Tsunami Mapping Related to Local Earthquakes on the French–Italian Riviera (Western

Mediterranean)

MANSOUR IOUALALEN,1 CHRISTOPHE LARROQUE,2 OONA SCOTTI,3 and CAMILLE DAUBORD4

Abstract—The Ligurian coast, located at the French–Italian

border, is densely populated as well as a touristic area. It is also a

location where earthquakes and underwater landslides are recur-

rent. The nature of the local tsunamigenesis is therefore a

legitimate question, because no tsunami warning system can

resolve tsunami arrival times of a few minutes, which is the case

for the area. As far as the seismicity of the area is concerned, the

frequent recurrent earthquakes are generally of moderate magni-

tude: most of them are lower than Mw 5. However, the relatively

large Mw 6.9 earthquake (Larroque et al., in Geophys J Int, 2012.

doi:10.1111/j.1365-246X.2012.05498.x) that occurred on the Feb-

ruary 23, 1887, offshore of Imperia (Italian Riviera) is quite

emblematic. This unusual event for the region merits a complete

study: the quantification of its rupture mechanism is essential (1) to

understand the regional active deformation, but also (2) to evaluate

its tsunamigenesis potential by deriving relevant rupture scenarios

obtained from our knowledge of the event; for that purpose the

event is extensively described here. The first point has been the

subject of quite a few studies based on the seismotectonics of the

area. The last documented approach has been completed by Lar-

roque et al. (Geophys J Int, 2012. doi:10.1111/j.1365-246X.2012.

05498.x) who proposed a rupture scenario involving a reverse

faulting along a north dipping fault and favoring a Mw 6.9 mag-

nitude. In the present paper (1) we study the accuracy of their

solutions in relation to the computational grid spacing and the

dispersive/nondispersive parameterization, (2) based on an uncer-

tainty on the recorded wave amplitude of the Genoa tide gauge they

used, we propose a Mw 6.7 earthquake magnitude solution for the

event (the kinematics is unchanged), co-existing with the Mw 6.9,

(3) we evaluate the tsunami coastal impact of the 1887 event, and

(4) we test a range of possible ruptures that local faults may

undergo in order to propose a synoptic mapping of the tsunami

threat in the area. The spatial distribution of the maximum wave

height (MWH) is provided with a tentative identification of the

processes that are responsible for it. This latter issue is imperative

in order to make our mapping as generic as possible in the

framework of our deterministic approach (based on realistic sce-

narios and not on ensemble statistics). The predictions suggest that

the wave impact is mostly local, considering the relatively mod-

erate size of the rupture planes. Although the present-day

seismicity in this region is moderate, stronger earthquakes

(M [ 6.5) have occurred in the past. The studied scenarios show

that for such events specific localities along the French–Italian

Riviera may experience very significant MWH related to the

shallow focal depth tested for such scenarios. We may reasonably

conclude that the tsunami threat is relatively significant and uni-

form at the Italian side of the Riviera (from Ventimiglia to

Imperia), while it is more localized (sporadic) at the French side

from Antibes to Menton with, however, higher local level of

inundation, e.g., Nice city center.

Key words: Earthquake, active fault, tsunami, Mediterranean,

Ligurian coast.

1. Introduction

The Ligurian coast (Fig. 1), located at the

boundary of the southwestern Alps on each side of

the French–Italian border, is a densely populated and

industrial area. The geological context is one of the

most active among the western Mediterranean, with

recurrent earthquakes, onshore and offshore land-

slides, and tsunami (EVA and RABINOVICH 1997;

LARROQUE et al. 2001; CAROBENE and CEVASCO 2011;

MIGEON et al. 2011). During the last 500 years, a

dozen of tsunamis have been reported (TINTI et al.

2004; LAMBERT and TERRIER 2011). Most of them

have been certainly triggered by earthquakes with the

noticeable exception of the October 16th 1979 event

which was followed by a shallow submarine landslide

located on the continental slope of the northern Lig-

urian margin, in the immediate vicinity of Nice city

international airport (Fig. 1).

1 Institut de Recherche Pour le Developpement, IRD, UMR

Geoazur 6526, CNRS-IRD-UNS-OCA, 250 Rue A. Einstein, 06560

Valbonne, France. E-mail: Mansour.Ioualalen@geoazur.unice.fr2 UMR Geoazur 6526, CNRS-IRD-UPMC-UNS-OCA, 250

Rue A. Einstein, 06560 Valbonne, France.3 Institut de Recherches et de Surete Nucleaire, BERSSIN,

B.P. 17, 92262 Fontenay-aux-Roses Cedex, France.4 Service Hydrographique et Oceanographique de la Marine,

13 Rue du Chatellier, 29228 Brest, France.

Pure Appl. Geophys.

� 2013 Springer Basel

DOI 10.1007/s00024-013-0699-1 Pure and Applied Geophysics

The northern Ligurian margin is peculiar because

it hosts a very narrow shelf and a steep continental

slope (Fig. 1). As a result, underwater landslides are

very recurrent in the area and numerous slide scars

may be observed (MIGEON et al. 2011). IOUALALEN

et al. (2010) estimated the tsunami hazard due to

recurrent landslides occurring in the area. The present

paper fills a gap in the tsunami hazard mapping by

proposing earthquake triggered tsunami mapping

along the Ligurian coast based on the recent study of

the seismotectonic activity in the area and on the

analysis of the strongest 1887 regional event.

Currently, the area encounters recurrent earth-

quakes, although most of them are of moderate

magnitude (BETHOUX et al. 1992; LARROQUE et al.

2001). Therefore, the local earthquake activity is not

likely to be tsunamigenic. Nevertheless, some strong

earthquakes were recorded in historical times, such as

the February 23, 1887, event that occurred offshore of

Imperia, Italy (FERRARI 1991), the magnitude of

which has been recently reappraised around 6.9

(LARROQUE et al. 2009, 2012, hereafter LSI12 for the

latter). This earthquake was quite damaging (FERRARI

1991; LAURENTI 1998) and induced a tsunami

observed along more than 200 km of the Ligurian

coast (DENZA 1887; EVA and RABINOVICH 1997).

Based on this information, the 1887 earthquake/

tsunami appears to be essential for the Ligurian

coastal area in terms of natural risk: it is the most

destructive earthquake ever recorded in the area and

this is the most significant tsunami triggered by

earthquakes recorded in the northern part of the

western Mediterranean. For these reasons, there is a

crucial need to study the event in detail. So far, the

most detailed study regarding the earthquake event

has been recently proposed by LSI12. They gathered

all the morphotectonic and macroseismic information

and used available hydrographic information (recor-

ded tsunami signal) to propose the best earthquake

scenario in terms of rupture parameters (fault location

and geometry and area of the rupture plane). For that

purpose, LSI12 analysis helped in reducing

Figure 1DEM of the northern Ligurian margin derived from high-resolution bathymetric data collected during the MALISAR geophysical surveys.

The Ligurian Faults system (LFS) is in white. The inset shows the location of the northern Ligurian margin at the junction between the Alps

and the western Mediterranean

M. Ioualalen et al. Pure Appl. Geophys.

considerably the range of parameters necessary to

explain the rupture process. Then the short-listed

ranges of rupture parameters were submitted to a

tsunami numerical modeling procedure in order to

tentatively identify the possible(s) mechanism(s) of

dislocation. For that purpose, the 1887 tide gauge

record at Genoa harbor has been processed along with

a tsunami modeling methodology. Another tide gauge

record was available at Nice harbor but was not used

because a sudden drop in pressure was difficult to

explain and subsequently filter.

We want to complete the previous work, and we

invite the reader to refer to the papers LARROQUE et al.

(2011) and LSI12 for a full description of the tec-

tonics and seismicity of the area. In the present paper,

we want to focus on the triggered tsunami. Firstly we

study the precision of the solutions proposed by

LSI12. Then we compute a further solution because

the amplitude of the Genoa tide gauge they used to

calibrate their solution is subject to uncertainty;

through the analysis of the theoretical tide, the

MAREMOTI technical report (2012) proposed that

the tide gauge overall wave amplitude exhibited in

EVA and RABINOVICH (1997) and used by LSI12 was

possibly over-estimated by a factor of 2.4. In that

context, and without the availability of the original

record, we have estimated a new solution with the use

of the newly processed wave amplitude. Conse-

quently, we propose a new solution, bearing in mind

that the possibly actual solution for the 1887 event is

bracketed between ours and the one of LSI12.

Finally, we have built an exhaustive set of earthquake

scenarios (including the 1887 scenario) and we

computed their tsunami-derived maximum wave

height (MWH) distribution. Note that the runup is the

altitude of the highest undated point versus the mean

sea level (MSL), while the wave height is the wave

elevation versus the MSL at any computational grid

point. Our computations of the tsunami mapping

provide the MWH, which is the highest wave height

encountered at each grid point during the entire wave

sequence. Eventually, our MWH at the highest

inundated point would correspond to the runup for a

grid spacing close to zero.

Note that we have considered here the solution of

LSI12 (not ours) because it is more extreme. Our

intention is to take into consideration realistic, even

extreme, scenarios and to identify the range of

potential tsunami signatures due to local seismicity,

i.e., the tsunami MWH mapping. In particular we

have considered the 1887 scenarios (S6 and S7 in

LSI12) and one scenario with similar earthquake

kinematics to the 1887 one, but which has been

shifted westward along the Ligurian Faults system

(LFS, Fig. 1). We conclude by proposing further

scenarios for which the entire LFS is rupturing. We

are not proposing a full study of the tsunami hazard.

Indeed, such study would require a large ensemble of

numerical simulations, the size of the ensemble

depending on the degree of constraints we have for

the LFS. Instead, we propose here specific scenarios

that may be representative of the ensemble and that

are based on the constraints we have of the LFS

(LSI12).

We want to stress here that we are more con-

cerned with tsunamis triggered by local earthquakes.

This is because, considering their arrival times (a few

minutes) the risk prevention cannot be handled by

existing tsunami warning systems. On the other hand,

remote tsunamis, like those triggered along the

Algerian Margin (for example on the May 21, 2003)

are the final objective of an ongoing project dedicated

to the development of a Mediterranean tsunami

warning system.

2. The Geological Context of the French–Italian

Riviera

The French–Italian Riviera is at the junction

between the southwestern Alps and the Ligurian

oceanic basin (LARROQUE et al. 2009). Onshore, the

Alpine belt corresponds to the Ligurian Alps and the

Nice-Castellane fold and thrust belts that were thru-

sted southward since 15 Ma (GIDON and PAIRIS 1992;

LAURENT et al. 2000; FOEKEN et al. 2003). The alpine

front of deformation is more or less co-extensive to

the Ligurian coast (SAGE et al. 2011). Offshore, the

northern Ligurian margin exhibits typical features of

a rifted continental margin, with tilted blocks boun-

ded by southeast-dipping listric faults (REHAULT et al.

1984; ROLLET et al. 2002). The Ligurian Basin is

assumed to have developed in a back-arc context

(DOGLIONI et al. 1997; JOLIVET et al. 2008) which

Tsunami Mapping Related to Local Earthquakes

opened from the late Oligocene to the early Miocene

times behind the Apulian subduction zone and partly

within the western alpine belt (WESTPHAL et al. 1976;

EDEL et al. 2001; GATTACCECA et al. 2007).

Currently, the overall Alpine and Ligurian regions

are characterized by low to moderate active defor-

mation (LARROQUE et al. 2001). The northern Ligurian

margin displays a compressive tectonic regime

(BETHOUX et al. 1992; LARROQUE et al. 2001; BIGOT-

CORMIER et al. 2004; BETHOUX et al. 2008; LARROQUE

et al. 2009). Offshore, the data collected during the

MALISAR marine geophysical surveys has produced

valuable images of the surface and subsurface struc-

tures along the northern Ligurian margin. These data

have allowed us to characterise Plio-Quaternary

compressive deformations, and particularly the Lig-

urian Faults System (LFS), an 80-km-long segmented

active fault system running at the foot of the conti-

nental slope (LARROQUE et al. 2011). The LFS has

been recently proposed as a candidate source for the

1887 Ligurian earthquake (LSI12).

3. Historical Tsunamis on the French–Italian Riviera

On October 16, 1979, following the Nice airport

submarine landslide, 2–3-m-high waves were repor-

ted (ASSIER-RZADKIEWICZ et al. 2000) in Antibes

(30 km west of Nice). The cause of this slide is still

unclear, and several studies have tried to explain it

(PIPER and SAVOYE 1993; HABIB 1994; MULDER et al.

1997; ASSIER-RZADKIEWICZ et al. 2000; DAN et al.

2007; IOUALALEN et al. 2010; LABBE et al. 2012). It

happened during the extension of the nearby harbor.

The slide itself caused 10 victim (workers on the

construction site) to be swept off by the water

depression at the instant of the slide. Both ASSIER-

RZADKIEWICZ et al. (2000) and IOUALALEN et al. (2010)

suggest that two distinct slides are necessary to

explain the recorded tsunami, i.e., the local slide itself

and another one that was triggered further down the

continental slope, although they did not address the

issue of the slides and tsunami chronology, which is a

crucial point.

IOUALALEN et al. (2010) reconstructed the most

significant ancient submarine landslide volumes

deduced from high-definition scars (up to several

km3) evidenced along the continental slope of the

Ligurian margin. Their tsunami computations/esti-

mations reveal that the slides might have triggered

wave heights as high as several meters, especially

near pronounced capes (Cap d’Antibes, Cap Ferrat,

Cap Martin, Fig. 1) through wave-focusing pro-

cesses. Similar coastal wave impacts have also been

reproduced by LABBE et al. (2012) through a one-slide

modeling taking water incorporation into account,

thus allowing an increase in volume along the slope.

However, the 1887 event attests that a significant

tsunamigenic earthquake may occur at return periods

of a century (SOLOVIEV 1990; SOLOVIEV et al. 2000;

TINTI et al. 2004). The 1887 Ligurian tsunami itself

was not as destructive as recent great subduction

tsunamis (IOUALALEN et al. 2007). The observed runup

from Marseille (France) to Livorno (Italy) (Fig. 2)

was small, but locally heights more than 2 m were

observed, with critical values in the vicinity of Im-

peria (DENZA 1887; and other references in LSI12).

The tide gauge at Marseille did not record a signifi-

cant signal, while the tide gauges at Nice and Genoa

harbors did. In any case, it is evidenced that local

active faults are able to produce tsunami waves that

break in a short time along the French–Italian

Riviera.

As far as distant tsunamis are concerned, only

earthquakes generated offshore of Algeria, at the

boundary of the Eurasia and Nubia plates, may affect

the area. However, the directivity of such tsunamis is

oriented northward and thus it is very unlikely to

affect the Ligurian coast. As an example, the Mw 6.8

Boumerdes earthquake (2003 May 21) generated

significant waves northward in the Baleares Islands,

but with few observations in the Ligurian sea (DE-

LOUIS et al. 2004; YELLES et al. 2004; ALASSET et al.

2006).

Several other historical earthquakes are supposed

to have been tsunamigenic (broadly discussed in

PELINOVSKY et al. 2002). This is the case of the 1564

‘‘nissard’’ earthquake (VOGT 1992): some fast sea

level fluctuations of few meters have been mentioned

in historical manuscripts between Nice and Antibes

(LAURENTI 1998). Nevertheless, the epicenter, likely

located inland 50 km north of Nice (LAMBERT et al.

1994; http://www.sisfrance.fr), implies that this tsu-

nami was certainly related to a submarine slide

M. Ioualalen et al. Pure Appl. Geophys.

induced by the earthquake. The February 23, 1818,

and the May 26, 1831, events are also discussed in

the literature, but runups for these events have not

been found (PELINOVSKY et al. 2002; TINTI et al. 2004;

LAMBERT and TERRIER 2011).

4. Seismotectonic Scenarios

The source characterization of the 1887 Ligurian

earthquake results from the test of a range of strikes,

dips and kinematics of faulting, together with the

tsunami numerical modeling based on the Genoa tide

gauge record provided in EVA and RABINOVICH (1997)

(LSI12). From the tsunami modeling point of view,

the best-fitting scenarios (S6 and S7, LARROQUE et al.

2012) correspond to (1) a N55�E striking fault plane,

with (2) a 16�-dipping northward or a 74�-dipping

southward fault plane, both with reverse kinematics,

(3) a 35-km-long and a 17-km-wide size, (4) a cen-

troid depth of 15 km, and (5) a Mw 6.8–6.9. Through

a morphotectonic analysis of the northern Ligurian

margin, we proposed that the optimal solution for the

Ligurian earthquake requires an active fault plane

with a low dip to the North (S7, LSI12).

Therefore in this paper, we first investigate the

inundation models related to the scenarios S6 and S7

(Table 1), directly related to the 1887 event. Never-

theless, the analysis of the earthquake-derived tsunami

hazard in this part of the western Mediterranean

requires testing other possible scenarios beyond the

known historical event. These scenarios must scan the

range of potential events in accordance with the geol-

ogy in order to allow a foresight discussion.

Figure 2(Modified from LARROQUE et al. 2012). Consequences of the February 23, 1887, earthquake observed along the coast from Marseille (France)

to Livorno (Italy). The dotted lines correspond to the distribution of the intensity of the tsunami [intensity scale from SIEBERG (1923) modified

by AMBRASEYS (1962), compilation by A. Laurenti]. The bars are local runup observations: 0 \ a \ 0.5 m; 0.5 \ b \ 1 m; 1 \ c \ 2 m.

Locations: 1 Marseille, 2 Frejus, 3 Cannes, 4 Antibes, 5 Nice, 6 St Jean-Cap-Ferrat, 7 Monaco; 8 Menton, 9 Ospedaletti, 10 San Remo, 11

Arma di Taggia, 12 Riva Ligure, 13 Imperia, 14 Oneglia, 15 Diano Marina, 16 Andora, 17 Alassio, 18 Genoa, 19 Bogliasco, 20 Recco, 21

Sestri Levante, 22 Livorno. B Bussana, P Pompeiana, C Cervo and San Bartolomeo del Cervo. The white rectangle represents our

computational domain throughout the study (43�050N–44�270N; 6�320E–9�150E)

Tsunami Mapping Related to Local Earthquakes

The three following scenarios are related to the

rupture of part of the Ligurian Faults system (LFS;

Fig. 1):

(i) S6 and S7 (Table 1) are the ‘‘historical’’ scenar-

ios. The rupture is located on the eastern segment

of the LFS: the strike, the 74� southward dip and

the 16� northward, and the kinematic parameters

of the faulting correspond to those determined by

LSI12 as the possible origin for the 1887 tsunami

(see discussion in LSI12).

(ii) S8 (Table 1) has same geometric and kinematic

parameters as S7 (see discussion in LSI12).

However, the centroid is shifted westward, filling

the western half-side of the total LFS.

Considering the length and the geometry of the

LFS, a second type of scenario can be envisaged in

which the entire LFS breaks as a roughly 80-km-long

rupture. Therefore, with the following scenarios (S9–

S12, Table 1) we test:

(iii) The influence of the strike: we consider N55�E

and N70�E which are the two extreme values for

the strike of the LFS at surface (LARROQUE et al.

2011).

(iv) The influence of the surface of the faulting: for a

length of 80 km, we take a width of 17 km, as

for the 1887 event, and a larger width of 27 km

in accordance with the low dip of the fault and

an assumed seismogenic thickness of 20 km

(BAROUX et al. 2001; EVA et al. 2001).

(v) The influence of the centroid depth: for each

surface faulting and strike we test a focal depth of

15 km, as for the 1887 event, and a shallower one

of 9 km.

5. Co-seismic Slip and Tsunami Modeling

The characterization of a seismogenic source for

the purpose of simulating a tsunami requires many

simplifying assumptions concerning the geometry of

the fault and the amount of slip that the fault may

undergo during an earthquake. In this paper, six

scenarios are explored; however, assumed geometries

and displacement estimates are each affected by large

uncertainties. Since the accuracy of the inferred

earthquake displacement is dependent upon the

quality of the assumed relationship between rupture

dimension and displacement, in the following para-

graph we briefly recall the main scaling laws that are

currently used in the seismic hazard community to

estimate displacements for simulated ruptures based

Table 1

Scenarios discussed in the text based on the varying parameters of OKADA’s (1985) dislocation method, each one corresponding to a specific

rectangular ruptured area

Scenario Kinematics / (strike) d (dip) L (km) W (km) D (m) d (km) Mo (N m) Mw (HK79)

S6 Reverse N55�E 74�S 35 17 1.3 15 2.55 9 1019 6.87

S7, S8 Reverse N55�E 16�N 35 17 1.5 15 2.95 9 1019 6.91

S7bis Reverse N55�E 16�N 26.30 15.85 1.1 15 1.51 9 1019 6.72

S9 Reverse N55�E 16�N 80 17 2 15 7.94 9 1019 7.24

S10 Reverse N70�E 16�N 80 17 2 15 7.94 9 1019 7.24

S11 Reverse N70�E 16�N 80 27 3.3 15 2.24 9 1020 7.51

S12 Reverse N55�E 16�N 80 27 3.3 9 2.24 9 1020 7.51

Scenarios S6 and S7 are proposed by LARROQUE et al. (2012) for the characterization of the 1887 Ligurian event offshore of Genova, Italy,

with a centroid located at (8.08�E, 43.70�N). The centroid is the geometrical center of the rupture plane, while the position of the hypocenter

(unknown) can be anywhere within the plane. It does not make a difference for the computation of the tsunami initial wave, because the

sequence of the rupture is ignored but the position of the centroid places the maximum of deformation. New scenario S8 is identical to S7, but

with a centroid shifted westward (7.55�E; 43.580�N). Scenario S7bis refers to a division by a factor of 2.4 of the tide gauge wave amplitude

reported in EVA and RABINOVICH (1997) (MAREMOTI Technical Report 2012). Other new scenarios S9–S10 and S11–S12 represent supposed

extreme scenarios for the area with a new centroid position (7.815�E, 43.640�N)

d represents the centroid depth, L is the length (along strike) of the rectangular fault plane and W is the width transversal to the fault plane. / is

the strike angle (counted CW from north) and d is the dip angle (counted positive from horizontal). The rake angle k is kept constant (90�)

(counted CCW from strike). D is the maximum fault slip. The medium shear modulus is taken as l = 3.3 9 1010 Pa. Mo is the seismic

moment (Mo ¼ lLWD) and Mw is the corresponding seismic magnitude (Mw ¼ log10ðMo � 9:1Þ=1:5) (HANKS and KANAMORI 1979)

M. Ioualalen et al. Pure Appl. Geophys.

on measured or inferred geometrical parameters of

the fault. The main point of this section is to under-

line the uncertainties that affect the parameters

chosen for the scenarios.

The most common scaling laws used in the seis-

mic hazard community are those proposed by WELLS

and COPPERSMITH (1994) and LEONARD (2010), here-

after named WC94 and L10, respectively. Other

scaling relations have been published but they will

not be considered here as they tend to concern narrow

classes of earthquakes (e.g., large strike-slip). The

scaling relations proposed by L10 are an improve-

ment compared to WC94 scaling laws in as much as

the author proposes alternative self-consistent scaling

laws: starting with one parameter (e.g., fault length) it

is possible to determine all of the other parameters

(area, width, displacement, moment), independently

of the choice of the initial parameter. Furthermore,

the distinction between interplate/and intraplate/sta-

ble continental region (SCR) earthquake proposed in

L10 is in agreement with the observation that SCR

earthquakes are often associated with relatively

higher stress drops and significantly larger slip dis-

placements compared to earthquakes in tectonically

active regions. Given that each simulated earthquake

requires fault dimensions, displacement, magnitude,

centroid, and orientation, accurate results require that

the rupture dimension, displacement, and moment be

consistent. In the case of our work however, both

WC94 and L10 are used to explore a range of pos-

sible scenarios and to choose reasonable

displacement scenarios within that range.

There are only two parameters available to

parameterize seismic scenarios for the Ligurian Fault

System (LFS): the mapped length of the fault seg-

ments in the LFS and the depth of the seismogenic

layer (LARROQUE et al. 2011). Given that each seg-

ment may break separately as 30–40-km-long

ruptures or simultaneously as a roughly 80-km-long

rupture, and that the LFS is a dip-slip structure, then

WC94 and L10 Fault Length-Displacement scaling

relations estimate mean displacements on such faults

in the range 0.76–2 m for a 35-km-long rupture (dark

red and yellow in Fig. 3a) and 0.97 and 3.78 m for an

80-km-long rupture (dark red and yellow in Fig. 3b),

depending on the kinematics of the fault, the statistics

performed and the indicator used. Assuming a dip of

the fault and knowing the seismogenic depth, then

scenarios based on the hypothetical area of the fault

can be proposed based on the L10 Fault Area-Dis-

placement relations. In this case, mean displacements

are in the range 1–2 m for a 35-km-long rupture (blue

in Fig. 3a) and 1.5 and 3.69 m for an 80-km-long

rupture, depending on the assumed width of the fault

(blue in Fig. 3b). If we limit the analysis to the most

commonly used indicators and we ignore the uncer-

tainty which underlies each one of these indicators

(not presented in the figure), the most important

decision for fixing a displacement for the seismic

scenarios concerns whether the LFS is a fault located

in an active region (low stress drop) or rather in a

stable continental region (high stress drop). As shown

in Fig. 3 (pink and red bars) and Table 1, the chosen

scenarios (1.3 m/1.5 m for the 35-km-long segments

and 2 m/3.3 m for the 80-km-long segments) are in

the higher range of estimated values, closer to L10

SCR estimates. The reader should bear in mind that

source scaling relations are empirically based models

and the displacement considered here are only mean

displacements. There is still significant stochastic

variability about these mean displacements. Last but

not least, in the simulations a rupture is assumed to

occur along a planar feature and the displacement is

uniformly distributed over the fault plane. Clearly,

for the purpose of tsunami hazard risk reduction

studies, a more exhaustive exploration of the uncer-

tainty would be necessary, which is beyond the scope

of the present paper.

6. The Rupture Parameters Solution of the February

23, 1887, Earthquake: Description and Precision

The best-fit solutions of LSI12 were obtained by

comparing the computed tsunami signal and the

recorded signal at the Genoa tide gauge provided by

EVA and RABINOVICH (1997) (e.g., scenario S6 in

Fig. 4). The tsunami fully nonlinear propagation

dispersive model, Funwave has been used (WEI and

KIRBY 1995; WEI et al. 1995). KENNEDY et al. (2000)

and CHEN et al. (2000) implemented a porous beach

method, used to keep the subaerial portion of the

model grid computationally active and to simplify the

calculation of runup on dry shorelines. The

Tsunami Mapping Related to Local Earthquakes

dissipation due to wave breaking and bottom friction

is represented with an eddy viscosity term (CHEN

et al. 2000). The computational grid covers a region

from west of Cannes, France, to east of Genoa, Italy,

i.e., 200 km (Fig. 2). Considering that the tsunami

model they used imposes a Cartesian uniform grid

(constant grid spacing), their work was able to be

completed only for grid spacing of 200 m. They used

a (1,150 9 776)—nodes grid along with an optimal

0.25-s time step to avoid numerical instabilities and

truncation errors. Each simulation of their 55-min

propagation (13,200 time steps) took 2 days of CPU

on their computer, i.e., a parallel version of Funwave

(POPHET et al. 2011) is implemented in a 8-core

cluster. For a 100-m grid spacing (used here), the

optimal time is found to be 0.125 s. Thus a typical

simulation presented here takes approximately

2 weeks of CPU with the same computer and Fun-

wave version.

In this paper, we take the time to discuss sensi-

tivity of modeling results to the grid spacing used.

The grid spacing is particularly important when per-

forming tsunami wave height mapping, which is very

sensitive to the precise description of the coastal

morphology. We focus here on scenario S6: a 400-,

200- and 100-m grid spacing are considered and the

signals are tested against the Genoa tide gauge

record. Figure 4 indicates a relatively good coher-

ency from 200 to 100 m with a slight over-estimation

of the wave amplitude for the latter. However, the

main jump in precision occurs between 400 and

200 m. Table 2 illustrates that the 200-m simulation

of LSI12 may be reasonably retained because the

solution did not deteriorate significantly from 200 to

100 m (a still very satisfactory 0.72 correlation

coefficient). A crucial point is the physical repre-

sentation of the Genoa harbour for which we used the

map provided by EVA and RABINOVICH (1997): At the

entrance of the harbour, we get eight ‘wet’ lateral

Figure 3Mean displacements predicted by the scaling relations of L10

(LEONARD 2010) and WC94 (WELLS and COPPERSMITH 1994).

ACTIVE faults located in active regions, SCR faults located in

stable continental regions NF Normal faults, RF reverse faults, ALL

all faults averaged together. Indicators used FL fault length, FA

fault area

Figure 4Convergence (related to the grid spacing resolution) test for the

1887 Genoa tide gauge record (in blue) and the simulated time

series for scenario S6 of LARROQUE et al. (2012) (Table 1) for

400-m (green), 200-m (red), 100-m (black) grid spacings and for

another 100-m grid spacing applied for a nonlinear shallow water

model (cyan blue). Time is in seconds and the wave height (WH) is

in centimeters

M. Ioualalen et al. Pure Appl. Geophys.

grid points for the 100-m grid and four points for the

200-m grid. The fact that there is no significant dif-

ference between the two simulations for (finally) a

few points (8 and 4) and a drastic doubling of the grid

resolution, indicate that the 100-m simulation can be

considered satisfactory. The simulations results do

not show any numerical instability or any misrepre-

sentation of local site effects.

An additional test consisted in the use of the

nonlinear shallow water configuration of Funwave

that does not take into account dispersive effects

(dispersion may occasionally drastically re-distribute

the wave amplitude and frequency content). Fig-

ure 4 shows that the gap between the solutions from

200 to 100 m and the one from the solutions with or

without dispersion is worse for the second case. In

particular, Table 2 indicates a slight deterioration of

the correlation coefficient (from 0.72 to 0.68). The

wave spectrum (Fig. 5) indicates that the lack of

dispersion gives too much credit to the lower fre-

quencies energy (853 and 1280 s). At the 200-m

level of precision, ignoring dispersive effects indu-

ces larger errors compared to the grid spacing

accuracy. We may then conjecture that, considering

the balance between the satisfactory numerical

model used (dispersive nonlinear Boussinesq model)

and their necessarily limited computational facilities,

the solution of LSI12 may be considered as a good

estimate (but only) with reference to the gauge

record provided by EVA and RABINOVICH (1997). In

that context, it can still be refined but the range of

parameters of the seismic source would not change

significantly.

However, through a detailed analysis of the Genoa

tide gauge record, the MAREMOTI Technical Report

(2012) showed that the wave amplitude provided by

EVA and RABINOVICH (1997) and used by LSI12 is

eventually over-estimated by a factor of 2.4 due to a

possible wrong scale (Figs. 6e, 7). This uncertainty

requires further analysis. We report here how the

Genoa tide gauge record has been processed by C.

Daubord in MAREMOTI technical report (2012)

benefiting from comparison to theoretical tide pre-

dictions: the record has been digitalized at a 1-min

sampling rate considering scale information on the

reproduction (Fig. 6a): height scale is in centimetres,

and time scale is in local time. It has been then changed

into current time system UT using the time correction

mean time in Genoa equal UT ? 36 min. A fictive

altimetric reference has been considered, as no

instrumental zero was mentioned, preventing us from

discussing the absolute heights. Comparing this digi-

talized signal to theoretical tide predicted signal on the

February 23 shows that the tidal range of the digital-

ized signal (58 cm) is far from the theoretical one

(24 cm) (Fig. 6b). This incoherency can be explained

neither by meteorological factors nor by defective tide

predictions. On one hand, it is unlikely that sudden

atmospheric forcing changes would have occurred

within less than 6 h. On the other hand, tide prediction

in Genoa has been computed using French official

harmonic constants, consistent with those of the Italian

Hydrographic Institute, and theoretical tide predic-

tions in neighbouring sites (Fig. 6c) of Genoa show

similar tidal ranges on the February 23, 1887 (Fig. 6d),

which strengthens the reliability of Genoa predictions.

Table 2

Convergence (related to the grid spacing resolution) for solution S6 of LARROQUE et al. (2012)

�gRec (cm) �gSol (cm) rRec (cm) rSol (cm) e (cm) q

S6_400 1.05 0.06 9.78 1.33 9.57 0.26

S6_200 1.14 8.70 6.14 0.79

S6_100 1.29 9.49 7.28 0.72

S6_NLSW 1.57 10.05 7.93 0.68

S7bis 0.42 0.45 3.91 3.61 2.90 0.71

Three grid spacings have been used: 400 m (S6_400), 200 m (S6_200 for scenario S6 of LARROQUE et al. 2012), 100 m (S6_100) and 100 m

without taking into account dispersion (S6_NLSW). The simulated signals (indexSol) are compared to the Genoa tide gauge record (indexRec).

The basic statistics are the mean wave height �g, the standard deviation r, the root mean squared error (RMSE) e and the cross-correlation

coefficient q. The usual RMSE representing the model skill is defined as

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

P

i ðgi;Rec � gi;SolÞ2=n

q

, where gi;Rec corresponds to the tide gauge

record points, gi;Sol is for the associated simulated one and n is the number of comparison points (every 10 s) for the respective time series

Tsunami Mapping Related to Local Earthquakes

Computing a linear regression between theoretical

tide predicted heights and recorded heights on Feb-

ruary 23 highlights a 2.4 scale factor (Fig. 6e). This

means that the scale indicated on the tide gauge

record published in EVA and RABINOVICH (1997) could

conduct to over-estimate heights by a factor of 2.4. If

we take into account this factor and re-evaluate tsu-

nami wave heights recorded at Genoa tide gauge, we

find that maximum wave heights are about 7 cm,

instead of 20 cm as proposed by EVA and RABINOVICH

(1997). The uncertainty on this bias could be reduced

considering the original tide gauge record and mete-

orological data, and performing a similar analysis on

more than one day. This result can indeed be con-

sidered as an alternative result till the viewing of the

original tide gauge record could be made available,

and could help settle the argument.

The statistics displayed in Table 2 are very

similar to the ones of scenario S6 and allows us to

validate this lower magnitude solution. Unfortu-

nately it has been impossible to obtain the original

record in order to definitely assess whether scenario

S7 or S7bis is the best-fit magnitude solution. We

conjecture that the S7bis scenario is more reliable,

while maintaining the solution S7 of LSI12. At this

stage, it is important to note that the uncertainty on

the tide gauge wave amplitude is not bracketed by

two scalars differing by a factor of 2.4, and thus that

the earthquake magnitude ranges from Mw 6.7 to

Mw 6.9. Rather, its value is either Mw 6.9 (LSI12) or

Mw 6.7 (the present paper). Consequently, this paper

proposes the two alternative solutions as being

credible with, however, a preference for the Mw 6.7

solution, regarding the tide gauge prediction that has

been processed carefully here. We believe the

present study is a necessary complement to our

LSI12 one in order to have a comprehensive over-

view of the event.

(1) The two magnitudes Mw 6.7 and Mw 6.9 are

relatively close, despite the 2.4 wave amplitude ratio.

(2) Also the shift from 200 m to 100 m of grid res-

olution for the LSI12 Mw 6.9 solution does not

change significantly the wave signal. At least, the

wave variation (in amplitude and phase) from one

simulation to another is far weaker than when

applying the ration 2.4 (from Mw 6.9 to Mw 6.7). Thus

we believe that an even better grid spacing would still

provide a very similar Mw 6.9 magnitude. Because of

(1) and (2), we believe it was important to process

again the tide gauge record (apply the 2.4 ratio) and

perform the grid spacing sensitivity tests because we

are able to conclude that the solution of LSI12 and

0

200

400

600

800

1000

1200

0 600 1200Period (s)

Wave spectrum

1280

853

640

513

320

TGS6_200S6_100S6_100_NLSW

Figure 5Wave spectra comparison for scenario S6 of Tables 1 and 2 between the Genoa tide gauge record (TG, in blue), scenario S7 for a 200-m grid

spacing (S7_200 in red, LARROQUE et al. 2012), a 100 m grid spacing (S6_100, in green) and a 100-m grid spacing without taking into account

dispersion (S6_100_NLSW)

M. Ioualalen et al. Pure Appl. Geophys.

our new solution are relatively close and then rela-

tively robust despite the uncertainties.

However, the tsunami signature computed below

will be based on the scenario S7, which is more

extreme than S7bis, although not differing signifi-

cantly, i.e., from Mw 6.7 to Mw 6.9 (Table 1).

7. The MWH Distribution for the February 23, 1887,

Earthquake-Derived Tsunami

We discuss here the computed MWH distribution

in order to map the tsunami for the proposed sce-

narios. The simulations are performed with a 100-m

Figure 6a Reproduction of Genoa tide gauge record, source: EVA and RABINOVICH (1997). b Comparison between Genoa tide gauge record from EVA

and RABINOVICH (1997) and Genoa tide prediction on the February 23, 1887. Tide Record filt is the recorded signal filtered by a low-pass

Butterworth filter. Recorded signals have a fictive vertical reference. Tide predicted signal is referenced to the hydrographic datum. c,

d Comparison of predicted tide amplitudes at neighbouring sites of Genoa on the February 23, 1887. Mean sea level has been forced to zero

for all sites. e Fit of a simple linear regression model between tide predicted signal and tide recorded signal. R2 indicates the determination

coefficient

Tsunami Mapping Related to Local Earthquakes

grid spacing and the use of Funwave Boussinesq

fully nonlinear model. Regarding the 1887 event,

there are very few records of witnessed wave

heights. DENZA (1887) [see discussion in FERRARI

(1991), EVA and RABINOVICH (1997) and Fig. 2]

reported a runup of the order of 1–2 m at Imperia,

San Remo (approx. 2 m) and Diano Marina (approx.

2 m) for instance. These runup observations are

consistent with our MWH where we find 3.20, 2.30

and 1.60 m at Imperia, San Remo and Diano Marina

respectively for scenario S6 (Figs. 8, 9 and 10a).

The simulation clearly identifies the exposed and

less exposed areas. This earthquake event is there-

fore quite significant for this area, considering the

relatively low extent of the active faults (compared

to great subduction areas; see IOUALALEN et al. 2007

for the 2004 Sumatra event). Significant wave

impacts are located in the immediate vicinity of the

earthquake location (Fig. 8), which is not surprising

considering the relatively short wavelength (approx.

17 km) and characteristic wave period (approx.

2 min) computed at the initial movement. As a

result, this event may represent the (local) worst-

case scenario in terms of tsunami impact for the

Italian Riviera. In that context, Figs. 9b and 10a (for

S6) may be considered as a local tsunami hazard

mapping. The processes that are responsible for the

coastal wave amplification are mainly site effects:

the MWH along the coast are located at focussing

areas (FC in Fig. 9). San Remo is centered around

two refraction-then-focussing areas (FC1 in Fig. 9),

explaining the two wave peaks. A second focusing

area (FC2) operates both in Riva Ligure and Ci-

pressa. Two other focussing areas are also identified

for Imperia (FC3) and Diano Marina (FC4) yielding

two wave peaks.

As far as scenario S7 (Table 1) is concerned, i.e.,

the south-dipping fault scenario of LSI12, the results

are quite similar to the S6 case (nearly the same

peaks) with sensibly weaker MWHs (Fig. 10a): 2.60,

1.90 and 1.50 m at Imperia, San Remo and Diano

Marina, respectively. Although the slip amount is

slightly higher for S7 (Table 1), the vertical seafloor

displacement is slightly larger for S6 (Fig. 8) because

of the steep dip of the fault. The wave impact is also

confined which is not surprising considering the very

similar initial waves. Then in terms of tsunami

earthquake-kinematics, the results of S6 and S7 are

very similar and, consequently, what is derived for

one scenario applies to the other.

Although, as expected, the impact of the S6-

and S7-type tsunami scenarios is local, it is

-15

-5

5

15

1200 1800 2400 3000Time (s)

WH (cm)

S7bis

Figure 7Best-fit solution for the 1887 Genoa tide gauge record (in blue) and the simulated time series obtained with scenario S7bis (Table 1) for a

100-m grid spacing (in red). Time is in seconds and the wave height (WH) is in cm

M. Ioualalen et al. Pure Appl. Geophys.

however quite significant, considering the rela-

tively short wave period and wavelength along

with a relatively small amount of slip. In order to

complete the tsunamigenic picture along the

French–Italian Riviera, it is therefore essential to

test additional scenarios along the western part of

the LFS (Fig. 1).

8. Rupturing the Western Part of the Ligurian

Faults System

First, an S7-type scenario was shifted further west

along the LFS in order to predict potential runup

within the French Riviera following a ‘‘Ligurian-

type’’ earthquake with an epicenter located westward,

Figure 8Simulated local maximum wave height (MWH) distribution (see the color scale in meters) along the coast for scenarios S6 to S11 of Table 1

for a dx = dy = 100 m grid spacing and the associated initial tsunami at the generation with the seafloor vertical deformation (dashed lines

are for subsidence and continuous lines are for uplift with 5 cm iso-contours) which has been taken as usual as the initial wave shape (the

seismic P-wave is considered to be fast, e.g.,[1 km s-1). The white star corresponds to the earthquake centroid location. The white rectangle

of S6 is the area enlarged in the next Fig. 9. The bathymetry is represented with 50-m iso-levels

Tsunami Mapping Related to Local Earthquakes

i.e., the scenario S8 in Fig. 8. The epicenter of the

proposed scenario S8 has been centered offshore

Nice–Monaco (Table 1, Fig. 8) in the area where

several earthquakes have been recently detected

(LARROQUE et al. 2011). The results (Fig. 10b) indi-

cate an overall range of WHM of several meters that

is similar to the one of S7.

Locally, significant spots appear, mainly due to

local topography: in particular more than 4 m runup

is predicted in the center of Nice (Nice-Massena)

which is due to a pronounced land slope of 4 %

compared to the neighboring 2 %, enhancing the

usual slope effect (1/4-powered Green’s law).

Another spot is located at Nice International airport.

Figure 9(Upper panel) Maximum wave height (MWH) distribution along the coast for scenario S6 within the white rectangle of Fig. 8. The

bathymetry is represented with 50 m iso-levels. (Lower panel): detailed (bars) of the MWH distribution with the entire computational domain.

Annotated locations are: France (FR): sm Sainte-Maxime, il Iles de Lerins (islands offshore Cannes), ca Cannes, as Antibes-la Salis beach, sl

Saint-Laurent du Var, na Nice International airport, nm Nice-Massena area, vlfr Villefranche-sur-mer, cf Saint-Jean Cap Ferrat, mc Monaco,

rcm Rocquebrune Cap Martin, me Menton. Italy (IT): ve Ventimiglia, bo Bordighera, os Ospedaletti, sr San Remo, rl Riva Ligure, ci Cipressa,

im Imperia, and ge Genoa

M. Ioualalen et al. Pure Appl. Geophys.

The reason is a focusing process: when the wave

propagates eastward, it encounters a bathymetric

shape that forces a pronounced refraction and a

subsequent focusing (see the shape of the 50 m—

isobath in Fig. 9). A significant MWH (2.60 m,

although less pronounced than the two other sites) is

also predicted at Antibes la Salis: it may be

interpreted as a growing-in-size wave due to a con-

fined seiche. We must notice that the location ‘‘La

Salis’’ was also the spot of the maximum MWH of

the 1979 tsunami (IOUALALEN et al. 2010). Surpris-

ingly, no significant MWH is predicted further south

at Cap d’Antibes (Fig. 10b). In that area, IOUALALEN

et al. (2010) computed the highest waves following a

(a)

(b)

(c)

Figure 10Distribution of MWH along the Ligurian land coast (in meters) for a scenarios S6 and S7, b S7, S8 and S9 and c S9, S10 and S11 of Table 1.

Here the reported MWH coincides with the runup (effective wave height added to the local land elevation of the highest inundated point).

Annotated locations are identical to those of Fig. 9

Tsunami Mapping Related to Local Earthquakes

tsunami triggered by submarine landslides collapsing

further south at the basis of the continental slope.

They explained that it was also due to refraction/

focusing processes when the wave comes directly

northward (see the 50 m—isobath 50 m surrounding

the cape, Fig. 9). This time it is not the case, because

the wave comes from the east and encounters straight

isobath shapes. Once again it is trivial to mention that

the wave directivity (and thus the location of the

tsunami triggering mechanism) is a crucial concern.

9. Rupturing the Entire Ligurian Faults System

In order to obtain a full picture of the MWH

distribution on the Ligurian coastal area, we must

consider a further rupturing of the entire LFS,

although there is no record of such event during

historical times. For that purpose we have built two

sets of scenarios: the first one (S9–S10) is based on

a 80 km 9 17 km fault plane with a 2-m co-seismic

slip while the second (S11–S12) represents a

80 km 9 27 km fault plane with a 3.3 m co-seismic

slip (Table 1). For each set we have considered two

strike angles based on the orientations of the fault

system from N55 to N70�E (Fig. 1). The testing of

the two different orientations (N55�E for S9, S12

and N70�E for S10, S11) is interesting because their

respective tsunamis have slightly different directiv-

ities and, eventually, the site effects may differ. The

focal depth is also tested (15 and 9 km for S11 and

S12 respectively; Table 1). For both sets of scenar-

ios we could eventually introduce heterogeneous

seafloor vertical movements, e.g., by using several

successive rupture segments. Such a representation

indeed applies for great subduction earthquakes (for

example for Sumatra or Andean faults, IOUALALEN

et al. 2007). But we cannot see evidence that a

potential rupture of the western side of the LFS

would be more energetic than the eastern side. For

sake of simplicity (and because of a lack of valuable

information) we consider a unique fault plane

without segmentation. Besides, considering the rel-

atively small size of the ruptured area (vs. great

subduction earthquakes) the rising time lag due to

the P-wave propagation would introduce only a few

seconds phase lag within the triggered tsunami,

which is not a critical issue.

As far as scenarios S9 and S10 are concerned

(Fig. 10c), we recover an overall same MWH peak

distribution, as expected for a larger slip rupture, but

greater MWH due to the simultaneous rupture of the

(aligned) eastern (S7 in Fig. 8) and western (S8 in

Fig. 8) segments. Consequently, the localized MWH

peaks are due to the same processes of focusing point

that were discussed previously. Several spots of more

than 4–5 m MWH are identified (Nice Massena,

Menton, Riva Ligure). The change in strike angle

between S9 and S10 may locally change MWH

amplitudes, but the overall spatial distribution is quite

conserved. When increasing the co-seismic slip (S11

and S12) from 2 to 3.3 m with the associated new

rupture sizes (Fig. 10c for S11), we still conserve the

MWH peaks distribution, with, however, larger

amplitudes and land inundation. Here MWH over

5 m occupy an area ranging from Nice, France, to

Imperia, Italy, which is a real threat considering its

rate of urbanization. A special issue is the Nice city

center. Like for scenario S8 (Fig. 10b) a noticeable

MWH occurs at Nice Massena (7 m) due to the

enhanced topography slope. However, the main fea-

ture is the 28 m MWH (Figs. 10c, 11) computed at

the edge of the ‘‘colline de Cimiez’’. The wave first

inundates Nice Harbor, then propagates northward to

‘‘Riquiez’’ (inundation of more than a 1 km) and

takes the pathway westward between two hills, i.e.,

the ‘‘colline de Cimiez’’ and the ‘‘colline du Cha-

teau’’, inundating downtown Nice (inundation of

nearly 2 km). For this scenario the incoming wave is

sufficiently high to overpass the northern part of the

harbor dock (of lower altitude than the lateral one)

and pursue its propagation inland. It is suggested that

within the harbor, a seiche-like wave is trapped and

builds up and inundates the lowest area. Here, the

large MWH and inundation process is purely a site-

and-threshold effect.

Note that the change of the strike angle and the

centroid depth from S11 to S12 does not significantly

change the MWH spatial distribution and amplitude

(not shown). The example of Nice downtown severe

inundation shows that the site-and-threshold effect

operates at first order.

M. Ioualalen et al. Pure Appl. Geophys.

10. Discussion

The French–Italian Riviera is potentially exposed

not only to submarine landslide tsunamis but also to

earthquake-triggered tsunamis which can cause

locally significant runups. Two main tsunamigenic

historical events have raised this concern: the 1887

February 23 earthquake offshore of Imperia, Italy,

which triggered a noticeable tsunami, and the tsun-

amigenic landslide of October 16, 1979, which

caused 11 victims. The French–Italian Riviera is

sufficiently densely populated, to necessitate

addressing the tsunami issue as completely as possi-

ble. Studies on tsunamis triggered by landslides have

been already carried out (e.g., ASSIER-RZADKIEWICZ

et al. 2000; IOUALALEN et al. 2010; LABBE et al. 2012).

However, tsunamis triggered by earthquakes have not

yet been addressed in sufficient detail. The main

reason is the lack of large present-day earthquakes:

the area experiences each year several tens of earth-

quakes of magnitude less than 5, i.e., not

tsunamigenic. However, the February 23, 1887,

earthquake is quite emblematic for the area because it

is unusual. Therefore, its occurrence forces to (1) try

to propose potential earthquake source scenarios and

(2) to evaluate their tsunami impact.

The most recent work has been proposed by

LSI12. Based on the study of the morphology and the

seismotectonic of the area and on an ensemble of

tsunami numerical simulations they short-listed

potential rupture modes along the Ligurian Faults

system. They determined a preferred solution for the

1887 event (S7, Table 1). In this study we tested the

sensitivity of the modeling results to the grid spacing

by improving the computational grid (the coastal

bathymetry and topography accuracy). We show that

Figure 11Inundation and MWH map in the vicinity of Nice Massena and the port of Nice (downtown Nice city) for scenario S11 of Table 1. The

background bathymetry and topography are plotted with 10 m iso-levels

Tsunami Mapping Related to Local Earthquakes

their solutions are quantitatively very reasonable. In

particular the possible imprecision due to the level of

accuracy of the coastal topography is weaker than

possible errors that can be generated if dispersive

effects are not considered. Further, since the tide

gauge record reported in EVA and RABINOVICH (1997)

and used by LSI12 for scenario S7, might over-esti-

mate the wave amplitude by a factor of 2.4

(MAREMOTI Technical Report 2012), we also pro-

posed an alternative solution (S7bis) of magnitude

Mw 6.7 (Mw 6.9 for the one proposed by LSI12

(Table 1). This magnitude evaluation is still consis-

tent with the plausible magnitudes proposed

discussed by LSI12.

Then, we have computed the tsunami MWH map

of the area. It is shown that the tsunami impact is

quite local, with some locations reaching MWH

higher than 2–3 m. The result in itself is interesting

considering the industrial, the residential and touristic

occupancy of the area. In order to have a full picture

of the tsunami MWH potential over the entire

French-Italian Riviera, we considered further sce-

narios (also confined and identical in terms of

faulting) further west along the active LFS. We found

very similar wave MWH as for the 1887 scenarios

with local 3–4 m spots due to local site effects in the

western part of the Riviera. Finally the rupturing of

the entire LFS with a 3.3-m co-seismic slip may

trigger very significant tsunami amplitudes with

MWH as high as more than 25 m in specific places,

such as within downtown Nice city.

11. Conclusion

The work presented here is the continuation of

LSI12 study concerning the emblematic Ligurian

earthquake that occurred on the February 23, 1887,

offshore of Imperia, northern Italy. This earthquake is

quite special because its magnitude (Mw 6.9) was

much larger than the present-day ones computed from

the instrumental seismicity and it produced a

noticeable tsunami. The tsunami modeling is based

on the record of the subsequently triggered tsunami

(mainly a tide gauge record at Genoa, Italy) and on

the results of a seismotectonic study of the 1887

event and its context. In this paper we investigated

the tsunami hazard induced by the 1887 earthquake,

completed by other offshore earthquake scenarios at

the scale of the entire French–Italian Riviera. Con-

sidering the urban occupancy of the area, such a

relatively high magnitude earthquake has caused

severe damages in the Imperia area. Also, since the

1887 earthquake was tsunamigenic, it is of primary

interest to estimate the wave impact locally. In the

present study, we are concerned with the relationship

between the faulting and the tsunami maximum wave

height (MWH) coastal distribution over the Ligurian

area.

First, we deal with the 1887 tsunamigenic earth-

quake. Due to necessarily limited computing

facilities, LSI12 decided to build a 200-m grid

spacing computational domain. Since the scenarios

proposed are the only ones indeed quantifying the

rupture parameters (S6 and S7 in Table 1), it is

necessary to test if the resolution is sufficient. The

grid spacing precision is a recurrent issue in tsunami

modeling. We demonstrated that grid spacing does

not affect significantly the conclusions of LSI12.

Then, appearing that the tide gauge record

reported in EVA and RABINOVICH (1997) and used by

LSI12 for scenario S7, might over-estimate the wave

amplitude by a factor of 2.4 (MAREMOTI technical

report 2012), we also proposed a further solution

(S7bis) of magnitude Mw 6.7 (Mw 6.9 for the one

proposed by LSI12 (Table 1). We conjecture that the

set of two alternative (and exhaustive) solutions is

representative of the event.

Finally, we computed the MWH distribution

along the French-Italian Riviera. A significant MWH

is ‘‘predicted’’ (up to 2–3 m) that is consistent with

the observations reported following the February

23rd 1887 tsunami. Although the tsunami impact is

quite local, this can be explained by the limited

spatial extension of the ruptured area and thus by its

limited wave period.

Further scenarios are proposed to map more

completely the tsunami signature in this area. We first

considered the same ‘‘Ligurian-type’’ scenario but

shifted to the western part of the Ligurian Faults

system (offshore Nice/Monaco). In this way, the new

scenarios (S9 in Table 1) complemented by the 1887

ones cover the potential active faults system. The

MWH signature is quite similar to the 1887 event

M. Ioualalen et al. Pure Appl. Geophys.

with identified local spots, however shifted westward.

Finally, we test the rupturing of the entire LFS with

larger fault surfaces and co-seismic slips of 2 and

3.3 m. These last scenarios yielded very significant

tsunami spreading along more than 200 km of coast.

They can be considered as the largest potential event

in the regional geological context and their threat in

urbanized areas is analyzed. Considering the large

and growing urbanization of the area, we believe the

proposed mapping can be crucial in terms of tsunami

risk assessment. Considering the multiple scenarios

of LFS rupturing and the associated MWH maps that

have been computed we may reasonably conclude

that the tsunami threat is relatively significant and

uniform at the Italian side of the Riviera (from

Ventimiglia to Imperia) while it is more localized

(sporadic) at the French side from Antibes to Menton

with however higher local level of inundation, e.g.,

Nice city center (see maps for more details).

The area may encounter remote tsunamis (trans-

Mediterranean), but the developing Mediterranean

warning systems in project are expected to resolve

them (SCHINDELE et al. 2012). We then focused our

study on locally generated tsunami, possibly trig-

gered by earthquakes and/or submarine landslides.

The issue is crucial because the active faults and the

unstable continental slope are very close to the

coastline due to the narrow continental shelf. Con-

sequently the tsunami arrival time is a matter of a few

minutes that could never be resolved by any tsunami

warning system. This study is also complementary to

the work of IOUALALEN et al. (2010) regarding the

mapping of local tsunami impact within the Ligurian

area.

Acknowledgments

The MALISAR geophysical surveys were funded by

the Commission Nationale Flotte et Engins (CNFE-

IFREMER). This work was partly funded by the

Agence Nationale de la Recherche through the ANR

projects QSHA (‘‘Quantitative Seismic Hazard

Assessment’’, http://qsha.unice.fr/) and TSUMOD

(ANR-05-CATT-016-02). Finally, the authors

acknowledge with thanks the contribution of the

anonymous reviewers who contributed to the

improvement of the first draft manuscript.

REFERENCES

ALASSET, P.-J., HEBERT, H., MAOUCHE, S., CALBINI, V., and MEG-

HRAOUI, M. (2006) (MThe tsunami induced by the 2003 Zemmouri

earthquake Mw = 6.9, Algeria): modeling and results. Geophys.

J. Int., 166, 213–226.

AMBRASEYS, N. N. (1962) Data for the investigation of seismic sea

waves in the Eastern Mediterranean. Seismological Society of

America Bulletin, 52, 895–913.

ASSIER-RZADKIEWICZ, S., HEINRICH, P., SABATIER, P.C., SAVOYE, B.,

and BOURILLET, J.F. (2000) Numerical modeling of a landslide-

generated tsunami: the 1979 Nice event, Pure appl. Geophys.,

157, 1707–1727.

BAROUX, E., BETHOUX, N., and BELLIER, O. (2001) Analyses of the

stress field in the southeastern France from earthquake focal

mechanisms. Geophys. J. Int., 145, 336–348.

BETHOUX, N., FRECHET, J., GUYOTON, F., THOUVENOT, F., CATTANEO,

M., EVA, C., NICOLAS, M., and GRANET M. (1992) A closing

Ligurian sea. Pure and Applied Geophysics, 139, 179–194.

BETHOUX, N., TRIC, E., CHERY, J., and BESLIER, M.O. (2008) Why is

the Ligurian basin (Mediterranean sea) seismogenic? Thermo-

mechanical modeling of a reactivated passive margin. Tectonics,

27, TC5011. doi:10.1029/2007TC002232.

BIGOT-CORMIER, F., SAGE, F., SOSSON, M., DEVERCHERE, J., FERRAN-

DINI, M., GUENNOC, P., POPOFF, M., and. STEPHAN, J.F. (2004)

Deformations pliocenes de la marge nord-Ligure (France) : Les

consequences d’un chevauchement crustal sud-alpin. Bulletin de

la Societe Geologique de France, 175(2), 197–211.

CAROBENE, L., and CEVASCO, A. (2011) A large scale lateral

spreading, its genesis and Quaternary evolution in the coastal

sector between Cogoleto and Varazze (Liguria–Italy). Geomor-

phology, 129, 398–411.

CHEN, Q., KIRBY, J. T., DALRYMPLE, R. A., KENNEDY, A. B., and

CHAWLA, A. (2000) Boussinesq modeling of wave transformation,

breaking, and run-up. II: 2D. J. Waterw. Port Coastal Ocean

Eng., 126(1), 48–56.

DAN, G., SULTAN, N., and SAVOYE, B. (2007) The 1979 Nice harbor

catastrophe revisited: trigger mechanism inferred from geo-

technical measurements and numerical modeling. Mar. Geol.,

245, 40–64.

DELOUIS, B., VALLEE, M., MEGHRAOUI, M., CALAIS, E. MAOUCHE, S.,

LAMMALI, K., MAHSAS,A., BRIOLE, P., BENHAMOUDA, F. and

YELLES, K. (2004) Slip distribution of the 2003 Boumerdes–Ze-

mmouri earthquake, Algeria, from teleseismic, GPS, and coastal

uplift data. Geophys. Res. Lett., 31, L18607. doi:10.1029/2004.

GL020687.

DENZA F. (1887) Alcune notizie sul terremoto del 23 febbraio 1887.

Torino.

DOGLIONI, C., GUEGUEN, E., SABAT, F., and FERNANDEZ M. (1997)

The western Mediterranean extensional basins and the alpine

orogen. Terra Nova, 9, 109–112.

EDEL, J.B., DUBOIS, D., MARCHANT, R., HERNANDEZ, J., and COSCA

M., 2001. La rotation miocene inferieure du bloc corso-sarde;

nouvelles contraintes paleomagnetiques sur la fin du mouvement.

Bulletin de la Societe Geologique de France, 172(3), 275–283.

Tsunami Mapping Related to Local Earthquakes

EVA, C. and RABINOVICH, A.B. (1997) The February 23, 1887 tsu-

nami recorded on the Ligurian coast, western Mediterranean.

Geophys. Res. Letters, 24, 2211–2214.

EVA, E., SOLARINO, S., and SPALLAROSSA, D. (2001) Seismicity and

crustal structure beneath the western Ligurian Sea derived from

local earthquake tomography. Tectonophysics, 339, 495–510.

FERRARI, G. (1991) The 1887 Ligurian earthquake: a detailed study

from contemporary scientific observations. Tectonophysics, 193,

131–139.

FOEKEN, J.P.T., DUNAI, T.J., BERTOTTI, G., and ANDRIESSEN P.A.M.

(2003) Late Miocene to present exhumation in the Ligurian Alps

(southwest Alps) with evidence for accelerate denudation during

the Messinian salinity crisis. Geology, 31, 797–800.

GATTACCECA, J., DEINO, A., RIZZO, R., JONES, D.S., HENRY, B.,

BEAUDOIN, B., and VALEBOIN F. (2007) Miocene rotation of

Sardinia: new paleomagnetic and geochronological constraints

and geodynamic implications. Earth and Planetary Science Let-

ters, 258, 359–377.

GIDON, M., and PAIRIS J.L. (1992) Relations entre le charriage de la

Nappe de Digne et la structure de son autochtone dans la vallee

du Bes (Alpes de Haute-Provence, France). Eclogae Geologicae

Helveticae, 85/2, 327–359.

HABIB, P. (1994) Aspects geotechniques de l’accident du nouveau

port de Nice. Revue Francaise de Geotechnique, 65, 2–15.

HANKS, T. C., and KANAMORI, H. (1979) A moment magnitude scale.

Journal of Geophysical Research, 84(B5). doi: 10.1029/

0JGREA0000840000B5002348000001. issn: 0148-0227.

IOUALALEN, M., ASAVANANT, J., KAEWBANJAK, N., GRILLI, S.T., KIR-

BY, J.T., and WATTS, P. (2007) Modeling of the 26th December

2004 Indian Ocean tsunami: Case study of impact in Thailand.

Journal of Geophysical Research, Oceans, 112, C07024. doi:10.

1029/2006JC003850.

IOUALALEN, M., MIGEON, S., and SARDOU, O. (2010) Landslide tsu-

nami vulnerability in the Ligurian Sea: case study of the 1979

October 16 Nice international airport submarine landslide and

of identified geological mass failures. Geophys. J. Int., 181,

724–740. doi: 10.1111/j.1365-246X.2010.04572.x.

JoLIVET, L., AUGIER, R., FACCENNA, C., NEGRO, F., RIMMELE, G.,

AGARD, P., ROBIN, C., ROSSETTI, F., and CRESPO-BLANC A.C.

(2008) Subduction, convergence et extension arriere-arc en

Mediterranee. Bulletin de la Societe Geologique de France,

179(6), 525–550. doi: 10.2113/gssgfbull.179.6.525.

KENNEDY, A. B., CHEN, Q., KIRBY, J. T., and DALRYMPLE, R. A.

(2000) Boussinesq modeling of wave transformation, breaking,

and run-up. I: 1D. J. Wtrwy, Port, Coast, and Oc.Engrg. ASCE,

126(1), 39–47.

LABBE, M., DONNADIEU, C., DAUBORD, C., and HEBERT, H. (2012)

Refined numerical modeling of the 1979 tsunami in Nice (French

Riviera): Comparison with coastal data. Journal of Geophysical

Research, 117, F01008. doi:10.1029/2011JF001964.

LAMBERT, J., MORONI, A., and STUCCHI, M. (1994) An intensity

distribution for the 1564, Maritimes Alpes earthquake. In

Materials of the CEC Project Review of Historical Seismicity in

Europe (eds. Albini, P. & Moroni, A.) (CNR, Milan) vol. 2,

143–152.

LAMBERT, J., and M. TERRIER (2011) Historical tsunami database

for France and its overseas territories. Nat. Hazard Earth. Syst.

Sci., 11, 1037–1046.

LARROQUE, C., BETHOUX, N., CALAIS, E., COURBOULEX, F., DES-

CHAMPS, A., DEVERCHERE, J., STEPHAN, J.F., RITZ, J.F., and GILLI E.

(2001) Active and recent deformation at the Southern Alps-

Ligurian basin junction. Netherlands Journal of Geosciences –

Geologie en Mijnbouw, 80, 255–272.

LARROQUE, C., DELOUIS, B., GODEL, B., and NOCQUET J.M. (2009)

Active deformation at the southwestern Alps – Ligurian basin

junction (France-Italy boundary): Evidence for recent change

from compression to extension in the Argentera massif. Tec-

tonophysics, 467(1–4), 22–34, doi:10.1016/j.tecto.2008.12.013.

LARROQUE, C., MERCIER DE LEPINAY, B., and MIGEON, S. (2011)

Morphotectonic and fault-earthquake relationships along the

northern Ligurian margin (Western Mediterranean) based on

high resolution multibeam bathymetry and multichannel seismic-

reflection profiles. Marine Geophysical Researches, 32(1–2),

163–179. doi:10.1007/s11001-010-9108-7.

LARROQUE, C., SCOTTI, O., and IOUALALEN, M. (2012) Reappraisal of

the 1887 Ligurian earthquake (western Mediterranean) from

macroseismicity, active tectonics and tsunami modelling. Geo-

phys. J. Int. doi:10.1111/j.1365-246X.2012.05498.x.

LAURENT, O, STEPHAN, J.F., and POPOFF M. (2000) Modalites de la

structuration miocene de la branche sud de l’arc de Castellane

(chaınes subalpines meridionales). Geologie de la France, 3,

33–65.

LAURENTI, A. (1998) Les tremblements de terre des Alpes-Mari-

times. Serre Editeur, 174 pp.

LEONARD, M. (2010) Earthquake fault scaling:self-consistent

relating of rupture length, width, average displacement, and

moment release. Seismol. Soc. Am. Bull. 100, 5A,

1971–1988.

MAREMOTI technical report (2012) WP1: tide gauges and sea

level coastal observations of tsunami, deliverables D1.1 and

D1.2, (ANR-08-RISKNAT-05).

MIGEON, S., CATTANEO, A., HASSOUN, V., LARROQUE, C., CORRADI,

N., FANUCCI, F., DANO, A., and MERCIER DE LEPINAY, B. (2011)

Morphology, distribution and origin of recent submarine land-

slides of the Ligurian Margin (North-western Mediterranean):

some insights into geohazard assessment. Marine Geophysical

Researches. doi:10.1007/s11001-011-9123-3.

MULDER, T., SAVOYE, B., and SYVITSKY (1997) Numerical modeling

of a mid-sized gravity flow: the 1979 Nice turbidity current

(dynamics, processes, sediment budget and seafloor impact).

Sedimentology, 44, 395–326.

PIPER, D.J., and SAVOYE, B. (1993) Process of late quaternary

turbidity current flow and deposition on the Var deep-sea fan,

Northwest Mediterranean Sea. Sedimentology, 40, 557–582.

PELINOVSKY, E., KHARIF C., RIABOV I., and FRANCIUS, M. (2002)

Modeling of tsunami propagation in the vicinity of the French

coast of the Mediterranean. Natural Hazards, 25, 135–159.

POPHET, N., KAEWBANJAK, N., ASAVANANT, J., and IOUALALEN M.

(2011) High grid resolution and parallelized tsunami simulation

with fully nonlinear Boussinesq equations. Computers and Flu-

ids, 40, 258–268.

REHAULT, J.P., BOILLOT, G., and MAUFFRET A. (1984) The western

Mediterranean basin geological evolution. Marine Geology, 55,

447–477.

ROLLET, N., DEVERCHERE, J., BESLIER, M.O., GUENNOC, P., REHAULT,

J.P., SOSSON, M., and TRUFFERT C. (2002) Back arc extension,

tectonic inheritance and volcanism in the Ligurian sea, Western

Mediterranean. Tectonics, 21. doi:10.1029/2001TC900027.

SAGE, F., BESLIER, M.O., THINON, I., LARROQUE, C., DESSA, J.X.,

MIGEON, S., ANGELIER, J., GUENNOC, P., SCHREIBER, D., MICHAUD,

F., STEPHAN, J.F., and SONNETTE, L. (2011) Structure and evolu-

tion of a passive margin in a compressive environment: example

M. Ioualalen et al. Pure Appl. Geophys.

of the south-western Alps-Ligurian basin junction during the

Cenozoic. Marine and Petroleum Geology, 28, 1263–1282.

doi:10.1016/j.marpetgeo.2011.03.012.

SCHINDELE, F., BOSSU, R., ALABRUNE, N., ARNOUL, P., DUPERRAY, P.,

GAILLER, A., GUILBERT, J., HEBERT, H., HERNANDEZ, B., LOEVEN-

BRUCK, A., and ROUDIL, P. (2012) The French Tsunami warning

center for the Mediterranean and North-East Atlantic (CENtre

d’ALerte aux Tsunamis, CENALT). Geophysical Research

Abstracts, Vol. 14, EGU2012-10166.

SIEBERG, A. (1923) Geologische, Physikalische und Angewandte

Erdbebenkunde. Edited by Gustav Fischer, Jena, 572 pp.

SOLOVIEV, S.L. (1990) Tsunamigenic zones in the Mediterranean

Sea. Natural Hazard, 3, 183–202.

SOLOVIEV, S. L., SOLOVIEVA, O. N., GO, CH. N., KIM, KH. S., and

SHCHETNIKOV, N. A. (2000) Tsunamis in the Mediterranean Sea

2000 BC-2000 AD, Vol. 13, Advances in Natural and Techno-

logical Hazards Research, Kluwer, Dordrecht.

TINTI, S., MARAMAI, A., and GRAZIANI, L. (2004) The new catalogue

of Italian tsunami. Natural Hazard, 33, 439–465.

VOGT, J. (1992) Le ‘‘complexe’’ de la crise sismique nissarde de

1564. Quaternaire, 3, 125–127.

WEI, G., and KIRBY, J. T. (1995) A time-dependent numerical code

for extended Boussinesq equations. J. Wtrwy Port Coast Oc.

Engrg, 121, 251–261.

WEI, G., KIRBY, J. T., GRILLI, S. T., and SUBRAMANYA, R. (1995) A

fully nonlinear Boussinesq model for free surface waves. Part 1:

Highly nonlinear unsteady waves. J. Fluid Mech., 294, 71–92.

WELLS, D. L., and COPPERSMITH, K. J. (1994) New empirical rela-

tionships among magnitude, rupture length, rupture width,

rupture area, and surface displacement. Bull. Seismol. Soc. Am.

84(4), 974–1002.

WESTPHAL, M., ORSINI, J., and VELLUTINI P. (1976) Le microconti-

nent corso-sarde, sa position initiale, donnees paleomagnetiques

et raccords geologiques. Tectonophysics, 30, 41–57.

YELLES, K., LAMMALI, K., MAHSAS, A., CALAIS, E., and BRIOLE, P.

(2004) Co-seismic deformation of the May 21st, 2003, Mw = 6.8

Boumerdes earthquake, Algeria, from GPS measurements. Geo-

phys. Res. Lett., 31, L13610. doi:10.1029/2004.GL019884.

(Received March 6, 2013, revised July 8, 2013, accepted July 18, 2013)

Tsunami Mapping Related to Local Earthquakes