Medieval forewarning of the 2004 Indian Ocean tsunami in Thailand

Vol 455 | 30 October 2008 | doi:10.1038/nature07373 LETTERS Medieval forewarning of the 2004 Indian Ocean tsunami in Thailand Kruawun Jankaew1, Brian F. Atwater2, Yuki Sawai3, Montri Choowong1, Thasinee Charoentitirat1, Maria E. Martin4 & Amy Prendergast5 Recent centuries provide no precedent for the 2004 Indian Ocean tsunami, either on the coasts it devastated or within its source area. The tsunami claimed nearly all of its victims on shores that had gone 200 years or more without a tsunami disaster1. The associated earthquake of magnitude 9.2 defied a Sumatra–Andaman catalogue that contains no nineteenth-century or twentieth-century earthquake larger than magnitude 7.9 (ref. 2). The tsunami and the earthquake together resulted from a fault rupture 1,500 km long that expended centuries’ worth of plate convergence2–5. Here, using sedimentary evidence for tsunamis6, we identify probable precedents for the 2004 tsunami at a grassy beach-ridge plain 125 km north of Phuket. The 2004 tsunami, running 2 km across this plain, coated the ridges and intervening swales with a sheet of sand commonly 5–20 cm thick. The peaty soils of two marshy swales preserve the remains of several earlier sand sheets less than 2,800 years old. If responsible for the youngest of these pre-2004 sand sheets, the most recent full-size predecessor to the 2004 tsunami occurred about 550–700 years ago. The 2004 Indian Ocean tsunami, cresting higher in Thailand than it did anywhere else east of Sumatra (Fig. 1b), rose as much as 20 m above sea level on Phra Thong Island7. The main wave, as observed from a hill on the island’s western shore (H, Fig. 1c), formed a relentless flood that rose stepwise to heights above treetops8. The tsunami ran more than 2 km inland across a Holocene plain composed of grassy beach ridges and intervening tree-lined swales. The flooding at Phra Thong, as on other Thai coastal plains9,10, produced local erosion and widespread deposition. The tsunami reamed out several drainages previously cut across beach ridges as much as 300 m inland (Fig. 1c, e, and Supplementary Fig. 1). In addition, it coated most of the island’s western half with a sheet of sand (Fig. 1d). This sheet is locally lineated with sand streaks that extend inland from the spoil piles of tin miners (Fig. 1e and Supplementary Fig. 1). Its mean particle size ranges from medium to very fine on a transect across the northern part of the island11. Horizontal bedding is common, as is overall upward fining to coarse silt. Building on a previous reconnaissance12, we sought pre-2004 sand sheets at Phra Thong by digging pits and augering holes into ridges and swales at more than 150 sites (Fig. 1d (dots) and Supplementary Table 1). At 20 of these sites we found pre-2004 sand interbedded with the peaty soils of swales that hold standing water most of the year (Fig. 1d, red dots). We found no pre-2004 sand beds in the quartzsand soils of the ridges or in the slightly organic soils of swales that are merely damp. We traced pre-2004 beds across each of two marshy swales near a place where the 2004 tsunami reportedly flowed about 10 m deep (Figs 1e and 2a, and Supplementary Fig. 3a). These swales formed about 2,500 years ago (bark and shell dates; Fig. 2b, c), when the area’s relative sea level was probably within 1–2 m of its present position13. Because beaches have built the island westwards, the more westerly of the swales (X) postdates its neighbour (Y). We assembled stratigraphic cross-sections from correlated pits, from auger borings and from a trench 35 m long, estimated particle size in the field, inferred a preliminary chronology from radiocarbon dating of individual plant remains and shells (Figs 2c and 3, and Supplementary Table 2), and made diatom analyses (Supplementary Figs 4 and 5). Peaty soil in swale X contains two sand sheets (B and C in Fig. 2b, d) that resemble the overlying 2004 deposit. Sheet C, the earlier, is commonly 10 cm thick. Coarse to very coarse sand forms a discontinuous basal layer that fills pre-existing pockets in the underlying soil. The rest of sheet C consists of very fine sand and coarse silt that contains horizontal laminae defined most visibly by leaf fragments (Fig. 3c). The entire sheet formed after 2,200–2,400 sidereal years ago, the age of an isolated piece of bark in the uppermost 1 cm of the underlying buried soil. Leaf fragments within the sheet gave ages that conflict with one another and that exceed the bark age by thousands of years (Fig. 3c). Sheet B, commonly 5 cm thick, typically fines upwards from fine sand to sandy silt. It conformably overlies peaty soil that contains a horizon of bark fragments in its uppermost 1 cm. Three of these fragments, collected separately, yielded ages between 530 6 40 and 570 6 40 radiocarbon years before AD 1950 (14C yr BP; Fig. 3b). If scarcely younger than these fragments, sheet B was deposited about 550–700 sidereal years ago (Supplementary Table 2). Three pre-2004 sand sheets alternate with peaty soil in swale Y (Fig. 2c, e). All three are similar in thickness to the overlying 2004 sand sheet, and all extend preferentially up the swale’s seaward side. All were formed after the swale ceased holding an intertidal flat that is marked by non-abraided molluscan shells 2,500–2,800 sidereal years old (Fig. 2c and Supplementary Table 2). The lowest two sheets, otherwise undated and thus left uncorrelated with swale X, consist mainly of very fine to fine sand. They lack sedimentary structures, probably because of bioturbation that blurs their contacts with the soils beneath (Fig. 2e). The highest pre-2004 sheet (B) retains a sharp base and tabular shape that extend the full length of the trench (Fig. 2f). This sheet typically fines upwards from basal fine or medium sand to parallel-laminated very fine sand that abounds in leaf fragments (Fig. 3a). It probably correlates with sheet B of swale X because each is the youngest pre-2004 sand sheet in its swale and because the leaf fragments in swale Y yielded ages too young for correlation with sheet C (Fig. 3a, c). Although the 2004 sand sheet abounds in brackish and marine diatoms, the earlier sand sheets in swales X and Y lack diatoms of 1 Department of Geology, Faculty of Science, Chulalongkorn University, Phayathai Road, Phatumwan, Bangkok 10330, Thailand. 2US Geological Survey at Department of Earth and Space Sciences, University of Washington, Seattle, Washington 98195-1310, USA. 3Geological Survey of Japan, National Institute of Advanced Industrial Science and Technology, Site C7 1-1-1 Higashi, Tsukuba 305-8567, Japan. 4Department of Earth and Space Sciences, University of Washington, Box 351310, Seattle, Washington 98195-1310, USA. 5Geoscience Australia, GPO Box 378, Canberra, Australian Capital Territory 2601, Australia. 1228 ©2008 Macmillan Publishers Limited. All rights reserved LETTERS NATURE | Vol 455 | 30 October 2008 any kind (Supplementary Figs 4 and 5). Marine and brackish-water diatoms aid in identifying tsunami deposits on temperate shores14. Perhaps their opaline silica valves do not last long in tropical warmth; in experiments, the dissolution of diatoms increases with temperature15. a b 95°E INDIAN OCEAN 100° THAILAND 15°N Andaman Islands 1000 km 1941 1881 Detail in c P Su 5° Me nd a N Tr en ch 0 Su m at ra c 10 20 Highest water of 2004 tsunami (m) d 98° 17′ 50′′ E 7 et Inl Rock Detail in e 10 H m 77 Phra Thong Island 9° 5′ N N Inl et Mangroves K 20 1 km Drainage reamed in 2004 Beach-ridge trend Angular discordance Sand sheets 2004 and earlier 2004 only None Sand streaks High water in 2004 (m) e 98° 15′ 30′′ Sand streak Spoil pile Drainage reamed in 2004 Cross-section in Fig. 2b X Pr of ile 9° 8′ 00′′ in Fi g. 2a ale Sw Y ale Sw e dg Ri N 100 m Cross-section in Fig. 2c Reported flow depth ~10 m at tree in Fig. 2a Preservation is also a problem for the sand sheets themselves. The pre-2004 sheets are distinct and sharply bounded where the swale soil is peaty, blurred by gradational contacts where the soil is just slightly organic, and totally absent in the sandy soils of beach-ridge crests. The 2004 tsunami deposit is already headed towards this fate: in wet swales it has a protective cap of organic matter as much as 5 cm thick, whereas on ridge crests it lacks any cover other than ejecta from burrows that tap the underlying sandy soil. Sheet B, if truly correlative between swales X and Y, initially spanned the intervening beach ridge for a total shore-normal length of no less than 100 m (Fig. 1e). Although sand sheets can record intense storms that drive waves over or through sandy beach berms16, the geographic setting limits Phra Thong’s exposure to such storms. Less than 10u from the Equator, this part of Thailand fringes the belt where the Coriolis minimum limits cyclonic winds17. Scores of twentieth-century cyclones originated in Indian Ocean waters to its west, but all these moved towards India, Bangladesh or Myanmar18 without producing a known storm surge in Thailand19. Tropical cyclones do strike Thailand from its Pacific side. However, such a storm loses strength during its overland crossing to the Indian Ocean (an example is given in Supplementary Fig. 6), and its anticlockwise winds can pile the sea against Thailand’s west coast only in the storm’s trailing-left quadrant. Phra Thong’s setting also disfavours sand-sheet deposition by river or wind. Tidal inlets separate the island from the nearest rivers (Fig. 1c). Aeolian dunes obscure little, if any, of the island’s delicate striping by beach ridges and swales (Fig. 1e and Supplementary Fig. 1). Chronology provides three further reasons to ascribe the pre-2004 sand sheets to tsunamis. First, the middle Holocene ages of the leaf fragments from sheet C (Fig. 3) imply scour into long-buried deposits beneath tidal inlets. The 2004 tsunami showed capacity for such scour by knocking down mangroves along an inner part of the inlet that bounds Phra Thong Island on the south (at K in Fig. 1d). Second, the sand sheets represent infrequent events: the soil between sheets C and B spans 1,500–1,850 years, although it may contain the bioturbated remains of an intervening sheet (Supplementary Fig. 2c); and the interval between sheet B and the 2004 tsunami lasted nearly 550– 700 years (ranges computed from calibrated ages in Supplementary Table 2). These time intervals are in the broad range of deductive estimates for the recurrence of giant earthquakes in the Sumatra– Andaman source region of the 2004 tsunami2–5. Third, sheet B, if little younger than AD 1300–1450 (Supplementary Table 2), may correlate with tsunami and earthquake evidence elsewhere. The youngest widespread pre-2004 sand sheet on a beach-ridge plain at Meulaboh, Sumatra (Fig. 1a, Me) overlies plant detritus dated to AD 1290–1400 (ref. 20). Two coral fragments on a marine terrace in the Andaman Islands gave ages in the range AD 1200–1650 (ref. 21). However, in accounts from Ibn Battuta (journey, AD 1325–1354)22 and the great Ming armadas (voyages, AD 1405–1433)23, we found no written evidence for a sheet-B tsunami on Sumatran and Sri Lankan shores that the 2004 tsunami would overrun. Figure 1 | Setting. a, Northern Sunda Trench and vicinity. Red lines show a modelled fault slip during the 2004 Sumatra–Andaman earthquake5 at 5 (outer contour), 10, 15 and 20 m. Green lines show pre-2004 rupture areas2,26. Bathymetry is shaded at 1-km intervals. Me, Meulaboh; P, Phuket. b, Heights of the 2004 tsunami on the eastern rim of the Indian Ocean7,27–30. The tallest bars for Thailand obscure dozens of height measurements below local maxima. c, Landforms of Phra Thong Island. White areas denote beach ridge plains (examples in e and Supplementary Fig. 1). The lines of angular discordance (repeated for reference in d) probably record shoreline retreat. H, Hornbill Hill8. d, Tsunami deposits on Phra Thong Island. K, knockeddown mangroves along southern inlet. The 2004 tsunami heights are maxima in metres above sea level (ref. 7 and Supplementary Fig. 1). e, Area of stratigraphic evidence shown in Figs 2 and 3. Post-tsunami image, probably taken early in 2005, from PointAsia.com. Trees and shrubs (dark green) delineate some of the swales. 1229 ©2008 Macmillan Publishers Limited. All rights reserved LETTERS WNW Indian Ocean Section in c Projected position of section in b Leaf fragments in sand 670 ± 40 14C yr BP 1,240 ± 40 Laminae 0 Surveyed point b VE ×10 500 m Sheet B, swale X (in pit of Fig. 2d) c b Swale X Inland Swale Y Trench in f 1m Sheet B, swale Y (in pit of Fig. 2e) Eyewitness’s tree 10 cm Elevation (m) Extreme tide range 2 m 10 a ESE Maximum tsunami heights Sand sheet Soil Sand, deposited chiefly on beaches Photo in d Base of sheet B 1 cm a NATURE | Vol 455 | 30 October 2008 Bark Photo in e New peaty soil Sheet A (2004) B Bark Soil Pulled apart Bark shed from Pandanus tree and preserved in soil 530 ± 40 14C yr BP 550 ± 40 570 ± 40 C Shells 2,500–2,800 yr old 20 m 20 m d VE ×20 e Sand sheet B C Stripes on ruler, 10 cm f Inland Sheet B Leaf fragments in sand 3,950 ± 40 14C yr BP 5,680 ± 40 6,150 ± 40 Piece of bark in soil 2,250 ± 40 14C yr BP Figure 3 | Parallel laminae and radiocarbon ages. a, b, Indistinct horizontal laminae in sheet B are highlighted by leaf fragments. The two fragments dated yielded ages that differ from one another by about 600 radiocarbon years (a). The younger of these ages overlaps with ages on detrital tree-trunk bark in the highest part of the underlying soil (b). c, Leaf-fragments from laminae in sheet C yielded widely scattered ages thousands of years greater than that of bark in the underlying soil, all sampled within the same pit (Fig. 2d). A A (2004) B Sheet C, swale X (in pit of Fig. 2d) 10 cm c Bark 2,200–2,400 yr old (Fig. 3c) 10 cm Soil Figure 2 | Cross-sectional shape of sand sheets. a, Topographic profile along the line in Fig. 1e. VE, vertical exaggeration. Maximum tsunami heights are from an eyewitness in the indicated tree (Fig. 1e and Supplementary Fig. 3a) and from a post-tsunami survey 1 km to the southwest (Fig. 1d and Supplementary Fig. 1b). b, c, Cross-sections in swales X (b) and Y (c) from inferred correlation between pits and auger borings (vertical grey lines). Soil peaty in swales (dark brown) and sandy on ridges (light brown). Ages are from Supplementary Table 2. d, e, Sand sheets alternate with dark peaty soils on the walls of pits in swales X (d) and Y (e). f, Lateral continuity of sand sheet B exposed in trench. What tsunami sources might Phra Thong’s pre-2004 sand sheets represent? Too little is known about the sheets’ landward extent on the island, let alone their potential correlates on other Indian Ocean shores, to require full-size predecessors to the 2004 Sumatra– Andaman earthquake. However, the sheets probably required ruptures larger than that of 1881 (Fig. 1a); no sand sheet from the 1881 tsunami, which crested less than 1 m high on Indian tide gauges24, is evident at Phra Thong Island in fibrous peaty soils that the 2004 tsunami failed to incise while covering them with sand (Fig. 2d and Supplementary Fig. 3c–g). The pre-2004 sheets may also require Sunda Trench earthquakes larger than magnitude 8.5 if, as estimated from numerical simulations3, such earthquakes would spawn Thai tsunamis only a few metres high—barely high enough to invade Phra Thong’s beach-ridge plain (Fig. 2a). Sand sheets of Phra Thong Island thus forewarn of infrequent catastrophe. They are already providing public officials and coastal residents with tangible evidence that the 2004 tsunami was not the first of its kind (Supplementary Fig. 7). Still to be determined is whether centuries dependably separate such outsize tsunamis of Sumatra–Andaman source, and whether these recur often enough to dominate Thailand’s probabilistic tsunami hazard. Tsunamis without precedent in written history may threaten Indian Ocean shores that face other parts of the Sunda Trench and the Makran subduction zone24,25. It can be hoped that natural warnings from recent geological history will help avert surprises from these additional tsunami sources. METHODS SUMMARY Landforms in Fig. 1c were traced from 1:50,000-scale aerial photographs taken in 1999 and from post-tsunami satellite images at PointAsia.com and Google Earth. In Fig. 2a the levelling had a closure error of 10 cm, and the tide levels refer to the Khura Buri gauge, 16 km northeast of Phra Thong. Diatom separation and analyses are described in Supplementary Figs 4 and 5. A separate suite of samples from sheet B yielded rare foraminiferal tests, all well preserved, that we do not interpret because of possible laboratory contamination with tests from the 2004 deposit. The radiocarbon age ranges reported in sidereal years span the 95% confidence interval from counting and calibration statistics (Supplementary Table 2). Received 6 March; accepted 27 August 2008. 1. 2. 3. 4. 5. 6. Dominey-Howes, D., Cummins, P. & Burbidge, D. Historic records of teletsunami in the Indian Ocean and insights from numerical modeling. Nat. Hazards 42, 1–17 (2007). Bilham, R., Engdahl, R., Feldl, N. & Satyabala, S. P. Partial and complete rupture of the Indo-Andaman Plate boundary 1847–2004. Seismol. Res. Lett. 76, 299–311 (2005). Lovholt, F. et al. Earthquake related tsunami hazard along the western coast of Thailand. Nat. Hazards Earth Syst. Sci. 6, 979–997 (2006). Stein, S. & Okal, E. A. Ultralong period seismic study of the December 2004 Indian Ocean earthquake and implications for regional tectonics and the subduction process. Bull. Seismol. Soc. Am. 97, S279–S295 (2007). Chlieh, M. et al. Coseismic slip and afterslip of the great Mw 9.15 SumatraAndaman earthquake of 2004. Bull. Seismol. Soc. Am. 97, S152–S173 (2007). Bourgeois, J. in The Sea Vol. 15 (eds Bernard, E. N. & Robinson, A. R.) (Harvard Univ. Press, in the press). 1230 ©2008 Macmillan Publishers Limited. All rights reserved LETTERS NATURE | Vol 455 | 30 October 2008 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. 24. Tsuji, Y. et al. The 2004 Indian tsunami in Thailand; surveyed runup heights and tide gauge records. Earth Planets Space 58, 223–232 (2006). Lyall, K. Out of the Blue: Facing the Tsunami (Australian Broadcasting Corporation, 2006). Hori, K. et al. Horizontal and vertical variation of 2004 Indian tsunami deposits; an example of two transects along the western coast of Thailand. Mar. Geol. 239, 163–172 (2007). Choowong, M. et al. 2004 Indian Ocean tsunami inflow and outflow at Phuket, Thailand. Mar. Geol. 248, 179–192 (2008). Fujino, S. et al. in Tsunamiites—Features and Implications (eds Shiki, T., Tsuji, Y., Minoura, K. & Yamazaki, T.) 123–132 (Elsevier, 2008). Fujino, S. et al. in Proc. Int. Symp. on Restoration Program from Giant Earthquakes and Tsunamis (ed. Kato, T.) 115–121 (Earthquake Research Institute, University of Tokyo, 2008). Horton, B. P. et al. Holocene sea levels and palaeoenvironments, Malay-Thai Peninsula, Southeast Asia. Holocene 15, 1199–1213 (2005). Hemphill-Haley, E. Diatoms as an aid in identifying late-Holocene tsunami deposits. Holocene 6, 439–448 (1996). Kamatani, A. Dissolution rates of silica from diatoms decomposing at various temperatures. Mar. Biol. 68, 91–96 (1982). Donnelly, J. P., Butler, J., Roll, S., Wengren, M. & Webb, T. III. A backbarrier overwash record of intense storms from Brigantine, New Jersey. Mar. Geol. 210, 107–121 (2004). Gray, W. M. in Storms Vol. 1 (eds Pilke, R. J. & Pilke, R. S.) 145–163 (Routledge, 2000). Pant, G. B&. Rupa Kumar, K. Climates of South Asia 320 (John Wiley & Sons, Chichester, England, 1997). Murty, T. S. & Flather, R. A. Impact of storm surges in the Bay of Bengal. J. Coast. Res. 12 (Spec. Iss.), 149–161 (1994). Monecke, K. et al. A 1,000-year sediment record of tsunami recurrence in northern Sumatra. Nature doi:10.1038/nature07374 (this issue). Rajendran, K. et al. Age estimates of coastal terraces in the Andaman and Nicobar Islands and their tectonic implications. Tectonophysics 455, 53–60 (2008). Ibn Battuta. The Travels of Ibn Battūta, A.D. 1325–1354 vol. 4 (transl. Gibb, H. A. R.) (Hakluyt Society, 1994). Dreyer, E. L. Zheng He: China and the oceans in the early Ming dynasty 1405–1433 (Pearson Longman, 2007). Okal, E. A. & Synolakis, C. E. Far-field tsunami hazard from mega-thrust earthquakes in the Indian Ocean. Geophys. J. Int. 172, 995–1015 (2008). 25. Heidarzadeh, M. et al. Historical tsunami in the Makran Subduction Zone off the southern coasts of Iran and Pakistan and results of numerical modeling. Ocean Eng. 35, 774–786 (2008). 26. Ortiz, M. & Bilham, R. Source area and rupture parameters of the 31 December 1881 Mw 5 7.9 Car Nicobar earthquake estimated from tsunamis recorded in the Bay of Bengal. J. Geophys. Res. 108 (B4) 2215, doi:10.1029/2002JB001941 (2003). 27. Satake, K. et al. Tsunami heights and damage along the Myanmar coast from the December 2004 Sumatra–Andaman earthquake. Earth Planets Space 58, 243–252 (2006). 28. Siripong, A. Andaman seacoast of Thailand field survey after the December 2004 Indian Ocean tsunami. Earthq. Spectra 22, 187–202 (2006). 29. Choi, B. H. Analysis and modeling of the distribution functions of runup heights of the December 26, 2004 earthquake tsunami in the Indian Ocean. Æhttp:// wave.skku.ac.kr/tsunami_survey_data/KEERC.report.pdfæ. 30. Hawkes, A. D. et al. Sediments deposited by the 2004 Indian Ocean tsunami along the Malaysia–Thailand Peninsula; Quaternary land–ocean interactions; sea-level change, sediments and tsunami. Mar. Geol. 242, 169–190 (2007). Supplementary Information is linked to the online version of the paper at www.nature.com/nature. Acknowledgements We thank B. Korsakun for logistical help; C. Tongjeen for permission for digging; V. Chutakositkanon, V. Jittanoon, T. Machado, T. Napradit, S. Pailoplee, S. Phantuwongraj, N. Rajeshwara Rao, S. Srinivasalu, P. Surakiatchai and A. Weerahong for contributions to Phra Thong field and laboratory work; Y. Fujii for providing bathymetric data; and S. Bondevik, M. Cisternas, H. Kelsey, A. Meltzner, K. Sieh, M. Tuttle and J. Woodruff for reviews. This report evolved from surveys supported by the Ministry of Natural Resources and Environment (Thailand), the National Science Foundation (USA), and the US Agency for International Development (participants are listed in Supplementary Table 3). Additional funding was provided by the Japan Society for the Promotion of Science (to Y.S.), the Thailand Research Fund (to M.C.), and Chulalongkorn University (through P. Charusiri). Author Contributions All authors participated in the fieldwork, led by K.J. (Supplementary Table 1). B.A. prepared most of the manuscript. Y.S. analysed the diatoms and prepared several of the figures. Author Information Reprints and permissions information is available at www.nature.com/reprints. Correspondence and requests for materials should be addressed to K.J. ([email protected]). 1231 ©2008 Macmillan Publishers Limited. All rights reserved