You are on page 1of 14

Soil Dynamics and Earthquake Engineering 24 (2004) 915928 www.elsevier.

com/locate/soildyn

Time probabilistic evaluation of seismically induced landslide hazard in Irpinia (Southern Italy)
V. Del Gaudioa,b,*, J. Wasowskic
` degli Studi di Bari, Campus, via E.Orabona 4, 70125 Bari, Italy Dipartimento di Geologia e Geosica, Universita b ` degli Studi di Bari, Bari, Italy Osservatorio Sismologico, Universita c Dipartimento di Ingegneria Civile e Ambientale, Consiglio Nazionale delle Ricerche-Istituto di Ricerca per la Protezione Idrogeologica, sezione di Bari, Politecnico di Bari-Campus, via E. Orabona, 4, 70125, Bari, Italy Accepted 12 June 2004
a

Abstract A recently proposed method, which incorporates the Newmark model to evaluate the earthquake-induced landslide hazard at regional scale, was applied to Irpinia, one of the most seismically active regions of Italy. The method adopts a probabilistic approach to calculate values of critical acceleration ac representing the minimum strength required for a slope not to fail at a xed probability level in a given time interval. Regional probabilistic hazard maps were generated for the two failure types most common in Irpinia (slumpearthows and rock falls). The results suggest that quite moderate critical acceleration (0.050.08 g) could sufce to keep the slope failure probability low. However, the available data indicate that potential slide surfaces with ac below these values could be common in Irpinia, where, perhaps in relation to particular geo-environmental conditions, a relative large number of marginally stable slopes might survive other destabilising actions and fail even on occasion of not particularly strong earthquake shaking. q 2004 Elsevier Ltd. All rights reserved.
Keywords: Arias intensity; Critical acceleration; Earthquake-induced landslides; Hazard; Newmark displacement; Southern Italy

1. Introduction It is widely recognised that earthquakes represent one of the most costly natural disasters in terms of human life and economic losses. Very frequently their damage effects are considerably enhanced by the concomitant occurrence of ground failures induced by seismic shaking within a certain distance from the seismogenic source. This is related to the active tectonics and geomorphic evolution characterising regions affected by earthquakes, which determine the spatial distribution of slopes prone to instability. Several methods have been developed for the evaluation of the combined type of hazard represented by seismically induced landslides. Some of these methods aim at a preliminary recognition of the areas exposed to major
* Corresponding author. Address: Dipartimento di Geologia e Geosica, ` degli Studi di Bari, Campus, via E.Orabona 4, 70125 Bari, Italy. Universita Tel.: C39-80-544-2279; fax: C39-80-544-2625. E-mail address: delga@geo.uniba.it (V. Del Gaudio). 0267-7261/$ - see front matter q 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.soildyn.2004.06.019

hazard and focus on a regional scale evaluation, rather than at predicting the behaviour of a specic slope [13]. In a recent paper [4], we presented a new timeprobabilistic approach to regional scale evaluation of landslide hazard triggered by earthquakes. This approach consisted in the production of a regional map representing the resistance to seismic shaking, required for slopes to keep failure probability below a xed value. It makes use of three basic parameters, i.e. Arias intensity (Ia) [5], expressed in m/s, critical acceleration (ac), expressed in g unit, and Newmark displacement (Dn) [6], expressed in cm, to represent, respectively, the level of seismic shaking, the slope resistance to failure and the conditions for seismic triggering of landslides. With regard to the last aspect, Wilson and Keefer [7] proposed to characterise landsliding conditions in terms of exceedance of Dn thresholds differentiated for different mass movement types, e.g. 10 cm for coherent slides (e.g. slumps, block slides, slow earth ows) and 2 cm for disrupted ones (e.g. rock falls, rock slides, rock avalanche).

916

V. Del Gaudio, J. Wasowski / Soil Dynamics and Earthquake Engineering 24 (2004) 915928

Taking into consideration the three mentioned parameters, the procedure proposed in Ref. [4] calculates, for a region of interest, a grid of values (Ac)x dening the minimum critical accelerations locally required for a slope to keep within a predened level (e.g. 10% in 50 yr) the probability that Newmark displacement exceeds a threshold x. For this purpose the number of events expected in, say, 50 yr for any possible value of Arias intensity Ia is estimated (e.g. following the Cornell approach [8] implemented through a code like SEISRISK III [9]). The exceedance probability of Dn for different critical accelerations ac is then derived by assuming a normal distribution for log Dn around the value provided by the following empirical equation proposed by Jibson et al. [1] log Dn Z a log Ia C b log ac C g (1)

where a, b and g are coefcients derived from a proper calibration. For Italy, Romeo [10] obtained, from a set of Italian accelerometric recordings, the values a Z1.891; bZK 2.621; gZ K 2.089 with a standard deviation slog DnZ0.562. The calculation of the (Ac)x value associated to a xed probability that Dn exceeds x, represents another way to measure seismic shaking expected with a certain exceedance probability, by expressing the shaking level in terms of the minimum critical acceleration required for a slope to resist this shaking. The scopes of the proposed technique, which avoids an impractical point by point evaluation of slope failure probability, are to provide: (i) a general overview of the areas where earthquake activity determines a major exposure to conditions of seismic triggering of landslides; (ii) reference values for evaluating whether a specic slope, once its actual critical acceleration is determined, has a signicant probability of failing in the future; (iii) the local safety factors required from engineering slope stabilisation works, to reduce future failure probability. Since the method is based on the Newmark model, it can be used within the limits of applications of that model (e.g. for relatively shallow landslides, but not for deep-seated mass movements). The new methodology was initially applied to Daunia, an area of Southern Italy located close to seismically active zones and frequently affected by slope failures even in static conditions: the results led to the conclusion, that because of the absence of important seismogenic structures within the Daunia region, seismicity is a less relevant as landslide triggering factor with respect to other causes (e.g. meteorological and anthropogenic). Two important natural hazard events, which occurred while paper [4] was in press, support this conclusion.

In particular, on 31 October 2002, an earthquake of local magnitude (ML) 5.4, whose epicenter was located 10 km NW of the Daunia area, produced a shaking of intensity VII on the MercalliCancaniSieberg (MCS) scale in the northern periphery of the study area, causing considerable damage to many buildings. However, our reconnaissance eld observations, combined with a helicopter overview of Daunia, documented only a few very minor rock falls in the northern-most part of the study area associated with this event. In contrast, very widespread shallow landslides were activated in the Daunia region after a three-day rain storm in late January 2003, following a relatively wet late fall-early winter period. These occurred after a few years of low mass movement activity characterised by a low average precipitation and high temperatures. In this study we have selected a different region to further test the applicability of the new approach to timeprobabilistic evaluation of seismically induced landslide hazard. It focuses on Irpinia, one of the Italian regions most severely devastated by past earthquakes. The last strong event occurred in 1980 (MSZ6.9), which triggered a large number of mass movements of different types (rock falls, slumps, earthows) and size (from a few cubic meters to a few tens of millions of cubic meters) [11,12]. The Irpinia region is close to Daunia (Fig. 1) but, whereas the latter is located at the eastern front of the Apennine mountain chain, Irpinia is situated well inside this orogenic belt and includes some of its most active seismogenic sources. This work rstly provides background information on the regional geological setting of Irpinia, its historical seismicity and the related landslide events. Secondly, the results of time-probabilistic evaluations of seismically induced landslide hazards are presented and their implications discussed.

2. Irpinia study area 2.1. Geological outline The described method was applied to an area between 40840 0 and 41810 0 of latitude North and between 14850 0 and 15830 0 of longitude East (Fig. 1). Almost the entire study area is located in the Irpinia region which occupies a central portion of the Southern Apennine mountains. The Southern Apennines represent a portion of an arc-shaped East-verging thrust-fold system which extends NWSE along the Italian peninsula. The current tectonic setting of the Southern Apennines originated from two major deformation phases [14]: (1) Middle MioceneUpper Pliocene, predominantly compressive tectonics, which resulted in crustal thickening and major uplift;

V. Del Gaudio, J. Wasowski / Soil Dynamics and Earthquake Engineering 24 (2004) 915928

917

Fig. 1. Location of the Irpinia (thick solid line square) and Daunia (thick dashed line rectangle) study areas in Southern Italy. Polygons in stipple represent the boundaries of the seismogenic zones according to Scandone [13].

(2) Quaternary, extensional tectonics taking over from the compressive deformation, following the gradual migration of the Apennine overthrust front towards the east. The ongoing tectonic extension is responsible for the high seismicity of the area, and the Irpinia region is rated among those with the highest earthquake hazard in Italy [15]. The general geological framework of the area has been inherited from Apennine tectonism, which produced large scale northeastward tectonic translations and the superposition of several highly deformed thrust sheets (e.g. Ref. [16]). The most striking structural features of the Southern Apennines are the Quaternary neotectonic normal faults which border the high relief massifs made of Mesozoic carbonate rocks. The intervening valleys include mainly Upper Cretaceous and Cenozoic terrigenous deposits of turbidite origin (ysch), containing variable amounts of clay materials. The carbonate rocks and turbidite deposits are tectonically deformed and locally sheared, and this makes them relatively prone to mass movements. Landslides of coherent and disruptive types are frequent, respectively, on valley slopes and steep mountain escarpments. The former include slumpearthows, earthows and composite (part rotational/part translational) slides, the latter are represented mostly by rock falls.

2.2. Historical seismicity Irpinia is located in a sector of the Apennine chain that has been affected by recurrent earthquakes of magnitude up to 7.0 and where frequent shocks are recorded by the national seismic network. Source parameters of major events were mostly obtained from macroseismic observations (CPTI Working Group [17]; Catalogue of Strong Italian Earthquake, CSIE [18]); in particular macroseismic magnitudes were estimated using methods exploiting the whole macroseismic eld (rather a simple relation magnitude-epicentral intensity) which proved to be quite reliable and provided results consistent with moment magnitude estimates [19]. For the major historical earthquakes, Fig. 2 shows the focal volumes, i.e. the volume of the relative sismogenic structure obtained following Bath and Duda [20], by means of the horizontal projection of an equivalent sphere. Fig. 2 also shows epicentres of shocks recorded after 1980, located by the seismic network of the Italian Istituto Nazionale di Geosica e Vulcanologia (INGV). Table 1 lists the major events which have affected the study area. Reported intensities, which in the original Italian historical catalogues are referred to the MCS scale, were converted to the MKS scale following [21]. For all the events macroseismic magnitudes are given, and for the most recent earthquakes (1962 and 1980) these coincide with

918

V. Del Gaudio, J. Wasowski / Soil Dynamics and Earthquake Engineering 24 (2004) 915928

Fig. 2. Historical and recent seismicity in Irpinia and surrounding regions. Circles represent the focal volumes calculated, following the formulation of Bath and Duda [20], for the earthquakes occurred until 1980 and reported by the CPTI catalogue [17]; dots mark the epicentre location of shocks recorded by the seismic network of the INGV from 1981 to 2001. Thick-line square marks the study area.

the instrumentally estimated values of MS. The largest earthquakes located inside the study area occurred in 1694 and in 1980, causing a large number of deaths (over 6000 and about 3000, respectively), in roughly the same area, i.e. extending from Irpinia to north-eastern Basilicata. Both events had a magnitude 6.9 and their epicentral coordinates were similar. Other highly damaging events had their epicentres slightly north of the previous ones. For example in 1732 a 6.6 magnitude event caused about 2000 fatalities throughout Irpinia, but especially in its northern part; in 1930 a 6.7 magnitude earthquake killed about 1500 people, mostly in Irpinia, but also to the east, in neighbouring Basilicata and Apulia provinces. Irpinia was also involved in one of the most disastrous events that hit Southern Italy, i.e. the 7.0 magnitude earthquake which occurred in 1456: this earthquake devastated a very large area including, from North to South, major portions of the Abruzzo, Molise and Campania regions, causing a high, though somewhat uncertain, number of deaths (around 10,000). The size and complex pattern of the macroseismic eld, reconstructed on the basis of historical documentation, suggested that in this case a few seismogenic structures were almost simultaneously involved in rupture processes. One of these possibly propagated southwards to northern Irpinia, generating one of the subevents (magnitude 6.6) forming the main shock. There are also documented cases of less energetic local events and of large but relatively distant source earthquakes, which produced less dramatic effects in Irpinia. Examples of

the rst kind include the Caposele event of 1853 (magnitude 5.9), which caused about ten fatalities, the Calitri event of 1910 (magnitude 5.8), which killed about 50 people, and the 1962 earthquake (magnitude 6.2), which hit some villages of northern Irpinia, killing 17 people. Among the strongest distant earthquakes, at least four cases deserve to be mentioned: (1) to the south of Irpinia, the 1561 earthquake (magnitude 6.4), which hit the Tanagro river valley at
Table 1 Dates and epicentral parameters of relevant historical earthquakes that affected the Irpinia area (data from CPTI [17]) Date 1456-12-05 1561-08-19 1694-09-08 1732-11-29 1805-07-26 1851-08-14 1853-04-09 1857-12-16 1910-06-07 1930-07-23 1962-08-21 1980-11-23 Epicentral coordinates 41.15N 14.87E 40.52N 15.48E 40.88N 15.35E 41.08N 15.05E 41.50N 14.47E 40.95N 15.67E 40.82N 15.22E 40.35N 15.85E 40.90N 15.42E 41.05N 15.37E 41.13N 14.97E 40.85N 15.28E Epicentral intensity X VIIIIX X X X VIIIIX VIII X VIIVIII X VIII X Macroseismic magnitude 6.6 6.4 6.9 6.6 6.6 6.3 5.9 7.0 5.8 6.7 6.2 6.9

MCS intensities, given by Ref. [17], were converted in MKS following Ref. [21]. The reported magnitude values are those derived from the macroseismic eld.

V. Del Gaudio, J. Wasowski / Soil Dynamics and Earthquake Engineering 24 (2004) 915928

919

the border between southern Campania and Basilicata, causing several hundred deaths; (2) to the north, the 1805 earthquake (magnitude 6.6), which devastated the Molise region, killing more than 5000 people; (3) to the east, the 1851 earthquake (magnitude 6.3), which hit the area around the extinct Vulture volcano, in northern Basilicata, resulting in about 1000 deaths; (4) also to the south, another of the major seismic catastrophes of southern Italy, i.e. the great 1857 earthquake (magnitude 7.0), which hit the Agri river valley in southern Basilicata, killing between 10,000 and 20,000 people. The historical seismic database shows that very strong events have occurred throughout the southern Apennine chain. The activation of large seismogenic sources appears to have been more frequent in the sector including Irpinia and its northern bordering area, compared to the southernmost sector (Fig. 2). 2.3. Seismically induced landslide events Several cases of ground failures have occurred in association with historical and recent earthquakes. Information on these is reported both in general seismic catalogues like CSIE [18] and in specic catalogues like the Catalogo nazionale degli Effetti Deformativi Indotti da Forti Terremoti (CEDIT, [22]); further data can be found in some works specically concerning Southern Italy [23,24]. Descriptions do not always allow the identication of the type of mass movement, but generally at least a gross distinction between coherent and disrupted landslides is possible. In particular, it is quite easy to recognise rock falls, whereas, with regard to landslides, in some cases they can be attributed to slumps or slump earthows, in others only to coherent type movements. In some cases descriptions also suggest that although ssures or cracks have affected soil slopes, the deformations have not evolved into a landslide. The most detailed information is available for the 1980 earthquake, which gave rise to several specic studies on seismically induced landslides (e.g. Refs. [11,12]). Table 2 reports localities and types of seismically induced ground failures documented in relation to different earthquakes, together with the indication of the MKS intensity at the site where phenomena were observed and the relative epicentral distance. Obviously this list cannot be considered a complete event record and for most localities, the ground failures induced by one or two earthquakes (one of which is usually the 1980 event) only are documented. Nevertheless, in some villages, where more specic studies were conducted (e.g. Refs. [25,26]), ground failures due to earthquakes were registered several times. In this region landslide triggering appears to have occurred under seismic shaking corresponding to site

intensity VVI MKS or higher, and generally within epicentral distances of a few tens of kilometres. Failures reported at Calitri, on the occasion of the Molise 1805 earthquake, seem to represent a special case, as this town was about 100 km distant from the epicentre location. However, the description of a contemporary chronicle [27] suggests that this earthquake did not induce landsliding, but only ground cracks, subsidence and/or liquefaction, probably in a at area close to the hill on which the town of Calitri is situated. More data are available for the last strong event, i.e. the 6.9 magnitude event of 23 November 1980. The published inventories of seismically triggered mass movements [12,22] appear more or less complete only for a limited area around the seismogenic source [28]. These studies demonstrated that failures consisted mainly of rock falls on steep carbonate slopes and slumpearthows on hillslopes in clay-rich turbiditic formations. These two types of landslides occurred in Irpinia under seismic shaking intensity equal to or exceeding VI degree MKS; this, however, simply reects the fact that nowhere in the study area was the intensity less than this value. Overall, the documented cases of rock falls appear fewer than those of the coherent slides. This seems to be due to the incompleteness of the historical records, especially with regard to smaller volume phenomena. Another reason may be that slopes most susceptible to rock falls are present in remote and sparsely populated mountainous areas. For example, better documented 1980 earthquake showed that rock falls represented the most numerous type of failure. Furthermore, the fact that all the pre-1980 reported cases involved localities where similar phenomena also occurred in 1980, suggests that disrupted slides recurrently affect the same zones whenever seismic shaking exceeds some signicant threshold. A further relevant aspect of the 1980 event is that whereas rock falls were generally co-seismic, most mass movements in soils were due to the reactivation of old landslides within the rst 24 h after the shaking [23]. This suggests that the earthquake induced movements were caused by a modication of time-dependent geotechnical and/or hydrogeological parameters, rather than by a direct triggering action. Post-seismic ground failures may thus also occur following shaking levels which are lower than those required for the immediate activation of mass movements.

3. Hazard evaluation 3.1. Estimation of Arias intensity for historical events The procedure for the evaluation of seismically induced landslide hazard described in Ref. [4] requires the denition of the maximum Arias intensity events that are relevant (in terms of inuence on the evaluation of

920

V. Del Gaudio, J. Wasowski / Soil Dynamics and Earthquake Engineering 24 (2004) 915928

Table 2 List of documented ground failures induced by historical earthquakes Event year 1456 1561 1694 Locality Apice Muro Lucano Bisaccia Calitri Colliano Ricigliano Teora Calitri Bisaccia Calabritto Caposele Senerchia Muro Lucano Castelgrande Calitri Andretta Aquilonia Ariano Irpino Bisaccia Calitri Flumeri Rocchetta SantAntonio Trevico Vallata Villanova del Battista Apice Melito Irpino SantArcangelo Trimonte Dist 6 26 15 7 18 26 8 104 25 4 1 9 54 60 1 14 12 26 5 18 19 9 11 10 18 4 6 5 Is X VIIVIII IXX X VII VIII IXX ? VVI VII VIII VII VVI VI VIII VI X VII VII VII VIII VIII VIII VII VIII VIIVIII VIIVIII VIII Type Cr Cr Co Sl Rf Cr Cr Cr Cr Rf, Co Rf, Cr Cr SlEf; Rf Cr Sl Cr Co Co Cr Co Cr Co Co Cr Co; Cr Co Co Co Event year 1980 Locality Acerno Andretta Bisaccia Calabritto Calitri Caposele Carife Castelfranci Castelgrande Colliano Conza della Campania Frigento Laviano Lioni Muro Lucano Nusco Oliveto Citra Pescopagano Rocca San Felice SantAndrea di Conza SantAngelo dei Lombardi Senerchia Solofra Teora Torella dei Lombardi Trevico Vallata Valva Volturara Irpina Dist 23 10 20 9 14 6 21 22 15 14 5 23 8 8 20 17 18 10 15 8 12 14 36 2 17 23 20 12 30 Is VII VI VI VIII VII VIII VIVII VI VII VII IXX VI X IXX VII VI VI VIII VII VIII IXX VIII VII VIII VII VI VI VII VII Type Rf Rt Cr; Rf SlEf; Rf Sl; Ef SlEf; Rf Rf SlEf SlEf; Rf Rf Sl SlEf; Rf Rf; Sl Rf Rf Rf; Ef SlEf; Rf SlEf SlEf SlEf Sl; Cr SlEf; Rf Rf SlEf SlEf; Rf Cr Cr Rf; Sl Cr

1805 1851 1853

1857 1910 1930

1962

Dist, epicentral distance (in km); Is, estimated site MKS intensity (? if unknown). Type of failure: Cr, ground cracks; Ef, earthow; Rf, rock falls; Sl, slumps; SlEf, slumpearthows; Rt, roto-translational slides; Co, coherent-type landslide (description does not allow a more precise classication).

(Ac)x and in relation to their frequency occurrence) for the estimates of Newmark displacement exceedance probability. In order to have some indication of Ia values produced in Irpinia by historical events, the Arias intensity expected for each earthquake reported in historical catalogues was calculated on a grid of equally spaced points covering the study area through an attenuation relation calibrated on Italian accelerometric data [29]. The relation used has the following expression p log Ia Z K4:066 C 0:911M K 1:818 log R2 C 5:32 C SG 0:397 (2)

where M is the magnitude (ML if less than 5.5, MS otherwise), R is the horizontal distance from the seismogenic fault and S represents the site geology effect, which assume values 0 for stiff rocks, 0.244 for shallow soils (up to 20 m thick) overlying a stiff

bedrock, and 0.139 for deeper soils. Calculations were carried out for the maximum amplication conditions (shallow soils); R was calculated as the minimum distance from seismogenic sources (represented as rectangular boxes), following the models reported by a specic catalogue and dened on the basis of geological, macroseismic and instrumental data [30]. In Fig. 3 the maximum Arias intensities provided for historical events by Eq. (2) at each node of the examined grid were contoured. These Ia values do not represent the maxima which actually occurred point by point, considering that Eq. (2) provides a median estimate. However, their space distribution suggests how seismically induced landslide hazard can be related to the location of historically known seismogenic sources. Ia values in Fig. 3 reach an absolute maximum of 14 m/s in the zones where the 1694 and 1980 events produced the most severe damage; relative maxima ranging from 78

V. Del Gaudio, J. Wasowski / Soil Dynamics and Earthquake Engineering 24 (2004) 915928

921

Fig. 3. Maximum Arias intensities provided by Eq. (2) for the past earthquakes in the study area (contour spacing is 0.1 m/s). Localities are shown where historically documented cases of seismically induced ground failures occurred (cf. Table 2). Black and white symbols indicate, respectively, types of ground failures occurred during historical earthquakes and 1980 Irpinia earthquake.

to 910 m/s, present in the northern part of the study area, are connected to the three 6.66.7 magnitude events of 1456, 1732 and 1930. The lowest values, found in the S-W corner, do not drop below 0.5 m/s, which, according to Keefer and Wilson [31], represents a minimum shaking threshold for seismic triggering of landslides. Thus, historical record indicates that the whole study area can potentially undergo shakings capable of inducing slope instability. Fig. 3 also shows the localities where seismically induced ground failures have been historically documented. Although the distribution of these sites is dominated by the more complete data set of the 1980 event, one can notice a certain concentration along a NWSE belt which coincides with the locations of the most active seismogenic zones. Some rare ground failure cases fall in the south-western zone characterised by the lowest Ia values, but these concern only minor mass movements triggered by the 1980 earthquake. 3.2. Selection of basic parameters The calculation of (Ac)x for a xed Newmark displacement exceedance probability requires a selection of some

specic parameters The rst is the relevant exceedance probability for which (Ac)x has to be computed. The choice of this value depends on the purpose of a hazard evaluation and on aspects of civil protection policy such as the social perception of risk tolerance. In this study we considered an exceedance probability of 10% in 50 yr, which is the value generally adopted for seismic building codes. This appears to be a reasonable choice where potential seismic ground failures represent a threat to buildings or infrastructure, and where engineering stabilisation works might be necessary. Indeed a rational approach to hazard management should require a uniform treatment of the risk linked to both direct and induced effects of earthquakes. However, it is recognised that the denition of the probability level used for building code and followed here is rather arbitrary and different choices could also be justied according to the specic scope of hazard assessment. The second parameter is the threshold x of Newmark displacement. Its value depends on the type of mass movement. Earthquakes have triggered both coherent and disrupted type landslides in the study area, and thus two different maps were produced. Following Wilson

922

V. Del Gaudio, J. Wasowski / Soil Dynamics and Earthquake Engineering 24 (2004) 915928

and Keefers indications [7], it was decided to map the spatial distribution of (Ac)10 (0.1, 50 yr) and (Ac)2 (0.1, 50 yr), i.e. the critical acceleration for which a 10% probability in 50 yr exists that Newmark displacement exceeds, respectively, 10 cm for coherent slides and 2 cm for disrupted ones. 3.3. Calculation of (Ac)x The calculation of (Ac)x requires the estimate of the expected number of events causing different Ia values at each node of the grid covering the study area, in a given time interval. It was obtained by the code SEISRISK III [9], whose application needs, in turn, the following basic ingredients: (a) a seismotectonic zonation, (b) the seismicity rates and the maximum expected magnitude for each of the pertinent seismogenic zones, (c) an attenuation relationship in terms of Arias intensity. Regarding the seismotectonic zonation, we adopted the one specically designed for hazard assessment by the Italian Gruppo Nazionale Difesa dai Terremoti (National Group of Defense from Earthquakes) [13]. The seismicity

rates of seismotectonic zones were estimated by applying the procedure described by Slejko et al. [15] to lists of historical events extracted from the catalogue CPTI [17]. Finally, for the attenuation relationship, Eq. (2) was employed taking into account the scatter of actual values around those predicted by it, through the inclusion of the standard deviation in the computation of the yearly number of events; the parameter S was assumed, respectively, equal to 0.139 for the cases of coherent slides in soils and to 0 for the rock falls on rock slopes. The results obtained are presented in a map format in Figs. 4 and 5. The two maps show a similar pattern characterised by an axial maximum present along the NWSE belt coinciding with the most active seismogenic zone, and a decrease both towards to the NE and to the SW corner, where (Ac)x (0.1, 50 yr) reaches its minimum values. However, the absolute (Ac)x values differ: for coherent slides the minimum critical acceleration required for a slope to keep failure probability in 50 yr below 10% reaches a maximum of 0.051 g and a minimum of 0.021 g, whereas for disrupted slides the corresponding values are higher (0.078 and 0.037 g, respectively). The lower values obtained for coherent slides in soils with respect to falls in rocks are due to the higher Newmark displacement required for mass movements to be triggered; this reduces the critical

Fig. 4. Values of (Ac)10 corresponding to a 10% probability in 50 yr that Newmark displacement exceeds 10 cm in soil slopes (contour spacing is 0.01 g). Black and white symbols indicate, respectively, types of ground failures occurred during historical earthquakes and 1980 Irpinia earthquake.

V. Del Gaudio, J. Wasowski / Soil Dynamics and Earthquake Engineering 24 (2004) 915928

923

Fig. 5. Values of (Ac)2 corresponding to a 10% probability in 50 yr that Newmark displacement exceeds 2 cm on rock slopes (contour spacing is 0.01 g). Black and white symbols indicate, respectively, types of ground failures occurred during historical earthquakes and 1980 Irpinia earthquake.

acceleration required for stability more than the increase due to the relative amplication of seismic shaking in soils in comparison to stiff rocks. Any inference regarding the major or minor occurrence probability of the two types of failures cannot neglect the actual geotechnical conditions of slopes susceptible to mass movements. Thus, for instance, even though in the same zone, rock slopes are found to need a higher critical acceleration than soft soil slopes to sustain seismic shaking without failing, this does not imply that the former have a major probability to undergo failure, considering that rock formations may actually have a higher shear strength.

4. Discussion 4.1. Comparison of results with data on critical acceleration The resistance required for Irpinian slopes to keep their failure probability comparable with the level of risk considered acceptable by building seismic codes does not appear very high. Wilson and Keefer [7] proposed 0.05 g as a minimum threshold of critical acceleration to characterise slopes susceptible to seismic landsliding. They claimed that

lower values of critical acceleration would imply such an unstable condition as to make more probable an activation of landslides by other more frequent non-seismic causes. Indeed, if so, the contemporaneous presence, at any time in a given region, of a large number of slopes having a very low critical acceleration and thus susceptible to simultaneous failure during an earthquake, would be unlikely. The (Ac)x (0.1, 50 yr) values obtained for the study area are only slightly higher than the 0.05 g minimum limit, but this does not mean that our results attribute a limited seismic landslide hazard to Irpinia. Indeed the 0.05 g threshold proposed by Wilson and Keefer on the basis of their observations in California, may not be automatically assumed valid everywhere and this threshold could vary in relation to the recurrence frequency of other destabilising phenomena (e.g. meteorological events). Therefore, a comparison between the calculated (Ac)x (0.1, 50 yr) values and spatial critical accelerations on slopes in the study area is needed for a better characterisation of actual hazard. Unfortunately, as regards Irpinia, the literature reports only a limited number of specic cases where critical accelerations have been evaluated for presumed slide surfaces. Table 3 presents seven cases of pre-existing landslides which underwent seismic shaking: they all refer to coherent slides in soils and the range of estimated values

924

V. Del Gaudio, J. Wasowski / Soil Dynamics and Earthquake Engineering 24 (2004) 915928

Table 3 Critical acceleration ac calculated for slide surfaces and (Ac)10 values estimated for the corresponding landslide slopes in relation to seismic shocks occurred (see Fig. 4 for location) Slide surface Andretta landslide (1980), lower part [32] Andretta landslide (1980), upper part [32] Calitri landslide (1980), surface A [33] Calitri landslide (1980), surface B [33] Senerchia Acquara landslide (1980) [34] Senerchia Vadoncello landslide (1995) [35] Senerchia Vadoncello landslide (1996) [35]
a

ac (g) 0.020.04 0.080.10 0.05 0.02 0.010.15 0.01 0.12

(Ac)10 (g) 0.141 0.084 0.095 0.002 0.003a

The 1996 shock did not reactivate the landslide.

reects mainly the uncertainty of the hydrogeological conditions, but also, in some cases (especially the Andretta and Calitri landslides), the uncertainties of the subsurface failure geometry. One can notice that the critical accelerations are generally rather low, often not greater than 0.05 g. A calculation of the critical acceleration representing the resistance limit required for a slope not to undergo Newmark displacement as large as 10 cm, according to the Jibson formula (1), was also made using the level of seismic shaking from the cases above. The estimates given in Table 3 should be viewed only as an indicative evaluation, considering that the relation (1) is affected by various sources of uncertainty and actually provides Dn as having a 50% probability of being exceeded under a given seismic shaking of Arias intensity Ia. Moreover, only in the case of the Calitri landslide was a site-specic instrumental measurement of Arias intensity available. In other cases, Ia was derived by applying the expression (2) which introduces a further source of uncertainty. However, comparing the estimated ac values with the (Ac)10 values corresponding to the seismic shaking observed or estimated, a reasonably good consistency was found and, in general, a given landslide was triggered according to whether (Ac)10 exceeded or not the critical acceleration. The three Senerchia cases deserve some further consideration. In the rst case, a very large uncertainty affects the ac estimate for the slope involved by the huge (2.5 km long) Serra dellAcquara landslide which occurred following the 1980 seismic event. About 2/3 of the range of the possible ac values (0.010.15 g) are below the minimum resistance predicted to avoid failure (0.095 g). The second case regards a subsidiary slumpearthow reactivated in 1995 on the left ank of the 1980 landslide, in a zone named Acquara-Vadoncello. Although the re-activation process was apparently gradual, marked initially by minor falls and slides in the main scarp area during early winter 1995, the rst larger volume headscarp failure took place in February 1995 [36], nine days after a seismic sequence of small magnitude (%3.6ML) events. In this case, (Ac)10 was slightly below the estimated critical ac, but the exact role of seismic shaking in landslide reactivation remains uncertain and the occurrence of a weak shock before the start of the process may have been fortuitous. The third case

also regards the Acquara-Vadoncello landslide and the 4.3 ML earthquake event the following year. Field observations the day after the shock indicated there were no macroscopic effects on slide activity. These observations are consistent with the estimate of ac whose value has considerably increased. This seems to be a consequence of intervening major slope modications caused naturally by active landsliding of 1995, which produced a more stable prole [35,36]. Some additional indications were obtained by examining the documented cases of failures induced by past earthquakes and evaluating rst, from relation (2), the local Arias intensity and then the (Ac)x values for which the relation (1) gives Newmark displacement equal to the thresholds proposed by Wilson and Keefer for landslide triggering. Considering that in all these cases mass movements actually occurred, the critical acceleration of the involved slopes should have been, in most of them, less than the calculated (Ac)x values. Tables 46 show the results obtained, respectively, for the cases of coherent slides and ground cracks in soft soils and for disrupted slides on rock slopes.
Table 4 Arias intensity and the corresponding (Ac)10 values estimated at sites where coherent slides in soils occurred during earthquakes Locality Andretta Apice Aquilonia Ariano Irpino Bisaccia Calabritto Calabritto Calitri Calitri Calitri Calitri Caposele Castelfranci Castelgrande Conza della Campania Frigento Laviano Melito Irpino Muro Lucano Nusco Oliveto Citra Pescopagano Rocca S.Felice Rocchetta S. Antonio S. Andrea di Conza S. Angelo dei Lombardi S. Arcangelo Trimonte Senerchia Teora Torella dei Lombardi Trevico Valva Villanova del Battista
*

Event 23-11-1980 21-08-1962 23-07-1930 23-07-1930 23-11-1980 09-04-1853 23-11-1980 08-09-1694 07-06-1910 23-07-1930 23-11-1980 23-11-1980 23-11-1980 23-11-1980 23-11-1980 23-11-1980 23-11-1980 21-08-1962 16-12-1857 23-11-1980 23-11-1980 23-11-1980 23-11-1980 23-07-1930 23-11-1980 23-11-1980 21-08-1962 23-11-1980 23-11-1980 23-11-1980 23-07-1930 23-11-1980 23-07-1930

Ia (m/s) 2.854 1.818 1.394 0.383 0.935 0.911 3.299 3.822 1.286 0.760 1.639 5.042 0.797 1.559 6.362 0.704 3.972 1.249 0.181 1.230 1.083 2.726 1.516 1.960 3.988 2.061 0.220 1.674 9.595 1.199 1.507 1.999 0.654

(Ac)10 (g) 0.141 0.102 0.084 0.033* 0.063 0.062 0.157 0.174 0.079 0.054 0.095 0.213 0.056 0.091 0.252 0.051 0.179 0.078 0.019* 0.077 0.070 0.137 0.090 0.108 0.180 0.112 0.022* 0.096 0.339 0.076 0.089 0.109 0.049*

Cases in which (Ac)10 was less than 0.05 g.

V. Del Gaudio, J. Wasowski / Soil Dynamics and Earthquake Engineering 24 (2004) 915928 Table 5 Arias intensity and the corresponding (Ac)10 values estimated at sites where the opening of ground cracks in soils occurred during earthquakes Locality Andretta Apice Bisaccia Bisaccia Bisaccia Calitri Caposele Castelgrande Flumeri Muro Lucano Ricigliano S. Angelo dei Lombardi Senerchia Teora Trevico Vallata Vallata Villanova del Battista Volturara Irpina
*

925

Event 23-07-1930 05-12-1456 14-08-1851 23-07-1930 23-11-1980 26-07-1805 09-04-1853 16-12-1857 23-07-1930 19-08-1561 08-09-1694 23-11-1980 09-04-1853 08-09-1694 23-11-1980 23-07-1930 23-11-1980 23-07-1930 23-11-1980

Ia (m/s) 1.136 2.580 0.181 4.790 0.935 0.024 1.292 0.151 0.673 0.193 0.552 2.061 0.379 3.257 0.767 1.833 0.883 0.654 0.439

(Ac)10 (g) 0.073 0.131 0.019* 0.205 0.063 0.005* 0.080 0.017* 0.050 0.020* 0.043* 0.112 0.033* 0.155 0.055 0.103 0.061 0.049* 0.037*

Cases in which (Ac)10 was less than 0.05 g.

The values of (Ac)10 for coherent slides are rather low: for nearly 1/2 of the documented cases (16 out of 33) the minimum critical acceleration required to resist instability is below 0.09 g and for about 1/3 of cases (10/33) it does not exceed 0.07 g; in four cases (about 12%) the estimated (Ac)10 is even less than 0.05 g. These results imply that a number of seismically induced landslides in soils affected
Table 6 Arias intensity and the corresponding (Ac)2 values estimated at sites where rock falls in carbonate rocks occurred during earthquakes Locality Acerno Bisaccia Calabritto Calabritto Caposele Caposele Carife Castelgrande Colliano Colliano Frigento Laviano Lioni Muro Lucano Muro Lucano Nusco Oliveto Citra Senerchia Solofra Torella dei Lombardi Valva
*

marginally stable slopes. For a comparison, the same calculation was carried out for the cases of ground cracks that did not evolve into landslides. These can be considered examples of situations where Newmark displacement did not reach a sufcient value to cause a complete slope collapse. In these cases the strength of slopes involved has been apparently greater than the minimum resistance required to avoid landsliding. Indeed, the calculated (Ac)10 values are rather low, with about 40% of cases (8 out of 19) below 0.05 g. The (Ac)2 values obtained for the disrupted landslides are slightly higher than the (Ac)10 values of the coherent slides: the median is 0.11 g and only in 2 out of 21 cases (Ac)2 is less than 0.05 g. Thus, the rock falls generally seem to have affected slopes not as weak as in the case of the coherent slides. This could be connected, in part, to the differences in mechanism and timing of mass movements. Earthquakeinduced rock falls are generally co-seismic and are triggered directly by the action of inertial forces. In contrast, the coherent slides observed in Irpinia are often reactivations of old landslides, where movements have been delayed by several hours with respect to the main shock. This implies that in some cases, the actual critical acceleration of the slope must have been slightly higher than the minimum value required to resist the shaking. The post-seismic failure followed as a consequence of a decrease in slope stability, possibly due to changes of hydraulic conditions (e.g. build up of pore water pressures following seismic loading). Thus, in the case of slopes prone to coherent type landslides, the estimates of the critical acceleration required for a slope to keep its future seismic failure probability below a xed limit, should be suitably increased with respect to the value provided by a simple analysis of the mechanical actions of shaking. 4.2. Inuence of the estimate of seismic shaking recurrence An additional aspect examined regards the inuence of methods chosen to evaluate temporal recurrence of seismic shaking on the nal hazard estimate This question involves the yet open problem of a sound validation of seismic hazard assessment methods. Indeed, when dealing with relatively long-term probabilistic estimates, it is difcult to discriminate between alternative methods, unless observations are available for long time after the hazard estimate. It should be stressed that the Cornell-type approach, which is the internationally most widely used method of seismic hazard assessment, does not necessarily provide the most conservative hazard estimates. The method assumes that the event generation probability within a single seismogenic zone be homogeneous (seismotectonic probabilism approach, according to Muir-Wood [37]). It is well known that this assumption generally tends to produce an effect of areal spreading of the hazard sources with respect to the actual spatial distribution of historical events. This can result in a certain underestimate of hazard, at least in some

Event 23-11-1980 23-11-1980 09-04-1853 23-11-1980 09-04-1853 23-11-1980 23-11-1980 23-11-1980 08-09-1694 23-11-1980 23-11-1980 23-11-1980 23-11-1980 16-12-1857 23-11-1980 23-11-1980 23-11-1980 23-11-1980 23-11-1980 23-11-1980 23-11-1980

Ia (m/s) 0.545 0.679 0.662 2.396 0.938 3.661 0.642 1.132 0.764 1.240 0.511 2.884 2.581 0.131 0.647 0.893 0.786 1.216 0.236 0.870 1.451

(Ac)2 (g) 0.079 0.093 0.091 0.230 0.117 0.312 0.089 0.134 0.101 0.143 0.075 0.263 0.243 0.028* 0.089 0.113 0.103 0.141 0.043* 0.111 0.160

Cases in which (Ac)12 was less than 0.05 g.

926

V. Del Gaudio, J. Wasowski / Soil Dynamics and Earthquake Engineering 24 (2004) 915928

areas, in comparison with the older historical probabilism approach (cf. Ref. [37]), which deduces the space-time distribution property of source activation from historical seismicity, without considering seismotectonic modelling. To test the implication of using different approaches in the evaluations of the expected number of events, we compared the estimates obtained in this study with our earlier results based on a historical probabilism approach applied to the occurrence of seismically induced rock falls in the Sele river valley [38]. This valley is orientated North South and its upper parts consist of steep limestone slopes particularly prone to rock falls. Following a technique described by Berardi et al. [39], we estimated the mean return time of macroseismic intensities relevant for the triggering of rock falls in this area. On the basis of historical evidence, shaking thresholds of VVI degree MKS have been examined. The mean return times of events exceeding such thresholds were calculated for several sites located on the two anks of the valley. The results indicated a general increasing trend of return time (which implies a decrease of hazard) from NE to SW, related to the more frequent recurrence of earthquakes in the northernmost axial sector of Southern Apennine chain. This is also consistent with the outcomes of the present work (Figs. 4 and 5). In order to obtain a more direct comparison between the results of the previous (historical) and the present (seismotectonic) method, we repeated the calculations of the historical approach in terms of Arias intensity return time. Following Keefer and Wilson [31] we adopted 0.11 m/s as a threshold characterising the minimum conditions for rock fall triggering. Table 7 shows a comparison of the results obtained with those reported in our previous work [38]. The events exceeding the selected threshold appear to have a return time intermediate between those relative to V and VI degree MSK. This does not imply a general equivalence between an Ia value of 0.11 m/s and a VVI MSK intensity. In a case study of an area in the south-central Apennines, the results of Ia estimates showed a better agreement with return times obtained for intensity VI or even higher (Authors unpublished data). This is probably related to a systematic

difference in the rate of attenuation with distance existing between Arias and macroseismic intensity, which in turn reects the different seismic wave properties that inuence these two parameters (e.g. the spectral band on which energy is mostly concentrated). Thus, their exceedance probability is differently inuenced by local attenuation conditions and by distances from dominant seismogenic structures. Table 7 also shows the probability of exceedance in 50 yr of the same Ia threshold and the critical acceleration (Ac)2 which determines a 10% probability in 50 yr that Newmark displacement exceeds 2 cm. These values are compared with those obtained using the seismotectonic approach. It is evident that the latter produces lower hazard estimates with respect to those based on the historical approach. The modelling of seismic activity according to seismotectonic zonation provides values of I a (0.180.26 against 0.560.66 m/s) and of (Ac) 2 (0.050.06 against 0.130.24 g) that are lower by a factor three or four with respect to those obtained directly from historical record. There is a general agreement that the historical approach provides a rather rough hazard estimate [37], which can result in overestimation or underestimation of the actual risk. It is apparent, however, that, with the present state-of-the-art, all the methods of seismic hazard assessment suffer from a rather large margin of uncertainty (e.g. Refs. [4042]). The new method proposed to estimate the hazard of seismically induced landslides can incorporate without difculty any technique, which evaluates the expected number of earthquakes. Thus, according to the purpose of hazard evaluation, one can consider employing different well established techniques and select a more or less conservative approach. 4.3. Developments and perspectives The proposed method has close analogies with approaches commonly used for seismic hazard assessment In both cases attention is paid to the spatial distribution of shaking level having an acceptable probability not to be exceeded within a time-window allowing for the denition

Table 7 Comparison between hazard estimates obtained with the historical and seismotectonical approaches at sites of the Sele river valley: return time (in years) for events exceeding the Arias intensity IaZ0.11 m/s; probability of exceedance of the same threshold in 50 yr; (Ac)2 value determining a 10% probability in 50 yr that Newmark displacement exceeds 2 cm Sites Historical approach Return time Castelnuovo di Conza Laviano Colliano Caposele Calabritto Senerchia Oliveto Citra 46.5 47.7 49.7 50.3 52.3 57.2 60.6 Exceed. prob. (in 50 yr) 0.659 0.649 0.634 0.630 0.616 0.583 0.562 (Ac)2 (0.10, 50 yr) 0.229 0.210 0.158 0.239 0.214 0.159 0.127 Seismotectonical approach Return time 167.7 155.5 181.8 166.9 222.6 196.4 257.9 Exceed. prob. (in 50 yr) 0.258 0.275 0.240 0.259 0.201 0.225 0.176 (Ac)2 (0.10, 50 yr) 0.060 0.058 0.053 0.057 0.055 0.049 0.045

V. Del Gaudio, J. Wasowski / Soil Dynamics and Earthquake Engineering 24 (2004) 915928

927

of optimal prevention actions. In seismic hazard this shaking level is expressed as a function of some parameter (e.g. peak ground acceleration) related to strength properties required for buildings or other engineering works to have acceptably low probabilities of damage. In the case of seismically induced landslide hazard, shaking can be expressed as a function of a parameter (critical acceleration (Ac)x) representing the slope resistance to a deformation (Newmark displacement) required for slope failures. In both cases the actual evaluation of the expected damages would require the integration of hazard parameter with other parameters representing the actual presence of potential damage targets (exposure) and their susceptibility to be damaged (vulnerability). In the case of seismic landslide hazard, the exposure could be expressed by the number or the areal frequency of slopes exceeding a minimum dip angle pertinent for the landslide type considered, whereas the vulnerability could be represented by the number or areal frequency of slopes having the actual critical acceleration lower than the pertinent (Ac)x values. Whereas the exposure could be obtained quite easily from numerical models of topography through GIS-based techniques, the vulnerability is hard to deal with at regional scale, because it requires a large amount of detailed site-specic information on lithological and hydrogeological characteristics. Nevertheless, this information could be collected at local scale only for the areas where the product of shaking hazard and slope exposure is high, especially if combined with the presence of relevant human and economic settlements.

5. Conclusions The application of a recently proposed method for the evaluation of seismically induced landslide hazard to Irpinia, one of the most seismically active region of southern Italy, provided further indications on its limitations and potential. The results obtained, expressed in terms of critical acceleration required to keep slope failure probability low, indicate that, the seismic activity levels observed in Italy (with magnitudes rarely reaching or exceeding 7.0) would not imply high levels of seismic landslide hazard outside active seismogenic zones. This is consistent with the relation found by Keefer [43] and Rodriguez et al. [44] on the maximum distance at which mass movements can be expected, which, for events of magnitude less than 7.0 is about 100150 km. Therefore in Italy the contribution to hazard from distant seismogenic structures can be relevant only in the case of highly susceptible slopes. On the whole, the critical acceleration values required to keep the probability of slope failures in Irpinia low seem rather small (at the most about 0.05 and 0.08 g, respectively,

for landslides in soil slopes and in rock slopes). However, the available data indicate that potential slide surfaces with ac below these values could be common. For example this is the case of coherent slides in soils, which are often found on slopes with ac values implying conditions of marginal stability. Such an inference is supported by the calculations of ac values for slopes involved in recent seismically induced failures (Table 3) and also by the estimates of the ac values compatible with the seismic triggering of documented mass movements induced by historical earthquakes (Table 4). This suggests that, in the study area, perhaps in relation to local hydrogeological and meteorological conditions, a considerable number of marginally stable slopes can survive non-seismic mobilising actions and fail due to earthquakes, even those not producing particularly strong shaking. The results of the applications of the proposed method are strongly inuenced by the seismic hazard assessment technique employed to calculate the expected number of earthquakes. However, any method capable of providing such an estimate can be easily integrated in the described procedure, and one can choose the most suitable technique according to the hazard assessment purposes (e.g. preliminary location of potentially unstable areas, estimate of expected damage, denition of safety factor for stabilisation works). The method proposed provides basic information for the denition of a risk index in relation to landslide triggering during earthquakes. This index can integrate other analogous risk indexes (e.g. expected number of buildings damaged, expected number of people involved) commonly considered in seismic risk assessment. Taking into account the differences existing between the types of landslides (both in terms of triggering conditions and practical consequences), distinct risk indexes should be evaluated (e.g. for coherent-type landslides and for disrupted ones).

Acknowledgements This work was supported in part by the Italian Ministry of Education, University and Research and by the European Community (contract no. EVGI-CT-2001-00055-project LEWIS). Figures were obtained employing the G.M.T. freeware package by W.H.F. Smith and P. Wessel. We thank two anonymous reviewers and Paul Gostelow for their thoughtful comments and suggestions.

References
[1] Jibson RW, Harp EL, Michael JA. A method for producing digital probabilistic seismic landslide hazard maps. Eng Geol 2000;58: 27189.

928

V. Del Gaudio, J. Wasowski / Soil Dynamics and Earthquake Engineering 24 (2004) 915928 [24] Pordo S, Esposito E, Vittori E, Tranfaglia G, Michetti AM, Blumetti M, Ferreli L, Guerrieri L, Serva L. Areal distribution of ground effects induced by strong earthquakes in the southern Apennines, Italy. Surveys Geophys 2002;23:52962. [25] Fenelli GB, Picarelli L, Silvestri F. Deformation process of a hill shaken by the Irpinia earthquake in 1980. Proceedings of the French Italian Conference on Slope Stability in Seismic Areas, Bordighera 1992 p. 4762. [26] Hutchinson JN, Del Prete M. Landslides at Calitri, Southern Apennines, reactivated by the earthquake of 23rd November 1980. Geol Appl Idrogeol 1985;20(1):938. [27] Poli GS. Memoria sul tremuoto de 26 luglio del corrente anno 1805, Napoli 1806. [28] Wasowski J, Del Gaudio V, Pierri P, Capolongo D. Factors controlling seismic susceptibility of the Sele valley slopes: the case of the 1980 earthquake Irpinia re-examined. Surveys Geophys 2002; 23:56393. [29] Sabetta F, Pugliese A. Estimation of response spectra and simulation of nonstationary earthquake ground motions. Bull Seism Soc Am 1996;86:33752. [30] Valensise G, Pantosti D (editors). Database of potential sources for earthquakes larger than M 5.5 in Italy. Annali di Geosica 2001;44:180. [31] Keefer DK, Wilson RC. Predicting earthquake-induced landslides, with emphasis on arid and semi-arid environments. In: Sadler PM, Morton DM, editors. Landslides in a semi-arid environment, vol. 2. Inland Geological Society; 1989. p. 11849. [32] DElia B. Dynamic aspects of a landslide reactivated by the November 23, 1980 Irpinia earthquake. Proceedings of the French Italian Conference on Slope Stability in Seismic Areas, Bordighera 1992 p. 2632. ` di un pendio [33] Crespellani T, Madiai C, Maugeri M. Analisi di stabilita in condizioni sismiche e post-sismiche. Rivista Italiana di Geotecnica 1996;1:5061. [34] Maugeri M, Motta E, Sorriso-Valvo M. The Senerchia landslide triggered by the 23 November 1980 earthquake. Proceedings of the Fourth International Congress IAEG, New Delhi 1982 p. 13949. [35] Del Gaudio V, Trizzino R, Calcagnile G, Calvaruso A, Pierri P. Landsliding in seismic areas: the case of the Acquara-Vadoncello landslide, Southern Italy. Bull Eng Geol Environ 2000;59:2337. [36] Wasowski J. Inclinometer and piezometer record of the 1995 reactivation of the Acquara-Vadoncello landslide, Italy. Proceedings of the Eighth International Congress IAEGE, Vancouver, Canada, 2125 September 1998 p. 1697704. [37] Muir-Wood R. From global seismotectonics to global seismic hazard. Annali di Geosica 1993;36(3/4):15368. [38] Del Gaudio V, Wasowski J. Seismic rockfall hazard in the Sele river valley, Southern Italy. In: Bromhead E, Dixon N, Ibsen M, editors. Landslides in research, theory and practice. Thomas Thelford; 2000, vol. 1, p. 42732. [39] Berardi R, Petrungaro C, Zonetti L, Magri L, Mucciarelli M. Mappe di ` per larea italiana.: ISMESIENEL; 1994. sismicita ` delle stime di pericolosita ` [40] Albarello D, DAmico V. La validita sismica in Italia Atti 198 Convegno GNGTS. Roma: CNR; 2000 [CD-ROM]. [41] Petersen MD, Cramer CH, Reichle MS, Frankel AD, Hanks TC. Discrepancy between earthquake rates implied by historic earthquakes and a consensus geologic source model for California. Bull Seism Soc Am 2000;90:111732. [42] DAmico V, Albarello D. The role of data processing and uncertainty management in seismic hazard evaluations: insight from estimates in the Garfagnana-Lunigiana area (northern Italy). Nat Hazards 2003;29: 7795. [43] Keefer DK. Landslides caused by earthquakes. Bull Geol Soc Am 1984;95:40621. [44] Rodriguez CE, Bommer JJ, Chandler RJ. Earthquake-induced landslides: 19801997. Soil Dyn Earthquake Eng 1999;18:32546.

[2] Miles SB, Ho CL. Rigorous landslide hazard zonation using Newmarks method and stochastic ground motion simulation. Soil Dyn Earthquake Eng 1999;18(4):30523. [3] Mankelow JM, Murphy W. Using GIS in the probabilistic assessment of earthquake triggered landslide hazards. J Earthquake Eng 1998; 2(4):593623. [4] Del Gaudio V, Pierri P, Wasowski J. An approach to time probabilistic evaluation of seismically-induced landslide hazard. Bull Seism Soc Am 2003;93:55769. [5] Arias A. A measure of earthquake intensity. In: Hansen RJ, editor. Seismic design for nuclear power plants. Cambridge, MA: MIT Press; 1970. p. 43883. [6] Newmark NM. Effects of earthquakes on dams and embankments. Geotechnique 1965;15:13960. [7] Wilson RC, Keefer DK. Predicting the areal limits of earthquakeinduced landsliding. In: Ziony JI, editor. Evaluating earthquake hazards in the Los Angeles regionan earth science perspective. US Geological Survey, Professional Paper 1360, 1985. p. 31645. [8] Cornell CA. Engineering seismic risk analysis. Bull Seism Soc Am 1968;58:1583606. [9] Bender B, Perkins DM. SEISRISK III: a computer program for seismic hazard estimation. US Geol Survey Bull 1772 1987;48. [10] Romeo R. Seismically induced landslide displacements: a predictive model. Eng Geol 2000;58:33751. [11] Cotecchia V. Ground deformations and slope instability produced by the earthquake of 23 November 1980 in Campania and Basilicata. Geol Appl Idrogeol 1986;21(5):31100. [12] Carrara A, Agnesi V, Macaluso T, Monteleone S, Pipitone G. Slope movements induced by the southern Italy earthquake of November 1980. Geol Appl Idrogeol 1986;21(2):23750. [13] Meletti C, Patacca E, Scandone P. Construction of a seismotectonic model: the case of Italy. Pure Appl Geophys 2000;157:1135. [14] Cinque A, Patacca E, Scandone P, Tozzi M. Quaternary kinematic evolution of the Southern Apennines. Relationships between surface geological features and deep lithospheric structures. Annali di Geosica 1993;36(2):24960. [15] Slejko D, Peruzza L, Rebez A. Seismic hazard maps of Italy. Annali di Geosica 1998;41:183214. [16] Mostardini F, Merlini S. Appennino Centro Meridionale. Sezioni geologiche e proposta di modello strutturale. Mem Soc Geol It 1986; 35:177202. [17] CPTI Working Group. CPTI Working Group (INGV, GNDT, SGA, SSN) Catalogo parametrico dei terremoti italiani. Bologna: Editrice Compositori; 1999. Also available at: http://emidius.mi.ingv.it/CPTI/ home.html. [18] Boschi E, Guidoboni E, Ferrari G, Mariotti D, Valensise G, Gasperini P (editors). Catalogue of strong Italian earthquakes from 461 B.C. to 1997. Annali di Geosica 2000;43:609868. [19] Gasperini P, Ferrari G. Deriving numerical estimates from descriptive information: the computation of earthquake parameters. Annali di Geosica 2000;43:72946. [20] Bath M, Duda SJ. Earthquake volume, fault plane area, seismic energy, strain, deformation and related quantities. Annali di Geosica 1964;17(3):35368. [21] Levret A, Mohammadioun B. Determination of seismic reference motion for nuclear sites in France. Eng Geol 1984;20:2538. [22] Romeo R, Delno L. Catalogo nazionale degli effetti deformativi del suolo indotti da forti terremoti Rapporto Tecnico SSN/RT/97/04, Roma 1997. Also available at: http://www.serviziosismico.it/RT/ rt9704/frameset.html. [23] Esposito E, Gargiulo A, Iaccarino G, Pordo S. Distribuzione dei fenomeni franosi riattivati dai terremoti dellAppennino Meridionale. Censimento delle frane del terremoto del 1980. Proceedings of International Conference on Prevention of Hydrogeological Hazards: the Role of Scientic Research, Alba, Italy 1996 p. 40927.

You might also like