Reconstructions of the coastal impact of the 2004 Indian Ocean tsunami in the Khao Lak area, Thailand

Reconstructions of the coastal impact of the 2004 Indian Ocean

tsunami in the Khao Lak area, Thailand

Johanna Mard Karlsson,1,2 Alasdair Skelton,1 Michael Sanden,1 Mansour Ioualalen,3

Narongrit Kaewbanjak,4 Nuttita Pophet,5 Jack Asavanant,5 and Axel von Matern1

Received 13 May 2009; revised 4 July 2009; accepted 17 July 2009; published 27 October 2009.

[1] Khao Lak, SW Thailand was severely affected by the tsunami on 26 December 2004.Here we present reconstructions of its coastal impact in this area. These are basedon (1) eyewitness reports alone and (2) eyewitness reports supported by videos andphotos of the tsunami and the damage it caused, field measurements, and satellite imagery.On the basis of eyewitness reports, we estimated that the sea began retreating at1000 local time (LT) and, based also on photos, that the tsunami arrived at 1026–1029 LT.On the basis of videos of the tsunami, we estimated an offshore wave direction of083 ± 3� and on the basis of the paths by which eyewitnesses were carried, we estimatedan onshore direction of 088 ± 6�. On the basis of videos, we calculated that thevelocity of the wavefront on its final approach was 33 ± 4 km/h. We obtained tsunamiheights of 7.3 ± 0.8 m (relative to ground level) on the basis of eyewitness reports and8.0 ± 0.6 m (relative to mean sea level) on the basis of field and photographic data.On the basis of eyewitness reports and photos, we concluded that Khao Lak experienced atleast two main waves with a period >40 min. From eyewitness reports and satelliteimagery, we measured maximum inundation �0.5 km in the southern part of the area,which is confined by a steeply sloping hinterland, and �1.5 km in the more gently slopingnorthern part. Comparison between these reconstructions supports the reliability ofeyewitness reports as a source of quantitative data, and comparison with the numericalsimulation by Ioualalen et al. (2007) supports the validity of the simulation.

Citation: Mard Karlsson, J., A. Skelton, M. Sanden, M. Ioualalen, N. Kaewbanjak, N. Pophet, J. Asavanant, and A. von Matern

(2009), Reconstructions of the coastal impact of the 2004 Indian Ocean tsunami in the Khao Lak area, Thailand, J. Geophys. Res., 114,

C10023, doi:10.1029/2009JC005516.

1. Introduction

[2] The 26 December 2004 Indian Ocean Tsunami causedenormous loss of life and major infrastructural damage. Itwas sourced along the Andaman-Sunda trench off the westcoast of Sumatra, where the Indo-Australian plate subductsbeneath the Burma and Sunda subplates (Figure 1) [Lay et al.,2005]. The event started at 0058:53 UTC with a magnitude9.1 earthquake at 3.32�N and 95.85�E, and propagatednorthward for�1300 km parallel to the trench with a velocityof 2.5 km/s [Ammon et al., 2005; Lay et al., 2005]. Thiscaused the seabed to be uplifted by several meters, whichdisplaced 30 km3 of water and triggered a series of tsunami

waves that propagated across the Indian Ocean with avelocity of 500–950 km/h [Kawata et al., 2005; Grilli etal., 2007]. The approximately north–south orientation of thedisplaced segment of the fault plane caused stronger tsunamiwaves to propagate westward toward Sri Lanka and India andeastward toward Indonesia and Thailand [Kawata et al.,2005]. Tidal gauge measurements and satellite images haveshown that the first westward-propagating wave was anelevation wave whereas the first eastward-propagating wavewas a depression wave, and that the wave period, whichvaried with location, was typically 25–40 min [Nagarajan etal., 2006; Merrifield et al., 2005; Rabinovich and Thomson,2007].[3] High-resolution satellite imagery (IKONOS, 1 m) and

posttsunami surveys of, e.g., water marks, sea sand or mudmarks on buildings, structural damage, objects trapped intrees, boats carried onshore and soil erosion, have shown amaximum tsunami inundation of 4 km and a maximumtsunami runup of 39 m in Banda Aceh, Sumatra [Kawata etal., 2005; Choi et al., 2006; Tsuji et al., 2006; Rossetto et al.,2007; Gillespie et al., 2007; Matsutomi et al., 2008; Satakeet al., 2008; Earthquake Engineering Field InvestigationTeam (EEFIT), The Indian Ocean tsunami of 26 December2004: Mission findings in Sri Lanka and Thailand, online

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 114, C10023, doi:10.1029/2009JC005516, 2009

1Department of Geology and Geochemistry, Stockholm University,Stockholm, Sweden.

2Department of Physical Geography and Quaternary Geology,Stockholm University, Stockholm, Sweden.

3Geosciences Azur, Observatoire Oceanologique, Villefranche-sur-Mer,France.

4Faculty of Resources and Environment, Kasetsart University, Si Racha,Thailand.

5Advanced Virtual and Intelligent Computing Research Center, Facultyof Science, Chulalongkorn University, Bangkok, Thailand.

Copyright 2009 by the American Geophysical Union.0148-0227/09/2009JC005516

C10023 1 of 14

report, Institution of Structural Engineers, London, 2006,http://www.eefit.org.uk].[4] The tsunami caused up to 300,000 fatalities in over

12 countries [Kawata et al., 2005; Emergency EventsDatabase (EM-DAT), Universite Catholique de Louvain,http://www.emdat.be, last accessed 17 January 2008;EEFIT, online report, 2006]. Bandah Aceh, Sumatra, wasthe worst affected area with over 230,000 people listed asdead or missing. Thailand was the second most severelyaffected coastal region. The number of fatalities reported inThailand varies, but according to the EM-DAT, the numberof people reported dead or missing was 8345. Of these, 2448were foreigners (EM-DAT, http://www.emdat.be, 2008).[5] Mitigating similar consequences of future tsunamis

requires a better understanding of tsunami propagation, coastalimpact and inundation. Numerical simulations have becomean important tool for achieving this goal. These simulations arenow an essential component of efforts aimed at targetingdisaster mitigation during the short, yet critical, time intervalwhen nomore than the approximate location andmagnitude ofa potentially tsunamigenic earthquake is known.[6] Ideally, numerical simulations can be used to forecast

and provide an estimate of the arrival time, runup height,coastal impact and inundation of a tsunami at a given location.These simulations are complex and based largely on earth-quake parameters, the spatial distribution of initial seafloormovements and seafloor morphology. With accurate determi-nation of the earthquake source parameters, simulation oftsunami generation and propagation has been shown to bereliable [e.g., Ioualalen et al., 2007]. However, simulation oftsunami runup height, coastal impact and inundation haveproven more challenging, requiring a combination of high-resolution bathymetric and topographic data and consideration

of non linear wave interactions, beach slope, wave breaking,energy dissipation and other effects which are specific to agiven coastal region [Segur, 2007]. Testing these simulations isessential and can be achieved by comparison with reconstruc-tions of the time at, and direction from, which a tsunamiarrived, and its maximum wave height, runup and inundation.These data can be acquired from tidal gauge measurements,satellite imagery, posttsunami surveys and eyewitness reports.[7] Unique for the 26 December 2004 tsunami is the large

number of detailed eyewitness reports which have beencompiled, not only by scientific and other official organiza-tions [e.g., Kawata et al., 2005; EM-DAT, http://www.emdat.be, 2008; EEFIT, online report, 2006], but also bymembers of the public usually for the purpose of providing aplatform for communication among survivors (e.g., www.radarheinrich.de and www.visomfinns.se). In this respect,Khao Lak in southwestern Thailand is one of the most welldocumented areas. Eyewitness reports from the Khao Lakarea include personal video recordings of the tsunami,photos of the tsunami and the damage it caused, and detaileddescriptions written by survivors of their experiences. Thistype of data has occasionally been used to support measure-ment data of, e.g., runup height and inundation [Tsuji et al.,2006; Choi et al., 2006; Rossetto et al., 2007; Matsutomi etal., 2008, Satake et al., 2008], but has in many casesremained unpublished. The aim of this study is thereforeto build reconstructions of the coastal impact of the 2004tsunami in the Khao Lak area on the basis of (1) eyewitnessreports alone and (2) eyewitness reports combined with videorecordings of the tsunami, photos of the tsunami and thedamage it caused, field measurements made together witheyewitnesses, and satellite imagery. We will then comparethese reconstructions with a location-specific numerical sim-ulation of the tsunamis propagation and its coastal impact.

2. Study Area

[8] The 2004 tsunami reached the southwest coast of Thai-land at the southern tip of Phuket Island, ca. 500 km from theepicenter, approximately 1 hour 40 min after the earthquakeoccurred [Grilli et al., 2007]. The first wave to arrive was adepression wave, which caused the sea to withdraw by 500–1000 m even though the tsunami occurred during high tide[Kawata et al., 2005]. This phenomenon is a key sign of atsunami’s arrival, but was unfortunately not recognized inmany places. There are eight tidal gauge stations on the coastof the Andaman Sea in Thailand, from Tummarang in thesouth (Satun province) to Ranong in the north (Ranongprovince). The Thai Marine Department operates five ofthese tidal gauge stations, and the others are operated by theHydrographic Department of Royal Thai Navy. Seven of thestations were in operation and recorded the sea level changethroughout the event [Tsuji et al., 2006]. The records show alowering of sea level with duration of 30–60 min before thearrival of the first positive wave. However, it should be notedthat the accuracy of time resolution of the tide gauge data fromThailand regarding tsunami arrival times is somewhat unreli-able as most gauges were analog devices and not designed tomeasure tsunami waves [Tsuji et al., 2006].[9] Several studies report measurements of the height of the

tsunami along the coast of southwestern Thailand. Thesemeasurements include (1) ‘‘tsunami height’’ which is the

Figure 1. Tectonic setting and location of the epicenter ofthe M 9.1 earthquake which caused the tsunami. Earthquakedata was obtained from the database of the U.S. GeologicalSurvey National Earthquake Information Center (http://earthquake.usgs.gov).

C10023 MARD KARLSSON ET AL.: RECONSTRUCTING THE 2004 TSUNAMI

2 of 14

C10023

height of thewave at impact with the shoreline, (2) ‘‘maximumwater level’’ which is the difference in elevation between thehighest watermark and the shoreline, and (3) ‘‘maximumrunup’’ which is the difference in elevation between themaximum inundation and the shoreline [IntergovernmentalOceanographic Commission (IOC), 1998]. These measure-ments are reported inmeters above sea level (masl).Maximumrunup, measured along the coast of southwestern Thailand,mostly ranged between 2 masl on the eastern side of PhuketIsland and up to 6 masl on the western side of Phuket Islandand on Phi Phi Island. The Khao Lak area experienced thehighest runups in Thailand. These mostly ranged from 6 to10 masl, but runups of 11–19.5 masl, corresponding tomaximum inundation up to 2 km, were measured in severalplaces [Kawata et al., 2005; Tsuji et al., 2006; Choi et al.,2006; Rossetto et al., 2007; Matsutomi et al., 2008; Satake etal., 2008; EEFIT, online report, 2006]. Only in Banda Aceh,Sumatra, were higher runups (5–39 masl), corresponding tomaximum inundation of up to 4 km, reported [Kawata et al.,2005;Grilli et al., 2007]. In this study, wewill report measure-ments of tsunami height. It is important to note that these arenot directly comparable with maximum runup.[10] Khao Lak, together with Phuket and Phi Phi Island,

are all popular tourist areas and the tsunami occurred duringtheir peak season. This is reflected in the high number offatalities among foreigners. Damage to buildings in theseareas was mainly caused by the wave impact, but largeobjects such as boats and cars, carried by the tsunami alsocontributed to this damage. Damaged or destroyed structureswere mainly 1–3 storey buildings of reinforced concrete andwooden or bamboo frame bungalows. Phuket and Krabi had17% of their tourist facilities damaged, whereas Khao Laklost 80% of its hotel capacity, making it the most devastatedarea in Thailand. The Khao Lak area had the highest numberof casualties and missing people in Thailand (�70%) andmortality rate among tourists (�50% [Rossetto et al., 2007;EEFIT, online report, 2006]). The Phang Nga province,where Khao Lak is situated, is reported to have lost 25%of its inhabitants because of the tsunami with almost 50fishing villages being severely damaged [Kawata et al.,2005; Rossetto et al., 2007]. There are several contributoryfactors that caused the widespread coastal destruction in theKhao Lak area. Because the area is popular for tourism it hadhotels and resorts densely positioned close to the shore.Gently sloping nearshore bathymetry may have been part ofthe reason for the tsunami waves being more devastating inKhao Lak than in other parts of Thailand. For example,Kawata et al. [2005] estimated that the wave velocity atimpact in Khao Lak was 21.6–28.8 km/h in comparisonwith 10.8–14.4 km/h in Phuket [Kawata et al., 2005;Rossetto et al., 2007; Warnitchai, 2005]. Local topographyand vegetation may also have influenced the amount ofdevastation caused by the tsunami. In some regions wherepopulated areas were separated from the shore by vegetatedland, some energy dissipation was reported [Kawata et al.,2005], but because many resorts in the Khao Lak area weredirectly exposed to the sea, such protection was negligible.[11] A numerical simulation of tsunami propagation and

coastal impact was constructed for the Khao Lak area, fromthe simulation by Ioualalen et al. [2007], which spans thecoast of southwestern Thailand. A complete description ofthis simulation is beyond the scope of this paper and the

interested reader is referred to Grilli et al. [2007]. In brief,the Boussinesq water wave model, FUNWAVE, which isdescribed by Kirby [2003], was used to build the simulation.This model allows for bottom dissipation and wave break-ing and thus avoids artificial amplification at the coast.The model accommodates land inundation by incorporatinga moving shoreline algorithm. FUNWAVE has been cali-brated in several regional landslide tsunami case studies[e.g., Day et al., 2005; Ioualalen et al., 2006; Watts et al.,2003; Waythomas and Watts, 2003].[12] Grilli et al. [2007] performed tsunami simulations in

the Indian Ocean using FUNWAVE with a five-segmenttsunami source using a 10 � 10 model grid for the entireBay of Bengal. The model successfully predicted largestrunups near Banda Aceh, western Thailand and along theeastern sides of India and Sri Lanka. Predicted runup values,including the extreme values for Banda Aceh and Khao Lak,show good correlation with observations made in posttsu-nami field surveys. Using FUNWAVE with a five-segmenttsunami source obtained iteratively by Grilli et al. [2007],Ioualalen et al. [2007] performed tsunami simulations usinga finer 0.250 � 0.250 grid to study coastal impact along theAndaman coast of Thailand. Figure 2 shows the propagationof tsunami waves near the Khao Lak coastline computedfrom this simulation. Figure 2 shows the tsunami wavesapproaching the southwestern coast of Thailand, betweenTakua Pa (in the north) and Thai Muang (in the south), withthe Khao Lak area in the center. The sequence of still imagesillustrates the tsunami’s approach at 5-min intervals, startingat 1010 local time (LT), which is 130 min after the earth-quake. The simulation predicts that the wavefront (which ishighlighted for clarity) approached the Khao Lak area with adirection of 080 ± 2� and an approximate velocity of 32.5 ±1.8 km/h. It further predicts that the sea was receding in KhaoLak when the wavefront had reached the Laem Ao Khampeninsula at 1020 LT and that the wavefront reached KhaoLak at 1030 LT. Free surface elevation was computed fromthe Boussinesq equations implemented in FUNWAVE [seeIoualalen et al., 2007] for the tsunami’s arrival at Bang NiangBeach. This is shown as a simulated time series of waterlevel, �500 m offshore with a time step of 0.5 s (Figure 3).This indicates clearly that the sea started to recede at�0950 LT with minimum water level at �1015 LT. Thecalculated arrival time of the first positive wave at BangNiang Beach is �1030 LTwith maximum tsunami height of�8.2 masl at 1040 LT. The calculated arrival time of the nextmajor positive wave is �1110 LT with a maximum tsunamiheight of 4.2 masl at �1130 LT. The predicted wave periodwas 40–60 min but the simulation predicts several parasiticwaves with shorter periods. Model tsunami heights wereestimated for the Khao Lak coastal area using the location-specific simulation presented by N. Pophet et al. (Newapproach on higher grid resolution for tsunami simulationusing parallelized fully nonlinear Boussinesq equations,submitted to Computers and Fluids, 2009) with a 0.1250 �0.1250 grid. Input data for this simulation are wave eleva-tions, velocities and their gradients, computed from thestudy by Ioualalen et al. [2007]. Detailed description ofthis technique can be found in the auxiliary material.1 The

1Auxiliary materials are available in the HTML. doi:10.1029/2009JC005516.


3 of 14

C10023

model tsunami heights are included in Figure 7b. In thisstudy, we will build a reconstruction of the impact of thetsunami in the Khao Lak area, which we will then comparewith the range of parameters that are predicted by thesesimulations. These include tsunami arrival time, tsunamiheight, waveform, wave velocity and wave off- and onshoretravel directions.

3. Methods

[13] The reconstruction of the coastal impact of the 2004tsunami in the Khao Lak area presented in this study isbased on eyewitness reports, photographs of the tsunamiand the damage it caused, video recordings of the tsunami,satellite imagery and field measurements made together

with eyewitnesses. These data were compiled in a GISdatabase to build the reconstruction.

3.1. Eyewitness Reports

[14] Eyewitnesses from Sweden were invited to partici-pate in a 1-day workshop, which was held in Stockholm inMay 2007. These persons were contacted via an Internet-based forum, which was established by survivors as aplatform for communication. Participants in the workshopwere provided with the following questionnaire for consid-eration prior to this meeting: (1) How many waves did youexperience? (2) How high were these waves (please com-pare with an object, e.g., a tree or a building)? (3) Howstrong were these waves (please list items which the wavewas carrying)? (4) Did you observe differences in strength

Figure 2. Numerical simulation of the tsunami approaching the Khao Lak area. This simulation wasextracted from the numerical simulation published by Ioualalen et al. [2007]. Still images are shown at(a) 1010, (b) 1015, (c) 1020, (d) 1025, and (e) 1030 LT (in Thailand).

Figure 3. Simulated free surface elevation �505 m offshore of Bang Niang Beach (circle in Figure 2).


4 of 14

C10023

and size of the waves? (5) How long time was it betweeneach wave? (6) Where were you when the first wavereached the beach? (7) If you were carried by the wave,please mark your path on the map (see auxiliary material)and provide any extra information. They were also providedwith a map showing the Khao Lak area as it was on26 December 2004 (see auxiliary material). This was to

help them identify their whereabouts at the time of thetsunami. At the workshop, each eyewitness gave an accountof his/her experience of the tsunami, guided where possibleby the questions listed above. Eyewitnesses were asked tofocus on their own experiences and avoid being influencedby how others have said that they experienced the tsunami.Notes were taken by two persons and cross-checked after

Figure 4. (a) Location map of the Khao Lak area with (b–d) insets. Locations A–E are GPS referencedpositions of objects seen in videos recordings of the tsunami (Figure 10). The location of the MukdaraResort is shown. This is where the photo sequence shown in Figure 8 was taken. The paths by whicheyewitnesses were carried by the tsunami are illustrated. These paths are extrapolated between GPSreferenced start, end, and occasional midpoints. The region protected from the prewave by the Khao HinShao Cape is shown. Maximum inundation based on satellite imagery is also shown.


5 of 14

C10023

the meeting. These were then circulated among the eye-witness participants to ensure their accuracy. These notesare included in the auxiliary material. Guided by theoutcome of this meeting, we contacted eyewitnesses whohad published firsthand accounts of the tsunami on aGerman Internet forum (www.radarheinrich.de) invitingtheir participation in the study. This communication washandled by e-mail. During our field study, we approachedadditional eyewitnesses, living in Khao Lak. In total,30 eyewitnesses who experienced the tsunami in the KhaoLak area participated in the study.

3.2. Analysis of Photo and Video Database

[15] To help build a reconstruction of the tsunami on thebasis of eyewitness data, we have used some of the over10,000 photos and videos taken before, during and after the

tsunami in Khao Lak and other affected areas, which arepublished online at www.radarheinrich.de. These includetwo amateur video recordings that show the tsunami as itapproached and impacted the shoreline at Khao Lak. Thesetwo video recordings were taken from the Sunset ViewpointRestaurant, which overlooks the southern end of SunsetBeach, and from the Garden Beach Resort, which is locatedon Nang Thong Beach (Figure 4). They were used to getestimates of the direction and velocity of the tsunamiwavefront as it approached the Khao Lak shoreline. Thiswas done by timing the engulfment of a sequence ofidentifiable features by the tsunami waves and then GPScoordinating their positions in the field. We used stillimages from these video recordings and photos taken duringthe event to constrain the waveform. We also comparedphotos taken after the tsunami showing the damage itcaused with photos taken during the event to confirm fieldestimates of maximum tsunami height (Figure 5).

3.3. Field Study in the Khao Lak Area

[16] We conducted a field study of the Khao Lak area forthe purposes of (1) gaining a more accurate understandingof the eyewitness reports, (2) estimating maximum tsunamiheights in meters above ground level (magl) and metersabove mean sea level, and (3) GPS coordinating featuresfrom video recordings and the paths along which eyewitnesswere carried by the tsunami wave/s. For part of this fieldstudy, eyewitnesses of the tsunami accompanied us. Thisfield study was carried out during the summer of 2008.[17] Tsunami height measurements were made relative to

the ground level using a clinometer, laser and/or tapemeasure. Because many buildings and resorts have beenrepaired or rebuilt since the event, we used a combination ofeyewitness descriptions and photos of the tsunami and thedamage it caused to relocate the height reached by thetsunami on new/repaired buildings. This was possiblebecause most owners chose to rebuild their resorts exactlyas they had been before the tsunami (e.g., Figure 5).[18] We obtained tsunami heights relative to mean sea

level (masl) from measurements made relative to the groundlevel (magl) by leveling. These measurements were maderelative to the waterline elevation at the time of measure-ment and tide tables were used to convert the elevationsobtained to masl. The leveling surveys were performedalong transects from the waterline to the positions at whichmaximum tsunami heights were measured, using a levelerand GPS for position. Control measurements were carriedout in the opposite direction, and the maximum toleratederror was 0.25 m in accordance to the IOC standards.

3.4. Inundation Based on Satellite Imagery

[19] High-resolution IKONOS satellite images (1 m)acquired 3 days after the tsunami by the Centre for RemoteImaging, Sensing and Processing at the National Universityof Singapore (www.crisp.nus.edu.sg/tsunami/tsunami.html)were used to identify tsunami affected areas and estimatemaximum inundation.

4. Results From Eyewitness Reports

[20] In total, 30 eyewitnesses participated in the study.Their reports encompassed the six main resort areas of

Figure 5. This photo sequence of the Briza Resort, whichis situated on Khao Lak South Beach, shows how tsunamiheights were measured in 2008 from buildings which(c) were rebuilt identical to how they were before thetsunami, by comparison with (a) photos taken during thetsunami and (b) photos of damage caused by the tsunami.(Photos in Figures 5a and 5b courtesy of John Tompson.)


6 of 14

C10023

the Khao Lak area. These are, from north to south;Pakarang Cape, Khuk Khak Beach, Bang Niang Beach,Nang Thong Beach, Sunset Beach and Khao Lak SouthBeach (Figure 4). These reports provided several types ofinformation. These are tsunami arrival time, number ofwaves and wave period, tsunami height and strength, andtransport paths of eyewitnesses who were carried by thewave/s. These results are summarized in Figures 4, 6, and 7and discussed in sections 4.1–4.4.

4.1. Tsunami Arrival Time, Number of Waves,and Wave Period

[21] Eyewitness reports suggest that the water beganreceding at 1000 (LT) and that the main wave impacted at1030 LT (see auxiliary material). The majority of eyewit-nesses experienced one or two waves (Figure 6a). Eye-witnesses experiencing more waves were carried offshore or

toward the sea (see auxiliary material). Most of the eye-witnesses who were located near Nang Thong Beach and afew of the eyewitnesses who were located near Bang NiangBeach described a smaller prewave, which arrived a fewminutes before the main wave. The estimated wave periodwas also bimodal, with peaks at 0–5 and 15–20 min(Figure 6c). The shorter period was estimated by thoseeyewitnesses who were referring to the time between theprewave and the main wave, whereas the longer period wasestimated by those who referred to the time between the mainwave and/or subsequent wave/s (see auxiliary material). Oneeyewitness, who was carried offshore at Nang Thong Beachshortly after 1030 LT, reports being carried back to the shoreat Sunset Beach by the fifth wave at 1205 LT (see auxiliarymaterial). This suggests a slightly longer wave period of20–25 min.

4.2. Tsunami Height and Strength

[22] The majority of eyewitnesses estimated tsunamiheight by comparison with palm trees or buildings. Weobtained numerical estimates by comparing either with thebuildings with which they made these comparisons or withaverage heights of resort buildings and palm trees measuredduring the field survey. This approach is clearly not idealbecause it introduces considerable uncertainty, comparedwith, e.g., posttsunami damage surveys. The wide range ofestimated tsunami heights (5–15 magl) is thus expected.Their distribution is bimodal with peaks at 6 and 10 magl,and a mean value of 7.3 ± 0.8 magl (Figure 6b). The 6 maglpeak corresponds with the approximate height of two-storybuildings in the Khao Lak area and the approximate heightto the base of the crown of typical palm trees. We attributethe 10 magl peak to water spray as the main waves hit palmtrees and buildings at the shoreline and a tendency, whenproviding numerical estimates, to round off to either 5 or10 magl. We were only able to obtain descriptive informa-tion regarding the strength of the waves. Eyewitnessesreported major structural damage to concrete buildingsand the total destruction of wooden buildings. Descriptionsof an extreme pressure at the moment of impact, whichcaused buildings to explode, were reported by many eye-witnesses. They also describe the water carrying largeamounts of heavy debris, such as cars, trees, furniture,pylons and building debris. Eyewitnesses consistentlyreported that the prewave was weaker than the main waveand that the main wave was much stronger than subsequentwaves. They reported severe turbulence as the incomingwaves interfered with the outflow and many eyewitnessesnoted that this outflow was stronger than the inflow, withthe consequence of sucking everything out to the sea.

4.3. Transport Paths

[23] Twenty eyewitnesses who participated in the studywere transported by the wave/s. They were carried totaldistances ranging from a few hundred meters inland in thesouthern part of the Khao Lak area up to approximately1.5 km in the northern part of the area (Figure 4). Thesetransport distances give a minimum estimate of tsunamiinundation, which compares well with satellite imagery(Figure 4). Five eyewitnesses, who were at Nang ThongBeach, were carried offshore (Figure 4). The reconstructedtransport paths shown in Figure 4 are based on start and end

Figure 6. Responses to eyewitness questionnaire shownas histograms of (a) number of waves, (b) tsunami height,and (c) wave period.


7 of 14

C10023

points and in a few cases one or more reference points onroute. Much of each path remains uncertain because eye-witnesses were pushed under water for prolonged periods oftime and were struggling for survival among the debriscarried by the wave and therefore disoriented. In Figure 7a,we have plotted compass directions for these transport paths,calculated from GPS referenced start and end positions, as afunction of latitude. This plot shows two groups of data, withmean transport directions of 88 ± 6� and 62 ± 7�. Performinga two-tailed t test allowing for unequal variance on thesetwo groups gives a p value of 0.000036, which supports theinterpretation that these are discrete populations. The lattergroup is restricted to early observations from the Nang ThongBeach area (Figure 7a, see also auxiliary material). This iswhere most eyewitnesses reported a smaller prewave pre-ceding the main wave. We therefore interpret that the trans-port directions of 88 ± 6� and 62 ± 7� refer to the respectiveonshore compass directions of the main and prewaves. Thisinterpretation is further supported by several eyewitnessreports where a ‘‘change of direction’’ between the prewaveand the main wave is described (see auxiliary material).Linear regression of these data suggests little variation inwave direction as a function of latitude. Estimated compassdirections for transport offshore toward the sea ranged from225� to 337�. We suggest that this greater variability is aconsequence of turbulent flow, particularly as the outflowmet incoming waves, and the obvious difficulty for eye-witnesses to determine their positions out at sea.

4.4. Summary

[24] In summary, on the basis of eyewitness reports,we conclude that the sea began receding at approximately1000 LT and that the tsunami arrived at approximately1030 LT. The main wave was preceded by a smaller andweaker prewave at Nang Thong and Bang Niang Beaches.Most eyewitnesses describe two waves, which may or maynot include the prewave. The time between arrival of theprewave and the main wave was estimated to 0–5 min andthe time between arrival of the main wave and subsequentwaves (where reported) was 15–20 min. Subsequent waveswere described as weaker than the main wave. The compassdirection of the prewave as it inundated the shore was 62 ±7� and the corresponding direction of the main wave was88 ± 6�. The outflow was more turbulent with estimatedcompass directions ranging from 225 to 337�. In section 5,we will supplement this reconstruction, which is basedsolely on eyewitness reports with additional data compiledfrom video recordings and photos of the tsunami, photos ofdamage caused by the tsunami, field measurements, andsatellite imagery.

5. Reconstruction

[25] In this section we obtain estimates of tsunami height,wave period and form, wave velocity and direction, andinundation. These estimates are based on eyewitnessreports, video recordings and photos of the tsunami, photos

Figure 7. Plots showing (a) compass directions in which eyewitnesses were carried by the tsunami, (b)tsunami heights estimated from damage to buildings (open diamonds) and obtained from the location-specific numerical simulation from the 0.1250 grid model (solid line), and (c) a location map for thesemeasurements.


8 of 14

C10023

of damage caused by the tsunami, field measurements, andsatellite imagery.

5.1. Tsunami Height

[26] Eyewitnesses reports of tsunami height relative to theground level yielded a mean value of 7.3 ± 0.8 m and amode of approximately 6 m. However, estimates rangedfrom 5 to 15 magl. This variability could be due to theeyewitnesses’ location in relation to the shoreline and/orbreaking effects as the wavefront hit palm trees and build-ings. In this section, we report tsunami heights estimated ina field survey conducted in 2008. Estimates were maderelative to both ground level and mean sea level. Becauseover 3 years had elapsed since the tsunami hit Khao Lak andbecause extensive rebuilding work has been completed, we

compared photos of the tsunami and the damage it causedwith buildings which had survived the tsunami and newbuildings, which had been rebuilt at the same location andidentical to those which had been damaged or destroyed(e.g., Figure 5). Comparison between photos of the tsunamiwith photos showing the damage it caused allowed us toconfirm that observed structural damage correlated withtsunami height and not breaking effects as it hit thesebuildings.[27] In total 41 measurements of tsunami height were

made close to the shoreline of the Khao Lak area duringfield surveys in 2008 (Figures 7b and 7c). These gave amean tsunami height relative to ground level of 4.9 ±0.6 magl. Combining these measurements of tsunami heightrelative to ground level with ground elevation estimates

Figure 8. Sequence of timed photos taken from the raised courtyard of the Mukdara Beach Resort.Photos were taken at camera times of (a) 1029, (b) 1029, (c) 1032, (d) 1036, (e) 1037, (f) 1037, (g) 1059,and (h) 1110 LT. The accuracy of the camera timer is unknown. However, comparison with eyewitnessreports and metadata obtained from photos taken from Khao Lak South Beach and Nang Thong Beachsuggests that the camera timer was up to 2 min slow. (Photos reproduced from the forum atwww.radarheinrich.se with permission from Heinrich Grosskopf; credit for photos goes to an anonymousphotographer.)


9 of 14

C10023

obtained by leveling gave a mean tsunami height (relative tomean sea level) of 8.0 ± 0.6 masl. Figure 7b indicates thatmaximum tsunami heights were experienced at Nang ThongBeach. Here, tsunami heights up to 11.5 ± 0.2 masl weremeasured.[28] The mean tsunami height relative to ground level of

4.9 ± 0.6 magl estimated in the field was lower than themean of 7.3 ± 0.6 magl and mode of 6 magl estimated byeyewitnesses. This could relate to the fact that estimatesbased on structural damage are minimum estimates oftsunami height. This is because its maximum height ifsustained only a short period of time might not have causedmeasurable damage. There might also be a tendency foreyewitnesses to overestimate tsunami height, particularlywhere numerical values rather than comparisons with otherobjects were given.

5.2. Number of Waves, Wave Period, and Wave Form

[29] Most eyewitnesses who were carried onshore by thetsunami experienced one or two waves, whereas eyewit-nesses who were carried offshore experienced more waves.Both groups estimated wave period of 0–5 min for the timeinterval between the prewave and the main wave and 15–20 min for the time interval between the main wave andsubsequent wave/s. One eyewitness who was carried off-shore recorded the time when he was brought back onshoreby the fifth wave. This gave an estimated wave period of20–25 min. Most eyewitnesses describe the extreme forceof the main wave at the moment of impact. This isconsistent with a steep wavefront [Didenkulova et al.,2007]. In this section, we use a sequence of timed photosof the tsunami (Figure 8) to constrain the wave period andthe form of the main wave. This photo sequence was takenfrom the elevated courtyard of the Mukdara Beach Resort,which is situated near Bang Niang Beach (Figure 4). It wasused to build a reconstruction of the wave form (Figure 9).Water levels seen on this photo sequence were calculated byGPS leveling in the field. Times given in this section are‘‘camera times.’’ Comparison with eyewitness reports andmetadata for photos taken from Khao Lak South Beach andNang Thong Beach suggest that the camera timer may have

been up to 2 min slow. The photos taken at the camera timeof 1029 LT show the water level at 3.8 ± 0.2 masl and 4.1 ±0.2 masl (Figures 8a and 8b). The next photo which wastaken approximately 3 min later at camera time 1032 LTshows that the water level had increased by 2.5 m to 6.6 ±0.2 masl (Figure 8c). This confirms the steepness of thewavefront (Figure 9). The next photo which was takenapproximately 4 min later at camera time 1036 LT showsthat the water level had increased by an additional 0.2 m to6.8 ± 0.2 masl. (Figure 8d). The maximum recorded waterlevel of 6.9 ± 0.2 masl was recorded after �1 min at cameratime 1037 LT (Figures 8e and 9). Then, in less than 1 min,the water level dropped by 0.4 m to 6.5 ± 0.2 masl at cameratime 1037 LT (Figure 8f). Photos taken at camera times1059 and 1110 LT of the pool area, which is at a lower level,show that the water level had decreased by 2.8 and 0.2 m to3.7 ± 0.2 masl and 3.5 ± 0.2 masl, respectively (Figures 8gand 8h). The photos taken at 1059 and 1110 LT show strongturbulence near the shoreline. This could be caused byoutgoing water from the main wave meeting the next waveas it approached the shoreline. This suggests a minimumwave period of 30–40 min (Figure 9). This is consistentwith estimates acquired from tide gauge records elsewherein the Indian Ocean [Nagarajan et al., 2006; Merrifield etal., 2005]. The inferred wave period is however consider-ably longer than the 15–20 min suggested by the eyewit-ness reports (Figure 6c). We suggest that the reason for thisdiscrepancy is that eyewitnesses not only experience themain wavefronts, but also those of parasitic waves. Even ifthe amplitudes of these parasitic waves were small incomparison to the main waves (e.g., 2–3 m), eyewitnesses,who were struggling for survival among floating debris inturbulent water, would very likely experience these asseparate waves.[30] Finally, the prewave, which was described by eye-

witnesses at Nang Thong Beach, can be seen ahead of themain wave on the videos taken from Sunset ViewpointRestaurant and Garden Beach Resort (Figures 10a and 10b).This is consistent with the reported transport directions forthe prewave and main wave. Interference between theprewave and the main wave might further explain the

Figure 9. The wave form, shown as a tsunami height versus time plot and estimated from the timedphoto sequence taken at the Mukdara Beach Resort (Figure 8).


10 of 14

C10023

greater wave heights experienced in Nang Thong and BangNiang Beaches.

5.3. Wave Velocity and Direction

[31] Wave velocity was not estimated from eyewitnessreports. The onshore directions of the prewave and mainwave were estimated at 62 ± 7� and 88 ± 6�, respectively. Inthis section, we use video recordings of the tsunami tocalculate the velocity and direction of the wavefront as itapproached and impacted with the shoreline of the KhaoLak area. We used two amateur videos taken from SunsetViewpoint Restaurant and Garden Beach Resort (near A andat E, Figure 4). Both videos were recorded with 25 framesper second. By analyzing these frames individually we wereable to estimate the time on the video at which a set ofidentifiable features were hit or engulfed by the tsunamiwaves. During the field survey, we GPS located the posi-tions of these features. We were then able to calculatevelocity vectors for the advancement of the wavefront usingthe expression

Vxy ¼zxy

txy; ð1Þ

where Vxy is the average velocity vector for the wavefronttraveling from location x to location y, zxy is the distancefrom location x to y and txy is the time taken for thewavefront to travel from location x to y. We were able toobtain three velocity vectors on the basis of timing thearrival of the wavefront at five positions. These were rocksA, B and C seen on the video taken from Sunset ViewpointRestaurant (Figures 4 and 10c–10e), engulfment of PoliceBoat 813 (D, Figure 4), seen on both videos and located bytriangulation from compass directions estimated in the field,and the impact of the main wave with the restaurantbuilding at the Garden Beach Resort (E, Figure 4) which isheard on the video taken from Garden Beach Resort andconfirmed by several eyewitnesses. The locations of rocksA, B and C are shown in Figure 4. These were engulfedby the main wave at video times 0644:96, 0710:96, and0748:24. This gives tAB = 26.00 ± 0.04 s and tBC =37.28 ± 0.04 s. Distances zAB and zBC obtained from thepositions of rocks A, B and C are 387 ± 14 m and 1113 ±14 m, respectively. For these values, we obtained thevelocity vectors VAB = 53.6 ± 2.0 km/h and VBC = 107.5 ±1.5 km/h, from equation (1). The position at which PoliceBoat 813 was engulfed by the main wave (D, Figure 4)was estimated by triangulation as follows. In the video taken

Figure 10. Video frames showing (a) the prewave and main wave from Sunset Viewpoint Restaurantand (b) the prewave and main wave from Garden Beach Resort, (c–e) rocks A–C from Sunset ViewpointRestaurant, (f) Police Boat 813 from Sunset Viewpoint Restaurant, and (g) Police Boat 813 from GardenBeach Resort. (Video frames reproduced from the forum at www.radarheinrich.se with permission fromHeinrich Grosskopf; credit for video frames goes to an anonymous videographer.)


11 of 14

C10023

from Sunset Viewpoint Restaurant, Police Boat 813 is seenin front of a wooded area, west of a distinct break in the treeson Pakarang Cape approximately 5 min before it was hit bythe tsunami (Figure 10f). The compass direction of thiswooded area (and therefore the police boat) from SunsetViewpoint Restaurant, estimated during the field survey, is352 ± 2�. It can be seen in the video taken from the GardenBeach Resort that Police Boat 813 was moving parallel tothe shore (and therefore directly toward Sunset ViewpointRestaurant) before it was hit by the main wave. Thus itscompass direction from Sunset Viewpoint Restaurant whenit was hit by the main wave five minutes later was probablyunchanged. In the video taken from Garden Beach Resort,Police Boat 813 is seen behind the rocks identified inFigure 10g at the moment it is hit by the main wave. Thecompass direction of these rocks from the position in front ofthe Garden Beach Resort from which eyewitnesses haveconfirmed that the video was taken is 300 ± 2�. Theapproximate location of the police boat (D, Figure 4) wascalculated from these compass directions and the positionsfrom which the videos were taken by applying the sphericallaw of cosines. The police boat was hit by the main wave atvideo time 0333:84 and the sound of the wave hitting therestaurant building of Garden Beach Resort is heard at videotime 0509:36. This gives tED = 95.52 ± 0.04 s. This isconsistent with timed photos taken from near the MukdaraBeach Resort (not shown) which show the tsunami engulfingthe police boat at a camera time of 1024 LT and reaching theGarden Beach Resort at a camera time of 1026 LT. Thedistance zDE was obtained from the position of the policeboat when it was hit by the main wave and the formerlocation of the restaurant building at Garden Beach Resort(E, Figure 4). This was 1065 ± 154 m. For these values, weobtained the velocity vector VED = 40.2 ± 5.8 km/h fromequation (1).[32] The average velocity (V) and average direction of the

wavefront for the time interval during which features A–Ewere engulfed by the main wave can be obtained from thethree velocity vectors VAB, VBC, and VDE and their respec-tive compass directions using the expression

V ¼ VAB: cos qAB ¼ VBC: cos qBC ¼ VDE: cos qDE; ð2Þ

where qxy is the angle between the compass direction of thewave normal and the compass direction from location x to y.This expression was satisfied for average velocity, V = 33 ±4 km/h and the compass direction 083 ± 3�. This compassdirection for wave travel offshore was similar to thecompass direction of 088 ± 6� for wave travel onshore,estimated from eyewitness transport paths. These wavedirections are further supported by studying damage to treeson the shoreline near Tap Lamu (south of the main studyarea). Trees along the northern part of this shoreline werepartly or totally damaged by the tsunami, whereas treesalong the southern part of this shoreline were undamaged.We suggest that the Laem Ao Kham peninsula protectedthese trees. This would imply that the compass direction ofthe incoming wave was approximately 080�.

5.4. Inundation

[33] On the basis of eyewitness transport paths, we caninterpret a minimum inundation of a few hundred meters in

the southern part of the Khao Lak area and between 1 and1.5 km in the northern part of the area. These estimatescompare well with the extent of damage seen on high-resolution IKONOS images over Khao Lak area acquired3 days after the event (Figure 4). Local inundation of up to3 km, which is indicated by satellite imagery, was notdetected using eyewitness reports.

5.5. Timeline

[34] We will conclude this section by presenting a recon-structed timeline for the sequence of events that occurred inthe Khao Lak area on the morning of 26 December 2004.5.5.1. At 1000 LT[35] The sea begins receding. This is confirmed by

eyewitness reports. One of these eyewitnesses confirmedthe time because he noticed that the sea was receding at thesame time as the breakfast buffet at his hotel was closing(see auxiliary material). There might be some time delaybetween the time that the sea begins receding and the timethat it is noticed by eyewitnesses. However, we suggest thatthis discrepancy will be small because of the shallowgradient of the seabed at Khao Lak. This estimate is furtherconsistent with measurements made at seven tidal gaugestations along the coast of southwestern Thailand [Tsujiet al., 2006].5.5.2. At 1005–1020 LT[36] The wavefront is visible on the horizon. This is

described as a white stripe on the horizon. It was observedthroughout the Khao Lak area, but it did not appearthreatening. The time range is confirmed by eyewitnessreports and on the video taken from Garden Beach Resort.For the line of sight of a person standing between theshoreline and the resort, which would be from 2 to 6 masl,and ignoring the offshore height of the wave above the seasurface, we estimate that the wavefront would have firstbecome visible when it was 5–9 km offshore. This gives avelocity for the incoming wavefront of 32 ± 20 km/h for awave direction of 083 ± 3�.5.5.3. At 1024–1027 LT[37] The wavefront hits Police Boat 814, which was

located 0.9 ± 0.1 km offshore. This was estimated fromthe time between the wavefront hitting the police boat andthe restaurant, recorded on the video taken from GardenBeach Resort, and the velocity and direction of the wave-front calculated in this study. It assumes that the wavereached the shoreline between 1026 and 1029 LT (seesection 5.5.4).5.5.4. At 1026–1029 LT[38] The wavefront hits the shoreline at Nang Thong and

Bang Niang Beaches. This range is confirmed by eyewit-ness reports and metadata from photos taken from Khao LakSouth Beach, Nang Thong Beach and Bang Niang Beach. Itis consistent with tidal gauge measurements from elsewherealong the Andaman coast of Thailand [Tsuji et al., 2006].The wave direction as it approached the shoreline was 083 ±3� and its velocity was 33 ± 4 km/h. This was calculatedfrom amateur videos of the tsunami taken from SunsetViewpoint Restaurant and Garden Beach Resort. The wavedirection as it traveled up on land was 088 ± 6�. This wasestimated from the start and endpoints of the paths alongwhich eyewitnesses were transported. Estimates of theheight of the main wave are 4.9 ± 0.6 magl (or 8.0 ±


12 of 14

C10023

0.6 masl), on the basis of damage to buildings, and 7.3 ±0.8 magl (mode = 6 magl), on the basis of eyewitnessreports. Eyewitnesses describe an explosive force exertedby the main wave at the moment it impacted with theshoreline (see auxiliary material). This is consistent with thesteepness of the wavefront, which is implicit from thesequence of photos taken from the raised courtyard of theMukdara Beach Resort. Eyewitnesses also described asmaller prewave, which was 1–2 magl in height and arrivedshortly before the main wave at Nang Thong Beach andalong parts of Bang Niang Beach (see auxiliary material).Its compass direction was 062 ± 7�. The prewave wasneither experienced by eyewitnesses nor was it seen onthe video at Sunset Beach. We suggest that this beach wasprotected by the Khao Hin Shao Cape (Figure 4). Eye-witnesses at Khuk Khak Beach and Pakarang Cape did notexperience the prewave. We suggest that it had beenovertaken by the main wave before it reached the northernpart of the area. Eyewitnesses who describe the main waveriding up on top of the prewave support this interpretation.This interference between the prewave and main wave mayexplain the extremity of damage and maximum tsunamiheights experiences between Nang Thong and Bang NiangBeaches.5.5.5. At 1037–1039 LT[39] Maximum water height (6.9 ± 0.2 masl) at the

Mukdara Beach Resort. This was shown by timed photostaken from the raised courtyard of the resort.5.5.6. At 1030–1045 LT[40] Many eyewitnesses were carried a few hundred

meters inland in the southern part of the Khao Lak areaand up to 1.5 km inland in the northern part of the area. Thisgives a minimum estimate for inundation which is consis-tent with measurements based on satellite imagery. Eye-witnesses who were at Nang Thong Beach when thetsunami arrived were transported both onshore and offshore.The video taken from Sunset Viewpoint Restaurant showeda strong current flowing in a southerly direction and parallelto the shoreline in this area. This outflow, which we suggestwas guided by the relative steepness of the hinterland in thesouthern part of the Khao Lak area, probably explains whyso many eyewitnesses were carried offshore in this area. Wefurther suggest that the longer transport distances of eye-witnesses in the northern part of the area relates to its moregently sloping nearshore topography. The steeper hinter-land, which is set back �1.5 km from the shore in this area,coincides with maximum inundation. Where this hinterlandwas breached by the tsunami, water was channeled intodeeply incised valleys. This channeling was experienced bytwo eyewitnesses (see auxiliary material).

5.5.7. At 1110–1112 LT[41] This is the earliest time at which the next main wave

could have arrived at Bang Niang Beach. This is suggestedby timed photos taken from the raised courtyard of theMukdara Beach Resort. This indicates a wave period of atleast 40 min, which is consistent with tidal gauge measure-ments [Tsuji et al., 2006]. The main tsunami waves weredecorated by parasitic waves, with an estimated wave periodof 15–20 min. This is suggested by the shorter wave periodinterpreted from eyewitness reports.5.5.8. At 1205 LT[42] Wave energy is reduced sufficiently so that it is

possible to find protection. This is confirmed by oneeyewitness who reports being carried offshore from NangThong Beach shortly after 1030 LT and carried backonshore at Sunset Beach at 1205 LT. He was able to protecthimself from subsequent waves.

6. Conclusion and Comparison With theNumerical Simulation

[43] Table 1 compares our reconstructions of the tsunamion the basis of (1) eyewitness data and (2) eyewitness datasupported by photos, videos, field measurements, andsatellite imagery, with a numerical simulation by Ioualalenet al. [2007] and a region-specific numerical simulation byPophet et al. (submitted manuscript, 2009).[44] On the basis of this comparison, we conclude the

following:[45] 1. The sea started retreating at 1000 LT in the

reconstruction, which was based solely on eyewitnessreports, whereas the numerical simulation suggests thatthe sea began retreating at 0950 LT (Figure 3). Thisdiscrepancy might be explained by the fact that the seawould need to have retreated some distance before it wouldbe noticed.[46] 2. The time at which the tsunami reached Nang

Thong Beach and Bang Niang Beach estimated by eye-witnesses was �1030 LT. Metadata from photos taken fromKhao Lak South Beach, Nang Thong Beach and BangNiang Beach give time estimates of 1026–1029 LT. Thenumerical simulation gives a time estimate of 1030 LT.[47] 3. The wave direction was 088 ± 6� (onshore) in the

reconstruction which was based solely on eyewitnessreports, 083 ± 3� (offshore) in the reconstruction whichwas based on video recordings of the tsunami, and �080 ±2� (offshore) in the numerical simulation. The small differ-ence between the offshore and onshore estimates mightrelate to refraction of the incoming wavefronts as they hitthe shoreline.

Table 1. Comparison of Tsunami Reconstructions With Numerical Simulation

Tsunami Parameters Reconstruction 1 Reconstruction 2 Simulation

Sea starts receding �1000 LT 0950 LTWave arrival time �1030 LT 1026–1029 LT 1030 LTWave direction 088 ± 6� (onshore) 083 ± 3� (offshore) 080 ± 2� (offshore)Wave velocity N/A 33 ± 4 km/h 32.5 ± 1.8 km/hWave period 15–20 min 40 min 40–60 minNumber of waves �2 �2 �2Maximum tsunami height 7.3 ± 0.8 magl 4.9 ± 0.6 maglMaximum tsunami height 8.0 ± 0.6 masl 7.97 ± 0.6 maslInundation (southern part of Khao Lak area) �0.5 km �0.5 kmInundation (northern part of Khao Lak area) �1.5 km �1.5 km


13 of 14

C10023

[48] 4. The wave velocity was 33 ± 4 km/h in thereconstruction, which was based on video recordings ofthe tsunami. This is similar to the velocity of �32.5 ±1.8 km/h, which is predicted by the numerical simulation.[49] 5. In the reconstruction, based on eyewitness reports,

we conclude that the Khao Lak area was affected by at leasttwo major waves. The first and strongest of these waves waspreceded by a weaker and smaller prewave at Nang ThongBeach and parts of Bang Niang Beach. This prewave can beseen in the videos taken from Sunset Viewpoint Restaurantand Garden Beach Resort (Figures 10a and 10b). Two majorwaves, smaller parasitic waves and the prewave can be seenon the simulated wave elevation plot for Bang Niang whichwas produced from the numerical simulation (Figure 3). Onthe basis of this simulation, we suggest that the prewavewas formed by reflection of the main wave off the Laem AoKham peninsula (Figure 2).[50] 6. The maximum tsunami height estimated by eye-

witnesses of 7.3 ± 0.8 magl is slightly higher than estimatesbased on field measurement of damage to buildings and thenumerical simulation. This could relate to a tendency forobservers to overestimate heights. There is also a potentialconfusion between tsunami height relative to ground leveland tsunami height relative to mean sea level and tsunamirunup [see IOC, 1998], which requires caution, particularlywhen comparing between eyewitness reports, field measure-ments and numerical simulations.[51] 7. Inundation in both reconstructions and the numer-

ical simulation are similar in both southern and northernparts of the Khao Lak area.[52] In general, we conclude that the similarity of the

reconstructions based on just 30 eyewitness reports and thereconstruction based on video recordings of the tsunami,photos of the tsunami and the damage it caused, fieldmeasurements and satellite imagery, emphasizes the valueof eyewitness reports in understanding complex, life-threatening events, such as tsunamis. That the region-specific numerical simulation yields such similar valueslends strong support to its validity.

[53] Acknowledgments. The Granholm Foundation is acknowledgedfor financial support. We thank Rolf and Carina Karlander and the othereyewitnesses who shared their experiences and contributed to the study. Wethank Ylva Sjoberg for assistance in the field, Andrew Mercer for help withfieldwork preparations, and Heinrich Grosskopf for access to the photo-graphic material and video footage used in this study. Des Barton and twoanonymous reviewers are thanked for thorough and constructive reviews.Martin Jakobsson and Benjamin Hell are thanked for helpful discussions.This study is dedicated to Paulina, John, Nina, Asa, and the many otherswho lost their lives in Khao Lak on 26 December 2004.

ReferencesAmmon, C. J., et al. (2005), Rupture process of the 2004 Sumatra-Anda-man earthquake, Science, 308, 1133–1139.

Choi, B. H., S. J. Hong, and E. Pelinovsky (2006), Distribution of runupheights of the December 26, 2004 tsunami in Indian Ocean, Geophys.Res. Lett., 33, L13601, doi:10.1029/2006GL025867.

Day, S. J., P. Watts, S. T. Grilli, and J. T. Kirby (2005), Mechanical modelsof the 1975 Kalapana, Hawaii earthquake and tsunami, Mar. Geol.,215(1–2), 59–92.

Didenkulova, I., E. Pelinovsky, T. Soomere, and N. Zahibo (2007), Runupof nonlinear assymetric waves on a plane beach, in Tsunami and Non-linear Waves, edited by A. Kundu, pp. 175–190, Springer, Berlin.

Gillespie, T. W., J. Chu, E. Frankenberg, and D. Thomas (2007), Assess-ment and prediction of natural hazards from satellite imagery, Prog. Phys.Geogr., 31(5), 459–470, doi:10.1177/0309133307083296.

Grilli, S. T., M. Ioualalen, J. Asavanant, F. Shi, J. T. Kirby, andP. Watts (2007), Source constraints and model simulation of the Decem-ber 26, 2004 Indian Ocean tsunami, J. Waterw. Port Coastal Ocean Eng.,133(6), 414–428, doi:10.1061/(ASCE)0733-950X(2007)133:6(414).

Ioualalen, M., B. Pelletier, M. Regnier, and P. Watts (2006), Numericalmodeling of the 26 November 1999 Vanuatu tsunami, J. Geophys.Res., 111, C06030, doi:10.1029/2005JC003249.

Ioualalen, M., J. Asavanant, N. Kaewbanjak, S. T. Grilli, J. T. Kirby, andP. Watts (2007), Modeling the 26 December 2004 Indian Ocean tsunami:Case study of impact in Thailand, J. Geophys. Res., 112, C07024,doi:10.1029/2006JC003850.

Intergovernmental Oceanographic Commission (IOC) (1998), Post-tsunamisurvey field guide, 1st edition, IOC Manuals and Guides 37, 62 pp., U. N.Educ. Sci. and Cult. Organ., Paris.

Kawata, T., et al. (2005), Comprehensive analysis of the damage and itsimpact on coastal zones by the 2004 Indian Ocean tsunami disaster,report, Disaster Prev. Res. Inst., Kyoto, Japan. (Available at http://www.tsunami.civil.tohoku.ac.jp/sumatra2004/report.html.)

Kirby, J. T. (2003), Boussinesq models and applications to nearshore wavepropagation, surfzone processes and wave-induced currents, in Advancesin Coastal Modelling, Oceangr. Ser., vol. 67, edited by V. C. Lakhan, pp.1–41, Elsevier, New York.

Lay, T., et al. (2005), The great Sumatra-Andaman earthquake of26 December 2004, Science, 308, 1127–1132, doi:10.1126/science.1112250.

Matsutomi, H., et al. (2008), The December 26, 2004 Sumatra earthquaketsunami, Tsunami field survey around Phuket, Thailand, report, DisasterPrev. Res. Inst., Kyoto, Japan. (Available at http://www.drs.dpri.kyoto-u.ac.jp/sumatra/thailand/phuket_survey_e.html.)

Merrifield, M. A., et al. (2005), Tide gauges observations of the IndianOcean tsunami, December 26, 2004, Geophys. Res. Lett., 32, L09603,doi:10.1029/2005GL022610.

Nagarajan, B., I. Suresh, D. Sundar, R. Sharma, A. K. Lal, S. Neetu, S. S. C.Shenoi, S. R. Shetye, and D. Shankar (2006), The Great Tsunami of26 December 2004: A description based on tide-gauge data from theIndian subcontinent and surrounding areas, Earth Planets Space, 58(2),211–215.

Rabinovich, A. B., and R. E. Thomson (2007), The 26 December 2004Sumatra tsunami: Analysis of tide gauge data from the World Ocean. Part1. Indian Ocean and South Africa, Pure Appl. Geophys., 164, 261–308,doi:10.1007/s00024-006-0164-5.

Rossetto, T., N. Peiris, A. Pomonis, S. M. Wilkinsin, D. Del Re, R. Koo,and S. Gallocher (2007), The Indian Ocean tsunami of December 26,2004: Observations in Sri Lanka and Thailand, Nat. Hazards, 42, 105–124, doi:10.1007/s11069-006-9064-3.

Satake, K., et al. (2008), The December 26, 2004 Sumatra earthquaketsunami, Tsunami field survey around Phuket, Thailand, report, DisasterPrev. Res. Inst., Kyoto, Japan. (Available at http://www.drs.dpri.kyoto-u.ac.jp/sumatra/thailand/phuket_survey_e.html.)

Segur, H. (2007), Waves in shallow water, with emphasis on the tsunami of2004, in Tsunami and Nonlinear Waves, edited by A. Kundu, pp. 3–29,Springer, Berlin.

Tsuji, Y., Y. Namegaya, H.Matsumoto, S. Iwasaki,W. Kanbua,M. Sriwichai,and V. Neesuk (2006), The 2004 Indian Ocean tsunami in Thailand: Sur-veyed runup heights and tide gauge records, Earth Planets Space, 58(2),223–232.

Warnitchai, P. (2005), Lessons learned from the 26 December 2004 tsunamidisaster in Thailand, Scientific forum on the tsunami: Its impact andrecovery, paper presented at Scientific Forum on Tsunami, its Impactand Recovery, Asian Inst. Technol., Thailand, 6–7 June.

Watts, P., S. T. Grilli, J. T. Kirby, G. J. Fryer, and D. R. Tappin (2003),Landslide tsunami case studies using a Boussinesq model and a fullynonlinear tsunami generation model, Nat. Hazards Earth Syst. Sci.,3(5), 391–402.

Waythomas, C. F., and P. Watts (2003), Simulation of tsunami generationby pyroclastic flow at Aniakchak Volcano, Alaska, Geophys. Res. Lett.,30(14), 1751, doi:10.1029/2003GL017220.

��J. Asavanant and N. Pophet, Advanced Virtual and Intelligent Computing

Research Center, Faculty of Science, Chulalongkorn University, Bangkok10330, Thailand.M. Ioualalen, Geosciences Azur, Observatoire Oceanologique, 2 Quai de

la Darse, B.P. 48, F-06230 Villefranche-sur-Mer, France.N. Kaewbanjak, Faculty of Resources and Environment, Kasetsart

University, Si Racha Campus, 199 Moo 6 Sukumvit Road, Si Racha 20230,Thailand.J. Mard Karlsson, M. Sanden, A. Skelton, and A. von Matern,

Department of Geology and Geochemistry, Stockholm University, SE-10691 Stockholm, Sweden. ([email protected])


14 of 14

C10023

Reconstructions of the coastal impact of the 2004 Indian Ocean tsunami in the Khao Lak area, Thailand

Documents