September 12, 2007 9:54 GeophysicalJournalInternational gji3571

Geophys. J. Int. (2007) doi: 10.1111/j.1365-246X.2007.03571.x

Simulations of strong ground motion in SW Iberia for the 1969 February 28 (M s = 8.0) and the 1755 November 1 (M ∼ 8.5) earthquakes – II. Strong ground motion simulations

,∗ , , , Rapha¨el Grandin,1 Jos´e Fernando Borges,1 2 Mourad Bezzeghoud,1 2 Bento Caldeira1 2 and Fernando Carrilho3 1Centro de Geof´ısica de Evora,´ Universidade de Evora,´ Evora,´ Portugal 2Departmento de F´ısica, Universidade de Evora,´ Evora,´ Portugal 3Instituto de Meteorologia, Lisbon, Portugal

Accepted 2007 July 30. Received 2007 July 30; in original form 2006 November 24

SUMMARY This is the second paper of a series of two concerning strong ground motion in SW Iberia due to earthquakes originating from the adjacent Atlantic area. The aim of this paper is to use the velocity model that was proposed and validated in the companion paper for seismic intensity modelling of the 1969 (M s = 8.0) and 1755 (M = 8.5–8.7) earthquakes. First, we propose a regression to convert simulated values of Peak Ground Velocity (PGV) into Modified Mercalli Intensity (MMI) in SW Iberia, and using this regression, we build synthetic isoseismal maps for a large (M s = 8.0) earthquake that occurred in 1969. Based on information on the seismic source provided by various authors, we show that the velocity model effectively reproduces macroseismic observations in the whole region. We also confirm that seismic intensity distribution is very sensitive to a small number of source parameters: rupture directivity, fault strike and fault dimensions. Then, we extrapolate the method to the case of the great (M = 8.5–8.7) 1755 earthquake, for a series of hypotheses recently proposed by three authors about the location of the epicentral region. The model involving a subduction- related rupture in the Gulf of C´adiz results in excessive ground motion in northern Morocco, suggesting that the source of the 1755 earthquake should be located further west. A rupture along the western coast of Portugal, compatible with an activation of the passive western Iberian GJI Seismology margin, would imply a relatively low average slip, which, alone, would could not account for the large tsunami observed in the whole northern Atlantic ocean. A seismic source located below the Gorringe Bank seems the most likely since it is more efficient in reproducing the distribution of high intensities in SW Iberia due to the 1755 earthquake. Key words: Iberian region, ocean–continent transition, strong ground motion, waveform modelling.

aries are not unequivocally defined by bathymetry, and where the 1 INTRODUCTION existence of an active subduction zone is not clearly supported by Most great tsunamigenic earthquakes are related to well-defined seismological evidence. Thus, despite the large size of the fault that interplate convergence zones: the circum-Pacific seismic belt, the is responsible for this extreme event (several hundred kilometres), Sunda arc, the Hellenic arc and the Antilles arc. An exception is the location of the epicentral region of the 1755 earthquake is poorly the massive earthquake that struck Iberia and Morocco on 1755 known; many hypotheses have been made by various authors for its November 1, was felt in a large part of Europe, and produced a source. Managing to discriminate between acceptable and unac- powerful tsunami that crossed the Atlantic Ocean. This earthquake ceptable scenarios for this earthquake is a fundamental step in the occurred along a passive margin, in a region where plate bound- improvement of seismic risk assessment. In this paper, we concentrate on the comparison of macroseismic observations with synthetic seismic intensity for a set of possible ∗Now at: Institut de Physique du Globe de Paris (IPGP), Laboratoire de source parameters. For this purpose, we have to take into account Tectonique et M´ecanique de la Lithosph`ere, 4 place Jussieu, 75252 Paris, the major heterogeneities that affect, at various scales, the crustal France. E-mail: [email protected] structure in the region, and that largely condition the distribution of

C 2007 The Authors 1 Journal compilation C 2007 RAS September 12, 2007 9:54 GeophysicalJournalInternational gji3571

2 Raphael¨ Grandin et al.

seismic intensities on land. A realistic velocity model, embedded quake source. This will enable us to determine which source fits best into a wave propagation code, is an appropriate way to study source the observed intensity distribution of the ‘Lisbon’ earthquake. parameters of the 1755 earthquake, provided there is sufficient data. The finite difference method is well adapted to this strategy, since it 2 SEISMO-TECTONIC SETTING is computationally efficient enough to enable a large number of tests to be run, and accurate in the low frequency range (f < 0.5 Hz), 2.1 Tectonic context which is the most significant for the study of the destructive effects of large earthquakes at a regional scale. The velocity model used The current tectonic regime at the boundary of the African and here was proposed and validated in the companion paper (Grandin Eurasian plates varies with longitude (Fig. 1), as a result of the et al. 2007). For this purpose, it is necessary to introduce a regres- rotation of Africa, with respect to Eurasia, around an Euler pole lo- sion between simulated values of Peak Ground Motion Parameter cated offshore Morocco, close to (20◦N; 20◦W) (Argus et al. 1989; (PGMP) and Modified Mercalli Intensity (MMI). The 1969 earth- DeMets et al. 1994). This boundary is commonly divided into three quake (M s = 8.0) constitutes the most obvious test to calibrate this sections (Buforn et al. 1988). To the west, between the Africa– regression: source parameters are reasonably well constrained, and Eurasia–North America triple junction (35◦W) to the eastern end the distribution of macroseismic observations is uniform and of good of the Terceira Ridge (24◦N), the regime is transtensional with an quality. extension rate of 4.4 mm yr–1 (Borges et al. 2007), and is respon- After a short introduction to the seismo-tectonic setting in SW sible for the active volcanism found in the Azores archipelago. In Iberia, we will establish this regression, and calibrate it using the the central section, the relative motion of the two plates seems to 1969 earthquake. Finally, we will extrapolate to the case of the 1755 be accommodated by a single right-lateral fault, the Gloria fault, earthquake, and test three hypothetical models for the 1755 earth- although significant seismic activity is observed in a broad region

Figure 1. Bathymetric map of the region of study. Inset shows the location of the region of study, and a sketch of the kinematic regime at plate boundaries. Isobaths are traced with a 1000 m interval below −1000 m, with a 200 m interval above (data provided by the British Oceanic Data Centre 2003). Dots show the location of earthquakes epicentres recorded from 1995 January to 2005 December (data provided by the Instituto de Meteorologia, Lisbon). Small light grey dots are earthquakes with 3 < M w < 4, medium dark grey dots are earthquakes with 4 < M w < 5, and large black dots are earthquakes with M w > 5. White triangles indicate the location of large historical earthquakes (the possible location of the source of the 1755 ‘Lisbon’ earthquake is discussed in the text). Major offshore active fault are also shown (from Hayward et al. 1999; Gr`acia et al. 2003a). Submarine canyons are identified by dashed lines.

C 2007 The Authors, GJI Journal compilation C 2007 RAS September 12, 2007 9:54 GeophysicalJournalInternational gji3571

Strong motion in SW Iberia: Part 2 3

off the fault (Lynnes & Ruff 1985). East of 16◦W, the bathymetric pattern of focal mechanisms, and the depth of earthquakes, between continuity of the Gloria fault cannot be followed, and the defini- 14 and 23 km, suggests that deformation is controlled by tectonic tion of the boundary is less clear. A transpressive tectonic regime activity deep in the basement (Fonseca & Long 1991). Seismic dominates, with a very low convergence rate of 4 mm yr–1 (Argus activity in the region of Evora´ is associated with unmapped struc- et al. 1989; McClusky et al. 2003) trending NW–NNW, consistent tures (Borges et al. 2001). In the Monchique range, the steady with the observed maximum horizontal stress direction (Ribeiro microtremor activity probably involves pre-existing tectonic fea- et al. 1996; Borges et al. 2001; Stich et al. 2003; Carrilho et al. tures located at depth (Dias 2001). Further south, the activity of the 2004). Deformation is distributed over an increasingly large area Portimao˜ Fault can be inferred from the N to S trending active zone that can reach a N–S width of 300 km near the continental mar- (Carrilho et al. 2004), consistent with average focal mechanisms gin of Iberia (Chen & Grimison 1989). In this section, seismicity is showing N–S strike-slip fault-planes (Bezzeghoud & Borges 2003), scattered, most events concentrating along a 100-km wide band that as suggested by Terrinha (1997). In the Guadalquivir foreland basin, trends ESE–WNW from 16◦Wto9◦W (Fig. 1). A progressive shift seismicity is low, and seems to be limited to the blind Subbetics of focal mechanisms, from strike-slip mechanisms in the west, to frontal thrust. predominant reverse faulting in the east, has also been reported by Offshore, the poor azimuthal coverage and the lack of stations at Buforn et al. (1988), and can be interpreted as an increasing plunge short distances from epicentres do not allow the depth distribution of the minimum compression axis. In the Gulf of Cadiz,´ seismicity of earthquakes, even close to the coast, to be constrained. However, is denser to the north, around the Guadalquivir Bank. East of the most events occurring between 14◦W and 7◦W are located at depths Strait of Gibraltar, in the Alboran´ domain and in Betic-Rif Arc, seis- between 5 and 25 km, and can reach 90 km (Buforn et al. 1988) in the micity (Calvert et al. 2000) and GPS data (Fadil et al. 2006) both oceanic domain that lies west of 10◦W. The widening of the diffuse suggest that active deformation is spread over several hundreds of plate boundary goes together with an increase of the maximum depth kilometres. In this study, we focus on the central section, which is of earthquakes: from 35 km at the eastern termination of the Gloria believed to be the host of the great 1755 earthquake. fault (Grimison & Chen 1986), to 60–90 km in the Gulf of Cadiz´ In the area, a series of topographic structures trending (Buforn et al. 2004), and 150 km east of the Strait of Gibraltar, WSW–ENE occur (Fig. 1). The Horseshoe scarp and the Marquesˆ de with occasional deep earthquakes (650 km) below the Alboran´ Sea Pombal scarp, parallel to the Sao˜ Vicente canyon, have experienced (Calvert et al. 2000). deformation since, at least, the Miocene (Zitellini et al. 2001; Gracia´ The largest instrumental earthquake, M s = 8.0 (USGS), occurred et al. 2003b; Terrinha et al. 2003). The most prominent seamount is on 1969 February 28 in the eastern Horseshoe Abyssal Plain. From Gorringe Bank, an uplifted thrust block of crustal and upper-mantle the study of depth phases (pP,sP and pwP), Grimison & Chen (1986) rocks that reaches 25 m below sea level and was emplaced during determined a focal depth of 50 km for a strong aftershock (mb = the Miocene (12 Ma) by plate breakage, differential loading and 5.5) that occurred on 1969 May 5. They also deduced that the old NW–SE shortening of the Jurassic (150 Ma) oceanic lithosphere Jurassic oceanic lithosphere (155 Ma) has a brittle thickness of ap- (Hayward et al. 1999). North of the Coral Patch Seamount and proximately 40–70 km, limited by the 600 ◦C isotherm, thus jus- Coral Patch Ridge, scarps affect the seafloor, with the same trend, tifying the intermediate depth seismicity observed in the area. In and large amplitude folds involve the sedimentary cover down to the Gulf of Cadiz,´ seismicity concentrated around the Guadalquivir the basement (Hayward et al. 1999). The transition between the lo- Bank, as well as the occurrence of the 1964 earthquake (Ud´ıas & calized and diffuse plate boundary may be due to a change in the Lopez-Arroyo´ 1970), with magnitude M w = 6.2 (Buforn et al. nature (and age) of the lithosphere (Jimenez-Munt´ et al. 2001). On 1988), suggest that this structure, that is clearly identified on seismic the other hand, by analogy with the Macquarie Ridge Complex at profiles and free-air gravity anomaly maps (Gracia´ et al. 2003a), is the Australia–Pacific Plate boundary, Massel et al. (2000) suggest active. that the proximity of the Europe–Africa pole of rotation may induce In the eastern Azores-Gibraltar Fault Zone, the largest events have significant changes in boundary conditions over small timescales (a thrust-type focal mechanisms, with a small strike-slip component. few million years) thus resulting in a transitional state of deforma- However, since the 1969 sequence, only one earthquake has reached 1 tion. The latter scenario is supported by the occurrence of unusually a magnitude of 5 (2003 July 29, M w = 5.3, discussed in detail in large oceanic earthquakes inside the area of scattered seismicity in the companion paper); hence, most events are too small to be of 1969 (M s = 8.0; Fukao 1973) and 1975 (M s = 7.9; Lynnes & Ruff regional significance. 1985). Moreover, focal mechanisms of these two events shared no obvious resemblance with bathymetric structures in the epicentral area. 2.3 Historical seismicity In the Alboran region, seismic activity is scattered, and strike-slip The Lower Tagus Valley was struck by several important his- mechanisms are consistent with a NW–SE compression, and hor- torical earthquakes: the 1531 January 26 earthquake (Justo & izontal extension in the NE–SW direction (Bezzeghoud & Buforn ∼ ◦ Salwa 1998), with magnitude M 6.5–7.0 (Justo & Salwa 1998; 1999). East of 1 E, as the least compressive axis shifts to a verti- Vilanova 2003) and the 1909 April 23 earthquake (Moreira 1985; cal direction, seismicity becomes more localized along the Algerian Teves-Costa et al. 1999), with magnitude M w = 6.0 (Teves-Costa coast, and the compressive stress regime is expressed by somewhat et al. 1999), are the best known cases. The source of the 1858 simpler thrust mechanisms (Bezzeghoud & Buforn 1999). November 11 Setubal´ earthquake, with magnitude M ∼ 7.1 (Vilanova 2003), might be a blind thrust fault, trending NNE–SSW, located below to the Setubal´ canyon, in the Sado Valley (Ribeiro 2.2 Instrumental seismicity

1 On land, seismicity is scattered, with a maximum focal depth of 15– We would like to acknowledge the occurrence of a M w = 6.1 earthquake 20 km (Fig. 1), and a large variety of focal mechanisms (Borges et al. on the 12th of February, 2007, while the present paper was under review. A 2001). In the Lusitanian Basin, microseismicity reveals a complex special issue of the EMSC Newsletter is dedicated to this event.

C 2007 The Authors, GJI Journal compilation C 2007 RAS September 12, 2007 9:54 GeophysicalJournalInternational gji3571

4 Raphael¨ Grandin et al.

2002), possibly connecting with an E–W fault on land (Moreira velocity is assumed to be constant over the fault plane. To prevent 1985). In the Algarve, the Loule´ Fault, which accommodated a sig- high-frequency noise from entering the radiated spectrum, and thus nificant part of the shortening across the Algarve basin, due to the generate a smooth source time function, one has to make sure that presence of halokinetic structures in its footwall, could be respon- rupture on each subfault is initiated before rupture on the previous sible for the 1856 December 1 earthquake (Moreira 1985; Terrinha adjacent subfault has stopped. We chose to use a Brune signal as 1997), with magnitude M ∼ 5.5 (Vilanova 2003). The source of the the elementary source time function for rupture on each subfault tsunamigenic 1722 December 27 Tavira earthquake, with magnitude (Brune 1970). The displacement rate u˙, as a function of time t,is M ∼ 7.0 (Vilanova 2003), might be located offshore the southern given by: Portuguese coast, possibly near the Guadalquivir bank area (Baptista t u˙(t) = e−t/τ , (1) et al. 2000). τ 2 In short, seismic activity is low on the continent, and thrust faults where τ is the characteristic time of subevent rupture, that sets both are often blind, with long recurrence times, leading to large uncer- the rise time and the duration of the rupture (Beresnev and Atkinson tainties. Offshore, activity is more significant, but epicentral loca- 1997). In the case of a finite source, rupture velocity is fixed to tions are, obviously, even more uncertain. The existence of major 2.5 km s–1, and the grid spacing is 1 km. By setting τ to 0.4 s, canyons, aligned with clearly mapped onshore faults, has led many rupture initiates on a subfault when the previous subfault has reached authors to associate some historical earthquakes with such struc- approximately 25 per cent of the total displacement. tures; however, these morphological observations can be misleading We additionally assume a uniform seismic moment distribution (Dinis 2006). In this context, it is not surprising that the location of over the fault plane. Thus, the slip is not uniform due to the variation the tsunamigenic 1755 November 1 earthquake, with magnitude M of rigidity modulus with depth inside the velocity model. However, ∼ 8.5–8.8 (Abe 1979; Johnston 1996), is still heavily debated (John- in the epicentral distance range considered here (d > 100 km), we ston 1996; Zitellini et al. 2001; Gutscher et al. 2002; Terrinha et al. have checked that this condition does not induce significant differ- 2003; Vilanova et al. 2003; Ribeiro et al. 2006). Similarly, historical ences from a uniform geometric moment distribution. Furthermore, information about the large tsunami generated by the 1761 March 31 we set the depth to the top of the fault so that coseismic rupture does earthquake, with magnitude 7.5 (Martins & Mendes V´ıctor 2001), not extend beyond the seismic basement. This prevents the occur- are too sparse to allow a determination of the epicentral region (Bap- rence of supershear rupture velocities in the shallow sedimentary tista et al. 2006). cover, but assumes that sediments are not involved in significant seismic wave generation. From a seismological point of view, this hypothesis is valid. Indeed, the total seismic moment is given by: 3 SEISMIC INTENSITY MODELLING M = (µb)dS, (2) 3.1 Numerical method and source implementation 0 To model the propagation of seismic waves in a 3-D media, we used where µ and b are, respectively, the rigidity and slip at an elementary the code E3D, an explicit 2-D/3-D elastic finite-difference wave surface dS, located on the fault plane . For large earthquakes, the propagation code (Larsen & Schultz 1995), based on the work of low rigidity of shallow sedimentary layers (two times lower than Madariaga (1976). The method and computational issues related average crustal rigidity), and their small thickness (a few per cent of to the finite-difference scheme are presented in details in the com- the thickness of the brittle Jurassic oceanic crust) ensure that their panion paper (Grandin et al. 2007). This numerical method was contribution to the total seismic excitation is low. This is valid for the successfully applied in 3-D by a large number of authors for gen- case of a uniform coseismic slip, or, inversely, of a uniform seismic erating synthetic seismograms in the USA (e.g. Olsen & Schuster moment distribution. 1992; Frankel & Vidale 1992; Olsen & Archuleta 1996; Larsen et al. Todetermine the stability of our results, we have performed a large 1997; Wald & Graves 1998; Stidham et al. 1999; Dreger et al. 2001; number of simulations to evaluate the importance of the parameters Pitarka et al. 2004; Hartzell et al. 2006) and in Japan (e.g. Pitarka we have arbitrarily fixed. In the case of the extended rupture, we et al. 1994, 1996, 1998; Sato et al. 1999, Satoh et al. 2001; Kagawa have verified that dividing or multiplying τ by a factor of 2 does et al. 2004). However, the size of the region studied in the present not affect significantly the radiated wavefield. Modifying the value v paper (∼400 × 400 km) is exceptionally large, and our knowledge of r within reasonable bounds only slightly affects the amplitude of the crustal structure in the region of interest is more limited than of ground shaking, by enhancing or weakening the directivity ef- in the above studies. Therefore, we expect only the broad features fect. Finally, we have checked that a complex geometry, involving a of the wavefield to be reproduced (Grandin et al. 2007). We recall slow, shallow rupture on a plane extending inside the sedimentary that simulations are only reliable at low frequency, the maximum cover does not affect the results significantly, while the additional frequency being proportional to the minimum velocity in the grid, number of parameters, and thus the risk of generating spurious re- and to the inverse of grid spacing. In this paper, a grid spacing of sults, increases dramatically. On the other hand, variations of focal 1 km is used, thus yielding a maximum frequency of 0.3 Hz. parameters, fault plane geometry and rupture directivity have strong In the case of the two larger earthquakes studied here (7.5 < effects on the resulting radiated wavefield. M w < 8.5), the finiteness of the dimension of the fault and of the In the rest of the paper, the computational domain has a grid spac- duration of rupture cannot be ignored. Following the source im- ing of 1 km, and extends to a depth of 70 km. Anelastic attenuation plementation scheme of E3D, we model these extended sources by is included (Table 1 in companion paper). superimposing a large number of point sources over a rectangular fault plane that shares the same strike and dip as the individual 3.2 Regression of Peak Ground Velocity versus Modified subevents. The kinematics of rupture, namely rupture propagation Mercalli Intensity direction, is simulated by triggering rupture on each subfault at the right time: rupture can nucleate at a certain location on the fault, Four sources of information exist for the seismicity that occurs off- and then propagate radially until the fault edge is reached; rupture shore SW Iberia. First, seismological data is available for recent

C 2007 The Authors, GJI Journal compilation C 2007 RAS September 12, 2007 9:54 GeophysicalJournalInternational gji3571

Strong motion in SW Iberia: Part 2 5

Table 1. Regressions of Peak Ground Velocity (PGV) measured in m s–1 versus Modified Mercalli Intensity (MMI) discussed here. Parameters a and b are defined in eq. (2). Reference ab Region of study Frequency range (Hz) Intensity range (MMI) Trifunac & Brady (1975) 4.00 10.52 Western United States 0.07–25.0 IV–X Wald et al. (1999) 3.47 9.29 Western United States – V–IX Wu et al. (2003) 2.35 8.01 Taiwan – III–IX Kaka & Atkinson (2004) 1.79 9.33 Northeastern America – II–VIII This study 3.50 10.50 Southwestern Iberia 0.10–0.50 IV–X

instrumental earthquakes. Second, macroseismic data has been re- ues of velocity and acceleration, versus MMI, in a study focused on trieved from inquiries carried out immediately after major events, the Western United States. Other authors provided alternative rela- or a posteriori, by the study of historical information. Third, several tions for the Western United States (Wald et al. 1999), Taiwan (Wu recent studies aimed at imaging the relation between the complex et al. 2003) or Eastern North America (Kaka & Atkinson 2004). bathymetry and the structure of the sediment cover, using geophysi- The relations referenced here are of the form: cal techniques, to locate active tectonic structures. Fourth, near-field MMI = a log(PGMP) + b, (3) tsunami data has been deduced from available tide-gauges and his- torical observations. Additional observations have been made in the where PGMP is the Peak Ground Motion Parameter (displacement, case of the great 1755 earthquake (far-field tsunami, near and far- velocity and acceleration). field hydrological effects, landslides, seiches and seismic activity The value of a mostly affects the geometric attenuation of inten- triggering), but, since they are not available for any other earth- sity,while b is a constant. Wald et al. (1999) and Wu et al. (2003) sug- quake, this type of information will not be discussed here. gest that, for high intensities (>VII), Peak Ground Velocity (PGV) In the area, all four types of information are reliably available for correlates better with MMI than Peak Ground Displacement (PGD) only one major earthquake, namely the 1969 February 28 Horse- or Peak Ground Acceleration (PGA). The values of the parame- shoe Abyssal Plain earthquake. More recent earthquakes have been ters a and b used by a series of authors are presented in Table 1, much smaller, and, as a consequence, were barely felt on land, and with PGV expressed in m s−1. Since they are clearly region-specific, probably did not produce surface ruptures, nor a detectable tsunami. these relations should be used with caution when applied to synthetic Older earthquakes of comparable magnitude are badly constrained ground motion in SW Iberia: they concern different tectonic and ge- from a seismological point of view (excluding the 1964 earthquake, ological environments, and were obtained using higher frequencies which occurred in the eastern Gulf of Cadiz)´ and macroseismic ob- (>1 Hz) than frequencies available in this study (<0.3 Hz). Thus, servations, as well as tsunami observations, are less reliable. This they provide bounds to the regression that may apply to the specific is obviously the case of historical earthquakes. Our major objec- case of SW Iberia. tive is the computation of synthetic seismic intensity maps, in order Our method to determine a regression of the form of eq. (3) is to assess seismic risk associated with sources located offshore SW the following. As a preliminary step, we performed a large number Iberia. In this context, correctly reproducing ground motion dis- of simulations for the 1969 event, and tested the dependence of the tribution for the 1969 earthquake is the first step towards the pre- results on variations of the source parameters in order to evaluate diction of isoseismal maps due to any hypothetical source (Levret whether small to moderate uncertainties on fault parameters can re- 1991; Johnston 1996; Paula & Oliveira 1996): the 1969 event is the sult in large uncertainties of the final result. We concluded that, at only event of comparable magnitude for which source parameters the regional scale, and in the case of an extended fault, the distribu- are well constrained, and macroseismic information is reliable and tion of PGV is, at the first order, mainly controlled by the magnitude dense enough to ensure a safe regression. However, we must bear in of the earthquake and source location (this result is not surprising, mind that seismic intensity heavily depend on local effects, which and will enable us, in the final stage of this study, to discriminate be- explains the large standard deviation of observations made in the tween realistic and unrealistic locations for the epicentre of the 1755 field; such effects are beyond our level of resolution. Therefore, we earthquake). Rupture directivity (controlled by the rupture initiation will only consider the broad features of the isoseismals, deduced location), strike direction and fault dimensions are critical factors by comparing the ‘smoothed’ intensity values provided by various for the azimuthal distribution of the maximum amplitude oscilla- authors. tions. In contrast, the computation is little sensitive to other source Laws relating the intensity of an earthquake (I 0) to epicentral parameters, such as the dip, the rake, rupture velocity and depth to distance in SW Iberia are available from a series of studies (Levret the top of the fault or the duration and shape of the source time 1991; Johnston 1996; Baptista 1998; Lopez´ Casado et al. 2000; function. Mart´ınez-Solares & Lopez-Arroyo´ 2004); these studies suggest that, Thus, we concluded that the problem can be partitioned: in a first in the area, attenuation is very low for large magnitude earthquakes. step, we select an average value for source parameters of the 1969 We use a different approach, where we relate the level of shaking, earthquake, which enables us to do the followings. obtained by a quantitative simulation of wave propagation through 1. Fix the value of b , which controls the decay of PGV with the crust and upper mantle, to the macroseismic intensity. Thus, in epicentral distance. addition to geometric attenuation, the effects of source finiteness, 2. Fix the value of a , which controls the average value of PGV. directivity, and heterogeneities of the propagation medium are also included. Because of the linearity of the equations used for the computation The observed correlation between the level of seismic intensity of ground motion in this study,the simulated seismic moment has the and the peak value of ground motion has been described at length same scaling effect as parameter b, and there is a trade-off between by Trifunac & Brady (1975), who gave regressions of the peak val- the two parameters that can only be resolved by the study of at least

C 2007 The Authors, GJI Journal compilation C 2007 RAS September 12, 2007 9:54 GeophysicalJournalInternational gji3571

6 Raphael¨ Grandin et al.

two earthquakes. We find that a relation similar to that of Trifunac (Fukao 1973), but, due to the 10-fold difference of seismic moment & Brady (1975) gives good results (Table 1). calculations, Buforn et al. (1988) and Grimison & Chen (1988) dis- In a second step, we discuss the effects of the uncertainties on agree on the value of fault width, with 20 and 50 km, respectively. source parameters, which have second order effects on the distribu- Fukao (1973), and Grimison & Chen (1988) both suggested that the tion of intensities. We especially focus on source directivity effects rupture propagation was directed towards the NE. The computed and basin amplification. This step is necessary to the validation of epicentral depths are 22 km (Lopez-Arroyo´ & Ud´ıas 1972), 33 km the values of a and b, since source parameters estimates of the 1969 (Fukao 1973) and 32 km (Grimison & Chen 1988). earthquake made by various authors are somewhat contradictory Macroseismic data for the 1969 earthquake in Iberia has been (see discussion below). compiled and published by Lopez-Arroyo´ & Ud´ıas (1972), Mendes (1974), Mart´ınez-Solares et al. (1979), Paula & Oliveira (1996) and Mart´ınez-Solares & Lopez-Arroyo´ (2004). Using data provided by 4 THE 1969 EARTHQUAKE Paula & Oliveira (1996) for Portugal, and by Mezcua (1982) for Spain, we built a new map of the observed seismic intensities for 4.1 Seismic source characterization the 1969 earthquake (Fig. 2). Authors cited above report the largest The focal mechanism and seismic moment of the 1969 February intensities (VII–VIII) along the southern coast of Portugal, and un- 28 earthquake has been studied by (among others) Lopez-Arroyo´ & derline the occurrence of moderate ground motion amplification in Ud´ıas (1972), McKenzie (1972), Fukao (1973), Buforn et al. (1988) the Guadalquivir Basin (V–VI). A N–S elongated region, located and Grimison & Chen (1988). They concluded that the mechanism on the western coast of Portugal, from the Algarve to the Setubal´ is a thrust, with a strike of 40–70◦, and a small left-lateral compo- area, also had intensity VI. nent. Grimison & Chen (1988) suggested that the earthquake was No anomalously high intensities are explicitly reported in the composed of two subevents, the second one having an almost pure Lower Tagus Basin, which has lead several authors to claim that strike-slip mechanism. Lopez-Arroyo´ & Ud´ıas (1972) suggested no site amplification was observed during the 1969 earthquake in that the SE dipping nodal plane should be the fault plane, but Fukao this area (Vilanova et al. 2003). Therefore, one would favour sce- (1973), and subsequent publications by other authors, favoured the narios that produce the same level of shaking in the Neogene and NW dipping plane. Lopez-Arroyo´ & Ud´ıas (1972) calculated a seis- Mesozoic basins. In the rest of the Iberian peninsula, a simpler ge- mic moment of 6.3 × 1019 Nm, while Fukao (1973) and Grimison ometric attenuation seems to have occurred. However, as can be & Chen (1988) computed a seismic moment nearly 10 times larger, seen in Fig. 2, no data points are available in a wide area located of 6.0 × 1020 Nm and 8.0 × 1020 Nm, respectively. Fault length NE of the Tagus Lagoon. The whole zone is covered by thick Neo- was estimated to be 85 km (Lopez-Arroyo´ & Ud´ıas 1972) or 80 km gene alluvial deposits (Carvalho et al. 2006). This ‘hole’ in the data

Figure 2. Compiled isoseismal map of the 1969 February 28 earthquake, using data provided by Paula & Oliveira (1996) for Portugal (the size and colour of dots varies with seismic intensity), and by Mezcua (1982) for Spain. The location of the epicentre, and the approximate fault dimensions and configuration are also shown.

C 2007 The Authors, GJI Journal compilation C 2007 RAS September 12, 2007 9:54 GeophysicalJournalInternational gji3571

Strong motion in SW Iberia: Part 2 7

is surprising, and will necessitate further investigations, but does not allow us to discuss whether amplification, or attenuation, oc-

curred in the central part of the Neogene Lower Tagus Basin due to ), case 3 the incoming long-wavelength seismic waves excited by the 1969 ◦ earthquake (M = 8.0). Similarly, observations made by Pereira de s 241/64.0 245/83.0 Sousa (1919) in the case of the 1755 earthquake (M ∼ 8.5) are bewildering: in the Neogene Lower Tagus Basin, large buildings (parish churches) located at the same epicentral distance and above comparable sedimentary column thicknesses suffered significantly different levels of shaking (for instance, IX–X in Benavente versus a b ), case 2 Strike/rake (

VI in Alcochete). However, in the case of the 1969 earthquake, we ◦

must admit that, should amplification have occurred, it must have (1988) and Fukao (1973), respectively. been very localized. Heinrich et al. (1994), Guesmia et al. (1996), Miranda et al. et al. (1996) and Gjevick et al. (1997) conducted studies of the tsunami associated with the 1969 event, and concluded that Fukao’s (1973) different, and we test three values of the strike. The source parameters allow the tsunami observations (traveltimes and of Buforn

amplitudes) in southern Iberia to be reproduced relatively well. How- ), case 1 Strike/rake ( ◦ ever, the synthetic mareograms do not succeed in reproducing the traveltime and polarity of the primary wave observed in Casablanca, Morocco. Gjevick et al. (1997) suggested that a slightly rotated strike of the source (from 55◦ to 73◦) provides a better agreement with the observations. These conclusions support the hypothesis of a complex 1969 tsunami source.

4.2 Results and discussion The main disagreement between authors about the 1969 earthquake lies in the determination of the seismic moment, and therefore, of

fault width and coseismic slip. We chose to test the two sources Pa) proposed by Fukao (1973), and Buforn et al. (1988), with fault 10 10

widths of 20 and 50 km, respectively. We also tested a third source, × with fault dimensions and focal parameters fixed at intermediary values. For each of the three sources, the fault plane was rotated by 10◦, clockwise and anticlockwise, in order to account for the uncertainties on the strike. A total of nine tests were performed (Table 2). Rupture is nucleated SW of the fault centre (at a distance of 20 km along strike), at two-thirds of the fault width (distance mea- sured along dip, from fault top), and propagates radially at a velocity –1

of 2.5 km s . We have checked that a bilateral rupture gives sim- Nm) magnitude rigidity (

ilar results, with slightly lower intensities, while a rupture directed 20 towards SW results in small intensities on land (maximum of VI), 10 × incompatible with observations (maximum of VIII). Due to the difficulty in estimating the depth to the top of the fault,

this parameter is not assessed by any author. In our study, we noticed moment ( that the results were little dependent on this depth, provided that the ) Seismic Moment Average Average slip (m) Strike/rake ( top of the fault is located below the seismic basement. Therefore, ◦ we fixed this parameter to 8 km. The middle of the top of the fault is located at (36.0◦N; –10.45◦E). The value of the seismic moment was adjusted for each hypothetical source in order to obtain a similar level of synthetic seismic intensity for all cases. The resulting synthetic intensity maps are shown in Fig. 3. The tested values of the seismic moments are 3.6 × 1020, 6.0 × 1020 and 1.0 × 1021 Nm for models A, B and C, respectively (Table 2). These values are in good agreement with the magnitudes provided

by Fukao (1973) and Grimison & Chen (1988). On the other hand, (1988). we tested a value of the seismic moment of 6.3 × 1019 Nm, follow- et al. ing the proposition of Lopez-Arroyo´ & Ud´ıas (1972), and we ob- Source parameters used for the computation of synthetic isoseismal maps of the 1969 earthquake. For each scenario, fault dimensions and dip angle are tained much lower intensities (except by choosing an unrealistically Buforn

high value of parameter a), suggesting that the seismic moment is Fukao (1988). a b BC 82.5 80.0 35.0 50.0 49.5 52.0 6.0 10.0 7.8 8.0 5.0 6.5 4.16 3.85 223/53.5 225/63.0 233/63.5 235/73.0 243/73.5 Table 2. Scenario Length (km)A With (km) Dip ( 85.0 20.0 47.0 3.6 7.7 4.0 5.29 2214/44.0 231/54.0 significantly underestimated. Lopez-Arroyo´ & Ud´ıas (1972) had value of the rake is adjusted so as to maintain approximately the direction of the axis of maximum compression. Source A2 and C2 correspond to the models

C 2007 The Authors, GJI Journal compilation C 2007 RAS September 12, 2007 9:54 GeophysicalJournalInternational gji3571

8 Raphael¨ Grandin et al.

Figure 3. Synthetic isoseismal maps of the 1969 February 28 earthquake, using source parameters summed in Table 2 (see text for discussion). The focal mechanism of each source is shown.

already expressed doubts about the compatibility of this value with Model A2, extrapolated from Lopez-Arroyo´ & Ud´ıas (1972) and the high stress drop of the 1969 earthquake, inferred from the Buforn et al. (1988), reproduces well the observations, with the high- surprisingly small number of aftershocks. est intensities in the Algarve (VIII) and a large ‘L-shaped’ region All 9 models succeed in reproducing the ‘L-shaped’ macro- of lower intensities (VI) extending from the western Guadalquivir seismic field, characterized by the same level of intensities along Basin to the Lisbon Area. The rotated mechanism (model A3) yields the southwestern coast of Portugal and Spain, with a decrease similar features. On the other hand, model C2, which corresponds to when moving inland. The largest intensities are found either in the solution proposed by Fukao (1973) gives much higher intensities the eastern Algarve (southern Portugal), or in the Lower Tagus in the Lower Tagus Valley Basin (IX) than in the Algarve (VII). A Valley Basin. For model A, because of the smaller fault width, rotation of the fault plane (model C3) does not improve this result directivity is stronger, and small variations of the fault strike in- significantly. However, a reduction of the fault width (B2) increases duce appreciable differences in the distribution of intensities. On the directivity effect, and re-establishes a balance between intensi- the other hand, model C is less sensitive to such variations. In ties calculated in the Algarve, in the Guadalquivir Basin and the all cases, solutions with a fault plane rotated anticlockwise give Lower Tagus Valley Basin. larger intensities in the Lower Tagus Valley Basin, and thus can be For each model, we can calculate the average displacement using excluded. the eq. (2), with a uniform rigidity modulus. Given the high rigidity

C 2007 The Authors, GJI Journal compilation C 2007 RAS September 12, 2007 9:54 GeophysicalJournalInternational gji3571

Strong motion in SW Iberia: Part 2 9

modulus at the fault location (Table 2), the calculated average dis- in the range of frequencies considered here. Second, synthetic in- placements (∼4 m) for models B and C are in good agreement with tensities in the Lusitanian Basin are much smaller that observed empirical regression moment magnitude versus average displace- intensities. This can be improved by a better modelling of the geo- ment (Wells & Coppersmith 1994). On the other hand, for models logical structure in the area, especially at shallow depth. A and B, the deduced value of the moment magnitude (M w = 7.7– 7.8) differs slightly from the surface wave magnitude (M = 8.0, s 5 THE 1755 EARTHQUAKE according to USGS), confirming the high stress drop character of the 1969 earthquake (Kanamori & Anderson 1975). We conclude 5.1 Source characterization that our preferred model is model B. In this part, we confirmed that the distortion of isoseismals, due The 1755 November 1 earthquake was the strongest earthquake to the variety of geological settings found in the region, could be ever reported in Europe, and was extremely destructive throughout correctly matched by using a relatively simple 3-D velocity model Portugal (Pereira de Sousa 1919), Spain (Mart´ınez-Solares et al. and appropriate source parameters. From the macroseismic study 1979) and Morocco (Levret 1991) (Fig. 4); the shock was even of the 1969 earthquake, we can draw three conclusions. First, a felt in Northern Germany, the Azores and the Cape Verde Islands clockwise rotation of the mechanisms can be excluded, because it (Reid 1914). The large size of the earthquake is further attested by would ‘transfer’ radiated energy from the Guadalquivir region to the observation of seiches in southern England, in Holland and as the Lisbon area, and generate significantly higher intensities in the far as Finland (Reid 1914). The large tsunami-waves generated by Lower Tagus Valley Basin. Second, the seismic moment calculated the earthquake also provoked extensive damage along the coasts of by Lopez-Arroyo´ & Ud´ıas (1972) is significantly lower than the Portugal, southern Spain and Morocco, and were even detected in seismic moment deduced from this study, which is, on the other the Lesser Antilles and southwestern England. Extensive geological hand, in good agreement with the value provided by Fukao (1973). evidence of tsunami deposits associated with the 1755 earthquakes Third, a fault width of 35 km gives better results than the fault widths have been reported in southern Portugal (Andrade 1992; Dawson of 20 and 50 km proposed by Buforn et al. (1988) and Fukao (1973), et al. 1995), southern England (Foster et al. 1993), on the Spanish respectively. coast (Luque et al. 2001), at the Tagus Estuary mouth (Abrantes However, two main discrepancies between observed and synthetic et al. 2005) and in the Cape da Roca-Cascais area west of Lisbon isoseismal maps persist. First, intensities at the NE border of the (Scheffers and Kelletat 2005). model are higher in the synthetics. This could be explained by the The problem of epicentral location has been addressed by various poor correlation between the lowest intensities and computed PGV early studies (Reid 1914; Pereira de Sousa 1919), and, since the

Figure 4. Compiled isoseismal map of the 1755 November 1 earthquake, using data provided by Pereira de Sousa (1919) and Moreira (1984) for Portugal, Mart´ınez-Solares et al. (1979) for Spain, and Levret (1991) for Morocco. The three tested fault models at Gorringe Bank (Johnston 1996), along the Marquesˆ de Pombal – Pereira de Sousa Fault Zone (Zitellini et al. 2001; Terrinha et al. 2003) and in the Gulf of Cadiz´ (Gutscher et al. 2002; Gutscher et al. 2006; Thiebot & Gutscher 2006) are shown (Table 3).

C 2007 The Authors, GJI Journal compilation C 2007 RAS September 12, 2007 9:54 GeophysicalJournalInternational gji3571

10 Raphael¨ Grandin et al.

beginning of the instrumental period, a consensus attributed the Platt & Houseman (2003) and Fonseca (2005) objected that the ab- origin of the earthquake to a structure located between the Gorringe sence of a Wadati–Benioff zone does not allow a straightforward Bank and the Coral Patch Ridge (Machado 1966; Moreira 1985; confirmation of the activity of a hypothetic subduction plane, nei- Johnston 1996). ther in the Gulf of Cadiz,´ nor along the western Portuguese margin. In the most recent hypothesis among several scenarios, Johnston Another objection comes from preliminary GPS data, which sug- (1996) suggested that a NE–SW trending fault, possibly outcropping gests that the western Betic-Rif belt and NW Morocco behave rigidly at the base of the NW flank of Gorringe Bank, could be responsible (Fadil et al. 2006; Stich et al. 2006), thus bringing into question the for the 1755 earthquake. Using isoseismal regression analysis, he connection of the southern ramp of the presumed subduction plane, deduced a seismic moment of 1.26 × 1022 Nm, and a fault area which lies in a region of low seismic activity (Fig. 1), with unmapped of 16 000 km2 compatible with an average slip of 12 m. The SE active structures onshore. dipping fault plane suggests that the Eurasian oceanic plate is being Another complex scenario has been proposed by Vilanova et al. overridden by the African oceanic plate, at least a shallow depth. (2003), who argued that the multiple shocks felt all over Iberia, and However, the recent re-evaluation of historical information avail- the high seismic intensities observed in the Lisbon area demonstrate able for the 1755 earthquake and tsunami (Baptista et al. 1995, the presence of a secondary source in the Lower Tagus Valley. The 1998a; Vilanova et al. 2003), as well as results from new seismic rupture of the inferred NNE–SSW striking, WNW dipping thrust surveys offshore SW Iberia (Zitellini et al. 2001; Gutscher et al. plane outcropping on the western bank of the Tagus river, at the 2002), has led a series of authors to propose alternative scenarios. frontier between the Mesozoic and Neogene basins (Fonseca et al. Baptista et al. (1995, 1998a) presented a review of historical 2000, 2001), would be due to stress triggering. Such a scenario would records of tsunami parameters that have been used as the basis for imply that strong motion in the Lisbon area is dominated by the a series of studies that aimed at determining the epicentral area existence of major local sources, rather than offshore sources, as the location by backward ray tracing of the tsunami (Baptista et al. epicentral location(s) of the 1755 earthquake(s) is (are) critical to our 1996, 1998b). A source located in Gorringe Bank, with a fault length understanding of the seismic hazard in the whole SW Iberia region of 120 km and a dip directed towards NE, was tested by Baptista (Montilla et al. 2002). However, as pointed by Matias et al. (2005), et al. (1996, 1998b), and led to simulated traveltimes systematically should clear evidence of the validity of this hypothesis be found, longer than observations in Iberia; the authors concluded that ‘the it is likely that the available historical information will never allow Gorringe Bank is a very unlikely source for the 1755 event’. Several a magnitude to be assigned to this secondary event. Consequently, subsequent publications aimed at testing the coherence of various we are unable to discuss the scenario proposed by Vilanova et al. hypothetical sources in terms of tsunami traveltimes (Baptista et al. (2003) in this study. 1996, 1998b, 2003; Gutscher et al. 2006). These studies concluded Seismic intensity modelling was performed by various authors, that a source compatible with an incipient subduction along the using probabilistic and stochastic methods (Mendes-Victor et al. western Portuguese margin (e.g. Ribeiro 2002) provides a better fit 1999; Baptista et al. 2003), or site-effect functions and attenuation between inferred and modelled tsunami traveltimes than a source relations (Gutscher et al. 2006). These methods suffer important located at Gorringe Banks. limitations, because they do not take into account physical con- At the same time, Zitellini et al. (2001) reported active thrusting siderations attached to the complexity of the seismic source, on at the base of a large submarine hill dubbed ‘Marquesˆ de Pombal’, one hand, and the propagation medium, on the other hand. Hence, located 80 km west of Cape Sao˜ Vicente. This inferred thrust zone it is not surprising that the conclusions of the studies cited above is associated with seismic activity, may have a listric geometry, and are contradictory: Mendes-Victor et al. (1999) back the hypoth- might be connected to a backthrust fault via a decollement´ surface esis of the two NNW–SSE striking sources (‘N160N135’) de- located at a depth of 16–18 km. However, the surface of the inferred scribed by Baptista et al. (1996, 1998b); Baptista et al. (2003) 2 fault is too small (7 000 km ) to generate a M w ∼ 8.5 earthquake. favour a ‘L-shaped’ rupture (‘MPTF/GB’) on two segments located Terrinha et al. (2003) subsequently proposed that the Marquesˆ de below Marquesˆ de Pombal and Guadalquivir Bank, respectively; Pombal Thrust should be extended further north, where, instead Gutscher et al. (2006) suggest that a source located at (−8.5◦W; of cutting directly to the surface, it would connect to a shallower 36.0◦N), possibly an asperity at the northwestern corner of the decollement´ level (depth of ∼8–10 km), and finally reach the base of subduction plane proposed by Gutscher et al. (2002), would ex- the continental slope at the eastern border of the TagusAbyssal Plain. plain most of the observed macroseismic intensities in Iberia and The complex geometry of the second segment would be responsi- Morocco. ble for the inversion of the normal faults of the Pereira de Sousa Generally, all authors cited above agree on a primary source lo- Fault Zone, and would yield a much larger surface (19 000 km2), cated offshore SW Portugal. compatible with the magnitudes of the earthquake and associ- ated tsunami. Additionally, Gracia´ et al. (2003b) showed that half of the NNE–SSW striking scarp created by this ESE dip- 5.2 Results and discussion ping fault, is buried below submarine landslide deposits, suggest- ing that slope failure processes could have contributed to tsunami In this study, we test, by forward modelling, three previously pub- generation. lished sources for the 1755 earthquake that can be considered as end- On the other hand, based on interpretation of other seismic pro- members of the set of proposed offshore seismic sources (Fig. 4). files, and gravimetry and thermal modelling, Gutscher et al. (2002; First (model G), a steep fault, dipping towards SE, and extending 2006) and Thiebot & Gutscher (2006) propose that the olistostrome down to a depth of 50 km below Gorringe Bank is tested (John- units of the western Gulf of Cadiz´ were accumulated as a result of ston 1996). A second source (model M) is implemented, consist- the ongoing westward displacement of a shallow east dipping fault ing of listric faults rooted in the middle crust, below the Marquesˆ plane, in a suduction context. This fault plane, currently locked, de Pombal – Pereira de Sousa fault zone, and distant of ∼50 km would have been responsible for the 1755 earthquake, at least as from the SW coast of Portugal, along the Alentejo margin (Zitellini a secondary source associated with a primary source further west. et al. 2001; Terrinha et al. 2003). Finally (model C), we simulate a

C 2007 The Authors, GJI Journal compilation C 2007 RAS September 12, 2007 9:54 GeophysicalJournalInternational gji3571

Strong motion in SW Iberia: Part 2 11

Table 3. Source parameters used for the computation of synthetic isoseismal maps of the 1755 earthquake. Latitude, longitude and depth are specified at the centre of the top of the fault. Gulf of Cadiz: source C Seismic moment (×1020 Nm) Rigidity, (×1010 Pa) Fault area (×103 km2) Average slip (m) Rupture velocity (km s−1) 127.9 3.5 36.6 10.0 2.5

Fault Longitude (◦E) Latitude (◦N) Depth b.s.l. (km) Length (km) Width (km) Strike (km) Dip (◦) Rake (◦) #1 −8.93 35.30 6.50 162.02 68.32 349.0 2.5 79.0 #2 −8.20 35.47 9.49 174.02 68.33 349.0 5.0 79.0 #3 −7.48 35.69 15.44 198.02 68.65 349.0 7.5 79.0

Marquesˆ de Pombal & Pereira de Sousa: source M Seismic moment (×1020 Nm) Rigidity (×1010 Pa) Fault area (×103 km2) Average slip (m) Rupture velocity (km s−1) 46.3 3.5 13.2 10.0 2.5

Fault Longitude (◦E) Latitude (◦N) Depth b.s.l. (km) Length (km) Width (km) Strike (km) Dip (◦) Rake (◦) #1 −9.79 36.66 18.00 55.66 40.00 20.0 0.0 60.0 #2 −10.10 36.75 5.00 55.66 31.96 20.0 24.0 60.0 #3 −9.61 37.05 18.00 37.15 20.00 20.0 0.0 60.0 #4 −9.78 37.10 11.00 37.15 17.21 20.0 24.0 60.0 #5 −10.27 37.24 9.00 37.15 46.43 20.0 2.5 60.0 #6 −9.60 37.49 11.00 55.66 17.21 20.0 24.0 60.0 #7 −10.58 37.78 7.00 55.66 92.86 20.0 2.5 60.0

Gorringe Banks: source G Seismic moment (×1020 Nm) Rigidity (×1010 Pa) Fault area (×103 km2) Average slip (m) Rupture velocity (km s−1) 126.0 6.0 16.0 13.1 2.7

Fault Longitude (◦E) Latitude (◦N) Depth b.s.l. (km) Length (km) Width (km) Strike (km) Dip (◦) Rake (◦) #1 −11.45 36.94 8.00 200.00 80.00 60.0 40.0 90.0

rupture along a shallow dipping subduction-related thrust in the Model C, with a fault located a few tens of kilometres away from Gulf of Cadiz´ (Gutscher et al. 2002; Gutscher et al. 2006; Thiebot the coasts of the eastern Gulf of Cadiz,´ systematically produces a & Gutscher 2006). similar level of ground shaking in southern Spain, southern Portugal Source parameters are shown in Table 3. The seismic moment and northern Morocco. Compared to the historical isoseismal map, for each model is deduced from eq. (2) using the value of slip and the highest intensities (X–XI) are shifted from the southern coast fault area proposed by the various authors, and the average rigidity of Portugal to the southwestern coast Spain. In the most favourable at the depth. For each source, fault plane geometry (dimensions, case (a rupture directed towards WSW), simulated intensities are strike, dip and depth) is set using available published information, still much higher than intensities deduced from historical studies but many other parameters (rake, rupture velocity and directivity) carried out in Morocco (IX–X versus VII–VIII) and in the Internal are, obviously, not constrained by historical information, and must Betics (VIII versus V) by Levret (1991) and Mart´ınez-Solares & be extrapolated from the study of similar instrumental earthquakes. Lopez-Arroyo´ (2004), respectively. Overall, this model fails to re- We assume that all fault segments act with a pure thrust mech- produce a key feature of the 1755 November 1 earthquake, that is anism; this hypothesis is supported by the observation of source the observation of higher intensities in Portugal than in Spain and mechanisms of recent great (M w > 8.0) shallow dip-slip earthquakes Morocco. (Harvard CMT Catalogue). In the case of model M, we observe high intensities in the Lisbon Rupture velocity was set to 2.5 km s–1 for the shallow rupture area (IX–X), in the Lusitanian Basin, in the Lower Tagus Valley models B and C, which is a typical value for dip-slip crustal earth- Basin, and in the Arrabida chain, that are compatible with obser- quakes. For model A, which involves faulting at larger depths, a vations. These intensities are conditioned by rupture directivity, be- slightly higher rupture velocity of 2.7 km s–1 was used. cause the area lies in the strike direction of the fault. Intensities in As emphasized by McGuire et al. (2002), the majority of large Algarve (IX) are also correctly reproduced, mainly because of the (M > 7) earthquakes exhibit a preferential direction of rupture, and proximity of the source. However, intensities computed in the west- the 1755 earthquake may be expected to have behaved in the same ern direction (VIII) are much higher than observations (VI–VII). In way. Hence, for sources A and B, in addition to a bilateral rupture the synthetics, the decrease of intensities with distance is very low, scenario, the two unilateral cases are tested. However, the low aspect with intensities VIII distributed over a 300-km wide zone. The re- ratio of source C is not typical of great tsunamigenic earthquakes, lation between PGV and MMI is not the cause of this phenomenon, and an along-dip directed rupture cannot be excluded. Therefore, since, in order to remove this effect, an unrealistically high value of for this earthquake, two additional nucleation points are considered, the parameter a, in eq. (3), would be required. The radiation pattern in addition to the three along-strike cases (Fig. 5). induced by the style of faulting can explain these discrepancies: in

C 2007 The Authors, GJI Journal compilation C 2007 RAS September 12, 2007 9:54 GeophysicalJournalInternational gji3571

12 Raphael¨ Grandin et al.

Figure 5. Synthetic isoseismal maps of the 1755 November 1 earthquake, using source parameters summed in Table 3. Stars denote epicentral location (see text for discussion). The first and second rows correspond to model C, the third row corresponds to model M, and the fourth row corresponds to model G.

C 2007 The Authors, GJI Journal compilation C 2007 RAS September 12, 2007 9:54 GeophysicalJournalInternational gji3571

Strong motion in SW Iberia: Part 2 13

20 the most remote areas, surface waves generated by the roughly E–W the seismic moment (M 0 = 6 × 10 Nm), and average slip (ap- coseismic slip have a maximum amplitude in the direction of the proximately 4 m). Finally, using the 1969 earthquake, we calibrated Algarve, the Guadalquivir Basin and the Betic-Rif Belt, but also a relationship between PGV and MMI. Using this relationship, the in the areas further north, in eastern Portugal and western Spain. case of the great 1755 earthquake (M w = 8.5–8.7) was examined, This effect might be reduced if a smaller fault width was used, but and we tested three end-member sources proposed by different au- would require a smaller seismic moment, and a lower tsunamigenic thors. We concluded that, using the methodology presented in this efficiency. paper, a primary source located at Gorringe Bank is the most realistic As for model G, most features of the observed macroseismic field hypothesis to fit the observed isoseismal pattern. seem to be correctly reproduced with a rupture directed towards SW: In this study, we have shown that unknowns lie in a realistic de- high intensities in the Algarve (IX–X) and in the Lisbon area (IX), scription of the seismic source in the area of study, but that only a associated with a moderate amplification in the Guadalquivir Basin few parameters are crucial when trying to simulate ground motion (VIII). We also observe a rapid decrease of intensities from south at low frequency. Obviously, the moment magnitude, focal mech- ◦ ◦ to north, at 8 W, and from west to east, at 39 N. The shape of the anism and source location influence the distribution of intensities limit between intensities IX and X is in very good agreement with in the near field. For large and great earthquakes, other details of observations, and the highest level of shaking is located at Cape Sao˜ the seismic source can alter dramatically this distribution. Rupture Vicente. On the other hand, simulated intensities at the SE corner directivity plays a particularly important role, and is conditioned ◦ ◦ of the computational grid (6.5 N; 39.5 N) are larger than intensities by various parameters, such as fault dimensions, epicentral location reported by Mart´ınez-Solares & Lopez-Arroyo´ (2004) in the area with respect to the fault centre, rupture velocity, and fault strike. (V–VI). However, it can be noted that this overestimation is also observed for models C (VII) and M (VII–VIII), but the error is much smaller for model G (VI). ACKNOWLEDGMENTS Magnitude estimates made from isoseismal area regression meth- ods (Johnston 1996) or tsunami strength (Abe 1979) were obtained We are grateful to Pascal Podvin who made this work possible, Paul assuming an epicentral location close to Gorringe Bank, and con- Tapponnier for fruitful discussions, and Geoffrey King whose re- strained using mostly near field data, as the quality of macroseismic marks and continuous support significantly improved the quality of and tsunami observations dramatically decreases with their mag- our work. The authors acknowledge helpful reviews of two anony- nitude. An eastward displacement of the source would obviously mous reviewers. We thank Shawn Larsen for supplying the code reduce the inferred moment magnitude. In this study, for both mod- E3D. Some figures were made using the GMT software written by els M and C, modelled intensities in the innermost areas of Iberia P. Wessel and W. H. F. Smith. (VII–VIII) are significantly higher than observed intensities (VI), This work has been developed with the support of the Fundac¸ao˜ even when an average slip of 10 m is used. This incompatibility be- para a Cienciaˆ e a Tecnologia, through project POCTI/CTE- tween the proposed fault locations of models M and C, and the ob- GIN/59994/2004: ‘Finite source modelization and strong mo- served macroseismic field, suggest that the seismic moment should tion prediction in Portugal’, and POCTI/CTE-GIN/59750/2004: ‘Seismic tomography of the continental lithosphere in Algarve be reduced, either by scaling down the average slip, or by reduc- (Portugal)’ and was largely carried out at Centro de Geof´ısica de ing fault dimensions. However, Baptista et al. (1998b, 2003) and Evora´ (Portugal). This is also IPGP contribution 2269. Gutscher et al. (2006) all admit that an average slip of ∼20mis needed to reproduce tsunami wave amplitudes and run-up values comparable with observation, in the case of a rupture located close to the coast. Such large average coseismic slips have only been ob- served for exceptional great (M > 9.0) mega-thrust earthquakes w REFERENCES (1957 Aleutian earthquake, 1960 Chile earthquake, 1964 Alaska earthquake and 2004 Sumatra earthquake), with rupture lengths an Abe, K., 1979. Size of great earthquakes of 1837–1974 inferred from tsunami order of magnitude larger than values considered here. Therefore, data, J. geophys. Res., 84, 1561–1568. for models M and C, a second incompatibility appears between pro- Abrantes, F., Lebreiro, S., Rodrigues, T., Gil, I., Bartels-Jonsd´ ottir,´ H., posed fault dimensions and the average slip needed to generate the Oliveira, P., Kissel, C. & Grimalt, J.O., 2005. Shallow-marine sediment large tsunami wave amplitude. cores record climate variability and earthquake activity off Lisbon (Por- tugal) for the last 2000 years, Quate. Sci. Rev., 24, 2477–2494. We conclude that a fault located below Gorringe Bank, with a Andrade, C., 1992. Tsunami generated forms in the Algarve barrier islands rupture directed towards SW, reproduces better the overall pattern (south Portugal), Sci. Tsunami Hazards, 10(1), 21–34. of macroseismic observations than a fault aligned along the Marquesˆ Argus, D.F., Gordon, R.G., DeMets, C. & Stein, S., 1989. Closure of the de Pombal – Pereira de Sousa Fault zone, or a subduction-related Africa-Eurasia-North America plate motion circuit and tectonics of the thrust fault in the Gulf of Cadiz.´ Gloria fault, J. geophys. Res., 94, 5585–5602. Baptista, M.A., Heitor, S. & Mendes-Victor, L., 1995. Historical review of the 1755 Lisbon tsunami. Evaluation of the tsunami parameters, in Proceedings of the XXIV General Assembly of the European Seismological 6 CONCLUSION Commission, Vol.3, 1790–1796. Baptista, M.A., Miranda, P.M.A.,Miranda, J.M. & Mendes Victor, L., 1996. We generated synthetic isoseismal maps for a large (M s = 8.0) Rupture extent of the 1755 Lisbon earthquake inferred from numerical earthquake that occurred 1969 in the region. The fault parameters modeling of tsunami data, Phys. Chem. Earth, 21(12), 65–70. provided by several previous studies were tested, and we obtained Baptista, M.A., 1998. Genese, Propagac¸ao˜ e Impacto de Tsunamis na Costa a very good fit of simulated and real intensity distributions for a Portuguesa, PhD thesis. University of Lisbon. ◦ ◦ ◦ focal mechanism (strike = 233 ; dip = 49.5 ; rake = 63.5 ) in good Baptista, M.A., Heitor, S., Miranda, J.M., Miranda, P.M.A. & Mendes agreement with existing solutions. Fault dimensions were also es- Victor, L., 1998a. The 1755 Lisbon earthquake; evaluation of the tsunam timated (fault length = 82.5 km; fault width = 35 km), as well as parameters. J. Geodyn., 25, 143–157.

C 2007 The Authors, GJI Journal compilation C 2007 RAS September 12, 2007 9:54 GeophysicalJournalInternational gji3571

14 Raphael¨ Grandin et al.

Baptista, M.A., Miranda, P.M.A.,Miranda, J.M. & Mendes Victor, L., 1998b. subcontinental lithospheric slab beneath the Rif Mountains, Morocco, Constraints on the source of the 1755 Lisbon tsunami inferred from nu- Geology, 34(7), 529–532. merical modelling of historical data on the source of the 1755 Lisbon Fonseca, J.F.B.D. & Long, R.E., 1991. Seismotectonics of SW Iberia: a tsunami, J. Geodyn., 25, 159–174. distributed plate margin? in Seismicity, Seismotectonics, and Seismic Risk Baptista, M.A., Lopes, C., Lopes, F.C.& Miranda, J.M., 2000. Analysis of the of the Ibero-Maghrebian Region, eds Mezcua, J. & Ud´ıas, A., Memoir 8, 1722.12.27 earthquake and the tsunami (Tavira-Portugal), 2a Assembleia Instituto Geografico Nacional, Madrid. Luso-Espanhola de Geodesia e Geof´ısica, Lagos, Portugal, 8–12 February, Fonseca, J.F.B.D., Vilanova, S., Bosi, V.B. & Meghraoui, M., 2000. Inves- Abstracts, pp. 153–154. tigations Unveil Holocene Thrusting for Onshore Portugal, EOS, 81(36), Baptista, M.A., Miranda, P.M.A.,Chierici, F.& Zitellini, N., 2003. New study 412–413. of the 1755 earthquake source based on multi-channel seismic survey data Fonseca, J.F.B.D., Vilanova, S., Bosi, V.B. & Meghraoui, M., 2001. Paleoseis- and tsunami modeling, Nat. Hazards Earth Sci. Syst., 3, 333–340. mological studies Near Lisbon: Holocene thrusting or landslide activity?, Baptista, M.A., Miranda, J.M. & Luis, J.F., 2006. In Search of the 31 March EOS, 82(32), 351–353. 1761 Earthquake and Tsunami Source, Bull. Seism. Soc. Am., 96(2), 713– Foster, I.D.L., Dawson, A.G., Dawson, S., Lees, J.A. & Mansfield, L., 1993. 721, doi:10.1785/0120050111. Tsunami sedimentation sequences in the Scilly Isles, south-west England, Beresnev, I.A. & Atkinson, G.M., 1997. Modeling finite-fault radiation from Sci. Tsunami Hazards, 11(1), 35–46. the n spectrum, Bull. Seism. Soc. Am., 87, 67–84. Frankel, A. & Vidale, J.E., 1992. A three-dimensional simulation of seismic Bezzeghoud, M. & Buforn, E., 1999. Source parameters of the 1992 Melilla waves in the Santa Clara Valley, California, from a Loma Prieta aftershock, (Spain, M w = 4.8), 1994 Alhoceima (Morocco, M w = 5.8), and 1994 Bull. Seism. Soc. Am., 82, 2045–2074. Mascara (Algeria, M w = 5.7) earthquakes and seismotectonic implica- Fukao, Y., 1973. Thrust faulting at a lithospheric plate boundary: the Portugal tions, Bull. Seism. Soc. Am., 89(2), 359–372. earthquake of 1969, Earth Planet. Sci. Lett., 18, 205–216. Bezzeghoud, M. & Borges, J.F., 2003.Mecanismos focais em Portugal con- Gjevik, B., Pedersen, G., Dybesland, E., Harbitz, C.B., Miranda, P.M., tinental, F´ısica de la Tierra, 15, 229–245. Baptista, M.A., Mendes-Vector, L., Heinrich, P., Roche, R. & Guesmia, Borges, J.F., Fitas, A.J.S., Bezzeghoud, M. & Teves-Costa, P., 2001. Seismo- M., 1997. Modeling tsunamis from earthquake sources near Gorringe tectonics of Portugal and its adjacent Atlantic area, Tectonophysics, 337, Bank sourthwest of Portugal, J. Geophys. Res., 102(C13), 27,931–27,949. 373–387. Gr`acia, E., Danobeitia,˜ J., Verges,´ J., Bartolome,´ R. & Cordoba,´ D. 2003a. Borges, J.F., Bezzeghoud, M., Buforn, E., Pro, C. & Fitas, A., 2007. The Crustal architecture and tectonic evolution of the Gulf of Cadiz (SW 1980, 1997 and 1998 Azores earthquakes and some seismo-tectonic im- Iberian margin) at the convergence of the Eurasian and African plates, plications, Tectonophysics, 435, 37–54, doi:10.1016/j.tecto.2007.01.008 Tectonics, 22(4), doi:10.1029/2001TC901045. British Oceanic Data Centre, 2003. Centenary edition of the GEBCO digital Gr`acia, E., Danobeitia,˜ J. & Verges,´ J., 2003b. Mapping active faults offshore atlas, www.bodc.ac.uk. Portugal (36◦N-38◦N): implications for the seismic hazard assessment Brune, J.N., 1970. Tectonic stress and the spectra of seismic shear waves along the southwest Iberian margin, Geology, 31(1), 83–86. from earthquakes, J. Geophys. Res., 75, 4997–5009. Grandin, R., Borges, J.F., Bezzeghoud, M., Caldeira, B. & Carrilho, F., 2007. Buforn, E., Ud´ıas, A. & Mezcua,´ J., 1988. Seismicity and focal mechanisms Simulations of strong ground motion in SW Iberia for the 1969 February in south Spain, Bull. Seism. Soc. Am., 78, 2008–2224. 28 (M S = 8.0) and the 1755 November 1 (M ∼ 8.5) earthquakes–I. Buforn, E., Bezzeghoud, M., Ud´ıas, A. & Pro, C., 2004. Seismic sources Velocitymodel, Geophys. J. Int., doi:10.1111/j.1365-246X.2007.03570.x. on the Iberia-African plate boundary and their tectonic implications, Pure Grimison, N. & Chen, W., 1986. The Azores-Gibraltar plate boundary: focal Appl. Geophys., 161, doi:10.1007/s00024-003-2466-1. mechanisms, depth of earthquakes and their tectonical implications, J. Calvert, A. et al., 2000. Geodynamic evolution of the lithosphere and upper geophys. Res., 91, 2029–2047. mantle beneath the Alboran´ region of the western Mediterranean: con- Grimison, N. & Chen, W., 1988. Focal mechanisms of four recent earth- straints from travel time tomography, J. geophys. Res., 105, 10 871–10 quakes along the Azores-Gibraltar plate boundary, Geophys. J. R. astr. 898. Soc., 92, 391–401. Carrilho, J., Teves-Costa, P., Morais, I., Pagarete, J. & Dias, R., 2004. Guesmia, M., Heinrich, P.& Mariotti, C., 1996. Finite element modelling of GEOALGAR Project: first results on seismicity and fault-plane solutions, the 1969 Portuguese tsunami, Phys. Chem. Earth, 21(12), 1–6. Pure Appl. Geophys., 161, 589–606. Gutscher, M.A., Malod, J., Rehault, J.P., Contrucci, I., Klingelhoefer, F., Carvalho, J., 2004. S´ısmica de alta resoluc¸ao˜ aplicadaa ` prospecc¸ao,˜ geotec- Mendes-Victor, L. & Spackman, W., 2002. Evidence for active subduction nica e risco s´ısmico, Ph.D. thesis. Univ. of Lisbon, 435 pp. ben(eath Gibraltar, Geology, 30, 1071–1074. Carvalho, J., Cabral, J., Goncalves,¸ R., Torres, L. & Mendes-Victor, L., 2006. Gutscher, M.-A., Baptista, M.A., Miranda, J.M., 2006. The Gibraltar Arc Geophysical methods applied to fault characterization and earthquake seismogenic zone. Part 2: constraints on a shallow east dipping fault plane potential assessment in the Lower Tagus Valley, Portugal, Tectonophysics, source for the 1755 Lisbon earthquake provided by ts Baptista et al., 2003 418, 277–297. unami modeling and seismic intensity, Tectonophysics, 426(1–2), 153– Chen, W.P. & Grimison, N., 1989. Earthquakes associated with diffuse 166. zones of deformation in the oceanic lithosphere: some examples, Tectono- Hartzell, S. et al., 2006. Modeling and Validation of a 3D Velocity Structure physics, 166, 133–150. for the Santa Clara Valley, California, for Seismic-Wave Simulations, Bull. Dawson, A.G., Hindson, R., Andrade, C., Freitas, C., Parish, R. & Bateman, Seism. Soc. Am., 96(5), 1851–1881. M., 1995. Tsunami sedimentation associated with the Lisbon earthquake Hayward, N., Watts, A.B., Westbrook, G.K. & Collier, J.S., 1999. A seismic of 1 November AD 1755: Boca do Rio, Algarve, Portugal, The Holocene, reflection and GLORIA study of compressional deformation in the Gor- 5(2), 209–215. ringe Bank region, eastern North Atlantic, Geophys. J. Int., 138, 831–850. DeMets, C., Gordon, R.G., Argus, D.F. & Stein, S., 1994. Effect of recent Heinrich, P., Baptista, M.A. & Miranda, P., 1994. Numerical simulations of revisions to the geomagnetic reversal time scale on estimates of current the 1969 tsunami along the Portuguese coasts, Preliminary results, Sci. plate motions, Geophys. Res. Lett., 21(20), 2191–2194. Tsunami Hazards, 12(1), 3–23. Dias, R.P., 2001. Neotectonica´ da Regiao˜ do Algarve, Ph.D. thesis. Univ. of Jimenez-Munt,´ I., Fern`andez, M., Torne, M. & Bird, P., 2001. The transi- Lisbon, 369 pp. tion from linear to diffuse plate boundary in the Azores-Gibraltar region: Dinis, J.L., 2006. Nazare´ do Canhao˜ a ` “falha”, da hipotese´ ao mito, paper results from a thin-sheet model, Earth. Planet. Sci. Lett., 192, 175–189. presented at VII Congresso Nacional de Geologia, Estremoz, Portugal, Johnston, A., 1996. Seismic moment assessment of earthquakes in stable 29 June – 13 July. continental regions – III. New Madrid 1811–1812, Charleston 1886 and Fadil, A., Vernant, P., McClusky, S., Reilinger, R., Gomez, F., Ben Sari, D., Lisbon 1755, Geophys. J. Int., 126, 314–344. Mourabit, T., Feigl, K. & Barazangi, M., 2006. Active tectonics of the Justo, J.L. & Salwa, C., 1998. The 1531 Lisbon earthquake, Bull. Seism. Soc. western Mediterranean: Geodetic evidence for rollback of a delaminated Am., 88(2), 319–328.

C 2007 The Authors, GJI Journal compilation C 2007 RAS September 12, 2007 9:54 GeophysicalJournalInternational gji3571

Strong motion in SW Iberia: Part 2 15

Kagawa, T., Zhao, B., Miyakoshi, K. & Irikura, K., 2004. Modeling of 3D Mendes-Victor, L., Baptista, M.A., Miranda, J.M. & Miranda, P.M.A.,1999. basin structures for seismic wave simulations based on available informa- Can hydrodynamic modelling of tsunami contribute to seismic risk assess- tion on the target area: case study of the Osaka Basin, Japan, Bull. Seism. ment?, Phys. Chem. Earth, 24, 139–144. Soc. Am., 94, 1353–1368. Miranda, J.M., Miranda, P.M.A.,Baptista, M.A. & Mendes-Victor, L., 1996. Kaka, S.I. & Atkinson, G.M., 2004. Relationships between instrumental A comparison of the spectral characterisitics of observed and simulated ground-motion parameters and Modified Mercalli Intensity in Eastern tsunamis, Phys. Chem. Earth, 21(12), 71–74. North America, Bull. Seism. Soc. Am., 94, 1728–1736. Montilla, J.A.P., Lopez´ Casado, C. & Henares Romero, J. 2002. Deag- Kanamori, H. & Anderson, D.L., 1975. Theoretical basis of some empirical gregation in magnitude, distance, and azimuth in the South and relations in seismology, Bull. Seism. Soc. Am., 65(5), 1073–1095. West of the Iberian Peninsula, Bull. Seism. Soc. Am., 92(6), 2177– Larsen, S.C. & Schultz, C.A., 1995. ELAS3D, 2D/3D Elastic Finite- 2185. Difference Wave Propagation Code, Lawrence Livermore National Lab- Moreira, V.S.,1984. Sismicidade historica´ de Portugal Continental, Revista oratory, UCRLMA-121792, 18 pp. do Instituto Nacional de Meteorologia e Geof´ısica, Lisbon, pp. 79. Larsen, S., Antolik, M., Dreger, D., Stidham, C., Schultz, C., Lomax, A. & Moreira, V.S.,1985. Seismotectonics of Portugal and its adjacent area in the Romanowicz, B., 1997. 3D models of seismic wave propagation; simu- Atlantic, Tectonophysics, 11, 85–96. lating scenario earthquakes along the Hayward Fault, Seism. Res. Lett., Olsen, K. & Schuster, G., 1992. Site amplification in the Salt Lake Valley by 68(2), 328. three-dimensional elastic wave propagation, EOS, Trans. Am. Geophys. Levret, A., 1991. The effects of the November 1st, 1755 “Lisbon” earthquake Union, 73(43) (Supplement), 338. in Morocco, Tectonophysics, 193, 83–94. Olsen, K.B. & Archuleta, R.J., 1996. Site response in the Los Angeles Basin Lopez-Arroyo,´ A. & Ud´ıas, A., 1972. Aftershock sequence and focal pa- from 3-D simulations of ground motion, SRL, 67(2), p. 49. rameters of the february 28th, 1969 earthquake of the Azores-Gibraltar Paula, A. & Oliveira, C.S., 1996. Evaluation of 1947–1993 macroseismic Fracture Zone, Bull. Seism. Soc. Am., 62(3), 699–720. information in Portugal using the EMS-92 scale, Ann. Geof. 39(5), 989– Lopez´ Casado, C., Molina Palacios, S., Delgado, J. & Pelaez,´ J.A., 2000. 1003. Attenuation of intensity with epicentral distance in the Iberian Peninsula, Pereira de Sousa, F.L., 1919. O terremoto do 1◦ de Novembro de 1755 em Bull. Seism. Soc. Am., 90(1), 34–47. Portugal e um estudo demogra´fico (Servicos¸ Geologicos´ de Portugal), Luque, L., Lario, J., Zazo, C., Goy, J.L., Dabrio, C.J. & Silva, P.G., 2001. Vol. I-IV. Tsunami deposits as paleoseismic indicators: examples from the Spanish Pitarka, A., Takenaka, H. & Suetsugu, D., 1994. Modeling strong motion in coast, Acta Geologica Hispanica, 36 (3–4), 197–211. the Ashigara Valley for the 1990 Odawara, Japan, earthquake, Bull. Seism. Lynnes, C.S. & Ruff, L.J., 1985. Source process and tectonic implications Soc. Am., 84, 1327–1335. of the great 1975 North Atlantic earthquake, Geophys. J. R. astr. Soc., 82, Pitarka, A., Suetsugu, D. & Takenaka, H., 1996. Elastic finite-difference 497–510. modeling of strong motion in Ashigara Valley for the 1990 Odawara, Machado, F., 1966. Contribuc¸ao˜ para o estudo do terremoto do terramoto de Japan, earthquake, Bull. Seism. Soc. Am., 86, 981–990. 1 de Novembro de 1755, Rev. Fac. Ciencias,ˆ Lisbon, C14, 19–31. Pitarka, A., Irikura, K., Iwata, T. & Sekiguchi, H., 1998. Three-dimensional Madariaga, R., 1976. Dynamics of an expanding circular fault, Bull. Seism. simulation of the near-fault ground motion for the 1995 Hyogo-Ken Soc. Am., 66 (3), 639–666. Nanbu (Kobe), Japan, earthquake, Bull. Seism. Soc. Am., 88, 428– Mart´ınez-Solares, J.M., Lopez, A. & Mezcua, J., 1979. Isoseismal map of 440. the 1755 Lisbon earthquake obtained from Spanish data, Tectonophysics, Pitarka, A., Graves, R. & Sommerville, P., 2004. Validation of a 3D velocity 53, 301–313. model of the Puget Sound region based on modeling ground motion from Mart´ınez-Solares, J.M. & Lopez-Arroyo, A., 2004. The great historical the 28 February 2001Nisqually earthquake, Bull. Seism. Soc. Am., 94(5), 1755 earthquake. Effects and damage in Spain, J. Seismol., 8(2), 275– 1670–1689. 294. Platt, J.P. & Houseman, G., 2003. Comment on: evidence for ac- Martins, I. & Mendes V´ıctor, L.A., 2001. Contribuic¸ao˜ para o Estudo da tive subduction beneath Gibraltar, by Gutscher, M.-A., J. Malod, Sismicidade da regiao˜ oeste da pen´ınsula iberica´ . Publicacao˜ no 25, Inst. J.-P. Rehault, I. Contrucci, F. Klingelhoefer, Mendes-Victor, L. and W. Geof. Infante D. Lu´ıs, Univ. de Lisboa, 67 pp., Instituto Geof´ısico do Spakman, Geology, 31(6), e22. (Online Forum http://www.gsajournals. Infante D. Lu´ıs, Lisboa. org/gsaonline/?request=get-static&name=i0091-7613-31-6). Massel, C. et al., 2000. Neotectonics of the Macquarie Ridge Complex, Reid, H.F., 1914. The Lisbon earthquake of November 1, 1755, Bull. Seism. Australia-Pacific plate boundary, J. geophys. Res., 105(B6), 13 457–13 Soc. Am., 4(2), 53–80. 480. Ribeiro, A., Cabral, J., Baptista, R., Matias, L., 1996. Stress pattern in Por- Matias, L.M., 1996. A sismologia experimental na modelac¸ao˜ da estru- tugal mainland and the adjacent Atlantic region, West Iberia, Tectonics, tura da crusta em Portugal continental, Ph.D. thesis. Univ. of Lisbon, 15, 641–659. 398 pp. Ribeiro, A., 2002. Soft Plate and Impact Tectonics, Springer Verlag, Berlin Matias, L., Ribeiro, A., Baptista, M.A., Zitellini, N., Cabral, J., Terrinha, Heidelberg, Germany, 324 pp. P., Teves-Costa, P. & Miranda, J.M., 2005. Comment on “Lisbon 1755: Ribeiro, A., Mendes-Victor, L., Cabral, J., Matias, L. & Terrinha, P., 2006. a case of triggered onshore rupture?” by Susana P. Vilanova, Catarina F. The 1755 Lisbon earthquake and the beginning of closure of the Atlantic, Nunes, and Joao F.B. D. Fonseca, Bull. Seism. Soc. Am., 95(6), 2534–2538, Eur. Rev., 14, 193–205. doi:10.1785/0120040023. Sato, T., Graves, R.W. & Somerville, P.G., 1999. Three-dimensional finite- McClusky, S., Reilinger, R., Mahmoud, S., Ben Sari, D. & Tealeb, A., 2003. difference simulations of long-period strong motions in the Tokyo GPS constraints on Africa (Nubia) and Arabia plate motions, Geophys. J. metropolitan area during the 1990 Odawara earthquake (Mj 5.1) and the Int., 155, 126–138. great 1923 Kanto earthquake (Ms8.2) in Japan, Bull. Seism. Soc. Am., 89, McGuire J., Zhao, L. & Jordan, T.H., 2002. Predominance of unilateral 579–607. rupture for a global catalog of large earthquakes, Bull. Seism. Soc. Am., Satoh, T., Kawase, H., Sato, T. & Pitarka, A. 2001. Three-dimensional finite- 92 (8), 3309–3317. difference waveform modeling of strong motions observed in the Sendai McKenzie D., 1972. Active Tectonicsof the Mediterranean Region, Geophys. Basin, Japan, Bull. Seism. Soc. Am., 91, 812–825. J. R. astr. Soc., 30, 109–185. Scheffers, A. & Kelletat, D., 2005. Tsunami relics on the coastal land- Mezcua, J., 1982. Catalogo general de isosistas de la Peninsula Iberica, scape west of Lisbon, Portugal, Science of Tsunami Hazards, 23(1), 3– Instituto Geographico Nacional, Madrid. 16. Mendes, A.S., 1974. Considerac¸oes˜ sobre o sismo de 28 de Fevereiro Stich, D., Ammon, C.J. & Morales, J., 2003. Moment tensor solutions for de 1969, Publ. Servico¸ Meteorologico´ Nacional, Portugal, RT1194, small and moderate earthquakes in the Ibero-Maghreb region, J. geophys. GEO158. Res., 108, 2148, doi:10.1029/2002JB002057.

C 2007 The Authors, GJI Journal compilation C 2007 RAS September 12, 2007 9:54 GeophysicalJournalInternational gji3571

16 Raphael¨ Grandin et al.

Stich, D., Serpelloni, E., Mancilla, F. & Morales, J., 2006. Kine- Trifunac, M.D. & Brady, A.G., 1975. On the correlation of seismic intensity matics of the Iberia-Maghreb plate contact from seismic mo- scales with the peaks of recorded strong ground motion, Bull. Seism. Soc. ment tensors and GPS observations, Tectonophysics, 426, 295–317. Am. 65, 139–162. doi:10.1016/j.tecto.2006.08.004. Ud´ıas, A. & Lopez-Arroyo,´ A., 1970. Body and surface wave study of source Stidham, C., Dreger, D., Antolik, M., Larsen, S. & Romanowicz, B., 1999. parameters of the March 15, 1964 Spanish earthquake, Tectonophysics, 9, Three-dimensional structure influences on the strong motion wavefield 323–346. of the 1989 Loma Prieta earthquake, Bull. Seism. Soc. Am., 89, 1184– Vilanova, S.P., 2003. Sismicidade e Perigosidade S´ısmica do Vale Inferior 1202. do Tejo, Ph.D. thesis. Univ. of Lisbon, 259 pp. Dreger, D., Dolence, D., Stidham, C., Baise, L. & Larsen, S., 2001. Evaluat- Vilanova, S.P., Nunes, C.F. & Fonseca, J.F.B.D., 2003. Lisbon 1755: a case ing 3D earth models for the purpose of strong ground motion simulation, of trigerred onshore rupture?, Bull. Seism. Soc. Am., 93, 2056–2068. Seism. Res. Lett., 72(2), 278. Wald, D. & Graves, R., 1998. The seismic response of the Los Angeles basin, Terrinha, P., 1997. Structural geology and tectonic evolution of the Algarve California, Bull. Seism. Soc. Am., 88, 337–356. basin, South Portugal, Ph.D. thesis. Univ. of Lisboa, 425 pp. Wald, D.J., Quitoriano, V., Heaton, T.H., Kanamori, H., Scrivner, C.W. & Terrinha, P. et al. 2003. Tsunamigenic-seismogenic structures, neotectonics, Worden, C.B., 1999. TriNet “ShakeMaps”: rapid generation of instru- sedimentary processes and slope instability on the southwest Portuguese mental ground motion and intensity maps for earthquakes in southern Margin, Mar. Geol., 195, 55–73. California, Earthq. Spectra, 15, 537–556. Teves-Costa, P., Borgis, J.F., Rio, I., Ribeiro, R. & Marreiros, C., 1999. Wells, D. & Coppersmith, K., 1994. New empirical relationships among Source parameters of old earthquakes: semi-automatic digitalization of magnitude, rupture length, rupture width, rupture area and surface dis- analog records and seismic moment assessment, Nat. Hazards, 19, 205– placement, Bull. Seism. Soc. Am., 84, 974–1002. 220. Wu, Y.-M., Teng, T.-L., Shin, T.-C. & Hsiao, N.-C., 2003. Relationship Thiebot, E. & Gutscher, M.-A., 2006. The Gibraltar Arc seismogenic zone between peak ground acceleration, peak ground velocity, and intensity (part 1): constraints on a shallow east dipping fault plane source for the in Taiwan, Bull. Seism. Soc. Am., 93, 386–396. 1755 Lisbon earthquake provided by seismic data, gravity and thermal Zitellini, N. et al., 2001. Source of the 1755 Lisbon earthquake and tsunami modeling, Tectonophysics, 426(1–2), 135–152. investigated, EOS Trans. Amer. Geophys. U., 82, 285.

C 2007 The Authors, GJI Journal compilation C 2007 RAS