Simulating Earthquake Scenarios in the European

Project LESSLOSS: The Case of Lisbon

G. Zonno, A. Carvalho, G. Franceschina, A. Akinci, A. Campos Costa, E. Coelho,

G. Cultrera, F. Pacor, V. Pessina and M. Cocco

1 Introduction

This paper is a completion of a previous paper published in the proceeding of the1755 Lisbon Earthquake Conference and contain the results after the EuropeanProject LESSLOSS Risk Mitigation for Earthquakes and Landslides (Lessloss)regarding the evaluation of earthquake scenarios for the Metropolitan Area ofLisbon (MAL).

In the framework of the subproject 10, SP10 Earthquake disaster scenarioprediction and loss modelling, finite-fault seismological models have beenproposed to compute the earthquake scenarios for three urban areas: Istanbul(Turkey), Lisbon (Portugal) and Thessaloniki (Greece). The overall aim ofSP10 is to create a tool, based on state-of-the art modelling software, to providestrong quantified statements about the benefits and costs of a range of possiblemitigation actions, to support decision-making by city and regional authoritiesfor seismic risk mitigation strategies. First element of this analysis process is thedefinition of the input as hazard parameter (Fig. 1).

The generation of earthquake ground motion scenarios involves both theparticular choice of earthquake sources with associated fault rupture para-meters, and the ensuing ground motion field calculated by an appropriatenumerical tool, or empirically estimated, at a set of selected points within theurban area of interest.

For SP10 the aim was to define ground shaking hazard associated withparticular scenario events, defined as earthquakes with a given magnitude andlocation, taken to be the worst event which would take place with a given returnperiod, 50, 100 or 500 years. Defining that earthquake requires a close study ofthe capable faults, of the earthquake recurrence on those faults, and may alsorequire a de-aggregation analysis of the local effects from all faults capable ofgenerating damaging ground motions.

G. Zonno (*)Istituto Nazionale di Geofisica e Vulcanologia, sezione di Milano – Pavia, Milan, Italye-mail: zonno@mi.ingv.it

L.A. Mendes-Victor et al. (eds.), The 1755 Lisbon Earthquake: Revisited,Geotechnical, Geological, and Earthquake Engineering 7,DOI 10.1007/978-1-4020-8609-0_14, � Springer ScienceþBusiness Media B.V. 2009

233

Approaches to ground shaking hazard depend on the type and level of the

analysis to be undertaken. For a Level 1 assessment, the site groundmotion needs

to be defined in terms of a macroseismic intensity, derived from the given

magnitude and location of the scenario earthquake using intensity attenuation

relationships. This approach was not used in SP10. For a Level 2 assessment

specific parameters of surface ground motion are needed as inputs to the struc-

tural vulnerability, and these are derived from a two-step process. In the first step

a set of expected bedrock ground motions is determined (with as fine a grid of

points as needed for the structural analysis). This is obtained either by the use of

empirical groundmotion attenuation relationships, orwith greater sophistication

by the use of simulation techniques in which the mechanics of ground motion

transmission from source to site is simulated. The bedrock groundmotion is then

used to determine the expected surface ground motion based on an understand-

ing of the typical soil profile at each location. Again this can be done in different

ways, either using either dynamic soil-column analysis, or using standard soil

amplification coefficients applicable to the given soil type and depth.In the case of Lisbon, the site effects are evaluated by means of an equivalent

stochastic non-linear one-dimensional ground response analysis of stratified

soil profile using the LNECloss system (Sousa et al., 2004). Given a seismic

scenario (magnitude and location), the Bedrock Seismic Input and the Local

Soil Effect software modules of the system, allow computing the Power Spectral

Density Function (PSDF) of the strong ground motion at bedrock and at

surface level of any site, respectively. A data base is available, containing

information on stratified soil profile units for the region under analysis: in the

framework of the project conducted by the portuguese civil protection

Fig. 1 Processes to achieve the definition of ground shaking hazard

234 G. Zonno et al.

authority (SNPC), it was carried out a geological – geotechnical survey thatallowed the characterization of stratified soil profile units for MAL. Thecomputer algorithms developed and implemented inside LNECloss systemintroduce some major improvements to take into account site effects due tosoil dynamic amplification in rather efficient way (Serra and Caldeira, 1998).

2 Numerical Approaches

The first important distinction among recommended approaches for creatingground shaking scenarios is between simplified and advanced methods. Simpli-fied methods make use of empirical attenuation relations of ground motionparameters and of local geological data. Within LESSLOSS project, advancedmethods are extensively applied because of their capability of physically repre-senting the ground motion. Indeed, finite-fault effects and directivity couldassume a very important role. Moreover, the high resolution of ground motionscenario can match with the complexity of geotechnical characterization, vul-nerability data and exposure factors involved in the urban level lossesestimations.

Ground motion simulations were performed in the frequency band of engi-neering interest (0.5–20 Hz) by two numerical methods: a hybrid stochastic-deterministic approach (DSM-Deterministic-Stochastic Method; Pacor et al.,2005), used for all the investigated cases, and a non-stationary stochastic finitefault simulation method (RSSIM; Carvalho et al., 2004, 2007), applied in thecase of Lisbon. Both methods allow computing synthetic time series for directS-wave field at bedrock sites, and are suitable to generate shaking scenariosnear an extended fault whereby the direct S wave-field is generally dominantin amplitude with respect to the reflected and superficial phases. Differentrupture propagation models on the selected faults can be hypothesized and,even when input data regarding earthquake source, propagation medium, andsite characteristics are of a very schematic nature, the complexity of near-source ground motion can be adequately reproduced. Extended fault simula-tions performed with different earthquake rupture models generally produce ahigh variability in the ground motion, mainly dependent on the assumedposition of the hypocenter on the fault plane which controls the rupturedirectivity. During the project, a sensitive study was performed using differentinput parameters and different approaches in order to give the basic informa-tion to evaluate the range of uncertainty in seismic scenarios.

3 Definition of Scenario Earthquakes

Probabilistic seismic hazard analysis (PSHA) carries out integration over thecontributions to the hazard fromall the sourceswithin a region, for certain rangesof magnitude M and distance R. It provides the estimate of a strong-motion

Simulating Earthquake Scenarios in the European Project LESSLOSS 235

parameter with a specified confidence level, during a given exposure period.Moreover it is able to account for the uncertainties associated with estimationof the seismicity and of the attenuation characteristics of the region. This latterproperty is taken into account by an additional integration over a given numberof standard deviations, ", of the adopted ground motion attenuation law(Bazzurro and Cornell, 1999).

Because of its integrative nature, PSHA does not provide a representativeearthquake, in terms of magnitude and source-to-site distance, that can beused for engineering analyses and decision making. However this can beachieved through the de-aggregation of the PSHA results. The fundamentalobjective of this analysis is to compute the contribution to hazard, at aspecific site, of every possible source S= (M,R, ") considered in compositePSHA calculations. It allows to have a controlling earthquake, in terms ofmagnitude M and source-to site distance R, and to identify the location ofthe most probable source contributing to hazard at a well-defined site.Distance R can eventually be substituted by geographical coordinates andthe measure of the deviation of the ground motion from the predicted value,", can also be considered in the de-aggregation process. Information aboutthe controlling source can be used to generate a scenario earthquake, whichis the basis of the deterministic hazard assessment (Harmesen and Frankel,2001). Because deterministic scenarios are associated to representative earth-quakes, they can be performed by advanced ground motion simulationmethods allowing the reproduction of specific source effects like the extendedfault properties, the earthquake rupture propagation, or the asperities dis-tribution on the fault plane. It is noteworthy that if seismic source isaccounted by only magnitude and distance parameters, none of these effectsare considered.

4 The Case of Lisbon

TheMetropolitan Area of Lisbon (MAL) has been historically struck by scarce,though intense, earthquakes, such as the 1755 Lisbon earthquake estimated asmagnitude M=8.7. The earthquake scenarios were defined based on de-aggre-gation of the probabilistic seismic hazard analysis (PSHA) for different returnperiods. For Lisbon area, LNEC (Campos Costa et al., 2002, 2006) performedthe de-aggregation process in two stages. Both de-aggregation processesallowed independently to assess offshore and inland seismic sources (Fig. 2).Based on the revised deaggregation of PSHA it was possible to conclude thatseismic hazard in theMetropolitan Area of Lisbon (MAL) is dominated by longdistance scenarios (offshore sources), only for return periods greater than50 years. The two previously defined scenarios with magnitude M 5.7 andM 7.6, associated with inland and offshore sources have been substituted by onlyone offshore scenario with magnitude M=7.9, adopted for 475 yr RP.

236 G. Zonno et al.

Seismological studies dealing with the possible seismogenic structures associated

withthe1755Lisbonearthquakeaimedtodefineplausible faultparameters for this

source. Indeed, adirect displayonamapof locations dominating thehazard as far

as the knowledge of the most likely magnitude, allowed defining a specific set of

earthquakes that present the greatest hazard to the site. Themost recent studies on

tectonic structures that could be associatedwith the 1755Lisbon earthquakewere

employed to define plausible parameters for the simulation of offshore sources,

while seismotectonic studies on the Lower Tagus Valley were used to define the

simulation parameters of inland sources (Baptista et al., 2003; Cabral et al., 2004).Since August 2005, when the selection of 50 year and 500 year scenarios for

Lisbon Metropolitan Area (MAL) was accomplished, the work on seismic

action scenarios selection and on the assessment of seismic motion achieved

several progresses. These progresses resulted in adjustments in the magnitude

and location of the scenarios initially studied by INGV. More specifically,

seismic action scenarios were revised based on modal values derived from 3D

deaggregation analyses in M and (X, Y) (magnitude and coordinates of bin

source) (Sousa, 2006). However, for the 500 year return period, the revised

scenario is still located offshore, validating the INGV preliminarily conclusions

Fig. 2 Surface projections of the faults represented in Table 1. Left: position of the offshoresourcesMPTFM 7.6 andMPTFM 7.9 respect to theMAL (black box).Right: position of theinland sources LTVFM5.7 (4.7) and LTVFM4.4 (black box and black circle, respectively). Aspot of the LTVF M 5.7 source with the assumed nucleation points is shown on the top left.The 277 parishes where time synthetic time series were computed are also shown (black littlesquares), together with the 5 test parishes selected for the sensitivity analysis (sites S021, S268,S250, S137 and S174)

Simulating Earthquake Scenarios in the European Project LESSLOSS 237

about the spatial variability of the ground shaking in MAL. For the 50 yr RP,the deaggregation analysis indicates a short distance scenario corresponding toinland sources of magnitude M=4.4. The corresponding source was locatedbeneath the Lower Tagus Valley (LTV), on the east side of MAL, and fromprevious study allows to infer both geometrical parameters and focal mechan-ism for a plausible source (Cabral et al., 2004).

4.1 Finite-Fault Model Parameters

Following this revised methodology the dominant scenarios found for Lisboncity are presented in Fig. 3. For the earthquake scenarios, the earthquake sourcelocations, its geometrical, kinematical structure and parameters and the start-ing point of fault rupture are very important issues. Therefore, the finite-faultsimulations require specification of the fault-plane geometry (length and thewidth, etc.), source (stress drop) and the crustal properties of the region (geo-metrical spreading coefficient, quality factor, etc.) and the site-specific soilresponse information. The fault dimensions as a function ofmoment magnitudeare calculated using the empirical relations of Wells and Coppersmith (1994).The fault geometry and the source mechanisms are given in Table 1.

The model parameters calibration has been done with a dataset that includeshorizontal components of ground acceleration records (hard sites) obtain by thenational digital accelerometer network of Lisbon. (Carvalho et al., 2004). Theother parameters, including crustal properties and sources calibration para-meters to complete the finite fault simulation have been defined in previouspapers (Zonno et al., 2005; Carvalho et al., 2007).

In this paper, we will present a bedrock ground motion simulation of accel-eration (DSM-Deterministic-Stochastic Method; Pacor et al., 2005) for thescenario II: the inlandLTVF, with length of 8.4 km andwidth of 6.0 correspond-ing toM 5.7 (see Table 1). TheDSMmethod is a modern and sophisticated toolsare applied to predict ground motions capturing the complexity of near-sourcegroundmotion, evenwhen input data regarding earthquake source, propagationmedium, and site characteristics are of a very schematic nature.

We will present also a bedrock and surface ground motion simulation ofacceleration using a non-stationary stochastic finite fault simulation method(RSSIM; Carvalho et al., 2004) for the scenario V: the offshore MPTFfault, with length (along the strike) of 166 km and width (down-dip) of30 km corresponding to an event of M 7.9 (see Table 1). The RSSIMmethod is a finite-fault method capable of capturing the offshore attenua-tion law calibrating the data of from national digital accelerometer networkof Lisbon.

During the European Project both methods, DSM and RSSIM, have beenused allow computing synthetic ground motion for direct S-wave field at bed-rock sites and are suitable to generate shaking scenarios near an extended fault.

238 G. Zonno et al.

Nevertheless it is authors believe that FINSIM is worldwide known and a

reference tool for a comparison of new approaches. A sensitivity analysis

has been done performed using both adopted methodologies and FINSIM

(Carvalho et al., 2007).

Fig. 3 PGA maps obtained in MAL through extended fault simulations using the M=5.7inland source (LTVF). Nucleation points of the NE/SWunilateral (1, top left), bilateral (2, topright) and SW/NE unilateral (3, down right) rupture propagations are indicated together withthe surface projection of the fault (down, left). A rupture velocity VR= 2.7 km/s was adoptedfor all cases. Black dots indicate the position of the test sites used to analyse the variability ofresults

Simulating Earthquake Scenarios in the European Project LESSLOSS 239

5 Evaluation of Seismic Scenarios

5.1 Onshore Analysis

The ground motion was computed at bedrock level through extended fault

simulations, and then amplified by considering the site effects. The character-

ization of local soil effects was taken into account by computing the Power

Spectra Density Function at the surface level and considering, for each parish,

the non linear behaviour of the available stratified geotechnical soil profile

(Zonno et al., 2005; Carvalho et al., 2007).In Fig. 3 are shown the PGAmaps obtained inMAL throughDSM extended

fault simulations using the M 5.7 inland source (LTVF). Nucleation points of

the NE/SW unilateral (1), bilateral (2) and SW/NE unilateral (3) rupture. As

evidenced, the unilateral rupture propagation (cases 1 and 3), produces the

Table 1 Fault parameters of seismic sources used to perform the finite fault simulations in thecase study of Lisbon

Name ofseismicscenario

FaultName

Mw Returnperiod(year)

Length (km) �Width (km)

Dip(degree)

Strike(km)

Ztop(km)

Scenario I LTVF 4.4 50 1.4 � 2.3 55 220 0.5

Scenario II LTVF 4.7 75 2.2 � 2.8 55 220 2.0

Scenario III LTVF 5.7 200 8.4 � 6.0 55 220 0.5

Scenario IV MPTF 7.6 200 110 � 24 24 20 4.5

Scenario V MPTF 7.9 500 166 � 30 24 20 4.5

Fig. 4 DSM Time series Acceleration obtained at 2 test sites (S174 and S268) in the Metro-politan Area of Lisbon for the inland source of M=5.7 with different rupture models. Leftpanel: unilateral nucleations (NE/SW, green lines; SW/NE, blue lines) and bilateral nucleation(red lines) with velocity rupture of 2.7 km/s

240 G. Zonno et al.

highest variability of the ground motion. Sites experiencing forward directivity

(located South and East to the source in cases 1 and 3, respectively) are

characterized by large amplitudes (and short durations).In Fig. 4 are shown the DSMTime series Acceleration obtained at 2 test sites

(S174 and S268) in the Metropolitan Area of Lisbon for the inland source of

M=5.7 with different rupture models. Left panel: unilateral nucleations (NE/

SW, green lines; SW/NE, blue lines) and bilateral nucleation (red lines) with

velocity rupture of 2.7 km/s.The PGA variability due to uncertainties in both the nucleation point posi-

tion and the rupture velocity is generally present over the entire simulated

frequency band (f > 1 Hz) (Carvalho et al., 2007).

5.2 Offshore Analysis

Generally, ground shaking scenarios obtained through extended fault simula-

tions present high variability mainly due to directivity effects. In fact, ground

motion maps are very sensitive to the position of the nucleation point on the

fault plane and unfortunately this is probably the most uncertain seismolo-

gical information that we have to introduce in simulations. Otherwise,

extended fault simulations are necessary in order to properly reproduce the

earthquake ground motion from large and/or near sources. For the offshore

analysis we have used the RSSIM method. In the Fig. 5 are shown the PGA

maps obtained in the MAL for the 475 yr RP. Ground motion maps

presented here show higher values of PGA in comparison with the accelera-

tion maps obtained in the initial analysis (Zonno et al., 2005). Moreover,

Fig. 5 evidences that the nonlinear soil amplification is quite pronounced,

Fig. 5 PGA maps obtained at bedrock and surface level are compared. Ground motionamplification are evident especially in the South margin of Tagus River, where soft soils are

Simulating Earthquake Scenarios in the European Project LESSLOSS 241

particularly in the south margin of Tagus River where soft soil sites are put

in over bedrock (in this area peak ground accelerations range between 250

and 300 gal).

6 Conclusions

From the seismological point of view, many equally probable ground shaking

scenarios were simulated by varying the nucleation point over the fault plane,

the rupture propagation velocity and the slip distribution. The following cri-

teria were used in performing the final choice:

� The proposed scenarios are those more compatible with the standard ana-lysis of hazard in terms of peak values of strong ground motion parametersand spectral ordinates (empirical ground motion models and PSHA).

� In general, they should be suitable for application over the whole metropo-litan area. In a second step, as in the case of large extension of the city(Istanbul), the scenarios have been finalized on a small zone.

� According to the final application, either the mean or the worst scenariosmight be used. (for instance, in Lisbon area, the worst case was selected withpurpose of generating precautionary scenarios only). The choice depends onthe demand of the local administrators, the sensibility to the earthquakephenomena and, above all, the level of the seismicity of the area.

Acknowledgments This study was performed with the financial support of the EuropeanCommission through the LESSLOSS FP6 Integrated Project Risk Mitigation for Earth-quakes and Landslides, No.: GOCE-CT-2003-505488.

References and Further Readings

Baptista MA, Miranda JM, Chierici F, Zitellini N (2003) New study of the 1755 earthquakesource based on multi-channel seismic survey data and tsunami modelling, NaturalHazards and Earth System Sciences, 3, pp. 333–340

Bazzurro P, Cornell CA (1999) Disaggregation of Seismic Hazard, Bulletin of SeismologicalSociety of America, 89, pp. 501–520

Cabral J, Ribeiro P, Figueiredo P, Pimentel N, Martins A (2004) The Azambuja fault: Anactive structure located in an intraplate basin with significant seismicity (Lower TagusValley, Portugal), Journal of Seismology, 8, pp. 347–362

Campos Costa A, SousaML, Carvalho A, Bile Serra J, Carvalho EC (2002) Regional SeismicRisk Scenarios on hazard deaggregation, Proceedings, 12th European Conference onEarthquake Engineering, London, paper 470

Campos Costa A, Sousa ML, Carvalho A, Coelho E (2006) Seismic loss scenarios based onhazard de-aggregation. Application to the Metropolitan Region of Lisbon, Portugal,Chapter 21 in the book Assessing andManaging Earthquake Risk, Editors: C.S. Oliveira,A. Roca and X. Goula, publisher: Springer

242 G. Zonno et al.

Carvalho A, Campos Costa A, Oliveira CS (2004) A stochastic finite-fault modeling for the1755 Lisbon earthquake, 13th World Conference on Earthquake Engineering Vancouver,B.C., Canada August 1-6, 2004 Paper No. 2194

Carvalho A, Zonno G, Franceschina G, Bile Serra J, Campos Costa A (2007) Earthquakeshaking scenarios for the Metropolitan Area of Lisbon. Soil Dynamics and EarthquakeEngineering; doi: 10.1016/ j.soildyn.2007.07.009

Harmesen S, Frankel A (2001) Geographic desaggregation of seismic hazard in the UnitedStates, Bulletin of Seismological Society of America v1, 13–26

Lessloss (2007) http://www.lessloss.org/Pacor F, Cultrera G, Mendez A, Cocco M (2005) Finite fault modeling of strong ground

motion using a hybrid deterministic – stochastic method, Bulletin of Seismological Societyof America, 95, 225–240

Serra Bile J, Caldeira L (1998) Dynamic behaviour of geotechnical structures by means of astochastic approach, conference Applications of Computational Mechanics in Geotech-nical Engineering, Portugal

Sousa ML (2006) Risco sısmico em Portugal Continental, PhD thesis, IST, LisboaSousa ML, Campos Costa A, Carvalho A, Coelho E (2004) An automatic seismic scenario

loss methodology integrated on a geographic information system, 13th World Conferenceon Earthquake Engineering Vancouver, B.C., Canada August 1-6, 2004 Paper No. 2526

Wells DL, Coppersmith KJ (1994) New empirical relationships among magnitude, rupturelength, rupture width, rupture area, and surface displacement, Bulletin of SeismologicalSociety of America, 84, pp. 974–1002

Zonno G, Carvalho A, Franceschina G, Akinci A, Campos Costa A, Coelho E, Cultrera G,Pacor F, Pessina V, Cocco M (2005) Simulating earthquake scenarios using finite-faultmodel for the Metropolitan Area of Lisbon (MAL), 250th anniversary of the 1755 Lisbonearthquake, 1–4 November 2005, Portugal

Simulating Earthquake Scenarios in the European Project LESSLOSS 243