A Finite – Fault Modeling of the 1755 Lisbon

Earthquake Sources

A. Carvalho1, A. Campos Costa, C. S. Oliveira

Introduction

In Portugal, being located at a moderate/low seismicity intraplate area, insufficient accelerograms

have been recorded to satisfactory undertake any regional empirical study. The number of
accelerograms is not only small but also refers to low-magnitude earthquakes located in only some
parts of its entire seismic area. For that reason, most prediction techniques of ground motion in
Portugal have been based on international empirical laws and not on regional data to quantify the
characteristics of ground motions. However, differences in the regional geology can led to
variations in ground motions characteristics and the use of empirical laws of other regions is
questionable and may not be appropriate for Portugal. As prediction cannot be based on empirical
analyses, theoretical models must be used as the basis for the predictions of strong motion in
Portugal. The development of stochastic based ground motion synthesis associated to a
seismological finite-fault modeling is a worldwide approach that can be used for representation of

1
A.. Carvalho (*)
LNEC, Laborato´ rio Nacional de Engenharia Civil, Lisboa, Portugal
e-mail: xana.carvalho@lnec.pt
L.A. Mendes-Victor et al. (eds.), The 1755 Lisbon Earthquake: Revisited,
Geotechnical, Geological, and Earthquake Engineering 7,
DOI 10.1007/978-1-4020-8609-0_28, _ Springer ScienceþBusiness Media B.V. 2009

433
future large magnitude earthquakes occurring in Portugal, allowing the reproducing of specific
source effects like directivity and asperities distribution.

The strong ground motion prediction based on finite-fault simulation requires the identification of
the fault (strike, dip, length and width), source kinematics parameters (stress drop, velocity of
rupture, slip distribution), regional crustal properties (geometrical spreading, anelastic structure,
amplification and attenuation upper crust parameters) and the determination of amplification
effects due to the local site geology.

The model parameters calibration has been obtained with a dataset that includes horizontal
components of ground acceleration records (at rock sites) obtained by the Portuguese digital
accelerometer network. Validation and comparison are entirely in terms of 5% damped pseudo
absolute response spectra for acceleration.

The demonstrated agreement between model and data for low to moderate events in Portugal
provides strong grounds for accepting the stochastic-process model predictions for this type of
events and to use it as the basis for characterization of stronger earthquakes considering a finite
fault rupture modeled as a sum of a number of point sources distributed spatially and temporally.

Being so, the calibrated model is used to simulate ground motion acceleration for the 1755 Lisbon
earthquake, considering different source models proposed in literature. For each possible source
model several simulations were performed varying nucleation points. The soil effect was taken
into account considering a nonlinear behaviour of the stratified geotechnical sites conditions. The
simulated intensity is compared with the observed intensity values.

Numerical approach

The non-stationary stochastic finite fault simulation method, herein called RSSIM (Carvalho et al.,
2004; 2007) has been implemented starting from the classic simulation code FINSIM (Beresnev
and Atkinson, 1998), widely employed in the seismological literature for simulation of the ground
motion from both moderate and high magnitude earthquakes. However, it differs from the classic
FINSIM as it avoids the computation of acceleration time series representing the contribution of
each sub-fault, but synthesizes the ground motion due to the entire fault from the Power Spectral
Density Function (PSDF) radiated by each sub-fault, using the random vibration theory and the
extreme values statistics (ex. Vanmarcke, 1976; Boore & Joyner, 1984; Boore, 2003).

Like FINSIM, the RSSIM method assumes that the fault plane is a rectangle, subdivided into an
appropriate number of sub-faults, which are modelled as point sources characterized by an ω2
spectrum.

The method starts estimating, for each subfault, the amplitude of the acceleration Fourier
spectrum, A( , R) :

A( , R)   2  C  S ( )  G ( R)  An( , R)  P( )  FZ ( ) (1)

434
where C is a scaling factor including the free surface amplification factor, the radiaton pattern of
shear waves and the energy partition into the two horizontal components, S() is the amplitude
displacement source spectrum, G(R) is the geometric spreading factor, An(,R) is the anelastic
path attenuation factor, P() accounts for the upper crust attenuation and FZ() is a crustal
amplification function. The functional form of all these factors and the respective physical
meaning can be found elsewhere [eg. Boore, 2003; Carvalho et al., 2004; Ferrer& Sanchez-
Carratalá, 2004] and in the modeling parameters section of this text.

Taking into account the Fourier amplitude spectrum, A( , R) , and a given source duration, Ts, the
one-sided power spectral density function (PSDF) of acceleration can be derived by means of:

2
1 A( , R )
Sa( )  (2)
 Ts

The PSDF of the response of the oscillator with a circular frequency n and a damping ratio 
assuming a stationary process is calculated as:

2
S x ( , n ,  )  Sa( ) H x ( , n ,  ) (3)

in which H x (,n , ) is the complex displacement frequency response function relative to the
input absolute base acceleration and given by

n2
H x (, n ,  )  (4)
n2  2i  n    2

To cope with the non stationary of the intensity of the response of the oscillator, a intensity time-
modulating response function,  , is specified directly in a way that the evolutionary response
moment of order k is obtained by the modulating function and the stationary response moment of
order k as:

k  t ,,n ,    2  t ,n ,   k (,n , ) (5)

in which


k  , n ,      k S x ( , n ,  ) d
0 (6)

and  , the response modulating function, is obviously dependent of frequency and damping of the
one degree of freedom system and can be found in Duarte (1978).:

 0 t  t1

 2n  ( t t1 )
 ( t,n , )   1 e t1  t  t2 . (7)

 e 2n  ( t t2 )  e2n  ( t t1 )

t  t2

435
where t1 represents the starting time of the strong motion part of the accelerogram and t2 gives the
starting time of the decay part being t2 = T + t1 , T the duration of the strong motion.

For a finite source, subdivided into N sub-faults, and considering that stochastic process associated
to all the N subfaults are independent, the final evolutionary finite–fault response moment, k T ,
can be estimated as the sum of all subfault response moments. Therefore:

N
k T  t ,,n ,     2j  t ,n ,   k j  ,n ,  (8)
j 1

Considering the extreme values statistics and taking T as the duration of the earthquake, the
displacement non-stationary response spectrum, RS, that synthesizes the integration over the fault
of all the delayed sub-sources can be estimated (Vanmarcke, 1976):

 
RS T ,  , n ,     2  ln(2  f0  T ) 
0.577216
  0T T ,  , n ,   (9)
 2  ln(2  f0  T ) 
 

Where fo is the central frequency of the response written as

1
1  T (T , ,  ,  )  2
f0   2 n  (10)
2  0T (T , , n ,  ) 

Comparison of response spectra obtained with the non-stationary stochastic finite fault simulation
method, RSSIM, and with classic FINSIM can be seen in Zonno et al. [2005] and Carvalho et al.
[2007].

Once the non-stationary response spectra have been achieved, an equivalent stationary PSDF can
be iteratively estimated, following the classical theory of stationary random process. This approach
was adopted in an automatic seismic loss estimate methodology (LNECLoss – Sousa et al., 2004)
that was developed at LNEC. This PSDF will be the input for the 1D analysis of soil effects.

In order to define scenarios at locations where different geotechnical conditions prevail, an

adequate transfer function must be defined to cope with the filtering effect and dissipative nature
of local soil propagation medium. The complex transfer function H (  | h,  ,ς ,G )i from rock

outcrop acceleration to shear strain  of each layer i is obtained, which allows to write the shear
strain PSDF as function of the independent variable associated with the circular frequency :

S (  )i  H (  | h,  ,ς ,G )i  Sro(  )  H* (  | h,  ,ς ,G )i (11)

in which h,  ,ς and G are parameters vectors of n+1 dimension, representing the height, density,
viscous proportional damping ratio and the elasticity distortion modulus, of each stratified soil

layer i, respectively. H (  )i and H* (  )i are complex conjugate functions.

436
This last equation is only valid under the hypothesis of elastic linear behaviour of soil meaning
that, parameters ( ς,G)i should be kept constant even if they depend on the value of shear strain .

However, even for low amplitude input motion, soil tends to exhibit a non-linear behaviour. In this
case an equivalent iterative linear secant stiffness approach was adopted as it is frequently used in
other soil dynamics computer codes (SHAKE, FLUSH). For that purpose it will be necessary to

have estimates of the peak shear strain vector  i,peak

j
whose members i vary from 1 to n for each

soil layer, and j stands for the interaction cycle

Extreme Statistics determines the estimates of  jpeak based on the hypothesis that the PSDFs

S (  )i, j are a Gaussian zero mean random processes. In that case estimates of  i,peak
j can be

obtained by:

   2i, j  
T
 i,peak  2 0i, j ln    ln(ln( 2 ) (12)
j   2  0i, j  
   

where the stationary moments of 1 st and 2nd orders of the strain are defined as defined in eq. 6 but
for the PSDFs S (  )i , j defined in equation 11.

Once convergence is attained, the transfer function Ha (  ) between rock outcrop acceleration and
site specific acceleration is defined, and the PSDF of this acceleration can be computed as:

2
Sa (  )  Sro (  )  H a (  ) (13)

2
in which H a (  ) represents the transfer function between outcrop acceleration and the absolute

acceleration at the top of the 1D vertically S wave propagation column of stratified soil medium.

This approach allows a significant saving of calculation time, roughly a factor of twenty with
respect to the usual approach based on the traditional methods, can be achieved.

Resuming, each seismic input is defined by the rock outcrop power spectral density function of
acceleration and, given a stratified soil profile, the new PSDF is computed at the surface level
using procedure described above.

Modeling parameters

Finite-fault simulations require that the fault-plane geometry (length, width, strike, dip, number of
subfaults considered and depth to the upper edge), the source parameters (seismic moment, slip
distribution, stress drop, nucleation point, rupture velocity), the crustal properties of the region
(geometrical spreading coefficient and anelastic attenuation) and the site-specific soil response
information be previously specified.

437
Carvalho et al [ 2005, 2007] estimated most of this parameters using previous information from
several regional studies and a least square sense, comparing synthetic and observed response
acceleration spectra, to obtain a set of parameters that best fit data. In the last months a progress
has been made towards understanding the source and crustal characteristics, estimating source
parameters based on source spectra of S waves of 12 earthquakes records (corresponding to a total
of 25 records) and the upper crustal attenuation parameter from the high-frequency decay of the
acceleration spectra, after correcting observed spectra from geometrical spreading and anelastic
attenuation effect. Here we report the results of that study.

The dataset used includes horizontal components of 25 ground acceleration records obtained by
the portuguese digital accelerographic network, on rock sites. The regional distribution of the
earthquakes epicenters and magnitudes are illustrated in Figure 1. Some of the events were
recorded in more than one station.

11° 9° 7°

42° N 42°

40°
Mag nitud e Ú
Ê 40°
Ú
Ê 3.3 Ú #10
Ê
#5
Ú
Ê
Ú
Ê
Ú
Ê
3.4
3.6 Ú
Ê #3
#9

Ê
ÚÊ
Ê Ú #1

Ú
Ê 3.9 #2 Ú
Ê Ú
Ê #8

Ê
Ú 4.0
38°
#4 Ú
Ê 38°
Ú
Ê 4.1

Ú
Ê 4.2

Ú
Ê 4.4
#7

Ú
Ê
#12

Ú
Ê
4.5
36°
#11

ÚÚ
Ê
Ú
Ê Ê 36°
5.4
Ú
Ê #6

Ú
Ê 5.9
11° 9° 7°

Fig. 1- Epicentral distribution of earthquakes of portuguese accelerographic data base

As two physical mechanisms of earthquake generations exists in Portugal, namely events

originated by the movement between the Eurasian and African plates (here in called interplates
events) and events originated in faults inside the Eurasian plate (herein called intraplate events),
calibration of model parameters was done separately with data corresponding to intraplate and
interplates events.

438
The horizontal components of the accelerations at each station were windowed, using a time
window that included the main S-wave arrival and 90% of total energy, and then Fourier
transformed to obtain acceleration and displacement-amplitude spectra.

The observed spectra

In RSSIM approach, the source term (S(ω) in eq. 1)is assumed as an omega-square (ω2) model
specified by [Brune, 1971]:

Mo
S ( )  (14)
1   c 
2

where Mo and c stand for seismic moment and circular corner frequency respectively.

A visual inspection of the displacement-amplitude spectra of S-waves for the 12 events (see figure
2 for an example) leads to the conclusion that the spectral shapes for all events appear to be
similar, showing a plateau at lower frequencies, a decrease of the spectral amplitudes as f-2 at
higher frequencies and one corner frequency (the interception of both plateau and f-2 lines), as
predicted by the ω2 model.. However this feature seems to begin to change as magnitude nears 5.4
and the spectral shape of the 6.0 earthquake has two corner frequencies, where the spectral
envelope changes its general trend. This observational result implies that the shape of the source as
described by the ω2 model ceases to be valid as magnitude increases up to 5.5.

Another feature showed is that for frequencies greater than a certain frequency the spectral
amplitudes decay more rapidly as f-2. This decay is modeled as the high-cut filter P(ω) in eq. 1 and

in this study is described as P(  )  exp(   k 2 ) [Anderson & Hough, 1984], being k the

attenuation upper crustal parameter.

1.E-02 1.E-02
AF [cm.s]

AF [cm.s]

1.E-04 1.E-04

1.E-06 1.E-06

#4 #6
1.E-08 1.E-08
0.1 1 10 100 0.1 1 10 100
Freq. [Hz] Freq. [Hz]

Fig. 2 Observed displacement-amplitude spectra of S-waves (for earthquakes #4 and #6, as in

figure 1) and lines of ω2 model.

439
Path parameters:

0.7
W adopted the frequency-dependent quality factor, Q(f)=250f of Pujades et al. [1990] for the
inelastic attenuation An(, R)  exp( fR Q( f )  ) and the tri-piece-wise geometric attenuation
function described by Atkinson & Boore [1995] defined by assuming a crustal seismogenic
thickness of 31 km for the intraplate events [Jiménez-Munt et al, 2001] and 20 km for the
interplates events [Dezes & Ziegler, 2001].

Upper crustal parameter:

The term P( )  exp( f  ) is a high cut filter to account for near-surface attenuation effects,
which describes the observed rapid spectral decay at high frequencies. The k parameter is
estimated by fitting the high-frequency decay of the acceleration spectrum with a straight line in a
log-linear scale (Anderson & Hough, 1984), as illustrated in Figure 3. k was measured separately
for the two horizontal components of all the 25 records of our database. A suitable frequency band
(depending on quality of data, generally ranging from 5-10 Hz to 20-25 Hz) was chosen in order to
perform reliable regressions. A value of k=(0.032 ± 0.005) s was inferred from the analysis of all
records.

1.E+01 1.E-01

1.E+00 1.E-02
AF [cm/s]
AF [cm/s]

1.E-01 1.E-03

1.E-02 1.E-04
#7 #4
SVF_X SVF_X
1.E-03 1.E-05
0 5 10 15 20 25 0 5 10 15 20 25
Freq. [Hz] Freq. [Hz]

Fig. 3 Example of k estimations from the amplitude Fourier spectra of acceleration. SVF_X: x
horizontal component of the S.Vicente Fora (SVF) station

As regard to site-specific amplification, we expressed the factor Fz(ω) in eq. 1, as a function that
takes into account of density and S wave velocity changes respect to an homogeneous propagation
model:

2
Fz (  )    zi (15)
2
 
1   zi 
  

We employed the impedance function with fzi =0.5Hz, for interplates events, accordingly to
Carvalho [2007].

440
Source parameter

Having a better understanding of attenuation, we can correct the observed spectrum back to the
source, obtaining the source displacement-amplitude spectrum as:

 e f k  R  e R f /  Q ( f )
A( f , R ) A( f , R )
S( f )   (16)
(2 f ) 2 An( f , R)  G ( R)  P( f ) (2 f ) 2

where A(f,R) is the S-wave spectrum of ground acceleration observed, the two exponential terms
are used to correct both upper crust and path and attenuation and the terms Rα corrects the
amplitude decrease due to geometric spreading.

In the w2 model, the high-frequency level of the source model is controlled by stress drop, Δσ, that
is correlated with frequency corner frequency in terms of the relation

1
fc  4.91  106      3
M 0 
(17)


where Δσ is in bars, β in km/sec and M0 in dyne.cm.

As stress drop increases for an earthquake of a given moment, so does fc and the high-frequency
level of the source spectrum. Being so, the parameters that take values appropriate to each
earthquake source spectrum are seismic moment and stress drop.

On the basis of w2 model, the seismic moment and the corner frequency can be obtained by the
horizontal and f-2 lines from the Source displacement-amplitude spectra. Rovelli et al [1991]
suggests that one should avoid visual inspection or the simultaneous inversion of the ground
motion spectrum in terms of both seismic moment and corner frequency and computed the corner
frequency by means of an objective technique using the expression of Andrews (1986):


2
 V ( f )df
1
fc  0 (18)
2 
2
 D ( f )df
0

where V (f) and D (f) are the source velocity and displacement spectra. Knowing fc and the seismic
moment, the stress drop can be obtained using eq. 17.

Another technique is to estimate simultaneously seismic moment and stress drop using the
theoretical function of the source spectrum (eq 14) and fit it to the source spectra of S waves for an
assumed seismic moment with a constant stress drop (example in Figure 4). The seismic moment
is obtained from the source spectral amplitude at low frequency while searching to obtain Δσ at
high frequencies.

We have used the three techniques (visual inspection, eq. 18 and the fit of source theoretical
function to the source S-wave spectra) to obtain source parameters for each earthquake.

441
For the set of earthquakes the average stress drop is 101 bars for intraplate events and 66 bars for
interplates events.

1.E-01 1.E+00

1.E-02 1.E-01
AF [m.s]

AF [m.s]
1.E-03 1.E-02

1.E-04 #4 1.E-03 #6
SVF SVF
1.E-05 1.E-04
0.1 1 10 100 0.1 1 10 100
Freq. [Hz] Freq. [Hz]

Fig. 4 Example of the source spectra of two events. The smooth curve represents the theoretical
model of eq. 14 that was fit to the envelope of the source spectrum.

Duration:

The duration T of an earthquake signal at a hypocentral distance R can be represented as [Atkinson

& Boore, 1995]

T ( R)  T0  bR (19)

where T0 is the source duration (here taken as 1/fc) and b is a coefficient controlling the increase of
duration with distance. Carvalho [2007] obtained b=0.1 for intraplate events and b=0.015 for
interplates events.

Results of model parameters are represented in table 1 and figure 5 shows some examples of
response spectra comparison between synthetic and observed. In general, the synthetics are in
good agreement with observations for at least one of the recorded horizontal components.

Table1: Model parameters

Intraplate Interplate
Crustal thickness, D [km] 31 20
Quality factor, Q(f) 250*f0.7
1/R (R  1.5D km)
Geometric attenuation 1/R0 (1.5D km< R  2.5D km)
1/R0.5 ( R > 2.5D km)
Distance-dependent duration 0.1 R 0.015 R
k [s] 0.032
Shear-wave velocity,  3.5
[km/s]
Crustal density,  [g/cm3] 2.8
Stress drop,  [bars] 101 66
2
Fzi ( )   zi
2
Amplification function  
1   zi 
Fzi(ω) 
fzi=0.5HZ

442
 100 10 100

10 10

Acc [cm/s^2]
Acc [cm/s^2]
1
Acc [cm/s^2] 1 1

0.1
0.1 0.1
#3 #4 #8
SVF SVF BEN
0.01 0.01 0.01
0.01 0.1 1 10 0.01 0.1 1 10 0.01 0.1 1 10
Períod [s] Períod [s] Períod [s]

100 100 100

10 10 10

Acc [cm/s^2]
Acc [cm/s^2]

Acc [cm/s^2]

1 1 1

0.1 0.1 0.1

#6 #6 #7
SVF SVI BEN
0.01 0.01 0.01
0.01 0.1 1 10 0.01 0.1 1 10 0.01 0.1 1 10
Períod [s] Period [s] Period [s]

Fig. 5 Example of response spectra comparison between synthetic and observed data. Black lines
are the simulated results, gray lines are both horizontal components of the recorded accelerogram.
(simple line: component N-S; bold line: component E-W). Stations: SVF – S. Vicente Fora
(Lisbon); BEN – Benavente (Benavente); SVI – S. Vicente (Algarve).

1755 Lisbon earthquake

Source models

The 1755.11.01 earthquake, known as the 1755 Lisbon earthquake, generated the largest known
tsunami in SW Europe and its magnitude has been estimated as Mw= 8.5 – 8.9 by several authors
[e.g. Abe, 1979; Moreira, 1984]. The exact location remains controversial, even though the
earthquake epicentre is know to have been offshore. For the localization of the major seismogenic
zones in the SW Iberia Margin, please refer to figure 1 of Ribeiro et al. [2008] (this volume).

The several isoseismal maps published led to the conclusion that the source location of this event
was in the vicinity of Gorringe Bank (GBF) [Martinez Solares, 1979; Levret, 1991]. This location
was further supported by the occurrence of a tsunamigenic earthquake on 1969.02.28. Johnston
[1996] suggested that a NE-SW trending fault, at the base of the NW flank of Gorringe Bank,
could be responsible for the 1755 earthquake.

Baptista et al.[1998] performed a hydrodynamic modelling of the 1755 tsunami. Results from a
backward ray tracing simulations suggest a tsunami source located quite close to the Portuguese
443
coast. In order to precisely locate the 1755 seismogenic source, in 1998 the area between the
Gorringe Bank and the Cape St. Vicente has been surveyed within the framework of the European
BIGSETS projects (Big Sources of Earthquake and Tsunami in SW Iberia). One of the main
results [Zitellini et al., 2001] was the characterization of the active tectonic structure located
offshore Cape St. Vicente named as Marques Pombal Thrust fault (MPT) which, accordingly to
the authors, could be the generator of the 1755 Lisbon earthquake.

Terrinha et al. [2003] pointed out that additional rupture areas have to be associated with the MPT
system to generate such a destructive earthquake and that the 65 km long Pereira de Sousa Normal
Fault (PSNF) located along the Portuguese margin, to the north of MPT, could constitute the
northward prolongation of the MPT.

Baptista et al. [2003], based on new MCS (multi-channel seismic reflections surveys) data,
presented a new reappraisal of the 1755 source and proposed a possible composite source
including the Marques de Pombal Thrust fault and the Guadalquivir Bank (NGBF), which,
accordingly to authors, will act as a possible southeastward extension of the rupture area related to
the 1755 event.

Vilanova et al [2003] propose that, although the main shock was offshore, the resulting static
stress changes induced the rupture of the Lower Tagus Valley (LTVF), near Lisbon. They favour
this model, rather that site effects causing high intensities in the Lisbon area, because not only the
highest intensities show a negative correlation to soft soil but also because this local rupture can
explain other phenomena described in the eyewitness accounts like an internal tsunami in the
Tagus River, ground deformation affecting the course of the Tagus River, the spatial pattern of
damaging aftershocks, duration of the event and the number of shocks felt.

Gutscher [2004] proposed an alternative area for the 1755 event, a shallow east dipping fault plane
beneath the Gulf of Cadiz associated with subduction beneath Gibraltar.

Zitellini et al. [2004] and Ribeiro et al. [2006] proposed the Horseshoe Thrust Fault (HF) as an
additional source area to the Marquês de Pombal Thrust. They favour this solution for additional
area/slip for the 1755 event, given its orientation sub parallel to the MPT and the almost geometric
continuity between both fault zones that facilitates the strain/displacement transfer between them.

Gradin et al. [2007], using a velocity model valid for frequencies up to 0,3Hz and validated for the
1969.02.28 earthquake, proposed a regression to convert simulated values of Peak Ground
Velocity (PGV) into Modified Mercalli Intensity (MMI) in SW Iberia. They extrapolate the
method and the empirical relationship PGV-IMM to the case of the 1755 earthquake, and tested
three sources proposed by different authors, namely the Gorringe Bank solution, the Marques de
Pombal – Pereira de Sousa fault system and the subduction in the Gulf of Cádiz. The authors
concluded that rupture directivity plays a particularly important role and that a primary source
located at Gorringe Bank is the most realistic hypothesis to fit the observed isoseismal pattern.

444
Modeling geometry

Using the stochastic finite fault method explained above, we have tested the six different fault
source geometries for the source of the 1755 earthquake proposed by the different authors, and
mentioned in the previous section.

For all geometries, fault segments were divided into smaller subfaults, each one considered as a
point source. The slip distribution is randomized, and for each fault model three nucleation points
were considered, to take in account directivity studies. The geometry of each source model is
presented in Table 2 and the fault source geometries and the different nucleation points considered.
are presented in the next figures. The geometry named GBF considers the model of Johnston
[1996], MPT-PS considers the proposal of Terrinha et al. [2003], the model MPT-GqB considers
the study of Baptista et al. [2003], model off-LVTF, considers besides the main shock offshore
(Gorringe Bank or Marquês Pombal Fault), a second earthquake in Lower Tagus Valley Fault
(LTVF) as proposed by Vilanova et al. [2003], the MPT – HF considers the Horseshoe Thrust
Fault as an additional source area to the Marquês de Pombal Thrust Fault and the geometry model
GC considers the proposal of Gutscher [2004]. The numbers presented in each fault geometry
represent the different nucleation points considered in simulations.

Table2: Source parameters: fault geometry (strike, dip and dimensions) and magnitude considered

Source Strike Dip (L x W) [km] M

CBF N60ºE 40º 200 x 80 8.7
MPT N20ºE 24º 100 x 70 8.3
PSNF N180ºE 24º 100 x 120 8.45
MPT N21,7ºE 24º 100 x 55 8.3
GqB N70ºE 45º 96 x 55 8.3
HF N60E 45º 175 x 140
8.3
MPT N20ºE 24º 100 x 70
LVTF N38ºE 55º 40 x 28 6.7
2,5º 162 x 68 8.4
GC N11ºW 5,0º 174 x 68 8.45
7,5º 198 x 6.8 8.5

Surface modeling

For each Portuguese council, the response spectrum derived from each source model, was
transformed into Power Spectrum Density Function (PSDF), at the bedrock, using the classical
theory of stationary random process already described. Site effects were, then, evaluated following
the numerical approach explained elsewhere in this paper and using an available data base (herein
DB1), containing information about soil classification for Portugal (classified as 0, 1 and 2 for
rock, medium and soft soil respectively – Oliveira et al., 1995) and a data base (DB2) with
stratified soil profile units for the Metropolitan Area of Lisbon (MAL). In DB2, each soil unit
considers the thickness of shallow layers, shear waves velocity, density and plastic index. This

445
data base was built in the framework of the project conducted by the Portuguese civil protection
authority, for which it was conducted a geological - geotechnical survey that allowed the
characterization of stratified soil profile units for MAL with a fair level of detail. Each classified
soil type (0, 1 or 2, accordingly to DB1) for MAL correspond to a particular stratified soil profile
accordingly to DB2 and we extrapolated this soil profile units for every Portuguese counties that
had the same soil type classification.

Results and discussion

For each fault geometry seismic action was computed at the bedrock level and surface. Results for
each source model considered, for each nucleation point, are shown in figures 6 to 10. PGAb and
PGAs means Peak Ground Acceleration at bedrock level and surface, respectively, PGVs stands for
Peak Ground Velocity at surface and IMM for the Modified Mercalli Intensity. Synthetic
isoseismal maps were built using the regression of Gradin et al. [2007] that convert PGV into
IMM in SW Iberia.

It is emphasized the primordial effects of rupture directions on strong ground motions, as also
shown in Gradin et al. [2007], which remarks the importance of using methodologies that can
account for directivity effects as already mentioned in Carvalho et al. [2007].

An important conclusion of these figures is that the soil effects play a particularly important role in
the pattern of isoseismals. In fact, even when peak ground values maps show a radial pattern, the
surface maps reproduce most features of the observed macroseismic field, for most of the source
models.

Our results do not sustained a solution of a source located below Gorringe Bank, which produce
too low values of IMM in south of Portugal. The composite models that considers the Marques the
Pombal fault and a additional source area ( MPT-GqB, MPT-PSNF, MPT-HF) seems good
hypothesis, considering an upward rupture, towards Lisbon City, the Guadalquivir Bank source
model reproducing higher intensities in Algarve. The subduction zone in Gulf of Cádiz fails to
reproduce the isoseismal features of the 1755 earthquake in western Portugal.

Considering a triggered rupture in Lower Tagus Valley Fault, a M6.7 seems a too high magnitude
for the inland earthquake (similar PGA values in Lisbon and Algarve) but a smaller magnitude, in
order of M6.5 can give satisfactory PGA values in Lisbon. Nevertheless, PGV values in Lisbon are
caused by the offshore source and not by local source which means, considering PGV the ideal
choice among the ground-motion parameters for damage direct relation, and as a consequence, that
correlates best with intensity [Atkinson & Kaka, 2007; Gerstenberg et al., 2005], that there is no
need to look for a trigger solution to justify high intensities in Lisbon area.

446
PN PGAb PGAs PGVs IMM

PG A [c m /s ^2 ]

0 - 1 00 IMM
100 - 15 0 < 6,0
150 - 20 0
6,0 - 6, 5
200 - 25PG 0 A [c m /s ^2 ] IMM
250 - 30 0 6,5 - 7, 0
PG A [c m /s ^2 ] 0 -PG V [c m /s ]
1 00
>3 00 7,0 - 7,<5 6,0
0 - 1 00 100 - 150 0- 1 0 IMM
7,5 - 8,6,0
0 - 6, 5
100 - 15 0 150 - 2010-
0 15
< 6,0
150 - 20 0 200 - 25150 - 20 8,0 - 8,6,5
5 - 7, 0
6,0 - 6, 57,0 - 7, 5
200 - 25 0 250 - 30200 - 25 8,5 - 9, 0
250 - 30 0 300 - 45250 - 30 6,5 - 7, 07,5 - 8, 0
PG V [c m /s ] 9,0 - 9, 5
>3 00 >3 0 7,0 - 7, 5
0 - 10 9,5 - 108,0
,0 - 8, 5
7,5 - 8, 0
10- 1 5 8,5 - 9, 0
3
N
15 - 20 8,0 - 8, 5
9,0 - 9, 5
2 20 - 25 8,5 - 9, 0
9,5 - 10 ,0
1 25 - 30 9,0 - 9, 5
>3 0
9,5 - 10 ,0
Fig. 6. Synthetic PGAb, PGAs, PGVs and isoseismal maps, for nucleation points 1 to 3, using the
source model GBF.

447
 PGAb PGAs PGVs IMM

3
PG A [c m /s ^2 ]

0 - 1 00 IMM
100 - 15 0 < 6,0
150 - 20 0
6,0 - IMM
6, 5
200 - 25 0 PG A [c m /s ^2 ]
250 - 30 0 0 - 1 00 6,5 -< 7,
6,0
0
PG A [c m PG V [c m /s ]
>3/s00
^2 ] 100 - 15 0 7,0 -6,0
7, 5
- 6, 5
0 - 1 00 0 -0 1 0
150 - 20 IMM
7,5 -6,5
8, 0
- 7, 0
200 - 25
10-0 15
100 - 15 0 < 6,0
250 - 30
150 - 20 8,0 -7,0 - 7, 5
8, 5
150 - 20 0
300 - 45 0 - 6,-7,5
6,08,5 - 8, 0
59, 0
200 - 25 0 20 - 25
250 - 30 0 25 - 30 - 7,-8,0
6,59,0 - 8, 5
09, 5
3
PG V [c m
>3/s0]
>3 00 N - 7,-8,5
7,09,5 510-,09, 0
2
0 - 10 9,0 - 9, 5
7,5 - 8, 0
10- 1 5 9,5 - 10 ,0
1
15 - 20 8,0 - 8, 5
20 - 25 8,5 - 9, 0
25 - 30 9,0 - 9, 5
>3 0
Fig. 7. Synthetic PGAb, PGAs, PGVs and isoseismal maps, for nucleation points
9,5 -110
to,03, using the
source model MPT-PSNF.

448
 PN PGAb PGAs PGVs IMM

PG A [c m /s ^2 ]

0 - 1 00 IMM
100 - 15 0 < 6,0
150 - 20 0
6,0 - 6, 5
200 - 25PG0A [c m /s ^2 ] IMM
6,5 - 7, 0
PG A [c m /s ^2 ] 0
250 - 30
>3 00 0 - 1PG
00 V [c m /s ] 7,0 - <7,6,0
5
0 - 1 00 IMM
100 - 15 00 - 1 0
100 - 15 0 7,5 - 8, 0 - 6, 5
6,0
150 - 20 010- 1 5 < 6,0
150 - 20 0 200 - 25 015 - 20 8,0 - 6,5
8, 5 - 7, 0
6,0 - 6, 5
200 - 25 0 250 - 30 020 - 25 8,5 - 9, 0 - 7, 5
7,0
250 - 30 0 6,5 - 7, 0
300 [c m025
PG V- 45 /s ]- 30 9,0 - 7,5
9, 5 - 8, 0
>3 00 >3 0 7,0 - 7, 5
0 - 10 9,5 - 8,0
10 ,0- 8, 5
7,5 - 8, 0
10- 1 5 8,5 - 9, 0
N 8,0 - 8, 5
3 15 - 20
9,0 - 9, 5
20 - 25 8,5 - 9, 0
2 1 25 - 30 9,5 - 10 ,0
9,0 - 9, 5
>3 0
9,5 - 10 ,0

Fig. 8. Synthetic PGAb, PGAs, PGVs and isoseismal maps, for nucleation points 1 to 3, using the
source model MPT-GqB.

449
 PN PGAb PGAs PGVs IMM

PG A [c m /s ^2 ]

0 - 1 00 IMM
100 - 15 0 < 6,0
150 - 20 0
6,0 - 6, 5
200 - 25 0
PG A [c m /s ^2 ] IMM
250 - 30 0 6,5 - 7, 0
PG A [c m /s ^2 ] PG V [c m /s ]
>3 00 0 - 1 00 < 6,0
7,0 - 7, 5
0 - 1 00 IMM
100 - 15 0 0 - 10
6,0 - -8,6,05
7,5
100 - 15 0 150 - 20 0 10- 1 5 < 6,0
150 - 20 0 6,5 - -8,7,50
8,0
200 - 25 0 15 - 20
6,0 - 6, 5
200 - 25 0 250 - 30 0 20 - 25 7,0 - -9,7,05
8,5
250 - 30 0 25 - 30 6,5 - 7, 0
300
PG V- 45
[c m0 /s ] 7,5 - -9,8,50
9,0
>3 00 >3 0 7,0 - 7, 5
0 - 10 9,5
8,0 - -10
8,,05
3 7,5 - 8, 0
10- 1 5 8,5 - 9, 0
N 8,0 - 8, 5
15 - 20
2 9,0 - 9, 5
20 - 25 8,5 - 9, 0
25 - 30 9,5 - 10 ,0
9,0 - 9, 5
1 >3 0
9,5 - 10 ,0

Fig. 9. Synthetic PGAb, PGAs, PGVs and isoseismal maps, for nucleation points 1 to 3, using the
source model MPT-HF.

450
 PN PGAb PGAs PGVs IMM

2
PG A [c m /s ^2 ]

0 - 1 00 IMM
100 - 15 0 < 6,0
150 - 20 0 PG A [c m /s ^2 ] IMM
6,0 - 6, 5
200 - 25 0 0 - 1 00 < 6,0
250 - 30 0 100 - 15 0 6,5 6,0
- 7,- 06, 5
PG A [c m /s ^2 ] PG
150 V- 20
[c m
0 /s ]
>3 00 200 - 25 0 7,0 6,5
- 7,- 57, 0
0 - 1 00 IMM
250 - 0
30-0 1 0 7,0 - 7, 5
7,5 - 8, 0
100 - 15 0 300 - 10-
45 0 1 5 < 6,0 7,5 - 8, 0
150 - 20 0 15 - 20 8,0 - 8, 5
6,0 - 6,8,0
5 - 8, 5
200 - 25 0 20 - 25 8,5 8,5
- 9,- 09, 0
N 6,5 - 7, 0
250 - 30 0 25 - 30
2 PG V [c m /s ] 9,0 9,0
- 9,- 59, 5
1 >3 00 >3 0 7,0 - 7,9,5
5 - 10 ,0
0 - 10 9,5 - 10 ,0
7,5 - 8, 0
Fig. 10. Synthetic PGAb, PGAs, PGVs and isoseismal maps,10-for
1 5 nucleation points 1 and 2, using
15 - 20 8,0 - 8, 5
UP: the source model GC; DOWN: the source model Off-LVT
20 - 25 8,5 - 9, 0
25 - 30
PN PGAb PGAs PGVs IMM9,0 - 9, 5
>3 0
9,5 - 10 ,0

2
PG A [c m /s ^2 ]

0 - 1 00 IMM
100 - 15 0 < 6,0
150 - 20 0PG A [c m /s ^2 ] IMM

0 - 1 00
6,0 - 6, 5
200 - 25 0 < 6,0
250 - 30 0 100 - 15 0 6,5 - 6,0
7, 0- 6, 5
PG A [c m /s ^2 ] PG- V20[c0m /s ]
150
>3 00 7,0 - 6,5
7, 5- 7, 0
200 - 25 0 IMM
0 - 1 00 0 - 10 7,0 - 7, 5
250 - 30 0 7,5 - 8, 0
100 - 15 0 10-
300 - 45 0 15 < 6,0 7,5 - 8, 0
150 - 20 0 15 - 20 8,0 - 8, 5
8,0 - 8, 5
6,0 - 6, 5
200 - 25 0 20 - 25 8,5 - 8,5
9, 0- 9, 0
250 - 30 0 N
25 - 30 6,5 - 7, 0
PG V [c m /s ] 9,0 - 9,0
9, 5- 9, 5
>3 00 >3 0 7,0 - 7, 5
2 1
0 - 10 9,5 - 9,5 - 10 ,0
10 ,0
7,5 - 8, 0
10- 1 5
15 - 20 8,0 - 8, 5
451
20 - 25 8,5 - 9, 0
25 - 30 9,0 - 9, 5
>3 0
9,5 - 10 ,0
Conclusions

A non – stationary stochastic method was applied to test four different models that have been
taken as possible for the source of the 1755 earthquake. Source parameters of the model (corner
frequency, stress drop) were estimated based on source spectra of S waves of 12 earthquakes
records (corresponding to a total of 25 records) and the upper crustal attenuation parameter was
estimated from the high-frequency decay of the acceleration spectra, after correcting observed
spectra from geometrical spreading and anelastic attenuation effect.

Site effects were evaluated by means of an equivalent stochastic non-linear one-dimensional

ground response analysis of stratified soil profile using the LNECloss system and using an
available data base (herein DB1), containing information about soil classification for Portugal and
a data base with stratified soil profile units for the Metropolitan Area of Lisbon (MAL).

Analyzing the most complete study performed here and its results we conclude that (i) the soil
effects play a particularly important role in the pattern of isoseismals; (ii) the composite models
that considers the Marques the Pombal fault and a additional source area ( MPT-GqB, MPT-PSNF,
MPT-HF) seems good hypothesis, considering an upward rupture, towards Lisbon City, the
Guadalquivir Bank source model reproducing higher intensities in Algarve; (iii) The subduction
zone in Gulf of Cádiz fails to reproduce the isoseismal features of the 1755 earthquake in western
Portugal; (iv) The high intensities in Lisbon area, in Lusitania Basisn and Lower Tagus valley
Basin can be explained as site effects, together with a directivity effect (upward direction) from an
offshore source, considering that PGV correlates best with intensity.

Other considerations, namely the possibility of each source to produce a tsunami, the time arrivals
of tsunami, the geological evidences, the consequent damage and so on, are left to others.

Current methods of predicting ground motions for future earthquakes in Portugal will/should be
based on an assumed seismological model of source and propagation processes.

Acknowledgments

This study was performed in the framework of the European Commission through the LESSLOSS
FP6 Integrated Project Risk Mitigation for Earthquakes and Landslides, No.: GOCE-CT-2003-
505488.

References
Abe K. [1979] “Size of great earthquakes of 1837 – 1974 inferred from tsunami data.” J. Geophys.
Res., 84:1561-1568.
Anderson, J.; Hough, S.; [1984]: “A model for the shape of the Fourier amplitude spectrum of
acceleration at high frequencies”. Bull. Seism. Soc. Am, V. 74, 1969-1993.

452
Andrews, D. J.; [1986]: “Objective determination of source parameters and similarity of
earthquakes of different size”, in Earthquake Source Mechanics.. Maurice Ewing Series 6, S. Dias,
J. Boatwright, and C. H. Scholz (Editors); American Geophysical Union; Washington, D. C., pp.
259-267.
Atkinson, G.; Boore, D.; [1995]:”New ground motion relations for eastern North America”. Bull.
Seism. Soc. Am.; Vol. 85, pp. 17–30.
Baptista M. A., Miranda P. A., Miranda J. M., Mendes Victor L. A. [1998] “Constrains on the
source of the 1755 Lisbon Tsunami inferred from numerical modelling of historical data on the
source of the 1755 Lisbon Tsunami.” J. Geodynamics; 25(2): 159-174.
Baptista, M.A., Miranda, J.M., Chierici, F., Zitellini, N. [2003] “New study of the 1755 earthquake
source based on multi-channel seismic survey data and tsunami modeling,” Natural Hazards and
Earth System Sciences, 3, pp. 333-340.
Beresnev, I.A.; Atkinson, G.M. [1998] “FINSIM - a FORTRAN program for simulating stochastic
acceleration time histories from finite fault,” Seism. Res. Lett., 69, pp. 27-52.
Boore, D.M. [2003] “Simulation of ground motion using the stochastic method,” Pure Appl.
Geophys. 160, pp. 635-676.
Boore, D.; Joyner, W. [1984]. “A note on the use of random vibration theory to predict peak
amplitudes of transient signals”. Bull. Seism. Soc. Am, V. 74, 2035-2039
Brune J. N. [1970].“Tectonic stress and the spectra of seismic shear waves from earthquakes.” J.
Geophys. Res. 75: 4997-5009.
Brune J. N. [1971] “Correction.” J. Geophys. Res. 76: 5002.
Carvalho, A.; Campos Costa, A.; Oliveira, C. S. [2004] “A stochastic finite-fault modeling for the
1755 Lisbon earthquake,” 13th World Conference on Earthquake Engineering Vancouver, B.C.,
Canada August 1-6, 2004 Paper No. 2194
Carvalho, A.; Campos Costa, A.; Oliveira, C. S.; [2005]. “ A Finite-fault modeling of the 1755
Lisbon earthquake sources”. International Conference 250th Anniversary of the 1755 Lisbon
Earthquake, pp 578 – 583, Lisboa,1-4 November.
Carvalho A.; Zonno, G.; Franceschina, G.; Bilé Serra, J.; Campos Costa, A.; [2007]. “Earthquake
shaking scenarios for the metropolitan area of Lisbon”. Soil Dyn. Earthquake Eng. doi: 10.1016/
j.soildyn.2007.07.009
Dèzes, P.; Ziegler, P. A.; [2001]:”European Map of the Mohorovicic discontinuity”. 2nd EUCOR-
URGENT workshop; Mte. St. Odile; France.
Duarte, T.; [1978]: ”Uma teoria linear do comportamento dinâmico das estruturas sujeitas a acções
sísmicas”. Relatório integrado nos estudos do domínio do PIAP DE/LNEC; Laboratório Nacional
de Engenharia Civil; Lisboa.

Ferrer,I.; Sánchez-Carratalá, R.[2004] “Application of non-stationary seismological models to the

determination of stochastic response spectra”, 13th World Conference on Earthquake Engineering
Vancouver, B.C., Canada August 1-6, 2004 Paper No. 2280
Grandin, R., Borges, J.F., Bezzeghoud, M., Caldeira, B. & Carrilho, F., [2007]: ”Simulations of
strong ground motion in SW Iberia for the 1969 February 28 (MS = 8.0) and the 1755 November 1
(M ~ 8.5) earthquakes – II. Strong ground motion simulations”, Geophys. J. Int.,
doi:10.1111/j.1365- 246X.2007.0357.x.
Gutscher, Marc-André; [2004]:”What caused the Great Lisbon Earthquake?”. Science; Vol. 305,
pp. 1247-1248.
Jiménez-Munt, I.; Fernàndez, M.; Torne, M.; Bird, P.;[ 2001] : ”The transition from linear to
diffuse plate boundary in the Azores-Gibraltar region: results from a thin-sheet model”. Earth and
Planetary Science Letters; 192, pp. 175-189.
Johnston, Arch C.; [1996]:”Seismic moment assessment of earthquakes in stable continental
regions – III. New Madrid 1811-1812, Charleston 1886 and Lisbon 1755”. Geophys. J. Int.; Vol.
126, pp. 314-344.

453
Levret A. [1991] “The effects of the November 1, 1755 Lisbon earthquake in Morocco.”
Tectonophysics; 193: 83-94.
Martinez Solares J. M., Arroyo L., Mezcua J. [1979] “Isosseismal map of the 1755 Lisbon
earthquake obtained from spanish data.” Tectonophysics, 56(3): 301-313.
Moreira V. [1984] “Sismicidade histórica de Portugal continental” Sup. Revista Instituto Nacional
de Meteorologia e Geofísica, Lisboa (in portuguese).
Pujades, L. G.; Canas, J. A.; Egozcue, J. J.; Puigví, M. A.; Gallart, J.; Lana, X.; Pous, J.; Casas,
A.; [1990]:”Coda-Q distribution in the Iberian Peninsula”. Geophys. J. Int.; 100, pp. 285-301.
Ribeiro, A.; Mendes-Victor, L.; Cabral, J.; Matias, L.; Terrinha, P.; [2006]:”The 1755 Lisbon
earthquake and the beginning of closure of the Atlantic”. European Review; Vol. 14; N. 2, pp.
193-205.
Rovelli, A.; Cocco, M.; Console, R.; Alessandrini, B.; Mazza, S.; [1991]: ”Ground motion
waveforms and source spectral scaling from close-distance accelerograms in a compressional
regime area (Friuli, northeastern Italy)”. Bull. Seism. Soc. Am.; Vol 81; N. 1, pp. 57-80.
Ribeiro, A., Mendes-Victor, L., Matias, L., Terrinha, P., Cabral J., Zitellini, N. [2008]: “The 1755
Lisbon earthquake: a review and the proposal for a tsunami early warning system in the Gulf of
Cadiz”, this volume.
Sousa, M.L., Campos Costa A., Carvalho, A., Coelho, E. [2004] “An automatic seismic scenario
loss methodology integrated on a geographic information system,” 13th World Conference on
Earthquake Engineering Vancouver, B.C., Canada August 1-6, 2004, No. 2526
Terrinha P., Pinheiro L. M., Henriet J. P., Matias L., Ivanov M. K., Monteiro J. H., Akhmetzhanov
A., Volkonskaya A., Cunha T., Shaskin P., Rovere, M. [2003] “Tsunamigenic – seismogenic
structure, neotectonics, sedimentary process and slope instability on the southwest Portuguese
margin.” Marine Geology ; 195: 55-73.
Vanmarcke E. H. [1976] “Strutural response to earthquakes”, Seismic Risk and Engineering
Decisions, 1976; C.Lomnitz and E. Rosenblueth, Reditors, Elsevier.
Vilanova, S., Catarina, F., Nunes, Fonseca, J.F.B.D. [2003] “Lisbon 1755: A Case of Triggered
Onshore Rupture?,” Bull. Seismol. Soc. Am, Vol. 93, No. 5, pp. 2056-2068.
Zitellini, N., Mendes, L. A., BIGSETS Team [2001] “Source of 1755 Lisbon earthquake, tsunami
investigated,” EOS 82, No. 26, pp. 285–291.
Zitellini, N.; Rovere, M.; Terrinha, P.; Chierici, F.; Matias, L.; BIGSETS Team; [2004]: “Neogene
Through Quaternary Tectonic Reactivation of SW Iberian Passive Margin”. Pure and Apl.
Geophis.; 161, pp. 567-587.
Zonno, G.; Carvalho, A.; Franceschina, G.; Akinci, A.; Campos Costa, A.; Coelho, E.; Cultrera,
G.; Pacor, F.; pessina, V.; Cocco, M. [2005] “ Simulating erthquake scenarios using finite-fault
model for the Metropolitan Area of Lisbon (MAL)”,

454