NUMERICAL MODELLING OF EARTHQUAKE-INDUCED ROCK LANDSLIDES: THE 1783 SCILLA CASE-HISTORY (SOUTHERN ITALY) Francesca Bozzano, Associate Professor, Research Centre for Geological Risks (CE.RI.) – University of Rome “Sapienza”, P.le A. Moro 5, 00185 Roma (Italy), francesca.bozzano@uniroma1.it, tel:+390649914924 Eliana Esposito, Researcher, Consiglio Nazionale delle Ricerche – Istituto per l’Ambiente Marino e Costiero (CNR-IAMC), Calata Porta di Massa - 80133 Napoli (Italy), eliana.esposito@iamc.cnr.it Luca Lenti, Researcher, Paris East University, Laboratoire Central des Ponte set Chaussées (LCPC), 58 Boulevard Lefebvre 75732 Cedex 15 Paris (France), luca.lenti@lcpc.fr Salvatore Martino, Researcher, Research Centre for Geological Risks (CE.RI.) – University of Rome “Sapienza”, P.le A. Moro 5, 00185 Roma (Italy), salvatore.martino@uniroma1.it, tel:+390649914923 Alfredo Montagna, PhD student, Department of Earth Sciences – University of Rome “Sapienza”, P.le A. Moro 5, 00185 Roma (Italy), alfredo.montagna@uniroma1.it Antonella Paciello, Researcher, Ente per le Nuove Tecnologie l’Energia e l’Ambiente (ENEA) – Via Anguillarese 301, 00060 S.Maria di Galeria (Italy), antonella.paciello@casaccia.enea.it Sabina Porfido, Researcher, Consiglio Nazionale delle Ricerche – Istituto per l’Ambiente Marino e Costiero (CNR-IAMC), Calata Porta di Massa 80133 Napoli (Italy), sabina.porfido@iamc.cnr.it ABSTRACT The coastal M. Pacì rock-avalanche occurred on 6th February 1783, near the village of Scilla (southern Calabria) and involved a subaerial volume of about 5·106 m3. This landslide produced a tsunami wave responsible for more than 1500 human life losses near Marina Grande beach. A geomechanical characterization of both the intact rock and the rock mass outcropping in the landslide slope was performed in order to obtain an engineering-geology model of the landslide according to an equivalent-continuum approach. Dynamic numerical modelling by FDM code FLAC 6.0 was performed to back-analyse the landslide occurrence during the 1783 seismic sequence. At this aim reference synthetic accelerometric ground motions were derived from strong motion records, taking into account both source and energy features of the 5th an 6th February mainshocks and local expected response spectra. In order to force the numerical model, levelled-energy multifrequencial equivalent signals were obtained from these reference records by experiencing the new LEMA_DES approach. The results of modelling show a post-seismic trigger of the rock-avalanche, related to the second mainshock of the seismic sequence, and are in good agreement with the present-day field evidences of the landslide scar area. In addition, no landslide results in the present M. Pacì slope by applying an equivalent input derived for the 1908 Reggio and Messina earthquake. The performed numerical analyses demonstrate the reliability of: 1. an equivalent-continuum approach to derive geomechanical properties of intensely jointed rock masses for numerical modelling of rock landslides; 2. levelled-energy multifrequencial equivalent signals derived from reference recordings; 3. the experienced approach for previsional analyses of landsliding, involving rock masses in highly seismic areas. Paper No. 4.25b 1. INTRODUCTION Ground effects due to earthquake-induced landslides affecting both rock and soil slopes (Hutchinson, 1987; Sassa, 1996; Rodriguez et al. 1999; Esposito et al., 2000; Porfido et al., 2002; Sassa et al., 2005; Towhata et al., 2008) can be regarded as responsible for the greatest damages and losses due to earthquakes (Bird and Bommer 2004). Seismicallyinduced landslides are documented for some historical earthquakes, such as the 1783 earthquake in Calabria, Italy (Sarconi, 1784; De Dolomieeu, 1785; Vivenzio 1788;Cotecchia et al.,1986 ) and the 1786 earthquake in Kanding-Luding, China (Dai et al., 2005), while a large database of earthquake-induced landslides for recent earthquakes is available in the literature (Rodriguez et al., 1999). It is only in the past few years that some research has focused on reconstructing the mechanisms of seismically-induced landslides and on deterministic prediction of earthquakeinduced ground failure scenarios in specific case studies (Wasowsky and Del Gaudio, 2000; Havenith et al., 2002; Bozzano et al., 2008a; Martino and Scarascia Mugnozza, 2005; Gerolymos and Gazetas, 2007). In particular, the possible interactions between seismic input, slope and preexisting landslide mass were recently analysed for seismically-induced landslides (Bozzano et al. 2008c; Del Gaudio and Wasowsky, 2007; Esposito et al., 2000; Martino and Scarascia Mugnozza, 2005: Martino et al., 2007) in order to point out the possible role of the seismic input properties (i.e. frequency content, directivity, peak of ground acceleration) to characterise both landslide mechanism and trigger. At this regard, since the above mentioned effects have been commonly taken into account for explaining different seismic response effects in the frame of seismic microzonation studies, the here discussed case history of the 1783 Scilla rock avalanche proves that they can be responsible also for seismically induced landslides, occurring along fault zones and involving hudge volumes of intensely jointed rock masses. 1. THE SCILLA ROCK AVALANCHE The Scilla rock-avalanche is one of the main ground effects induced by the above 1783 “Terremoto delle Calabrie” seismic sequence and represents one of the most damaging landslides historically reported in Italy. The landslide produced a tsunami that killed more than1500 people in the neighbour Marina Grande beach (Gerardi et al, 2008; Graziani et al, 2006). The landslide occurred on February 6 1783 at 1:45 a.m., in the M. Pacì coastal slope very close to the village of Scilla, some 30’ after the mainshock (Hamilton, 1783;Torcia, 1783; Gallo 1784; Sarconi, 1784; Minasi, 1970, 1971, 1785) with estimated magnitude of about 5.9-6.1 Ms. The landslide is witnessed, among other historical records, by three engravings of A. Minasi, a XVIII century local painter, illustrating the M.Pacì slope before and after the landslide (Fig.1). Fig. 1. Engravings showing the M. Pacì slope before and after the 6 February landslide (from Minasi A., 1783). The Scilla landslide can be considered as a complex event (Cruden and Varnes 1996) characterized by an initial sliding mechanism, along structurally controlled joints, evolving into a proper rock avalanche (Fig.2) (Bozzano et al., 2008b). The landslide-induced tsunami wave occurred with a delay of about one minute to the landslide occurrence and it was about 19 m heigh at the Marina Grande beach. 2. ENGINEERING-GEOLOGY MODEL The landslide area is mainly characterized by gneiss rock masses, which has been related to the Scilla Metamorphic Unit (Fig.3). Paper No. 4.25b Fig. 2. View of the Scilla landslide scar area from the Tyrrhenian sea. In the upper part of the scar area white breccias of gneiss cemented by a calcitic matrix widely outcrop; these breccias are associated to a normal fault, circa parallel to the coast line, which main plain correspond to the scarp of the Mt. Pacì landslide. Similar calcitic breccias also outcrop downslope in a wide zone circa parallel to the coastline. The Holocene activity of these faults is recently quoted by Ferranti at al. (2007). Other evidences of the same breccias have been collected all over the slope and prove the complex structural setting of the area (Ferranti et al, 2007). Fig. 3. Geological map of the Scilla landslide area. Paper No. 4.25b Fig. 4. Geological section along Mt. Pacì slope and related engineering geology model (see Fig. 3 for the legend). The pre-landslide shape is also reported (white zones). The left flank of the landslide is characterised by intensely jointed to cataclastic gneiss; moreover, intensely foliated gneiss outcrop, representing the melanocratic facies of the Scilla Metamorphic Unit. The same flank corresponds to a main fault, bounding westward the Mt. Pacì-Mt. Bova horststructure (Ghisetti, 1984), which dip direction is nearly N50°E. In order to characterize the geomechanical properties of the material involved in the Scilla landslide, 16 geomechanical scanlines were performed by measuring orientation, spacing, aperture, type of filling on the main joint sets as well as JRC and Schmidt hummer values. In 29 additional geomechanical stations measures of Jv and Ib indexes were also obtained (ISRM, 1978). Three boreholes were also drilled: the first one within the landslide deposit and the second and third ones 1.5 km far from the Mt. Pacì slope. Point load and ultrasonic labtests were performed on the sampled cores of gneiss in order to evaluate both strength and deformational parameters of the intact rock. Starting from the in site geomechanical characterization a mapping of the rock mass geomechanical properties, was performed in the 3 following steps. 1 - A classification of the rock mass was obtained experiencing the equivalent continuum approach by Sridevi and Sitharam, 2000: at this regard, the two independent rock mass jointing parameters Jf (joint factor by Ramamurthy, 1994) and Jv (number of discontinuities for unit volume according to ISRM, 1978) were considered in agreement with each continuum equivalent approach, for the analysed geomechanical stations. The computed values were plotted in order to obtain some clusters to derive three rock mass classes. 2 - A class distribution for the rock mass in the landslide area was obtained by interpolating and contouring the Jf and the Jv values obtained for each geomechanical station. 3 - An overlapping of the contour lines for the two indexes was obtained by applying an intersection criterium to map the location of each class. The same equivalent continuum approach was also applied in order to obtain the distribution of the jointed rock mass stiffness, as well as to take into account the in site depthdependent stresses, starting from the Young’s modulus of the intact rock measured by ultrasonic lab-tests (Sridevi and Sitharam, 2000; Esposito et al, 2007). On the basis of the Young’s modulus values estimated by using the Schimdt hammer in the outcropping rock-mass and the stiffness parameter values obtained by ultrasonic lab-tests, an intact rock mass modulus was estimated in the range 30007210 MPa by applying the empirical equation of Sridevi and Sitharam, 2000 for different values of confining pressure (Tab.1). The highly jointed rock masses (class 3) strictly correspond to both the coastward dipping fault zones and to the eastward dipping fault zone which outcrops along the left flank; a moderate jointed rock mass (classes 1 and 2) widely outcrops along the right flank of the landslide. Paper No. 4.25b Table 1. Geomechanical parameters referred to the engineering-geology model of Fig.4 The obtained 3D engineering-geology model of the landslide proves the significant structural constrain on its kinematism, also justifying the main geomorphological evidences in the subaerial scar area (Fig. 4), such as the planar shape of the left flank and the subvertical cliff of the crown area. The evidence of a northeastward movement of the landslide in the detachment area related to the biplanar wedge-like shape of the detached volume, is consistent with the increasing of the rock mass quality, in terms of rock mass jointing, from the left to the right flank of the landslide. 3. SEISMICITY OF THE SCILLA AREA The seismicity of the Scilla area is strictly connected with the Siculo-Calabrian rift zone, one of the most seismically active areas of the Italian peninsula, characterised by several seismogenic sources capable of producing earthquakes with M=7 (CPTI04), intensity values I=X both on the MCS (CPTI04) scale and on the new ESI 2007 scale (Michetti et al., 2007, Porfido et al., 2008). Historical seismicity, as regards large earthquakes, occurred in this area is fairly well documented describing a very high-recurrence of events with at least 34 earthquakes with 8