Engineering Structures 27 (2005) 213–227

www.elsevier.com/locate/engstruct

he

paperesearchresbserved in

ry and pastl area is

s.

Seismic response of structures and infrastructure facilities during tLefkada, Greece earthquake of 14/8/2003

Christos Karakostas∗, Vassilios Lekidis, Triantafyllos Makarios, Thomas Salonikios,Issam Sous, Milton Demosthenous

Institute of Engineering Seismology and Earthquake Engineering (ITSAK), P.O. Box 53, GR 55102 Finikas, Thessaloniki, Greece

Received 5 March 2004; received in revised form 23 September 2004; accepted 29 September 2004

Available online 2 December 2004

Abstract

A strong(M = 6.2) earthquake struck the island of Lefkada (also called ‘Lefkas’) in western Greece on August 14, 2003. In thisthe seismic behaviour of the buildings and infrastructure in the region, as established by numerous in situ investigations by the rteam, is presented. For an integrated presentation, first some basic seismological issues are given, as well as a short description of failuof geotechnical nature. Buildings in the area are classified according to their structural system, and the response and damage oeach structural category is presented. The response spectra of the strong ground motion are compared with both contemporaGreek seismic codeprovisions, and their effect on damage observed is discussed. The distribution of damage in the meisoseismaalso presented and discussed. Analytical investigations on the seismic response of two representative buildings (atraditional and a reinforcedconcrete one) help in explaining observed damage and in gaining insight on various factors that mitigated the earthquake consequenceFinally, a short overview of the emergency management measures taken is also presented.© 2004 Elsevier Ltd. All rights reserved.

Keywords: Lefkada earthquake; Seismic response of structures; Structural damage

)

ake-

theeroinbynesstnd

elyFor-esd

e to-

of--C)hec-g

nd

1. Introduction

On August 14, 2003 at 08:15 local time (05:15 GMTa strong earthquake of magnitudeM = 6.2 occurred closeto the island of Lefkada in Western Greece. The earthquwas strongly felt in the rest of the Ionian islands (Cephalonia, Zakynthos, Ithaki etc) and in an extensive area ofmainland. The earthquake was recorded by several accelgraphs of the national permanent strong motion networkthe area, which were installed, serviced and monitoredthe Institute of Engineering Seismology and Earthquake Egineering (ITSAK). The Ionian islands area is one of thmost earthquake-prone in Greece, and was always clafied in Greek Seismic Codes among those with the highesseismic hazard. At the time of the seismic event, the isla

∗ Corresponding author. Tel.: +30 2 310 476 081/4; fax: +30 2 310 476085.

E-mail address: christos@itsak.gr (C. Karakostas).

0141-0296/$ - see front matter © 2004 Elsevier Ltd. All rights reserved.doi:10.1016/j.engstruct.2004.09.009

ats

-

-

i-

of Lefkada was estimated to be inhabited by approximat60,000 tourists, besides the 22,500 permanent residents.tunately no human loss occurred, while about 45 injuriwere reported, most of them from free-falling roof tiles another nonstructural elements. Serious structural damagbuildings was rather limited, given the intensity and proximity (∼12 km) of the earthquake epicenter to the townLefkada, the island’s capital, where a peak ground acceleration of ag = 0.42g was recorded at the accelerograph installed at the town hospital. One reinforced concrete (R/building collapsed, while damage to other structures in tarea was limited to local failures of structural and nonstrutural components that fortunately did not lead to buildincollapses. A considerable number of marine infrastructureswere affected moderately to heavily by the strong groumotion, while traffic on a large part of the road networkthe western part of the island was disrupted for several daydue to a significant number of landslides and rock falls.

214 C. Karakostas et al. / Engineering Structures 27 (2005) 213–227

redd,e tstedntalanto

thacopgs

terallyake

hesice

ynged

ns

lts,ofheinof

o

d

icaek

outai

14,the

ares.ap

d

asingC

ry

gy,e

lls,ri

ndrly

nda ofri

esadsi,

ndur

tage

From the very day of the main event and for mothan two months afterwards, scientists (seismologists ancivil engineers) from ITSAK were present at the islanrecording the aftershock activity and assessing damagbuildings, marine structures and road networks. The resultof these in situ investigations on structures are presenin this paper. For a better understanding of the evea short description of the most important seismologicand geotechnical issues pertaining to the earthquakeits consequences is given. A description of damagedifferent structural types of the existing building stock ispresented, along with explanations of the various factorscaused it. These evaluations are based both on macrosexaminations, and on dynamic analyses of typical buildinof the area, which were conducted at ITSAK for a betassessment of the seismic behaviour of structures. Finan overview of the emergency management measures tby the authorities is presented.

2. Seismological issues

The earthquake, of magnitudeM = 6.2, occurred onAugust 14, 2003 at 08:15 local time (05:15 GMT). Thypocenter coordinates, as computed by the GeophyLaboratory of Aristotle University of Thessaloniki, wer38.76◦ N 20.60◦ E and its focal depth ∼10 km. Longbefore the mainshock, digital accelerographs maintained bITSAK and belonging to the permanent National StroMotion Network were installed, seven of which recordthe mainshock [1]. The closest (R ∼ 12 km) to themain event accelerograph (station LEF1, seeFig. 1) wasthe one installed at the hospital of Lefkada town, wherea peak ground accelerationag = 0.42g was recorded(Fig. 2). In the transverse horizontal and vertical directiorespective PGAs of 0.34g and 0.19g were recorded. Theseismic sequence can be associated [2] with dextral strike-slip faulting along the Cephalonia Transform Fault [3],comprising the distinctive Cephalonia and Lefkada fauwith the latter running parallel to the western coastthe island of Lefkada. An extensive investigation of tseismotectonic properties of the area (seismic historythe region, mainshock mechanism, seismic sequencethe event etc) can be found in Karakostas et al. [2] Thelong duration of the event (estimated bracketed duration18 s) [1], combinedwith the high PGA values, list thisearthquake as one of the most intense ever recordeGreece.

3. Geotechnical issues

The earthquake had a significant number of geotechnimpacts, something rather uncommon in past Greearthquakes. The western part of the island is mountainwith steep slopes towards the western coast. The moun

o

,

d

tic

,n

al

f

in

l

s,n

Fig. 1. The island of Lefkada. The epicenter of the mainshock of August2003 is denoted by star. Permanent strong motion stations of ITSAK intowns of Lefkada (LEF1) and Preveza (PRE1) are denoted by dark squThe location of Lefkada island is denoted by the dark circle in the inset mof Greece.

slopes are less steep towards the eastern part of the islan(Fig. 1).

The soil structure beneath Lefkada town is classifiedlow quality (soil categories C and in some cases X accordto the Greek Seismic Code (EAK2000)–soil categoriesand E according to Eurocode 8/2002 [1,4]. Also, mainly inthe old historical district of the town the water table is vehigh (less than 1.5–2 m beneath ground surface).

Due to either poor soil conditions or steep morpholoseveral failures pertaining to geotechnical aspects werobserved [1], among which:

• Damage toport and marine structures (docks, seawabreakwaters, etc) in Lefkada town and at Lygia, Nydand Vasiliki villages due to ground settlement alateral spreading of loose surface soil layers and poocompacted fills behind the seawall facilities. (Fig. 3).

• Liquefaction and consequent loss of strength, grousettlements and lateral spreading in the broader arethe castle of the town of Lefkada as well as in Nydvillage.

• Landslides and rock falls on both natural and cut slopmainly at the western side of the island along the ronetwork joining the town of Lefkada with the villageof Tsoukalades, Agios Nikitas, Kathisma, KalamitsChortata, Dragano, Komilio, Aghios Petros, Vassiliki aPorto Katsiki beach. Thanks to the early morning hothat the mainshock occurred,slope failures did not causeany deaths. (Fig. 4).

• Lifeline failures were rather limited, with the mosserious one being the breakage of the main sew

C. Karakostas et al. / Engineering Structures 27 (2005) 213–227 215

da

totoese

n

teIn

tedte

by

nthtld

of

retd

Fig. 2. The Lefkada mainshock of August 15, 2003. Recordings at Lefkahospital (station LEF1).

pipeline in the town of Lefkada. It is interesting to notethat the failure occurred at a transition zone from softstif fer soils, a situation known from past earthquakesbe more prone to damage, due to difference in the strconditions on the pipeline in each soil type during thshaking.

• Ground settlement in some buildings along the waterfroin the town of Lefkada. Due to the more or less uniformsettlement under each building, any damage causedtheir load-bearing structural systems should be rathattributed to the shaking than the ground settlement.

s

t

or

Fig. 3. Ground settlement and lateral spreading of poorly compacfills behind the seawall in Vassiliki village caused cracking of concrepavement.

Fig. 4. Road connecting villages of Tsoukalades and Kathisma cut offlandslides (courtesy: A. Anastasiadis, Dpt. of Soil Dynamics, ITSAK).

a certain case of two adjacent buildings, the differedegree of settlement resulted in a differential heigdifference of approximately 5 cm between them, as coube macroscopically established by a common pipelinesystem thatbroke. (Fig. 5).A preliminary assessment of the geotechnical aspects

the earthquake can be found in Gazetas [5].

4. Seismic response of structures

4.1. Description of building types in Lefkada island

Buildings in Lefkada island can be classified in fivecategories, according to their load-bearing system (Table 1).

The town of Lefkada consists of the historic centdistrict (mainly with older, traditional buildings and sofsoil conditions), surrounded by the more recent Bei anNeapoli districts, with typically modern R/C buildings

216 C. Karakostas et al. / Engineering Structures 27 (2005) 213–227

l

Table 1Building types in the island of Lefkada

Category Description Remarks

A One- or two-storey stone masonrybuildings

Buildings more than 50 years old, with load-bearing system ofstone masonry. Sometimes they incorporate empirical lateraload-bearing systems, which are in general insufficient.

B One- to three-storey traditionalwood buildings

Met mainly in the town of Lefkada. Load-bearing system bywood frames, sometimes with clay bricks as infill.

C Buildings of special typology withdual structural system

A traditional structural typology met mainly in the oldtown district of Lefkada. Main load-bearing system of stonemasonry on ground floor level, complemented by a secondary(redundant) wood structural frame. Structural system of woodframes used in upper floor(s).

D Modern R/C buildings with one tofive storeys

The majority of houses, office and hotel buildings, up to fivestoreys high (in the town of Lefkada). Load-bearing systemwith cast-in-place reinforced concrete. Walls with hollowbricks used as infill.

E Middle-age and later-era monu-ments

Buildings of architectural heritage, with no special seismicprovisions. Mostly churches and some castles along theentrance from the mainland to the island.

en

fyf

g

,ryg

riay

1%ndtoofssost

yd-e

ottal

byof

lledsny

heiredinileofIn

avy

Fig. 5. Ground settlement resulted in 5 cm height difference betweadjacent buildings in Lefkada waterfront, as witnessed by broken pipelines(photo taken after repair).

on better quality, stiffer soils. The whole town has anestimated number of 2100 buildings. An in situ survey oa representative sample of the building stock (approximatel10% of the total stock) yielded the following distribution obuilding categories: 6%masonry buildings (categories A andE), 15% wood buildings (category B), 34% buildings withdual structural system (category C), while the remainin45% of the building stock consists of R/C buildings(category D). For the whole building stock of the islandthe only relevant data available are some preliminaresults from the 2001 National Census, but with a buildincategorization based rather on the construction matethan on the structural type. According to preliminardata from the Greek National Statistical Service site(http://www.statistics.gr), out of 15,683 buildings on the

l

island,38.87% are made of concrete, 1.04% of steel, 4.5of wood, 14.46% of bricks, 39.46% of stone masonry a1.66% of other materials. It should be noted that any effortfind aone-to-one correspondence with the structural typesTable 1 will most certainly introduce inherent uncertaintie(e.g., it is notcertain if, during the census, the buildingof category C were classified as masonry, as seems mprobable, or if some were classified as wood buildings).

(a) Masonry buildingsBuildings in category A are generally one- or two-store

old houses, many of them over 50 years old, with loabearing masonry walls made of stone or brick, weak lim(and rarely cement or clay) mortar, and usually, but nalways, without any seismic provisions such as horizonconcrete or wooden tie-belts. Floors are typically wooden,consisting of beams uniformly distributed and coveredwooden planking. The roofs in such buildings consistwooden trusses covered by wooden planking and tiles.

(b) Wood buildingsCategory B buildings are made of wood frames with

diagonal wood braces. In some cases, wood frames are fiwith clay bricks or covered by planks. This type of buildingexhibited a very satisfactory seismic behaviour, without adamage to the load-bearing wood frames.

(c) Buildings with dual masonry and wood frames systemCategory C buildings are mostly two-storey (and in

somecases three-storey) ones. As mentioned above, tspecial characteristic is the dual load carrying system uson ground floor level to handle seismic actions. The maload-carrying system consists of stone masonry walls, whthe redundant (secondary) load-carrying system consistswood frames at the inner perimeter of the masonry walls.case the masonry walls fail due seismic actions (e.g., he

C. Karakostas et al. / Engineering Structures 27 (2005) 213–227 217

Fig. 6. Traditional dual system building (inset photo) and mechanism of activation of secondary wood frame in case of failure of masonry load-bearingwall.

-r t

nadau

hegsal

s i

De

allyasose

eir. Tcreedaloftineruetn

cehe

tosiveith

ipis

ofset)reIns,uchofis

ssis-ed

9,inin

umall

gs)in

cracking and/or partial collapse leading to reduced loadbearing capacity), the wood frames are activated in ordecarry the upper floors’ loads (Fig. 6). On upper floors theload-bearing system consists of wood frames with diagotrusses andbrick infills. The dual bearing system, combinewith the relatively low mass of the upper storeys, exhibitsremarkably reduced vulnerability to earthquake actions. Dto poor soil conditions (and a high water table level), tmasonry walls are typically founded on extended footinmade up of horizontally placed tree trunks. An analyticinvestigation of the seismic behaviour of such structurepresented later in the paper.

(d) Reinforced concrete buildingsAs already mentioned, reinforced concrete (category

buildings comprise the majority of the building stock in thmore recent Bei and Neapoli districts. They are generone- to five-storey buildings, constructed during the ldecades. In many (especiallythose designed according tthe 1984 seismic code and thereafter), shear walls are uto resist seismic actions. In this category of buildings,an important characteristic, playing a key role in thresponse to earthquakes, is the type of the ground storeyGeneral Greek Building Code (not the code for the seismidesign) permits buildings to have an open ground sto(“pilotis”, to be used as car parking, playground or flowerbspace), without counting it in the maximum permitted totfloor area. This became very popular, but created a “sfirst storey, due to the drastic reduction of brick infillscomparison to the storeys above. A similar, but not as sevproblem is created with ground storeys used as shops, dthe elimination of several infill walls for creating large fronwindows and large continuous interior spaces. In the islaof Lefkada, an area with the highest seismicity in Greealthough use of pilotis is not uncommon (at least for t

o

l

e

s

)

t

d

he

y

”

e,to

d,

more recent buildings), in most cases care has been takenmitigate their undesirable consequences through extenuse of shear walls and design of regular space frames wsmall to medium spans. Also, the quality of workmansh(for both load-bearing and nonstructural components)in general higher than the one met in other regionsGreece. Carefully constructed infill walls and extensive uof horizontal R/C belts (up to two along the wall’s heighis a common practice. Various types of foundations amet, ranging from spread footings to mat foundations.the town of Lefkada, due to rather poor soil conditionproper measures are usually taken for the foundations, sas the use of concrete piles. An analytical investigationthe dynamic behaviour of a representative R/C buildingpresented later in the paper.

(e) Middle-age and later-era monumentsThe fifth category of buildings, (category E), include

middle-age and later-era monuments. Their seismic retance system, if any, was formed empirically by experienclocal technicians. Structuresof this category in the island ofLefkada are mainly churches and some castles.

4.2. Strong ground motion compared to seismic codeprovisions

The first Greek Seismic Code (AK) was issued in 195and was revised in 1984. A major new revision took place1992 (EAK1992), and upgraded versions were published2000 and 2003. Until 1992, design was based on maximallowable stresses, and thereafter on ultimate strength. Incodes, the area of Ionian islands (where Lefkada belonwas characterized as one with the higher seismicityGreece.

218 C. Karakostas et al. / Engineering Structures 27 (2005) 213–227

ic

ffi-as

dt,i-

edenn,

thano

al

datic0)e

s

eign

ic

ar

umwa

nlahe

h

uiltherat

gththesichor.ntsas

ularear

ckgesdof

ndas

sthegs

As5%e

to

eicage

inthe

hendingends)n,

ns

gsf

Fig. 7. Response spectra of horizontalcomponents of Lefkada mainshock(5% damping) compared with elastic design spectra of the Greek SeismCode (EAK2000) and the pre-1992 code (AK) provisions.

For Lefkada, the base shear seismic design coecient, according to the 1959 Greek Seismic Code wε = 0.08, 0.12 and 0.16, for firm, medium ansoft soils, respectively. This coefficient was constanindependent of the building’s period and applied unformly to all buildings. Since the 1959 Code was bason the allowable-stress design method, the coefficiis modified to correspond to ultimate strength desigleading to values ofε′ = 0.14, 0.21 and 0.27 [6].

For the seismic zone in which Lefkada belongs, withe highest seismic hazard, seismic codes from 1992onwards, establish a ground acceleration coefficientα = 0.36 and typical design spectra (with a spectrmagnification factorβo = 2.5). In Fig. 7, the responsespectra of the two horizontal components of the Lefkamainshock (for 5% damping) are compared with the elasdesign spectra of the Greek Seismic Code (EAK200provisions (for soil types B-medium and C-soft). In the samfigure, the pre-1992 provisions (AK) are also plotted.

In Greece, limit-state design was introduced in the codein 1992, and a reduction (‘behaviour’) factorq is appliedto the seismic actions in order to take into account thelastoplastic behaviour of the structure during the desearthquake. According to the post-1992 Greek SeismCodes, the maximum allowable behaviour factor isq =3.5 for reinforced concrete frames (with or without shewalls), q = 1.5 for masonry andq = 2 for wood frames.Before 1992, design in Greece was based on the maximallowable stress concept, i.e., essentially elastic designused. It is obvious that low-rise buildings with relativelysmall mass and fundamental period(T < 0.15–0.20 s),which comprise the majority of the building stock iLefkada, were not heavily stressed, due to the particushape of the response spectrum of the mainshock. Tductility demands imposed by this particular earthquak

t

df

s

re

on buildings in this specific range were not too hig(i.e., for 0< T < 0.20 s ductility demandsD are in therange of 1.6 < D < 2.25), thus explaining the limiteddamage observed. Higher buildings were in general baccording to modern code provisions, thus possessing higresistance to seismic actions. Moreover, it is noted thexisting buildings possess a substantial amount of strenreserves (depending mainly on their redundancy and onoverstrength of individual structural members), as well apossible additional energy dissipation mechanisms, whcontribute to a significant increase of their behaviour factExperience gathered from this and previous seismic evesuggests that seismic protection of Greek urban arerelies also on several alternative factors (such as regconfiguration of the structural system, extensive use of shwalls, properly built infill walls with horizontal R/C belts,highquality in materials and workmanship, etc.) [7].

4.3. Distribution of damage in the meisoseismal area

Most damage from the August 14, 2003 mainshooccurred in the island of Lefkada. Also less severe damawere reported in the nearby regions of Thesprotia anAitoloakarnania in the mainland, as well as in the islandCephalonia.

As established through in situ inspections in the islaof Lefkada, serious structural damage to buildings wrather limited, given the intensity and proximity of themain event. In Fig. 8 damage distribution in various partof the island is shown. Damage pies in the figure showpercentage of unharmed buildings (dark parts) vs. buildinthat presented some form of damage (light parts). Thepercentage of harmed buildings varies from 1% to 29%.can be seen from the figure, most damages (15% to 2harmed buildings) were observed in communities on thwestern part of the island, in a NE/SW direction parallelthe seismic fault that caused the mainshock (seeSection 2).Assuming a more or less equal vulnerability of low-risbuildings throughout the island, directivity of the seismmotion seems to have played an important role in the damdistribution. Extended damage was especially reportedthe villages of Lazarata, Asprogerakata, Kavalos andsurrounding area, most probably due to the soil conditionsand the basin morphology of the region that amplified tshaking. Some deviations from the general damage tre(as in the older villages of Poros and Marantoxori,the southeastern part of the island, with a high damapercentage or in the more recent villages of Kathisma aKalamitsi on the west coast with relatively few damagecan be attributed to building stocks with different thausual vulnerability levels: in recently created communitiesbuildings conform to the more strict provisions of moderseismic codes, while in old villages, most of the buildingwere built with no seismic provisions.

The area with the higher number of damaged buildinwas the townof Lefkada, in the northeastern part o

C. Karakostas et al. / Engineering Structures 27 (2005) 213–227 219

enngs

isity

eress.ofdul”,ens

d,

icot

is

newnhes ofdsherrse,ateitthles.

ofe

eedialedsntnotirce)

ck,essesss

ost

oodd-to

Fig. 8. Damage distribution of buildings in Lefkada island. The dark greportion of damage pies corresponds to unharmed, light yellow to buildipresenting some form of damage.

the island, where the majority of the building stockfound. Despite the intensity of the event, and the proximof the town to the epicentral area (∼12 km), only one,poorly designed R/C building collapsed. Most damages wconcentrated in the old town district, while minor damagwere reported to the more modern Neapoli and Bei districtThe intensity for the main event in the old, historic districtLefkada was estimated between VII and VIII of the modifieMercalli scale, with no apparent differentiations throughothe district. A two-level inspection of 3165 buildings, alover the island of Lefkada, resulted in 1544 “green1495 “yellow” and 126 “red”. These results confirmed thevaluations of the macroseismic intensity. The classificatioreferred to degree of damage (“red”, “yellow”, “green”) ibriefly described as follows:

“Green”: Original seismic capacity has not been decreasethe buildings are immediately usable and entry isunlimited.

e

t

,

“Yellow”: Buildings in this category have decreased seismcapacity and should be repaired. Usage is npermitted on a continuous base.

“Red”: Buildings in this category are unsafe and entryprohibited. Decision for demolition will be madeon the basis of more thorough inspection.

4.4. Observed damage to buildings

As mentioned previously, the main event produced oof the strongest motions recorded in Greece. In the toof Lefkada, at approximately 12 km from the epicentre, trecorded seismic motion had peak ground acceleration0.42g and 0.33g in the horizontal directions and a bracketeduration of 18 s [1]. Despite its severity, damage to building(and especially the most modern R/C ones) was ratmoderate, for reasons that are presented later. Of couthe seismic vulnerability of a building depends to a gredegree on its structural system. A notably common — albnonstructural — damage, met in all types of buildings witiled roofs, was the detachment of a large percentage of tiA more detailed description of damage observed for eachthe structural typologies met in the island of Lefkada (seSection 4.1) is presented below.

4.4.1. Observed damage to buildings according to theirstructural system

(a) Masonry buildingsTraditional stone masonry buildings (category A, se

Section 4.1) have usually one or two storeys; they constituta small percentage of the building stock in the island, anare mainly met in the villages. Damage (cracking or partcollapse of masonry walls) to such buildings was observmainly in villages in the western part of the island. Failureare to be attributed mainly to insufficient or nonexisteseismic resistance measures (almost all of them werebuilt according to any seismic code), as well as to thealready poor condition (old age, inadequate maintenaneven before the main event.

(b) Wood buildingsThese constitute a small percentage of the building sto

and are mainly met in the town of Lefkada. The wood framthat comprise the load-bearing system have diagonal trusto carry transverse loads. Due to their relatively small maand the flexibility of the wood frames, they presented almno structural damage.

(c) Buildings with dual masonry and wood frames systemTraditional buildings with a dual structural system

(category C, seeSection 4.1), behaved in a rathersatisfactory way, given the intensity of the main event. Insome cases partial collapse of the masonry walls took place(Fig. 9(a)), but the structural stability of the building wasensured by the activation of the secondary (redundant) wframe on the ground floor level. In the upper floors, the loabearing wood frames suffered no damage, but cracking

220 C. Karakostas et al. / Engineering Structures 27 (2005) 213–227

reer

llsforoned

edoff

o adke

of

f

eythead-

ereonelis

alheatentddede

dlly,lly

Fig. 9. Traditional dual system building in Lefkada town. (a) Local failuof stone masonry wall over door opening. Vertical loads from uppfloor carried by activation of redundant wood frame on the ground floor.(b) Typical observed damage of brick infills at midheight of upper floor.

the brick infills was observed (Fig. 9(b)). These cracks weredifficult to be noticed from outside, since the external waat the upper storeys are typically clad with zinc sheets (rain protection). Damage could thus be observed onlythe interior faces of the walls, which are usually plasterwith lime. Due to the use of the extended wood footingsdescribed inSection 4.1, and the relatively small massof these buildings, no foundation settlings were observdespite the poor soil conditions at the old town districtthe town of Lefkada, where the majority of such type obuildings is found. Severe or total damage observed tlimited number of buildings of this type can be attributeto old age and poor maintenance, with the earthquaaggravating their already poor condition.

,

Fig. 10. (a) Collapse of poorly designed R/C building in the townLefkada; (b) detail of corner column, with lack of stirrups.

(d) Reinforced concrete buildingsIt should be noted that the majority of the majority o

R/C (category D, seeSection 4.1) buildings (and especiallythe post-1985 ones) behaved in a satisfactory way. Thresponded essentially within the elastic range duringmainshock, presenting no or minimal damage to their lobearing system. Nevertheless, several local damages wobserved to structural and nonstructural elements. Onlycollapse of a three-storey building occurred in the Neapodistrict of the town of Lefkada (Fig. 10). The ground floorfailed totally, while the remaining two retained a structurstability enough to spare the lives of the inhabitants. Tbuilding was poorly designed with inadequate stirrupsthe columns while each storey was constructed in differyears (1969, 1975 and 1980) and a wooden penthouse asometime later. Apart from the obvious poor quality of thworkmanship, on the ground floor thick brick infill wallswere constructed at its backside while large openings existeat the façade, thus leading to torsional effects. Additionadue to the openings, short columns were unintentiona

C. Karakostas et al. / Engineering Structures 27 (2005) 213–227 221

for

tenk

n

fasd,nts

in

or

s.

tlls

b

eial

of

-f-

uee

Fig. 11. Shear failure of ground floor column of R/C building (town oLefkada). The non-existence of infill walls at the façade of the ground flostorey led to torsional phenomena.

created in the façade. The collapse can thus be attributo the overall poor seismic resistance of the building thathe intensity of the seismic motion. In fact, the collapse tooplace in an area where damage to other R/C buildings waslimited, and soil conditions were better than in other towdistricts.

Also, in several waterfront buildings in the town oLefkada. a more or less uniform foundation settlement wobserved. Since no differential settlements were involveany local damages to structural or nonstructural elemeshould be rather attributed to the shaking itself.

Among the most typical local failures observed in R/Cbuildings are the following:

– Severe damage to poorly designed columns and wallsbuildings with soft storey at ground level.

– Damage to vertical structural elements at ground flolevel due to non-symmetrical distribution of infill walls.

– Shear and/or flexure failure of poorly designed column(Fig. 11).

– Shear failure of short columns that were originally nodesigned to act as such (e.g., non-continuous infill waon either side due to openings) (Fig. 12).

– Shear failure of R/C shear walls due to inadequate wereinforcement.

– Flexural cracking at R/C beam ends. It should bnoted, however, that this type of failure has a benefic

d

Fig. 12. Short column failure in R/C building (built in 1983).

Fig. 13. Detachment of infill wall built out of the plane of the R/Cstructural frame. Architectural detail presenting a potential hazard in caseshaking.

contribution to the overall building stability, sinceformation of plastic hinges at columns is avoided.

– Cracking of infill walls (diagonal shear cracks, detachment of the wall from the surrounding frames, out-oplane collapses) (Fig. 13).

– A systematic oxidation of the steel reinforcement wasobserved at the base of the ground floor columns, dto the high underground water level. In some cases, th

222 C. Karakostas et al. / Engineering Structures 27 (2005) 213–227

f

or

oryiv

thetila

larrveingutithenthing

ofgeefed

datedin/C

ei

eral

rum

a,indend

uild-d

ha-filltyssof

asat

naleatert

n,ongeresn

e

s ang,ithnd

sn ofal

aed,

ceing

Fig. 14. Cracking of masonry walls of Agios Minas church (town oLefkada).

oxidation was very severe, a serious vulnerability factfor the structural safety.

(e) Middle-age and later-era monumentsAmong middle-age and later-era monuments (categ

E), churches suffered the most serious and extensdamage. More than 40 churches all over the island (withmajority in the town of Lefkada) were put out of service unrestoration measures were taken. Churches in Lefkadatypically built with stone masonry walls, have a rectangufloor plan and a wooden roof. Serious damage was obseat the perimeter walls as well as at the corners of adjoinwalls (Fig. 14). No collapse of bell towers was reported, bone should notice that most towers were of recent build (wR/C or steel elements), since the original ones had bedestroyed in past earthquakes. It should also be notedseveral churches had been already restored after sufferdamage during an earlierM = 6.5 earthquake in 1948.The middle-age castle of Agia Mavra, at the entrancethe town of Lefkada from the mainland, suffered no damato its exterior walls. In its interior, partial collapse of sombuilding ruins as well as somepermanent displacements osome stoneparapets and decorative elements was observ

4.4.2. Factors affecting damage to buildingsAs mentioned earlier, damage from this strong event in

Lefkada was not as severe as one might have expecteview of the intensity of the mainshock and the fact thpre-1985 buildings were built according to the outdat1959 seismic code. The majority of the building stockLefkada behaved in a nearly elastic manner (especially Rbuildings and, to a great extent, traditional dual systembuildings), since they did not present serious damage to th

e

re

d

at

.

in

r

load-bearing systems. This should be attributed to sevfactors, the most important of which are:

(1) Due to the particular shape of the response spectof the mainshock, buildingswith relatively smallfundamental period(T < 0.15–0.20 s), whichconstitutes the majority of the building stock in Lefkadwere not heavily stressed. Higher buildings weregeneral built according to modern (post-1985) coprovisions, thus possessing higher ductility levels aresistance to seismic actions.

(2) The substantial strength reserves possessed by bings, due mainly to their redundancy, the infill walls anthe overstrength of individual structural members.

(3) The existence of additional energy dissipation mecnisms (e.g., those provided by the cracking of the inwalls). The beneficiary effect of the usually good qualiof workmanship met in many of the buildings towardtheir seismic behaviour is also not to be neglected. Thifact probably stems from the long-lasting experiencethe local constructors with strong earthquakes.

(4) The earthquake motion was most probably notintense throughout the city as the record obtainedthe hospital building would indicate. Large motiovariations are typically observed at small epicentrdistances (also the case here) and become even grdue to variations inlocal soil conditions. The fact thapractically no damage was observed in sections of towwhere the quality of construction was statistically ndifferent from that in the damaged areas, is a stroindication of this effect. Site response analyses that wrecently performed at ITSAK for different locationin the town of Lefkada also support this conclusio(A. Anastasiadis, personal communication [8]).

4.5. Analytic investigation of seismic response of differentbuilding types

In order to gain insight into the seismic behaviour of thbuilding stock, a dynamic analysis of two buildings withdifferent structural types was performed. The first one irepresentative traditional dual system (category C) buildiwhile the second is an R/C (category D) building, both wdimensions and structural details typical of the ones fouin Lefkada. These two structural types comprise almost 80%of the building stock in the town of Lefkada. Since thiinvestigation is intended to serve as a general evaluatiothe overall dynamic behaviour of the examined structurtype and the various factors that affect it, and not asdetailed investigation of some specific damage observlinear elastic analyses wereperformed. This approach isfurther justified by the fact that the majority of the buildingstock in Lefkada behaved in a nearly elastic manner, sinthey did not present serious damage to their load-bearsystems. Damage observed to masonry buildings (categoriesA and E, seeSection 4.4.1) was to be expected, given the

C. Karakostas et al. / Engineering Structures 27 (2005) 213–227 223

n).

Fig. 15. Section of Finite Element model of traditional dual system structural building (façade elements not show

in

rechhe

ceh

inry

gt

ed

rgratsllyllytssheea

ofs

glof,

0)

asInisasss

sis3 snse

ea

lack of seismic protection measures and poor conditionthe majority of such buildings. On theother hand, damageto wood (category B buildings) was minimal (Section 4.4.1).Thus the analytical investigation was focused on the mointeresting cases of buildings of categories C and D, whiin general performed better than anticipated, given tproximity and intensity of the mainshock.

4.5.1. Traditional dual system buildingThe interesting concept in these buildings is the existen

of the redundant wood frame on the ground level, whicis fully activated after possible local collapse of the maload-bearing system, which consists of exterior masonwalls of thickness of0.50–0.70 m. As already mentionedbefore, the overall response of this type of buildings durinthe Lefkada earthquake was satisfactory, with no damagethe wood load-bearing frames which performed as plannin the few cases of masonry failures (Fig. 9(a)). The mostcommon damage was the failure(out-of-plane dislocationand/or in-plane cracking) of the infill walls on the uppestoreys. Buildings of this type constitute 34% of the buildinstock in the town of Lefkada and a significantly biggepercentage for the rest of the island. In the town of Lefkadthe majority of the buildings in the old historic center districbelong to this structural type. Due to poor soil conditionand a high water table level, the masonry walls are typicafounded on extended footings made up of horizontaplaced tree trunks. The wood frames (with diagonal struon the upper floors are usually filled by raw clay brickand are protected from rain by zinc sheets nailed on texterior faces. Floors are typically wooden, and given thrather small plan dimensions of the buildings, they provide

o

,

)

rather satisfactory diaphragmatic in-plane behaviour. Roare also made of wood and covered by tiles.

For the analysis of a two-storey representative buildinthe SAP2000 [9] structural analysis program was used. Shelelements were used for modeling the masonry wallsthe ground floor and the infill walls of the upper floorsand frame elements for the wood frame (Fig. 15). Forthe mechanical properties of the materials involved, thefollowing mean values were used for Young’s modulusE ,Poisson ratiov and densityρ, based on existing literatureand past experience of the research team:

– Wood E = 9000 000 kN/m2 v = 0.30 ρ = 0.5 t/m3

– Stone masonry E = 4325 000 kN/m2 v = 0.15 ρ = 2.7 t/m3

– Clay brick infillwalls

E = 1708 000 kN/m2 v = 0.15 ρ = 2.1 t/m3

The more recent Greek Seismic Code (EAK200prescribes the same viscous damping ratioζ = 5% for boththe wood frame and the masonry walls, so this value wused for the elastic analysis of the traditional building.buildings of this type, the largest portion of their massdistributed on the exterior walls. The ground floor storey hfive to six times the mass of a typical upper storey. This madistribution differs from that of an R/C building, where mostof the mass is found on the floor slabs. Eigenvalue analyyielded a fundamental period of around 0.02 s (around 0.0for three-storey buildings). As can be seen by the respospectrum of the mainshock (Fig. 7), in this period rangethere is almost no spectral amplification of the excitation.

Time history analyses were also performed, using threcordings of the mainshock in the town of Lefkad(station LEF1). Three different excitation directions (0◦,45◦ and 90◦) were assumed, inorder to get an envelope

224 C. Karakostas et al. / Engineering Structures 27 (2005) 213–227

F tho

ig. 16. Envelopes of tensile stresses in masonry elements from time-history analysis: (a) stressess11, (b) stressess22. Contours above typical tensile strengf masonry (>0.4 MPa) as well as at highest stress levels are denoted by arrows.

it

lss:ns

e

irndes

se

e,

--ala-

seehee

sa

ed

outn

eeasehere/Cof

gnistedis

theralts,--un-ldof

s-f

of the possible stress states of the building. InFig. 16,the envelopes of the tensile stresses are presented, wcontours indicating stress levelss11, s22 > 0.4 MPa (typicaltensile strength for Greek masonry). High tensile stressconcentrations are observed in midheight of the infill walof the upper floors, agreeing with in situ observationtypical failures observed in post-earthquake investigatioinvolved the detachment of the infill walls from theirsurrounding frames, crackings and out-of-plane falls of thinfill walls (Fig. 9(b)). The analyses also show a low stressconcentration on the ground floor masonry, due to thebig thickness, with an increase of tensile stresses arouopenings (and indeed, the relatively few damage casobserved for stone masonry involved local failures in theregions — seeFig. 9(a)). At the same time, wood framesremainin the elastic range, presenting no structural failura fact also confirmed from in situ investigations.

The analytical investigations agree with field observations of a satisfactory behaviour of buildings of this structural type during the Lefkada earthquake. It is a structurconcept developed before 1800 A.D. by the local popultion, that has proven its merit in many instances of severeexcitations in this earthquake-prone region: even in the caof very strong shaking, damage is usually limited to thmasonry walls, while the redundant wood frame assures tlife safety of theinhabitants and the rather easy repair to thoriginal state.

4.5.2. Reinforced concrete buildingIn order to evaluate the contribution of various factor

that affected the dynamic response of R/C structures,four-storey building with dimensions and structural details

hrepresentative of those found in Lefkada was analyz(Figs. 17and18). The structural frameconsists of columnsand shear walls in a rather dense layout (spans of ab3.5 to 4 m). Shear walls constitute a significant portioof the load-bearing system, asis rather typical in Lefkadabuildings. The �-shaped core is typically used to housthe stairs and/or lift of the building. In the analyses, throle of two factors that were deemed most important winvestigated, i.e., the role of infill walls and the compliancof the foundations. Given the long-lasting experience of tlocal population with strong earthquakes, infill walls ausually carefully constructed, with several horizontal Rbelts along their height, clearly affecting the responsethe structure, although their role, according to the GreekSeismic Codes, is not taken into account during the desiof a building. Also the foundation compliance, whichusually not taken into account in the design process, affecin a beneficial manner the overall building behaviour, assuggested by the analytical investigation that follows.

For the analysis the SAP2000 [9] structural analysis pro-gram was used. Shell elements were used for modelingshear walls, and beam elements for the rest of the structuframe. Infill walls were modeled by diagonal truss elemenaccording to well-established methods proposed in the literature [10]. For the simulation of the foundation compliance, vertical and horizontal springs were used at the fodation. Based on information of the soil profile in the odistrict of the town of Lefkada, a dynamic spring constantksdv = 2400 kN/m3 was estimated for the vertical springsandksdh = 1600 kN/m3 for the horizontal ones.

Eigenvalue analysis yielded the two fundamental tranlational periods in the rangeof 0.20–0.25 s for the case o

C. Karakostas et al. / Engineering Structures 27 (2005) 213–227 225

dhe

ye

e

ofgoin

rl

oysam

ly,

/Cce,

l

ayds

ts,

noningntno

ryndnger

usur,

s,ls

ck

elyeon

se

onsnd

Fig. 17. A typical R/C building in the town of Lefkada that respondeelastically during the mainshock. The inset photo shows the interior of tground floor. Note the small spans of the beams and the extensive usshear walls.

the bare (no infill walls), clamped at the base building. Tak-ing into account infill walls, the corresponding periods lain the 0.16–0.19 s range. The role of the soft soil was moraccentuated: fundamental periods lay in the 0.75–1.00 srange when springs were added to the original bare clampstructure.

Time history analyses performed, using the recordingsthe mainshock (station LEF1), confirmed the importance othe above factors in the seismic response of the buildinConsidering the clamped structure and not taking intaccount infill walls, shear stress concentrations surpassthe code limits(s12 > 1.9 MPa) were observed at the groundfloor level of the shear walls, and especially at the�-shapedcore (Fig. 19(a)). Tensile and compression stress fields wewithin the allowable limits, thus explaining the almost totaabsence of flexural type failures. Taking infill walls intoaccount results in a decrease of the stress field level by ab15% on average. The compliance of the foundations plaa more important role, leading to a decrease of the strefield by 35% on average, combined with maximum verticand horizontal displacements of the order of 4.5 and 3 crespectively (Fig. 19(b)).

Since no damage was observed to buildings conformingto the Seismic Code provisions, the above analyses indicate

eof

d

f

.

g

e

utssl

that the compliance of the foundations and, secondariinfill walls, might have played a significant role in mitigatingthe consequences of the severe Lefkada earthquake on Rbuildings. Of course, one should also not forget the existenof all other factors (attenuation of the seismic motionoverstrength and regular configuration of the structurasystem etc) that contribute to added safety margins of R/Cbuildings.

5. Emergency management

Emergency response to the disaster was underwimmediately after the earthquake. Fortunately, the demanwere less than those of previous similar intensity eventhanks to the fact that thisM = 6.2 eventdid not cause lifelosses or heavydamage to structures. At the time of the maievent, Lefkada Island had an estimated combined populatiof about 80,000 permanent residents and tourists. Accordto the 2001 National Census there are 22,506 permaneresidents in the island. Despite the dense population,deaths and about 45 injuries were reported.

The earthquake damaged mainly unreinforced masonand poorly designed R/C structures as well as roads amarine structures, and caused limited damage to drinkiand waste water systems as well as to the electrical powsupply network. In fact, disruption of the traffic networkon the western part of the island was the most serioinfrastructure problem caused by the earthquake. Sixty-focivil engineers from the Greek Ministry for the EnvironmentPhysical Planning and Public Works formed 32 inspectionteams to proceed with visual assessments of buildingproviding the basis for determining which structures wilrequire more thorough examination. The inspection waperformed in two stages, a rapid one and a more detailedone, and was completed about 20 days after the mainshofor all the buildings of the island. Deviating from usuallyfollowed practices in previous Greek earthquakes, in thfirst level inspection, buildings were characterized simpas inhabitable or not inhabitable. It was only in the mordetailed, second level inspection that the usual classificatiof the buildings in the known green/yellow/red categories(described inSection 4.3) was applied. The inspectionprocedure resulted in a more rapid first-level evaluation andits applicability to future events should not be ruled out.

6. Discussion and conclusions

The earthquake(M = 6.2) that struck the island ofLefkada on August 14, 2003 was one of the most intenever recorded in Greece. The maximum horizontal peakground acceleration wasag = 0.42g, and the correspondingestimated bracketed duration was 18 s. Several conclusican be drawn from the investigation of the earthquake aits effect on structures and infrastructure facilities:

226 C. Karakostas et al. / Engineering Structures 27 (2005) 213–227

Fig. 18. Typical plan and formwork of representative R/C building.

Fig. 19. Envelopes of shear stresses in shear walls from time-history analysis (without infill walls): (a) structureclamped at base, (b) structure onsprings tomodel foundation compliance. Contours above allowable stress level (>1.8 MPa) as well as at highest stress levels are denoted by arrows.

ofed

cengind

orer

th

.

her

nencial

e

• Due to either poor soil conditions or steep morphologythe island, the geotechnical failures observed exceedthose of most previous strong earthquakes in GreePhenomena of ground settlement, lateral spreadiliquefaction, landslides and rock falls were observedseveral areas of the island, leading to damage of port anmarine infrastructures and of a significant part of the roadnetwork in the western area of the island.

• Despite the strong shaking and, in some cases, the posoil conditions (such as in the town of Lefkada), damagto buildings was rather limited. Due to the particula

.,

shape of the response spectrum, low-rise buildings wismall (<0.20 s) fundamental period (which constitute themajority of the building stock) were not heavily stressedHigher buildings were in general built according tomodern seismic code provisions, thus possessing higresistance to seismic actions.

• The area of Lefkada is one of the most earthquake-proin Greece. Due to frequent shaking, the local populatiohas a lasting awareness of earthquakes, and takes specare in the quality of workmanship in buildings. It isthis experience, that led, even before 1800 A.D., to th

C. Karakostas et al. / Engineering Structures 27 (2005) 213–227 227

licfr

tesban

edicrell,

il

e

te

osivotnnd

dy

ein

enes

thec

gncy

,A,-,i-

,cialof

thencef

sics

is

ral].ece,nal

13,

95:on

,

nd

creation of traditional buildings in the area with a dua(masonry and wood frame) system to handle seismactions. Even in modern, R/C buildings, the quality oworkmanship is generally higher than that met in otheGreek regions.

• The various failures observed in reinforced concrebuildings (collapse of one building, and local failureof the load-bearing systems in other cases), canattributed rather to poor design and workmanship thexcessive shaking. Local failures observed in traditional(masonry and dual system) buildings can be attributto old age, poor condition and lack of adequate seismresisting mechanisms. In many cases, failures welimited to nonstructural components, such as brick infiwalls of R/C or wood load-bearing frames. Finallyseveral churches developed cracking in theirload-bearingmasonry walls, and were put out of service untrestoration measures were taken.

• Analytical investigationsof a traditional dual systembuilding and a representative R/C building werperformed. The analytical results agree with fieldinspectionsof damage, and also point out the significanrole of the compliance of the foundations and thextensive use of infill walls in the mitigation of theconsequences of the severe Lefkada earthquakebuildings. This earthquake, as well as previous onehas demonstrated the importance of several alternatfactors that contribute to increased safety marginsbuildings. Among such factors, which usually are notaken into account in the design, one can mentiothe attenuation of the seismic motion, overstrength aproper configuration of the structural system, qualityof workmanship, compliance of the foundations ansoil-structure interaction phenomena in general, properlplaced infill walls, etc.

• The emergency management of the event is deemsuccessful, especially since the earthquake took placea densely populated region. A two-level inspection of thbuilding stock was completed within 20 days of the maievent, resulting in reduced damage to economic activition the island.

Much has been learned from this and previous earquakes, and significant progress has been made in Gre

e

n,ef

d

-e,

especially during the last 20 years, in improving buildincodes and practices and in handling earthquake emergesituations.

Acknowledgements

The authors would like to thank Drs. Dimitriou PMargaris V, Papaioannou Ch, Theodoulidis N, SavaidisDpt. of Engineering Seismology, ITSAK and Drs. Anastasiadis A, Klimis N, and Makra K, Dpt. of Soil DynamicsITSAK, for providing data on seismological and geotechncal issues.

References

[1] Margaris B, Papaioannou C, Theodulidis N, Savvaidis A, AnastasiadisA, Klimis N et al. Preliminary observations on the August 142003 Lefkada Island (Western Greece) earthquake. EERI speearthquake report, November 2003 (Joint report by InstituteEngineering Seismology and Earthquake Engineering, NationalTechnical University of Athens & University of Athens).

[2] Karakostas VG, Papadimitriou EE, Papazaxos BC. Properties of2003 Lefkada, Ionian Islands, Greece, earthquake seismic sequeand seismicity triggering. Bulletin of the Seismological Society oAmerica 2004;94(5):1976–81.

[3] Louvari E, Kiratzi AA, Papazaxos BC. The cephalonia transform faultand its extension to western Lefkada Island (Greece). Tectonophy1999;308:223–6.

[4] Dimitriou P, Theodulidis N, Margaris B, Papaioannou C, SavvaidA, Anastasiadis A et al. The Lefkada earthquake(M = 6.2): strongground motion–effect of the earthquake on the built and natuenvironment. Technical Chamber of Greece; 2004. p. 62 [in Greek

[5] Gazetas G. Geotechnical aspects of the Ms 6.4 Lefkada island, Gre2003 earthquake: preliminary assessment. Proc. 5th internatioconference on case histories in geotechnical engineering. 2004.

[6] Anagnostopoulos SA, Rinaldis D, Lekidis VA, Margaris VN,Theodulidis N. The Kalamata, Greece earthquake of September1986. Earthquake Spectra 1987;3(2):365–402.

[7] Lekidis VA, Karakostas CZ, Dimitriu PP, Margaris BN, Kalogeras I,Theodulidis N. The Aigio (Greece) seismic sequence of June 19seismological, strong-motion data and effects of the earthquakesstructures. Journal of EarthquakeEngineering 1999;3(3):349–80.

[8] Anastasiadis A. Dpt. of Soil Dynamics, ITSAK, 2004 [personalcommunication].

[9] Computers & Structures Inc. SAP2000—structural analysis programver. 8.2, Berkeley, CA, USA; 2003.

[10] Paulay T, Priestley MJN. Seismic design of reinforced concrete amasonry buildings. John Wiley & Sons Inc.; 1992.