The 1755 Lisbon earthquake and the onshore LTV fault
by
Vilanova, Susana
Instituto Superior Tecnico
Coauthors: Joao Fonseca

Introduction. The 1755 Lisbon earthquake is the best-known natural catastrophic event that affected Portugal. With an estimated magnitude of M8.7 (Johnston, 1996) it damaged severely a large area including the city of Lisbon, the SW of Iberia and the Atlantic coast of Morocco. The location of the source has been widely discussed and due to apparent contradictions on the available historical data the different proposals spread over a relatively large region. While the tsunami arrivals and the damage on SW Portugal and NE Morocco suggest an oceanic location SW of Cape St. Vincent, the high intensities on the Lower Tagus Valley favour a northerly source. It was well established (e.g. Reid, 1914) that the Lisbon 1755 earthquake was a multiple event, composed by three sub-shocks separated by few minutes. Based on the analyses of contemporaneous reports, macroseismic data and old maps we propose a model in which the large offshore earthquake triggered an onshore rupture on the Lower Tagus Valley (LTV) fault. The LTV fault is a structure that has been postulated as responsible for the onshore earthquakes of 1334 and 1531, which affected severely the Portuguese capital, and the more recent 1909 that damaged several villages on the Tagus Valley. Although some geophysical and geological data indicated a deep crustal discontinuity and differential movement during the Quaternary period, it was only recently that the fault was identified on the surface deforming Holocene sediments. This structure seems therefore responsible for most of the seismic hazard in the Lower Tagus Valley.

Tectonic setting and zones of weakness. The Iberian Peninsula is situated in the Eurasian plate, in the vicinity of its southern boundary with the African Plate (figure 1). The Azores-Gibraltar section of the boundary shows a complex behaviour changing from a transtensional regime on the west to a transpressional regime on the east. The central section is characterised by a linear E-W fracture, the Gloria fault, exhibiting 4mm/yr of dextral strike-slip motion (Argus et al., 1990). West of –12 the plate boundary is characterised by a broad region of deformation under NNW-SSE horizontal compression, without a clear transition between interplate and intraplate seismic activity. The onshore active structures of Western Iberia and Morocco may be therefore regarded as the distal parts of this zone of deformation. As in most of the intraplate continental domains, the location of the actual deformation zones was determined by the geological evolution and correspondent crustal weaknesses. The LTV is a tardi-hercynian NNE-SSW strike-slip structure (Arthaud and Matte, 1977) which was reactivated several times under different tectonic regimes. During the extensional regime that led to the opening of the North Atlancic Ocean the fractures of the earlier shear zone were reactivated as normal faults and controlled the formation of the Early Jurassic basins. The Lusitanian Basin was formed during this first rift pulse and the LTV fault was probably it’s SW boundary. In the Mid Jurassic the continental rifting jumped towards the West and the Lusitanian Basin became an aborted rift.

Since the beginning of the Cretaceous period the Lusitanian Basin was structurally inverted, first due to the Pyrenean compression and later due to the Africa-Iberia convergence.

LTV fault system: geophysical, geological and seismological data. The available subsurface seismic data on the LTV consist of deep seismic refraction profiles and commercial seismic reflection exploration surveys. While the first led to the inference of a crustal fracture affecting the Moho (Mendes-Victor et al., 1980), the later provides an insight into the deformation of the sedimentary layers, although it does not illuminate the basement structures. Figure 1shows, in the upper right corner, the intra-basin fractures revealed by seismic reflection (GPEP, 1986). The broad lagoon in front of Lisbon seems therefore to be tectonically controlled. According to the model by Fonseca and Long (1991), the Alcochete segment acted as a releasing bend in a transpressional left-lateral regime, transferring shear from Vila Franca segment to the Arrábida Range.

The active tectonics study on the LTV (Fonseca et al., 2000) led to the identification of an approximately NE-SW oriented scarp on the right bank of the Tagus river. Geomorphologic analysis suggested two 20km segments: the segment Vila Franca de Xira–Azambuja and the segment Azambuja–Asseca. Towards the north although the reference scarp is still visible the signs of deformation are less evident. On the Azambuja–Asseca segment, for which the geomorphologic observations suggested the maximum vertical displacement (tributary profile convexity and anomalous terrace elevations) we excavated a trench and performed a paleoseismological analysis (Bosi et al., submitted). The trench, illustrated in figure 2 shows intense deformation with several inverse fault planes displacing Holocene C14 dated sediments. The N-S striations present in several fault planes indicate a left lateral component of motion. Some of the affected units had pottery fragments of Calcolithic (Copper Age) and Late Bronze Age (the Bronze Age in Portugal was aprox. from 4000 B.P. to 2800 B.P), corroborating the ratio-carbon dating. The most recent age for the deformed units is 365 AD – 655 AD (2-sigma calibrated age) and the units cut by the sub-horizontal fault plane displays approximately 3m of minimum displacement yield the dates 1395 - 1138 BC and 2875 - 2570 BC. It can therefore be calculated a preliminary slip-rate value of about 0.5-0.7mm/yr, nearly one order of magnitude higher that the proposed by Cabral (1995) from neotectonic considerations. Further geophysical prospecting performed at the trench site corroborated the tectonic origin of the deformation by revealing a contrast of resistivity aligned and consistent with the observations on the trenches. The active tectonics data is consistent with the model proposed by Fonseca and Long (1991).

The LTV had no significant seismic activity at present, and the small earthquakes located by the regional network are mostly below the Lusitanian basin and not along its flanking faults. However, the historical seismicity is noticeable and the higher intensity data points are distributed along the river valley and in good agreement with the scarp identified by the geomorphologic analysis.

1755 earthquake and tsunami data. Contemporaneous reports of the 1755 earthquake in Portugal are abundant and rich in information since an inquire was sent to all the parish churches with precise questions about the duration of shaking, geological phenomena, damage to buildings, etc. Sousa (1932) recollected most of these data together with other eyewitness accounts. A compilation by Nozes (1990) of accounts by British merchants and seafarers provides detailed information about the earthquake in Lisbon. Martinez Solares (2001) provide a compilation of the earthquake experience in Spain and Levret (1991) gives an account of the damage in Morocco. Figure 3a displays the high intensity data points and shows two lobes of high intensities, one in SW Portugal and the other in the LTV. The pattern of intensities for the 1755 earthquakes is similar to the combination of patterns for 1969 St Vincent earthquake and the 1531 LTV earthquake (fig3b and fig 3c). One alternative explanation for the high LTV intensities would be site effects: low-frequency waves arriving from a distant source could be amplified by the response of loose sediments. However, the damage on the parish churches does not correlate with the soft surface geology. In the 1969 (MS7.9) S. Vincent earthquake the higher intensities were restricted to SW Portugal and no amplification was observed on the LTV.

One historical report describes an alteration of the Tagus riverbed (the “rising of the land” that left the boats on dry land) affecting the river’s course, which suggests local intense deformation. Differences between maps drawn before and after 1755 corroborate this interpretation Some accurate reports on the earthquake in Lisbon indicate that the duration of the event was around 7 minutes with two small intervals, and that the second shock was the most violent. Also the strongest aftershocks reported in Lisbon are not in the same days as those reported in SW Portugal.

The tsunami travel-time to Cape St. Vincent was about 15 minutes, to Lisbon was about 90 minutes (Baptista et al., 1998) and to Safi (Morocco) about 30 minutes, suggesting that the source location was SW of Cape St. Vincent. There is however a severe incongruence in the 15 minutes of travel-time to Oeiras (15km W of Lisbon). Before the arrival of the tsunami several eyewitnesses also describe a disruption of the shallow (about 10m) waters of the Tagus Lagoon, in front of Lisbon, immediately after the earthquake. This disruption is described as a “ vast body of water forming at some distance like a mountain rushing towards the shore”. An English captain says that by then large ships were carried down the river and an English merchant clarifies the time of the events: “The rising of the water, I guess, happened at about eleven of the clock (…) It is probable that the whole bed of the river is altered; for during the first two shocks, and an hour before the rising of the water in so extraordinary a manner as I have described, several boats passing on the river were seen to whirl round as a whirlpool” (anon., in Nozes, 1990).

Discussion: the model proposed . We propose a model based on the idea that the stress changes associated with the occurrence of the large offshore earthquake can be significant even at large distances (300 km) to induce a local rupture on an independent structure as the LTV fault (Vilanova et al, 2003). The pattern of variation of the Coulomb failure stress shows positive correlation with the distribution of aftershocks (e.g. Stein et al., 1992; Harris et al., 1995). Particularly after the Landers 1992 earthquake (M7.4), followed 3hours after by Big Bear (M6.5), this mechanism of interaction has been identified and modelled (Stein et al., 1992). The program 3~def of Gomberg and Ellis (1992) allows the computation of stress tensor changes in an elastic and homogeneous half- space due to a displacement discontinuity on a fault. The Coulomb failure stress changes for a receiver fault is given by DsC=D |t|- m’DsN (t is the shear stress, sN is the normal stress and m is the apparent coefficient of internal friction that takes into account the effect of pore fluid pressure. Nunes (2000) shows that the rupture model proposed by Johnston (1996) for the 1755 earthquake (12m of displacement on a 200x80km fault on the Gorringe bank, figure 4) leads to positive values of DsC that range from 0.2bar (for m’=0) to 0.5 bar (for m’=0.75) on the SW end of the fault, bringing it closer to failure. These values are usually considered sufficient to cause the nucleation of the rupture if the fault was already close to failure. The rupture of the southwestern segment of LTV fault could cause a tilt on the footwall causing the water to rush towards the shore and allowing the emergence of dry land as described above. This model explains also the elevated intensities on the LTV and the larger damage to the churches in the consolidated Mesozoic rock than to those located on soft sediments and the different sets of aftershocks reported in general and at the LTV.

The mechanism of interaction between earthquakes, in particular in regions subjected to very large interplate earthquakes and moderate to large intraplate earthquakes, is very important for an accurate calculation of the seismic hazard. Extra geological data are crucial for a better constraint of the parameters used to characterise both the offshore and onshore seismicity.

References. Argus, D., R. G. Gordon, C. DeSmets and S. Stein, 1989. Closure of the Africa-Eurasia-North America Plate motion Circuit and Tectonics of the Gloria Fault, J. Geophys. Res., 94 (B5), 5585-5602.

Baptista, M.A., P.M. Miranda, J.M. Miranda and L. Mendes-Victor, (1998b). Constrains on the Source of the 1755 Lisbon Tsunami Inferred from Numerical Modelling of Historical Data on the source of 1755 Lisbon Tsunami. J. Geodynamics, 25, 159-174.

Bosi, V, S.P. Vilanova, M. Meghraoui and J.F.B.D. Fonseca, New Insights on the Onshore Earthquakes of the Lower Tagus Valley, submitted to Geophys. J. Int.

Cabral, E.D. (1790). Memória sobre os danos causados pelo Tejo sobre as suas ribanceiras, Memórias Económicas da Academia Real das Ciências de Lisboa, II, 177-204.

Fonseca, J.F.B.D. and R.E. Long (1991). Seismotectonics of SW Iberia: A Distributed Plate Margin? in Seismicity, Seismotectonics and Seismic Risk of the Ibero-Maghrebian Region, J. Mezcua and A. Udías (Editors), Memoir no. 8, Instituto Geografico Nacional, Madrid.

Fonseca, J.F.B.D., S.P. Vilanova, V. Bosi and M. Meghraoui (2000). Paleoseismological Investigations Unveil Holocene Thrusting Onshore Portugal, EOS Trans., 81 (36), 412-413.

Harris, R. A. (1998). Introduction to the special session: stress trigger, stress shadows and implications for seismic hazard, J. Geophys. Res., 103 (B10), 24347-24358.

Levret, A. (1991). The Effects of the November 1, 1755 Lisbon Earthquake in Morocco, Tectonophysics, 193, 83-94.

Martinez-Solares, J. M. (2001). Los effectos en España del terremoto de Lisboa, Instituto Geografico Nacional, Madrid.

Mendes-Victor, L., A Hirn and L. Veinante (1980). A seismic session across the Tagus Valley, Portugal: possible evolution of the crust, Ann. Geophys, 36 (4), 469-476 Nozes, J. (1990). O Terramoto de 1755: Testemunhos Britânicos; British Accounts of the Lisbon Earthquake of 1755, The British Historical Society of Portugal, Lisbon.

Nunes, A. C. (2002). Potencial para “stress triggering” de sismos ao largo da costa portuguesa: aplicação ao sismo de 1755, unpublished graduation thesis, IST, Lisbon.

Reid, H. F. (1914). The Lisbon Earthquake of November 1, 1755, Bull. Seism. Soc. Am., 4 (2), 53-80.

Sousa, F. L. P. (1932). O Terremoto Do 1º de Novembro de 1755, Serviços Geológicos de Portugal, Lisbon.

Stein, R., G. King and J. Lin (1992). Change in failure stress in Southern San Andreas fault system caused by the 1992 M=7.4 Landers earthquake, Science, 258, 1328-1332.

Vilanova, S.P., C. Nunes, and J.F.B.D. Fonseca. Lisbon 1755: A Case of Triggered Onshore Rupture?, Bull. Seism. Soc. Am., 93 (5), 2056-2068.

Figure Captions. Fig. 1 – Epicentral locations (black circles) proposed by different authors for the Lisbon 1755 earthquake. Mi - Milne (1841); R - Reid (1914); Ma - Machado (1966); Mo - Moreira (1989); Z - Zittellini et al. (1999). Structural features offshore according to Buforn et al., 1995. Lower Tagus Valley Fault and associated structures (inset) according to Fonseca and Long, 1991. Based on Vilanova et al. (2003);

Fig. 2 – a) Interpreted log of wall 1 in trench 1, Vila Cha de Ourique; b) View of the trench; Based on Bosi et al. (submitted);

Fig. 3 – A) High intensity data points for the 1755 earthquake, according to Moreira (1984), Martinez-Solares et al. (1979) and Levret (1991). B) High intensity data points (MMI equal or above VI) for the MS7.9 Gorringe Bank earthquake of February 1969. C) High intensity data points (MSK equal or above VI) for the 1531 Lisbon earthquake (M6.9) on the Lower Tagus Valley fault after Justo and Salwa (1998). Based on Vilanova et al. (2003);

Fig. 4 – Pattern of Coulomb failure-stress variation due to 12m of displacement on a 200km versus 80km thrust fault (40ºdip) on the Gorringe Bank. Based on Nunes (2002);

Date received: November 26, 2003