard deviation values of typical parameters adopted to represent and summarize the horizontal (EW and NS) components of the seismograms. Readers can refer to Kramer (1996) and Cosenza and Manfredi (2000) for the meaning of the listed parameters. From Table 2 it can be observed that most parameters are about coincident for the two horizontal components, with the NS component of the synthetic accelerograms showing a slightly higher mean value in terms of PGA, higher duration and spectral

velocity, and lower frequency content. The NS component of the synthetic accelerograms was, therefore, assumed as reference input motion at the bedrock formation of the Benevento subsoil. Figure 6a shows the acceleration response spectra (structural damping 5%) for the NS component of all the synthetic seismograms, together with the average and the standard deviation; the relatively large amplitudes at high periods and their smooth decay can be noted. Figures 6b and 6c respectively show the values of PGA and Arias intensity, again with their average and the standard deviation. To save computation time, a ‘clustering’ criterion (d’Onofrio et al. 2004) was established to reduce the number of site response analyses. Since representative ground motion parameters of the synthetic seismograms, such as PGA, PGV or Arias intensity, were distributed according to a Gaussian law, it was possible to extract from the whole set of signals a sub-set with a smaller dimension. A sample of 15 signals was then identified, trying to keep the mean and the standard deviation of the relevant ground motion parameters as close as possible to those pertaining to the whole data set (cf. the last two columns of Table2). Both the 2 test performed on PGA, PGV and Arias Intensity and the adaptability test on the PGA confirmed the validity in the selection of the reduced set of seismograms.

3.2 Reference input motion for San Giuliano di Puglia The seismic response analyses of the village of San Giuliano di Puglia reported in this paper were aimed at a retrospective assessment of the damage pattern of the MW 5.7 Molise earthquake in October, 31st, 2002 (Puglia 2008); therefore, the selection of the RIM was related to this specific event. The main shock activated twelve accelerometric stations (seven digital and five analog) of the Italian seismic network (RAN). Due to the low-seismicity presumed for that area, and to the relatively low density of RAN at that time, none of them was located in the near field, being the epicentral distance of the triggered instruments in the range from 26 to 190 km. The seismological studies characterized the earthquake source as a strike-slip mechanism, related to a couple of adjacent sub-vertical faults, generating the upper cited mainshock, a further strongmotion event (MW= 5.7) the day after, and a sequence of low-energy aftershocks in the following weeks. Table 3 reports the key geometrical characteristics of the faults causing the main shocks (after Basili and Vannoli 2005). In the first weeks after the mainshocks, two mobile accelerometric stations were installed on the flysch outcrop (CHI) and on the marly clay formation (SCL) at San Giuliano di Puglia (see Figure 5). Preliminary seismological studies provided synthetic accelerograms accounting for source, propagation and site effects (Gorini et al. 2004). These signals, however, were not used for the seismic microzonation of San Giuliano di Puglia, which was based on seismic response analyses directly computing the numerical convolution of a RIM defined as response spectra on rigid rock outcrop (Baranello et al. 2003). A subsequent retrospective study was developed as a teamwork research project (DPC-INGV, 2007) and is thoroughly reported by Puglia (2008) and by the papers collected in the special publication of Rivista Italiana di Geotecnica (2009). Within this project, specific regional attenuation laws for the area where calibrated (Luzi et al. 2006), estimating a reference PGA of 0.091g (on rigid bedrock) for the mainshock of October 31st, 2002. Thereafter, the Hybrid IntegralComposite (HIC) technique (Franceschina et al. 2006) was used, adopting the Basili and Vannoli (2005) source model, to derive the synthetic RIM at San Giuliano di Puglia due to the October 31st 2002 mainshock. Figure 7a reports the components along the longitudinal and transversal axis of the village (see the sketch in the figure) and the up-down motion, while the associated acceleration response spectra are plotted in Figure 7b. The frequency content of the synthetic RIM in the typical range of periods of the building stock at San Giuliano di Puglia was found similar to that of the most significant aftershocks recorded in the village with the mobile accelerometric station installed on the flysch outcrop (Puglia 2008).

4. THE CONSTRUCTION OF THE SUBSOIL MODEL For the reliability of any seismic response analysis, a key issue is the creation of an accurate subsoil model, based on the interpretation of well-calibrated experimental investigation. A strict relationship between the zonation methods and the scale, type and accuracy of geological/geophysical/geotechnical investigations needed is clearly stated in the guidelines by ISSMGE-TC4 (1999). Some specific issues will be discussed in the following, with particular reference to the casehistories of Benevento and San Giuliano di Puglia.

4.1 The subsoil model at Benevento The geotechnical characterization of the Benevento subsoil was achieved integrating data published in a previous research project (Marcellini et al. 1995a) with further information provided by local administrations (i.e. data from boreholes, Down-Holes and Standard Penetration Tests SPT) together with on-purpose site and laboratory investigations carried out by the University of Naples (Penna 2005; GNDT 2004; MIUR 2004). The characterization was aimed at performing a seismic zonation up to Grade-3, i.e. using numerical models. To this purpose, it is well known that at least the stratigraphy and mechanical properties of the most widespread soil formations in the studied area should be characterized. Mechanical properties should generally include the initial shear stiffness and damping ratio and their variation with the shear strain level. Therefore, the approach used to characterize the Benevento subsoil was the following: 1. all the relevant geotechnical data were collected; 2. a preliminary screening was done to discard the less significant or the unreliable information; 3. the remaining data were georeferenced and uploaded into a Geographic Information System (GIS). For each single vertical, the information was stored into a database connected with the GIS, containing the data at the relevant depth from the ground level; 4. the main geotechnical units were re-classified, slightly modifying the previous subsoil model for Benevento proposed by Improta (1998); 5. shear wave velocity values were attributed to the geotechnical units combining direct (Cross-Hole and Down-Hole) and indirect measurements (SPT); 6. a 3-D shear wave velocity model was created adopting an originally developed geostatistical approach (Penna 2005); 7. the VS model was validated using direct low-seismicity measurements (see for instance Improta et al. 2005; Di Giulio et al. 2008); 8. at each geotechnical unit, non-linear soil properties were assigned based on laboratory tests and literature data. All the subsoil profiles from the available boreholes were interpreted, synthesized and classified. Eventually, 263 vertical profiles were available for the analyses. To increase the number of locations where a shear wave velocity profile could be estimated, 30 Down-Hole profiles were compared to nearby results from Standard Penetration Tests, building up site-specific correlations between NSPT and VS (Penna et al. 2007). To produce a map of VS from the set of data available at scattered locations, an originally developed geostatistical approach was followed. The urban territory was discretized into squared elements of 50x50 m, and the shear wave velocity VS was extrapolated to 6.900 verticals, corresponding to about 210.000 data points. Figure 8 shows the 3-D VS model obtained by the application of the procedure above synthesized. It is possible to observe that:

1. the shear wave velocity model appears quite irregular, reflecting the complex nature of the Benevento subsoil; 2. a shallow soft layer of variable depth, characterized by values of shear wave velocity ranging between 180 and 350 m/s, covers a large part of the historical center; 3. stiff materials can be also found either at the ground level or at a limited depth; 4. the pliocenic clay (AGA), which constitutes the ‘geological bedrock’ formation, has a shear wave velocity ranging between 650 and 800 m/s; 5. the conglomeratic formation (CR), with a higher value of shear wave velocity (VS>800 m/s), sometimes overlays the AGA bedrock. The 3-D shear wave velocity model was used to locate the bedrock for the whole urban territory of Benevento. From a geological viewpoint, the pliocenic clay should be assumed as the bedrock formation; Figure 9a reports the contours of the depth of the top of this formation, which is found at more than 20 m in the Northern part of the town, while in the Southern area the depth ranges between 5 and 20 m. Due to the variable lithostatic stress state, the pliocenic clay is characterized by a different shear wave velocity, depending on the depth of the top of the formation. In detail, VS is equal to 550 m/s for depths between 5 and 15 m, while it increases up to 650 m/s where the top of the formation lies between 15 and 30 m, and reaches values as high as 800 m/s for a depth larger than 30 m. A completely different scenario is described in Figure 9b, which reports the depth of the socalled “seismic bedrock”, i.e. any formation characterized by VS larger than 800 m/s, as it is usually assumed in site response analyses. In the Southern part of the town, the top of the seismic bedrock corresponds to the stiffer layer of pliocenic clay, with a depth always larger than 30 m. Instead, the Northern part of the town, including most of the historical center, is characterized by shallow layers of soft soils lying above a seismic bedrock corresponding to the Rissian conglomerates, with shear wave velocity higher that the underlying pliocenic clay. Table 4 summarizes the physical and mechanical properties for each geotechnical unit of the Benevento subsoil. The table was created merging literature experience on comparable soils (Vucetic and Dobry 1991; EPRI 1993; Palazzo 1993; AnhDan and Koseki 2003) with specific laboratory investigation, consisting of resonant column-torsional shear tests and triaxial compression tests with local axial strain measurements, carried out on undisturbed samples at the University of Naples (GNDT 2004; MIUR 2004). Non-linear stiffness properties were summarized through the well-known Ramberg and Osgood curves (Ramberg and Osgood 1943); the increase of the damping ratio D with the strain level from its initial value, D0, was described by combining the Ramberg-Osgood model with the Masing rules (Hardin and Drnevich 1972), following the criteria described in Santucci de Magistris et al. 2004. The variability of non-linear properties with soil lithology is reflected by the wide range of values assumed by the linear and volumetric threshold strains, l and v.

4.2 The subsoil model at San Giuliano di Puglia An unusually large amount of geological, geophysical and geotechnical data was available to build up the subsoil model at San Giuliano di Puglia. Figure 10 reports the location of the 98 boreholes, which refer to three different investigation stages. The 16 locations labeled with ‘V’ refer to two surveys executed in 1992 by the San Giuliano di Puglia municipality, for urban planning purposes and in 1996 by the Molise Region; in these investigations, addressed to ordinary static design of buildings, the maximum boring depth was 39 m. Only eight out of the stratigraphic sequences were adopted for the subsoil characterization, being the remaining data unclear or not consistent with other information. A number of 15 boreholes, labeled with ‘S’, refers to an investigation jointly committed, soon after the earthquake, by the Department of the Civil Protection and the General Attorney. This investigation was planned for

both the seismic microzonation of the village and the inquiry on the collapse of the primary school. This time, the maximum depth of boreholes was as high as 70 m and the investigation included also CPT, Cross-Hole, Down-Hole and extensive laboratory tests on undisturbed samples. Finally, the locations labeled with ‘R’ refer to further 47 boreholes (with a maximum depth equal to 25 m) executed in the urban area for reconstruction purposes. The unusual density of borehole distribution allowed describing, in detail, the geometrical and mechanical properties of the Faeto flysch and especially of the Toppo Capuana marly clay (Silvestri et al. 2006; Puglia 2008; d’Onofrio et al 2009). Even though extremely dense, the boreholes did not allow to completely define the subsoil layering under the village, due to the lack of direct information about the geometry of the top of the flysch formation beneath and around the marly clay. To this purpose, deep geophysical investigations were carried out to support geologists and geotechnical engineers to complete the definition of a subsoil model. Mucciarelli et al (2009) proposed a 3-D large-scale model covering an area of 2 km on each side and extending to depths of about 1500 m from surface, as illustrated in Figure 11. Four main lithological units were considered: the Faeto Flysch(FF), the Toppo Capuana marly clays (TC), a ‘Mélange’of clayey and carbonatic materials(M) and underlying Pliocene carbonatic bedrock(P). According to this subsoil model, the interface between TC and FF (i.e. the bedrock depth) is located as deep as 300 m under the village, as foreseen by Puglia et al. (2007)back-figuring the frequency response measured during the aftershocks by the two mobile stations installed in the village (see Figure 5).

5. SEISMIC RESPONSE ANALYSIS AND MICROZONATION MAPS Although the methodological approaches for Seismic Microzonation seem well established (ISSMGE-TC4, 1999), some valuable information might be obtained by the seismic response analyses of Benevento and San Giuliano di Puglia. The use of Grade-2 zonation is very often based on the knowledge of the equivalent shear wave velocity, VS,30. This method of zonation is becoming widely adopted, being practically congruent with the subsoil classification criteria predicted by most seismic codes. However, it was already remarked in this paper that seismic zonation and building codes for seismic areas are not necessarily related. Hereafter, this point will be discussed with reference to Grade-2 and Grade-3 seismic zonation maps of the town of Benevento. Some remarks on the seismic response analysis of the village of San Giuliano di Puglia and on the interpretation of the damage induced by the simulated earthquake will contribute to understand which are the most suitable parameters to be represented in Grade-3 SM maps.

5.1 Grade-2 seismic zonation of Benevento Following the objective to link the SM procedure to the seismic codes for constructions, the most immediate approach for Grade-2 zonation consists of grouping areas having similar lithology and value of the equivalent shear wave velocity in the first 30 m,VS,30. The latter is defined as: Vs ,30 

30 hi  i 1 Vi N

(1)

where hi and vi denote respectively the thickness and the shear wave velocity of the i-th formation or layer. Following the groundwork of Borcherdt (1994), the equivalent velocity VS,30 has been widely adopted by national and international building codes, being associated to particular values of the stratigraphic amplification coefficients and to the shape of the normalized response spectra. Figure 12a shows the Grade-2 SM map of the town of Benevento, based on the evaluation of VS,30 on the whole urban territory (Santucci de Magistris et al. 2004). Homogeneous zones are identified according to the velocity ranges reported in Eurocode 8, which defines four different ground categories (A to D), each one associated to a different amplification factor and to a different normalized response spectrum. The intervals of the equivalent shear wave velocities associated to the ground categories are sketched in the same figure, where the classification is assumed independent of the bedrock depth. The majority of the investigated area belongs to class B (VS,30 = 360800 m/s), while only few sites pertain to C (VS,30 = 180360 m/s) or A (VS,30> 800 m/s) classes. It might be concluded that, apparently, site effects are not relevant in the city of Benevento, since an almost homogeneous distribution of acceleration response spectra was obtained using this approach. As already outlined by various authors (Cavallaro et al. 2006; Bouckovalas et al. 2006; Pitilakis et al. 2006), the EC8 ground types do not cover adequately and uniquely all soil conditions which are often encountered in practice, in terms of both variety of soil types and definition of the VS,30 parameter. This latter may be an overestimated indicator of soil stiffness in the case of soft but shallow soil profiles (H>30 m), or for profiles with an abrupt change of stiffness between 30 m of depth and a deeper underlying bedrock. As a matter of fact, the previous map is based on an interpretation of EC-8 site classification criteria which does not account for the depth of the bedrock formation and for the variation of soil properties with depth; on the other hand, in Eurocode 8, as well as in the Italian Technical Code, it is explicitly mentioned that a subsoil classification can be ambiguous for a shallow bedrock and it is reliable only for a subsoil showing an increase of mechanical properties with depth. In order to overtake the limitations and ambiguities in Eurocode classification, Bouckovalas et al. (2006) recommended to carefully accounting for bedrock depth in the site classification. Therefore, they suggested introducing a more generalized definition of an ‘elastic shear wave velocity’, VS,el, of a soil column down to a given depth H of a seismic bedrock with a shear wave velocity Vb, through the relationship:

VS ,el

 H ; H 30 m   30  30  H V V b   S ,30  a  V  H  ; H 30 m  S ,30  30  

(2)

In other words, VS,el is the equivalent shear wave velocity down to the bedrock if its depth is lower than 30 m, while it results from a power law extrapolation (with an exponent a) of VS,30 with depth, if the top of the bedrock lies at a depth higher than 30 m. The same authors suggested to refine the interpretation of EC-8 ground classes and to increase their number from five to seven (cf. Figure 12b), by: assigning class E to those sites with a bedrock depth between 5 to 30 m and VS