Você está na página 1de 8

Soil Dynamics and Earthquake Engineering 20 (2000) 8592

www.elsevier.com/locate/soildyn

Seismic hazard maps of Italy including site effects


R. Romeo a,*, A. Paciello b, D. Rinaldis b
a

Servizio Sismico Nazionale, Via Curtatone 3, Rome 00185, Italy b ENEA, C.R. Casaccia, Rome 00060, Italy

Abstract The seismic hazard of Italy was mapped adopting multiple models of seismic sources, recurrence rates and attenuation relationships, and investigating the inuence of site effects on assessing the earthquake hazard. This result was achieved by placing all the alternatives in a logic-tree diagram, and performing an uncertainty analysis of the main epistemic variables involved in probabilistic seismic hazard computations. Several ground motion parameters such as peak ground acceleration (PGA), velocity (PGV) and displacement (PGD), spectral accelerations at 0.2 and 1.0 s periods, and site intensities were computed. Maps of PGA and PGV values with a 10% chance of being exceeded in 50 years are shown as an illustration of the seismic hazard of Italy. q 2000 Elsevier Science Ltd. All rights reserved.
Keywords: Probabilistic seismic hazard analysis; Logic tree formalism; Uncertainty analysis; Site effects; Seismic hazard maps

1. Introduction The seismic hazard of Italy has already been mapped by Romeo and Pugliese [1] to formulate the proposal of a new seismic classication of the country matching the requirements of Eurocode 8 [2]. In that study the authors pointed out the need to account properly for uncertainties in seismicity models, activity rates and ground motion models, as well as for the inuence of the local site conditions on seismic hazard. In fact, neglecting uncertainty was demonstrated to misrepresent the contribution to hazard arising from different critical choices [3,4]. Moreover, site effects are recognized to strongly affect the response of structural systems to ground shaking, thus inuencing the results of a risk analysis. Keeping this in mind, we produced new seismic hazard maps of Italy investigating the inuence exerted by the main epistemic variables involved in probabilistic seismic hazard assessments (multiple models of seismic sources, recurrence rates and attenuation relationships), as well as incorporating site effects. This study was specically carried out for a project promoted by the Italian National Agency for New Technologies, Energy and Environment (ENEA), and the Italian Ministry of University and Research concerned with the protection of historical monuments against seismic risk.
* Corresponding author. Tel.: 139-06-4444-2276; fax: 139-06-4466579. E-mail address: fromeo@tiscalinet.it (R. Romeo).

2. Uncertainty analysis Based on the current state-of-knowledge on Italian seismicity, we investigated in detail three basic sources of uncertainty concerning four seismic source models, three methodological approaches to activity rates determination, and three different attenuation relationships. All the alternatives were treated in a logic-tree formalism, schematically shown in Fig. 1, avoiding any weighting operation so that a user can choose the most appropriate route for his target. Models of seismic sources included: Seismotectonic model (SM in Fig. 1, by Scandone et al. [5]) based on the structural and kinematic setting of the Italian peninsula. It consists of 81 zones with a homogeneous distribution of seismicity (neutral polygons in Fig. 2), which address in essence the uncertainty in locating future earthquakes, thus allowing for a spatial nonstationarity of seismicity. Spatially clustered seismicity model (SCS in Fig. 1) consisting of zones of concentrated seismicity drawn through the geographic clustering, at different hierarchical levels, of the observed seismicity. This model is based on the assumption that large earthquakes occurring in very small areas are likely to belong to the same seismogenic fault or to a homogeneous system of faults. This implies that seismicity is concentrated in small areas reaching high values of ground motion, thus reecting the typical behavior of most seismicity that occurs along the Apennine chain, as the recent Umbria-Marche seismic sequence [6] displayed. A threshold magnitude 5.8 was chosen, above

0267-7261/00/$ - see front matter q 2000 Elsevier Science Ltd. All rights reserved. PII: S 0267-726 1(00)00040-3

86

R. Romeo et al. / Soil Dynamics and Earthquake Engineering 20 (2000) 8592

Fig. 1. Logic tree formalism adopted in seismic hazard analyses of Italy.

Fig. 2. Models of seismic sources adopted in the seismic hazard of Italy. Small-size (hundreds to a few thousand of square kilometers) black and gray polygons: rst- and second-level, respectively, zones of concentrated seismicity (SCS model). Medium-size (thousands of square kilometers) neutral polygons: zones of homogeneous seismicity (SM model). Nine large-size (tens of thousands of square kilometers) zones of diffuse seismicity (BG model), encompassing the entire country, are shown with dashed lines.

R. Romeo et al. / Soil Dynamics and Earthquake Engineering 20 (2000) 8592

87

Fig. 3. Left: relation between surface rupture length and magnitude adopted in the present work compared with those proposed by Wells and Coppersmith [7] for various seismotectonic environments. Right: scheme for drawing zones of concentrated seismicity; rst-level zones (R1) assume a radius equal to half rupture length, whereas second-level zones (R2) assume a radius equal to the rupture length.

Fig. 4. NT4 seismic catalog: events above magnitude 4.7 since 1000 AD.

88

R. Romeo et al. / Soil Dynamics and Earthquake Engineering 20 (2000) 8592

Table 1 Variability in the beginning of completeness times for magnitude ranges Magnitude range Years (AD) .6.3 13001620 5.86.3 16001690 5.25.7 16901760 4.65.1 17801830 4.04.5 1870

which earthquakes occurring within a radius proportional to the rupture length are identied as belonging to zones of concentrated seismicity. This threshold, approximately, corresponds to an epicentral intensity of 89 in the MCS scale, above which old masonry buildings collapse. The relation between earthquake magnitude and rupture length adopted in the cluster analysis is shown in Fig. 3. It is derived from that proposed by Wells and Coppersmith [7] for all faults, truncated at magnitude 7.2 above which a rupture length not greater than 55 km is allowed. Truncation of the rupture length is suggested by paleoseismological evidences [8], according to which no surface-faulting event known to date exceeded this value in Italy. The radius for the earthquake clustering was rst xed to half rupture length (thus simulating a bilateral rupture), which allowed 13 rst-level zones of concentrated seismicity (black polygons in Fig. 2) to be identied. In a second iteration, secondlevel zones of concentrated seismicity were identied assuming a radius equal to the rupture length, thus simulating a unilateral rupture (15 zones marked by gray polygons in Fig. 2). Seismicity below the threshold magnitude was attributed to these zones according to the same criteria adopted to draw them, assuming an equally weighted partition of those earthquakes shared by two or more intersecting zones. This method is very useful when multiple seismic source models need to be analyzed simultaneously, allowing a user to attribute any arbitrary weight to the events that belong to overlapping zones. Zones of concentrated seismicity are a tentative to locate possible active faults, which still have not been identied clearly by paleoseismologists, with very few exceptions. Point sources model (HPS in Fig. 1) formed by the epicenters of historical earthquakes contained in the NT4 seismic catalog (Camassi and Stucchi [9]), which was specically assembled for hazard studies. In fact, the NT4 catalog lists only mainshocks, aftershock and foreshock sequences having been removed. This model of seismic sources assumes a full spatial stationarity of seismicity. Background model (BG in Fig. 1) based on nine large zones of diffuse seismicity was set to face any non-stationarity in the spatial distribution of earthquakes. The seismic catalog NT4, which integrates historical (since 1000 AD) and instrumental seismicity, is shown in Fig. 4. An alternative catalog, at least for large earthquakes (M $ 5.9), is now available in Italy issued by the National Institute of Geophysics (Boschi et al. [10]). The inuence on seismic hazard of changing seismicity parameters is currently under study. We performed a formal completeness analysis of the NT4

catalog according to a procedure described in Romeo and Pugliese [1], and based on a modied version of the methodology proposed by Tinti and Mulargia [11]. Table 1 shows the results of the completeness analyses expressed in terms of variability in the beginning of completeness times for magnitude ranges. Three models of recurrence rates were adopted: 1. The methodology proposed by Albarello et al. [12], in which the observed historic rates are taken for each zone without any assumption about their recurrence model. 2. An ordinary frequencymagnitude relation (Gutenberg Richter form), with the b-value specically computed for each zone (local b-value); the GR-relations were truncated at the maximum historical magnitudes. 3. A frequencymagnitude relation with the b-value computed for three large seismotectonic domains (Alps, Apennines, Sicily and Calabrian Arc: regional b-value), and the activity rates above the magnitude of engineering interest (M $ 4.0) determined specically for each seismic source zone. The computed b-values turned out to be very close to 1.0. Hence a b 1:0 was set for all source zones. A truncated double-exponential frequencymagnitude relation was adopted with the following functional form:

nm $ M 10a2bM nm $ M 10a2bM
{M0 # M # Mmax

{Mmin # M # M0 1 2 102bMmax 2M 1 2 102bMmax 2M0

where Mmax is the maximum magnitude and M0 the threshold magnitude above which the relation becomes truncated. To account for the uncertainty in magnitude determination, the method proposed by Veneziano and Van Dyke (in EPRI [13]) was implemented, calculating an equivalent magnitude M p:  M p M 1 1:2bsm 3 where b is the GR b-value and sm is the statistical variance of magnitude (set to 0.16 in the present study, according to the average standard deviation for magnitudes reported in the NT4 catalog). Empirical ground motion models included: 1. Italian-specic attenuation relationships [14];

R. Romeo et al. / Soil Dynamics and Earthquake Engineering 20 (2000) 8592

89

Fig. 5. Comparison of peak ground accelerations (in g) predicted by the three attenuation models used in this work. Model 1 is the Italian-specic attenuation law [14]; model 2 is the European attenuation law [15]; model 3 is the joint ItalianGreek attenuation law [16]. For model 3 an extensional seismotectonic environment was selected to draw attenuation curves.

2. European attenuation relationships [15]; and 3. joint ItalianGreek attenuation relationships [16], which enhanced the former two accounting for the proper seismotectonic environment. In Fig. 5 peak ground accelerations predicted by the three attenuation models are compared. Attenuation curves of model 3 were drawn for an extensional seismotectonic environment. Finally, geologic site conditions of 8100 Italian municipalities were determined by means of a GIS-ready geologic map of the whole country. The digitized geologic map of Italy refers to about 130 lithotechnical geologic units, which we reduced to three reference site conditions according to the available site amplication coefcients in the Italianspecic attenuation relationships [14]. The three reference site conditions are the following: Rock or stiff soil conditions (soil category A), with shear wave velocity greater than 800 m/s; Loose deposits (soil category B1), with an average shear wave velocity in the shallower 20 m below 400 m/s; and Deep medium-dense deposits (soil category B2), with an average shear wave velocity in the rst 2030 m between 400 and 800 m/s. According to soil category denition, Holocene deposits were classied as B1. Plio-Pleistocene normally consolidated or slightly overconsolidated sediments and poorly cemented soils were classied as B2. Plio-Quaternary strongly overconsolidated sediments, well-cemented soils, rocks and pre-Pliocene formations were classied as A. Then we computed, for each municipality, the percentage

of the territorial extent of the downtown area belonging to each one of the three soil categories. The proper site amplication coefcient in the attenuation relationship was attributed according to the prevailing soil category in the downtown area. This approach was preferred to other conventional seismic hazard analyses incorporating site effects [17], as it allows the direct evaluation of the impact of geologic conditions on historical heritage buildings and monuments, which all exist in an urban environment. 3. Seismic hazard maps The seismic hazard of Italy was mapped in terms of peak ground acceleration, velocity and displacement, site intensities, and spectral ordinates at selected frequencies (1 and 5 Hz), for 40, 10, 5 and 2% exceedance probability within a reference period of 50 years, for all the alternatives shown in the logic-tree diagram (Fig. 1). Ground motion values are shown for site conditions referred to as rock or stiff soils (category A). The effects of site conditions have been analyzed separately and are discussed at the end of this chapter. For the sake of simplicity only seismic hazard maps of peak ground acceleration and velocity with a 10% chance of being exceeded in 50 years are shown hereafter. They are compared for some of the most relevant alternatives shown in Fig. 1, stressing whenever possible the range of variability in the predicted ground motion values. For an easier comparison of the maps, peak ground acceleration (PGA) values have been grouped into four levels. Neglecting PGA values smaller than 0.05 g, the four levels

90

R. Romeo et al. / Soil Dynamics and Earthquake Engineering 20 (2000) 8592

Fig. 6. Peak ground acceleration values (PGA, in g) with a 10% chance of being exceeded in 50 years. Panels AD compare seismic hazard maps of Italy drawn according to some of the alternatives shown in the logic-tree diagram of Fig. 1.

are as follows: A low-hazard level, with PGA ranging between 0.05 and 0.15 g; An intermediate-hazard level, with PGA ranging between 0.15 and 0.25 g; A high-hazard level, with PGA ranging between 0.25 and 0.35 g; and A very high-hazard level, with PGA greater than 0.35 g.

They correspond, approximately, to the three plus one levels of seismicity recommended by Eurocode 8 [18] for the zoning of national countries. Fig. 6 is divided into four panels: the rst three panels (A, B and C) compare the effect on hazard exerted by the adoption of different recurrence rates and attenuation relationships, and using the same sources model (SM model); whereas the fourth panel (D) shows an alternative source model and the

R. Romeo et al. / Soil Dynamics and Earthquake Engineering 20 (2000) 8592

91

Fig. 7. Peak ground velocity values (PGV, in cm/s) with a 10% chance of being exceeded in 50 years, predicted without (A) and including (B) site effects.

subsequent adoption of consistent recurrence rates and attenuation relationships. The comparison of panels A, B and C shows that the major inuence on seismic hazard is exerted by the appropriate choice of the recurrence rates model. Particularly, the historic rates model provides the highest ground motion values, due to the very conservative way in which rates are computed, especially at high magnitudes. European attenuation relationships compute smaller ground motion values than Italian attenuation relationships, due to a stronger attenuation at intermediate and high magnitudes (see Fig. 5). Panel D shows PGA values predicted by the SCS model coupled with the BG model; they conjugate the location of possible active faults (zones of concentrated seismicity) with very large zones where the precise location of future earthquakes is quite impossible. Hence the latter are treated as zones of diffuse and homogeneous seismicity. Consistently, recurrence rates were computed using local b-values (model 2) for zones of concentrated seismicity and regional b-values (model 3) for zones of diffuse seismicity. Model 3 of attenuation relationships (ItalianGreek attenuation laws) differentiated by seismotectonic environment (Alps compressive, Apennines, Calabria and Sicily extensional), was implemented. The comparison of panels A and D shows that the SCS model predicts larger ground motion values than the SM model in more restricted areas, which appear to follow the trend of some active faults hypothesized by paleoseismological studies [8]. Peak ground acceleration, relating to unusual, quasi-

random peaks (spikes) at very high frequency, is a bad estimator of structural damage. On the contrary, since destructiveness is directly related to the energy transmitted by ground motions to man-made structures, a much better correlation can be expected with peak ground velocity (PGV). Thus, PGV allows one to account properly for the inuence of geologic site conditions in determining the increase of ground shaking on buildings. Moreover, PGV is insensitive to the depth of the alluvia, being related only to the stiffness contrast between rock and alluvium. This allowed the joining of deep and shallow alluvia, which are difcult to recognize from a large-scale geologic map, such as those used in this study. Fig. 7 shows the inuence of site conditions on seismic hazard. Site effects bring an increase of the predicted PGV-values of about 30%, related proportionally to an increase in the energy content of the expected ground shaking. A map including site effects resembles a patchwork, reecting the geologic pattern of alluvial deposits lling valleys and basins. The increase in PGV-values highlights new areas of high hazard such as intra-mountain basins, Tyrrhenian grabens and Apulian foredeep. The implementation of site effects modies ground motion predictions substantially in several municipalities, leading to an increase in the prediction of expected damages and human life losses. In fact, when seismic risk analyses are carried out incorporating site effects [19], an average increase of about 29% of the expected damages to dwellings, and about 43% of the population dead or injured, is obtained on a national scale.

92

R. Romeo et al. / Soil Dynamics and Earthquake Engineering 20 (2000) 8592 [6] Selvaggi F, Basili G, Boschi A, Couroulex E, Deschamps F, Gaffet S, Bittarelli G, Chiaraluce L, Piccinini D, Ripepe M. The 1997 UmbriaMarche, Italy, earthquake sequence: a rst look at the main shocks and aftershocks. Geophys Res Lett 1998;25:28614. [7] Wells DL, Coppersmith KJ. New empirical relationships among magnitude, rupture length, rupture width, rupture area and surface displacement. Bull Seismol Soc Am 1994;84(4):9741002. [8] Valensise, G., Pantosti, D., D'Addezio, G., Cinti, F.R., Cucci, L. L'identicazione e la caratterizzazione di faglie sismogenetiche nell'Appennino centro-meridionale e nell'arco calabro: nuovi risultati e ipotesi interpretative. Atti XII Convegno CNR-GNGTS, 1993;1:33142. [9] Camassi, R., Stucchi, M. NT4.1 un catalogo parametrico di terremoti di area italiana al di sopra della soglia del danno. CNR-GNDT, Milano, 1996. p. 186. http://emidius.itim.mi.cnr.it/NT.html. [10] Boschi E, Guidoboni E, Ferrari G, Valensise G, Gasperini P. Catalogo dei forti terremoti in Italia dal 461 a.C. al 1990. Rome: Istituto Nazionale di Geosica, 1997 (644pp.). [11] Tinti S, Mulargia F. Completeness analysis of a seismic catalog. Ann Geophys 1985;3:40714. [12] Albarello, D., Rebez, A., Slejko, D. Caratterizzazione della sismicita nel calcolo della pericolosita sismica. Atti XIV Convegno CNRGNGTS. 1995;2:5514. [13] EPRI, Electric Power Research Institute. Seismic hazard methodology for the Central and Eastern United States, vol. I, Methodology. EPRI Report NP-4726, 1986; section 4. [14] Sabetta F, Pugliese A. Estimation of response spectra and simulation of nonstationary earthquake ground motion. Bull Seismol Soc Am 1996;86(2):33752. [15] Ambraseys NN, Simpson KA, Bommer JJ. Prediction of horizontal response spectra in Europe. Earthq Engng Struct Dynam 1996;25:371400. [16] Rinaldis, D., Berardi, R., Theodulidis, N., Margaris, B. Empirical predictive models based on a joint Italian and Greek strong-motion database: I, peak ground acceleration and velocity. Proceedings XI European Conference on Earthquake Engineering, 1998; CNIT, Paris la Defense, France. [17] Petersen MD, Bryant WA, Cramer CH, Reichle MS, Real CR. Seismic ground-motion hazard mapping incorporating site effects for Los Angeles, Orange and Ventura Counties, California: a geographical information system application. Bull Seismol Soc Am 1997;87(1):24955. [18] ENV 1998-1-1. EUROCODE 8, Design Provisions for Earthquake Resistance of Structures. Seismic actions and general requirements of structures. CEN/TC 250, May 1994. [19] Di Pasquale, G., Orsini, G., Romeo, R. Sensitivity analysis in seismic risk assessment. Proceedings 6th International Conference on Seismic Zonation, 2000, Palm Spring, CA, November 1215; submitted for publication.

4. Conclusions The seismic hazard of Italy has been mapped implementing multiple models of seismic sources, recurrence rates and attenuation relationships, and investigating the inuence of site effects on assessing the earthquake hazard in the view of risk-oriented analyses. Results of these studies go beyond the aim of the specic project in which they have been realized, and allow separate choices of the input parameters; any user can select the most appropriate parameters according to his target. This result has been achieved putting all the alternatives into a logic-tree diagram, allowing, moreover, an uncertainty analysis of the main epistemic variables involved in probabilistic seismic hazard computations. In this way, simple logical or mathematical operations (such as maximum or average expected values) can be easily performed and the resulting seismic hazard maps produced. Acknowledgements The authors express their heartfelt thanks to an anonymous reviewer whose comments improved this paper greatly. References
[1] Romeo R, Pugliese A. Seismicity, seismotectonics and seismic hazard of Italy. Int J Engng Geol 2000;55(4):24166. [2] Pugliese, A., Romeo, R., Sano, T. Seismic zonation of Italy according to Eurocode 8. Proceedings 11th European Conference on Earthquake Engineering, September 611, 1998, CNIT, Paris la Defense, France. [3] SSHAC, Senior Seismic Hazard Analysis Committee. Recommendations for probabilistic seismic hazard analysis: guidance on uncertainty and use of experts. Lawrence Livermore National Laboratory, prepared for US Nuclear Regulatory Commission, US Department of Energy, E.P.R.I., NUREG/CR-6372, UCRL-ID-122160, vol. 1 1997. 256pp. [4] National Research Council. Probabilistic Seismic Hazard Analysis. Washington DC: National Academy Press, 1988. [5] Scandone P, Patacca E, Meletti C. Main recent deformation and seismotectonics in the Central Mediterranean region. International School Earth and Planetary, Workshop on Geodynamics of the Mediterranean area and implications on volcanic and seismic hazards, Siena (Italy), 1996.

Você também pode gostar