High-frequency Coastal Overwash Deposits from Phra Thong Island, Thailand

The 26th December 2004 Indian Ocean Tsunami (IOT) emanated from an Mw 9.2 earthquake that generated a 1600 km-long rupture along the Sumatran Megathrust and generated tsunami waves up to 30 m high. The IOT directly impacted the Bay of Bengal and east Africa, with over 283,000 people perishing. At the time, this catastrophic event was considered unprecedented and sparked intense investigations to test this claim. It is now believed that four pre-2004 IOT events have occurred in the last 2500 years, recurring every 550 to 700 years. Much of this information comes from Phra Thong Island, Thailand, where a sequence of four stacked sandsheets separated by organic units has been recognised and compared to the 2004 IOT event. Recently, ground-penetrating radar on Phra Thong Island identified a region that could not be explained by the known stratigraphy. The stratigraphy of the area was investigated from auger cores and pits, and several previously-unrecognised sandsheets were identified and compared to the known tsunami sandsheets. The proximity of the newly-recognised sandsheets to the palaeo-coastline of Phra Thong Island does not preclude the impacts of localised storms in sandsheet emplacement or that tsunamigenic earthquake recurrence may have been more frequent in the past.

The four older sand sheets preserved in Swales X and Y show strong sedimentological similarities (sharp erosional contacts, fining upward unimodal coarse sand to coarse silt, containing leaf or bark fragments) to the 2004 IOT deposit, but lack many of the sedimentary structures, microfauna and microflora 20,21,30 . The palaeo-sand sheets from Swale Y lack preserved diatoms and carbonate microfauna (foraminifera), likely due to temperature-affected dissolution 20,21,30,32 , although it is unclear why the same processes would not have affected the carbonate and unabraided shell material in the intertidal deposits preserved at the base of the swale 21,32 . Sandsheets B and C were correlated between Swales X and Y, but Sandsheet D was recognised in Swale Y only 21,36 . Three of the four palaeo-sand sheets from Swale VI also correlate with Sandsheets B, C and D in Swale Y 30 , reinforcing the spatial extent of the palaeo-sand sheet coverage across the northern beach ridge system on Phra Thong Island 21 .
The buried sand sheets were postulated to have been deposited following tsunami events like the 2004 IOT, which is supported by the similarity of the sedimentology of the older deposits. Furthermore, the trajectory of historical storms that cross the Thai-Malay Peninsular are incapable of generating the storm surge set-up to comprehensively inundate the island 21 . Between 1945 and 1996 nine storms passed within 180 nautical miles of Phuket, none of which generated sufficient set-up to inundate the west coast of Thailand 37 . However, on the 1 st May 2007 a tropical depression (TD2) formed in the northern Gulf of Thailand, crossed the Malay-Thailand Peninsular and moved into the Andaman Sea on the 2 nd May 38 . This storm produced sufficient set-up to overtop the youngest berm (ca. 1.3 m above mean sea level, masl) and deposit two small (~30 × 50 m and ~20 × 15 m) overwash fans (Fig. 1d). The smaller fan is located 490 m northwest of Swale Y and the sediments were deposited on top and behind the 2007 berm. The larger fan occurs 380 m west of Swale Y within the partially recovered 2004 IOT backwash scour. The poorly formed berm in front of the scour was easily breached by the storm and the accommodation space afforded by the recovering tsunami scour allowed further inland penetration and thick storm deposits to form (Fig. 1d). This implies that storms can deposit sediment on Phra Thong Island, but from this solitary example, fan development may be restricted to a few 10 s of metres from the modern shorelineunlike the kilometre-scale inundation and sedimentation caused by the 2004 IOT. The granulometric properties of the sediments collected from the washover fans are statistically indistinguishable from the modern-day beachface sediments and are geochemically similar to Sandsheets A and C, nearshore and beachface deposits suggesting a local and similar source of sediments supplying Sandsheets A and C and the storm deposit 35 .
To determine the recurrence interval of tsunami that have inundated Phra Thong Island and deposited the sand sheets, an intensive dating campaign has been undertaken 21,29,36,39 . In Swale Y, radiocarbon dates of bark, plant fragments and shells from the organic muds above and below the sand sheets were examined to obtain event bounding ages, and leaves collected from within the sand sheets to date the deposits 21 . Bulk radiocarbon dates on the organic muds above and below the sand sheets were also collected 29 . Quartz OSL dates from the Swale X and Y sand sheets and surrounding beach ridges were collected to constrain the timing of the development of the swales and deposition of the sand sheets in Swale Y 36 and Swale VI 30 . These findings all suggest that deposition of the palaeo-sand sheets on Phra Thong Island occurred from inundation events similar to the 2004 IOT event at approximately 550 to 700 year intervals 30,36 . This recurrence interval has been approximated from other studies of sand sheets and archaeological records recovered from beach ridge archives elsewhere from western Thailand 40 , Sumatra 22,41,42 and India 6,7 . Prior to 3800 cal. yr. BP, the recurrence interval determined from palaeotsunami sandsheets from near Aceh spans ca. 600 to 900 years 43 . However, precisely-dated (U/Th) coral microatolls from Simuelue Island, Indonesia suggest that two similar magnitude great earthquakes occurred within approximately 50 years at ca. 1390 and 1455CE 17 .
Here we present results from new auger cores and pits collected from the south-western section of Swale Y ( Fig. 1), where high-frequency ground-penetrating radar 44 identified a complex stratigraphy with more sand sheets than had been previously recognised from this swale. The new sand sheets may confound the consensus model presented above for palaeotsunami deposition on Phra Thong Island and have implications for Sunda Trench 2004 IOT-type tsunami recurrence 21,30,36 .

Results
The south-western section of Swale Y preserves at least three more sand sheets than has been described from elsewhere in Swale Y 21 and at least two more sand sheets than have been identified from Swale VI 30 . The stratigraphy recovered from Auger 10 and Pit 2 ( Fig. 2) are used to investigate the stratigraphy of the south-western section of Swale Y. The 89 cm of sediment recovered from Auger 10 penetrated through to the beach ridge sands underlying the swale sediments at 84 cm, and thus, there is no deeper sedimentary record of tsunami sediments within the swale (Fig. 2a). Pit 2, located less than 30 cm west of Auger 10, did not extend to the total depth of the swale due to the elevated groundwater table and only the upper 78.5 cm of the stratigraphy was examined here (Fig. 2b).
Much of the stratigraphy of Pit 2 and Auger 10 mirror the stratigraphy reported elsewhere from Swale Y (Fig. 3 21 ; Supplementary Information). Here we present the sedimentology of the new sand sheets and their stratigraphic placement recovered from Auger 10 (Fig. 2).
Two new sand sheets occur within the lowest stratigraphic unit above the base of Swale Y. This lowest unit, extending from 84 cm to 46 cm (Auger 10) and 45 cm (Pit 2) is a clastic medium silty to very coarse silty fine sand, with small black charcoal or diagenetically altered plant material. This unit has been interpreted as an intertidal sand elsewhere in Swale Y 21 . Within this intertidal unit in Auger 10, two thin, light grey, clean, quartz-rich, fine to medium sand layers occur from 80.5 to 80 cm and 77 to 75 cm depth. The lower contact of each layer is sharp, but the upper contact grades into the overlying sediments and no sedimentary structures are observed. Following from the nomenclature established previously 21 , these layers are named Sandsheets F and E, respectively.
Between 51 and 50 cm in Auger 10, is a dipping, gun-metal blue, quartz-rich, silty fine sand with sharp upper and lower contacts. This sand sheet is found in other auger cores collected from Swale Y 44 (Fig. 3). The stratigraphic positioning, wholly contained within the intertidal sands, and distinctive colouration suggests that this sand sheet is a correlative of Sandsheet D found in Swale Y 21 .
Sandsheets A, B and C recovered from Pit 2 and Auger 10 are identical to the correlative sandsheets elsewhere on Phra Thong Island 21,29,30 .
Auger 10 also records a further clear, tan fine sand sheet between 19.5 and 19 cm, and which has no correlative elsewhere in Swale Y 21 . This sand sheet is distinct from the stratigraphically higher 2004 IOT Sandsheet A (base at 18 cm in Auger 10) by a 1 cm thick, dark brown, organic-rich (11 to 25% organics), muddy sand to sandy mud, with a high content of plant stems, leaves and rootlets. This unit signifies a return to typical swale-like organic mud accumulation 21 . The sand sheet is stratigraphically identical to Sandsheet X from Swale VI 30 .

Discussion
The sedimentological and stratigraphic analysis of new sediment archives from Swale Y on Phra Thong Island 21 raises several questions about the identification of the causal mechanism of sand emplacement and the recurrence interval of sand sheets preserved at this site.
From closely spaced auger cores it is possible to trace the sand sheets, variations in thickness of each unit and variations in the stratigraphic placement of each sand sheet. Sandsheets A and C extend across the swale (Fig. 3) 21 . Sandsheets B and D are also found in Swale Y but have a patchy distribution (Fig. 3) 44,45 . Direct dating of the sand sheets using quartz OSL 30,32,36,39 , and bounding ages of the sandsheets using 14 C dating of bulk sediments 29 , plant and leaf fragments and buried shells 21 have helped constrain the recurrence interval of sand sheet deposition on Phra Thong Island. Sandsheets B to D were likely formed during tsunami inundation following great earthquakes from the Sunda Trench subduction zone, of which Sandsheet A is a modern analogue 21 . However, the discovery of a sand sheet immediately between the Sandsheets A and B 30 (Sandsheet X) and additional sand sheets found between Sandsheet D and the base of Swale Y in Auger cores 8, 9 and 10 44 (Fig. 3), imply a more complex overwash history than has thus far been presented.
Sandsheet X was observed in pit KPT 37 from Swale VI 30 , and quartz OSL dated the layer as forming at 216 ± 23 years. This unit was interpreted as a lower subunit of the 2004 IOT sand sheet as it contained carbonate and unidentified fossils, and reworked plant material separated this sand layer from the overlying 2004 IOT sand sheet 30 . From Auger 10, a similar thin sand sheet has been recovered from an identical stratigraphic location between Sandsheets A and B (Figs 2a and 3), but this sand sheet was not observed in any other cores or dated. However, the sand sheet is constrained to have been deposited prior to 2004 but after ca. 1455 CE 17 . Although Sandsheet X is thin and only found in Auger 10, it is unlikely to have been deposited by a storm as the location of the swale between 1455 CE and 2004 was likely to have been between 100 to 200 m from the 2012 shoreline (Fig. 1d). No modern analogues exist to suggest such an expansive storm deposit on Phra Thong Island to have deposited sand in Swale Y. The sand sheet is unlikely to have been formed by a historical tsunami caused by an earthquake occurring in the northern segment of the 2004 IOT rupture zone (e.g. 1847, 1881 and 1941; Fig. 1) 10,30,46 . It was postulated that Sandsheet X was deposited during the 2004 IOT and that the organic layer formed from settling of material that was eroded by the initial incoming wave 30 . However, it is difficult to reconcile that the first wave of the 2004 IOT deposited such a small volume of clastic sediment and then reworked plant material, the source of which was not specified, before later waves emplaced the decimetre thick Sandsheet A unit above 30 . An alternative explanation for Sandsheet X in Swale VI and the sand sheet recovered from Auger 10 is that the sand sheet was formed by aeolian reworking allowing sands from the surrounding beach ridges to be blown into the swale. If the sole quartz OSL date from Sandsheet X of 216 ± 23 years is correct 30 , this places the formation of this sand sheet within the timeframes of the Strange Parallels Drought (ca. 1756-1758AD) and the East India Drought (1792-1796AD) as determined from tree-ring records 47 . From the tree ring record, the Strange Parallels Drought severely affected the west coast of Thailand, whereas the East India Drought was less severe 47 . Both droughts resulted in significant rainfall declines that would have reduced plant coverage of the beach ridges and allowed the remobilisation of sand into the swales.
The complex geophysical signature recorded by the high-frequency GPR analysis within Swale Y 44 prompted a more detailed investigation into the deeper stratigraphy at this site from both Auger 10 and Pit 2 (Fig. 2).
The base of Swale Y has been variably dated using 14 C analysed from three unabraided molluscan shells to give a date between 2500 to 2800 cal. yr. BP 21 , and a quartz OSL date of the basal sediment of 2.54 ± 0.34 ka that confirms the 14  Of note, high-frequency GPR analysis has shown that the lee side of Ridge 2 has been scoured and sequentially rebuilt up to ca. 1 m after palaeo-tsunami inundation 44,45 . This was observed at locations where quartz OSL samples were collected between 0.4 and 0.6 m depth 36 . Also, there is extensive reworking of the sediments on the modern beach ridges by burrowing lizards and on the beach face by crabs, which may contribute to quartz OSL dating errors of beach ridges on Phra Thong Island 48,49 .
Sandsheet D has been quartz OSL dated at 2.1 ± 0.26 ka 36 in Swale Y, and 1.85 ± 0.21 ka, 2.04 ± 0.28 ka and 1.8 ± 0.2 ka 30 in Swale VI, suggesting that between the two swales the event was either diachronous, or if the sediments were deposited from the same event, the event occurred at ca. 2 ± 0.48 ka. Sandsheets E and F were deposited prior to Sandsheet D and after the formation of Swale Y, i.e. between 2.54 ± 0.34 ka and 2 ± 0.48 ka. If the outer errors are considered correct, then the postulated 550 to 700-year recurrence interval 21 for great earthquakes originating from the Sunda Trench and producing 2004 IOT-type inundation and sedimentological deposits on Phra Thong Island cannot be discounted. If the mean ages or inner errors between the formation of Swale Y and the deposition of Sandsheet D are considered correct, then the potential recurrence interval generating Sandsheets E and F may be more frequent than the 550 to 700-year recurrence interval. If this is correct, then non-linear recurrence intervals for tsunami-generated washover deposits 42,50,51 must occur for tsunami-type sediment deposition on Phra Thong Island. Further dating of Sandsheets E and F and surrounding sediments using current quartz OSL and 14 C AMS dating techniques are unlikely to constrain the ages of these units.
However, the presence of the May 2007 storm washover fans extending between 20 and 50 m inland from the present day berm 35 (Fig. 1) demonstrates that unusual storms can cause wave set-up to generate overwash and sediment deposition on Phra Thong Island. The age of Ridge 2 at ca. 2.16 ± 0.28 ka 36 implies approximately 370 metres of beach ridges seaward of Swale Y have accreted at an assumed linear rate of 0.17 m/yr. If this rate is correct, then the beach face was within 30 m of the swale when Sandsheet D was deposited, and closer during the deposition of Sandsheets E and F. Mean sea level was also likely to have been less than 1 meter higher than modern sea level along the Malay-Thai coastline at ca. 2 ka 27,52 . The distance between the ca. 2 ka shoreline and the swale on the lee side of the immature beach ridge/berm crest is within the distance that recent storms have emplaced sediment along the modern shoreline. The back-beach environment was accumulating intertidal sands in this period and retained sufficient accommodation space to preserve thin overwash deposits. The distinct packages of clean sand comprising Sandsheets E and F, and the lack of sedimentary structures does not preclude the deposition of storm or tsunami as both causal mechanisms can deposit massive, sand-rich units in protected embayments 53,54 .
Due to the background intertidal sedimentation occurring prior to Sandsheet C deposition in Swale Y, other mechanisms of sand sheet emplacement cannot be discounted for the deposition of Sandsheets E and F. The modern intertidal channel that Swale Y drains towards is currently subjected to tidal variations and swash bar sedimentation. The close proximity of Swale Y to swash bar accretion and tidal reworking at the time of Sandsheets E and F emplacement cannot be discounted as a potential causal mechanism for the formation of these sand sheets. Similarly, the role of intense droughts 47 resulting in sand reworked from the beach ridges also cannot be excluded. More recently, the role of meteotsunami has also been investigated 55 , but little work has been conducted to investigate this mechanism as potential cause of washover deposits. Each of these considerations remain at the forefront of palaeo-overwash investigation.

Methods
Ten auger cores were collected in March 2013 from the south-western portion of Swale Y on Phra Thong Island, Thailand 44 . These augers variably penetrated 62 to 89 cm of the swale sediments (Fig. 3) and in each auger core, the basal, clean, coarse sands of the underlying beach ridge were collected, ensuring a complete stratigraphic sequence 44 . Several of the augers preserved sandsheets that could not be correlated with the known sandsheets from the previous work conducted on the island. Auger core 10 was selected for further sedimentological investigation as it contained the most sandsheets that were easily identifiable and very well preserved. At each augering location, detailed topographic data was collected using differential Global Position System (dGPS) with a maximum vertical error of 4 cm and maximum horizontal error of 1.5 cm.
Each sample from Pit 2 was analysed by loss on ignition (LOI), where crushed samples were sequentially heated to 105 °C (2 hours) to remove water, 550 °C (4 hours) to remove organics, and 950 °C (2 hours) to remove carbonate content 56 . After each heating, the samples were allowed to cool to room temperature before they were weighed and the percentage of water, organic and carbonate content were determined. The remaining sediment corresponds to the percentage of clastic material in each sample.
Each sample from the sand layers subsampled from Auger 10, and each centimetre from the Pit 2 profile were analysed for grainsize parameters. Initially, bulk sediment samples (Untreated) were analysed from Pit 2. Auger 10 samples and samples from Pit 2 were subjected to HCl treatment to remove carbonates, and several H 2 O 2 treatments to remove organic material to determine the granulometry of the clastic component (Treated). This facilitates comparison between the clastic-only component of the sediment and the clastic + organic + carbonate components of the sediment (Supplementary Information). Grainsize was analysed on a Malvern Mastersizer2000 which uses laser obscuration to measure grainsize. Each sample was subjected to 1 minute sonication to further disaggregate the sediments. Each sediment analysis is replicated three times such that the relative standard deviation (RSD) of the mean grainsize (φ ) between replicates is < 1%, and the average of the three replicates is subsequently used. Where RSD is > 1%, the sample is re-analysed until RSD achieves the < 1% requirement. Each sediment analysis was analysed using GRADISTAT 57 reported using the terminology of Folk and Ward 58 .