1600 year-long sedimentary record of tsunamis and hurricanes in the Lesser Antilles (Scrub Island, Anguilla)

a Université de la Rochelle, UMR 7266 LIENSs, Bâtiment Marie Curie, avenue Michel Crépeau, 17042 La Rochelle cedex, 6 France. 7 b Université Savoie Mont Blanc, UMR 5204 EDYTEM, Bâtiment « Pôle Montagne », 5 bd de la mer Caspienne, 73376 Le 8 Bourget du Lac cedex, France. 9 c UNSW Sydney, Sydney 2252, NSW, Australia. 10 d Université de Angers, UMR CNRS 6112 LPG, 2 boulevard Lavoisier, 49045 Angers Cedex 1, France. 11 e Institut de Physique du Globe de Paris, Université́ de Paris, CNRS, 75005 Paris, France. 12 f CEA, DAM, DIF, F-91297 Arpajon, France. 13 g Laboratoire SYMME, domaine Universitaire BP 80439, 74944 Annecy le Vieux Cedex, France. 14 15 Abstract 16


36
Worldwide, an increasing number of people living along the coasts are threatened by natural hazards 37 (earthquakes, volcanoes, hurricanes, tsunamis and marine floods). To better constrain the risk due to 38 natural hazards in a region, it is crucial to determine the recurrence interval and the intensity of a given 39 hazard. Historical records of these events are of prime importance to this goal but they are often too This is a non-peer reviewed preprint submitted to EarthArXiv The manuscript is under review at Sedimentary Geology This is a non-peer reviewed preprint submitted to EarthArXiv The manuscript is under review at Sedimentary Geology its magnitude, this event may have triggered only a small tsunami which was observed only in English 134 harbour at Antigua (with a wave amplitude of 1.2 m) and Nevis (Shepherd, 1992). 135 136 Another historical tsunami occurred in the British Virgin Islands on 18 November 1867 AD (Reid and 137 Taber, 1920). The wave reached 4 to 6-m height according to the observations and exceeded 10m at 138 Guadeloupe, which is the highest reported value of tsunami height in the Caribbean Sea (Lander et al., 139 2002). The tsunami followed a strong earthquake felt in all the British Virgin Islands, but the origin of 140 the event remains poorly constrained (Reid and Taber, 1920;Zahibo et al., 2003). Islands, Atwater et al., 2012Atwater et al., , 2017. 14 C dating of coral boulders, shells and organic remains indicates 154 that these overwash occurred in the Middle age (between 1200 and 1480 cal. AD). This event has been 155 attributed to an extraordinary storm or a tsunami of nearby origin (rupture of the Puerto-Rico megathrust, 156 outer-rise normal faulting). In St Thomas (US Virgin Islands), four coastal ponds and a mangrove site 157 contain a record of a tsunami also dated with radiocarbon from plant remains between 1200 and 1450 158 cal. AD (Fuentes et al., 2017). 159 This is a non-peer reviewed preprint submitted to EarthArXiv The manuscript is under review at Sedimentary Geology 161 Numerous hurricanes have been reported in the Lesser Antilles Arc. They occur mostly between July 162

Past hurricanes in the Lesser Antilles
and November with frequency, magnitude and trajectory changes from one year to another (Garnier et 163 al., 2015). They are influenced by large regional climate oscillations such as the El Niño Southern 164 Oscillation (ENSO) (Jury et al., 2007), the Atlantic Multidecadal Oscillation (AMO) ( 2). This event is unprecedented in the historical record (Cangialosi, et al., 2018). It resulted in major 183 damage, destroying many towns and killing 47 people before continuing its route north-westwards, 184 where it impacted the Virgin Islands and Florida (Cangialosi et al., 2018). Other category 3 hurricanes, 185 such as Donna (1960 AD (Dunn, 1961) and Dog (1950 AD) (coast.noaa.gov/hurricanes, 2020) with 186 This is a non-peer reviewed preprint submitted to EarthArXiv The manuscript is under review at Sedimentary Geology south-west trajectories relative to Scrub Island (Fig. 2), strongly affected this part of the Lesser Antilles 187 (Table 1) in Scrub Island westernmost lagoon. Two one-meter-long sediment cores were sampled at 80 cm water 193 depth using an Uwitec gravity corer with a hammer. The location of cores was chosen along a transect 194 perpendicular to the coast. SCR1-18-01 was taken just behind the dune whereas SCR1-18-03 was taken 195 at the opposite side of the lagoon (150 m east of core SCR1-18-01), in distal position relative to potential 196 washovers. The cores were horizontally stored in a walk-in cooler in the laboratory, before being split, 197 photographed, described in detail and subsampled. 198 199

Sedimentological and mineralogical analyses 200
Grain-size analyses were carried out at the EDYTEM laboratory using a Beckman Coulter LS 201 13 320 XR particle size analyser using sonication, with a range size between 0.010 µm and 3000 µm. 202 The sampling step was about 5 mm through core SCR1-18-01 respecting facies boundaries. The mean 203 grain size curve and the grain size contour plot were produced using the Matlab software. 204 Core SCR1-18-01 was scanned at the SYMME laboratory (Annecy, France) using an X-ray 205 tomography system (Easytom XL 150 from RX SOLUTION society) at 150 kV and 200 μA. The core 206 was cut in two sections (110 to 70 cm and 70 to 0 cm), which were scanned one after the other. The 207 acquisition time was about 1 hour per core (for each section, 7 successive scans by translating along the 208 beam were necessary to cover the entire volume of the core). The reconstructed data are sets of 16 bits 209 Tiff image files representing cross-sections of the cores (horizontal slices from top to base). The image 210 resolution (voxel size) is 55 µm for the lower 40 cm and 85 μm for the upper 70 cm of the core. The 211 total number of images available for the core is 12,992 (6,048 + 6,944). For each section, the vertical cross-sections were reconstructed using the Image J software and including the Fiji distribution 213 extension (Schindelin et al., 2012). 214 X-ray diffraction (XRD) analyses were performed on core SCR1-18-01 with a sampling step of 215 Ten slabs were sub-sampled from core SCR1-18-01 and resin-embedded for analysis with a 240  were computed with the package 'serac' in R software (Bruel and Sabatier, 2020).
A total of 13 organic macro-remains (wood and leaves) were collected in the muddy layers. 14 C 267 measurements were performed by accelerator mass spectrometry (AMS) at the Poznan Radiocarbon 268 Laboratory and at the Laboratoire de Mesure 14C (LMC14) ARTEMIS at the CEA (Atomic Energy 269 Commission) Institute at Saclay. The 14 C ages were converted to 'calendar' years at two sigma using 270 the Intcal13 calibration curve (Reimer et al., 2013). An age-depth model was constructed using the R 271 code package 'clam' in R software (Blaauw, 2010) in the amount of organic matter. Some sediment layers show a green (olive) color due to the presence 282 of algal mats (at 16 cm depth for example, Fig. 3). Mean organic matter (LOI550) content is 24% (+/-283 8%) and mean carbonate (LOI950) content is 35% (+/-2%) (Fig. 3). The carbonated fraction is 284 composed of small calcitic shells and some patches of magnesian calcite (confirmed by the XRD data). 285 These shells are about 0.5 to 1 mm-long and mainly consist of bivalves and benthic foraminifera, such 286 as miliolids (Fig. 3), which are typical of tropical lagoon environments (Hallock and Glenn, 1986). 287 Layers of facies 1 display varying amounts of Miliolinella sp. tests (0 to 5 on the semi-quantitative scale) 288 that are thin, transparent and show only minimal amount of wear (Fig. 5). Facies 1 also contains very 289 small (less than 5 µm in diameter) angular quartz grains (Fig. 3). Some rare thin horizontal beds of NaCl 290 deposits (halite) are also visible (Fig. 3).
Facies 2 (F2) is a very pale brown unique deposit in the cores, between 46 and 40 cm depth in core 292 SCR1-18-01, and from 31 to 23 cm depth in Core SCR1-18-03. It is composed of 8% (+/-3%) organic 293 matter (LOI550) and 42% (+/-1%) carbonates (LOI950). It displays many whole and intact shells of 294 variable sizes (0.5 mm to 1 cm) and some larger sandy grains (about 1 mm), the latter consisting of 295 calcite and magnesian calcite (Fig. 3). Small bivalves (Fig. 3)   Some Elphidium discoidale (4 on the semi-quantitative scale) were also found between 92 and 93 cm 317 depth in core SCR1-18-01 (Fig. 5). Unlike the specimens of Miliolinella sp., the tests of Amphistegina This is a non-peer reviewed preprint submitted to EarthArXiv The manuscript is under review at Sedimentary Geology foraminifera represent larger symbiont-bearing species with a very robust shell, typical of high-energy 320 reef environments (Hallock and Glenn, 1986). 321 In rare F3 layers, very few Cerithideopsis costata shells have been found. Based on the presence or 322 absence of salt coating (halite) on grains, three sub facies are distinguished within F3: Subfacies 3.1 is 323 characterized by salt coating around all grains; subfacies 3.2 is characterized by an absence of salt 324 coating; subfacies 3.3 is characterized by both salt-coated and non-salt-coated grains (Fig. 3). 325

326
The main results obtained from XRF analyses of core SCR1-18-01 are presented in Fig. 4. Linear 327 regression tests on the Br and LOI550 results show a significant relationship with a R of 0.53 and p-328 value < 0.05 (0.01342). This suggests that Br XRF peak areas can probably be used as a proxy for 329 organic matter content, since Br is known to have a strong affinity for organic matter (e.g., Chagué-Goff 330 and Fyfe, 1996; Ribeiro Guevara et al., 2019). A strong relation between S and Br, in association with 331 organic matter, has been previously reported using PCA, explaining 61% of the variance (Chagué et al., 332 2018). However, Br and S also occur in higher concentrations in seawater than freshwater (Wedepohl, 333 1971), and this could also explain their strong relationship in the Scrub lagoon, although it appears that 334 their affinity with organic matter mostly explains their distribution. By labelling samples with the facies information in the PCA, it was possible to establish a map of the 344 geochemical data distribution (Fig. 4C). Facies F1 is dominated by the "organic matter", "silicate" endmembers and more or less influenced by the "salt" endmember. Facies F2 seems to be correlated 346 with the "carbonates" endmember, due to the presence of Sr-rich shell fragments (e.g., Sabatier et al.,347 2010; Chagué-Goff et al., 2017 and references therein). Facies F3 is also linked to the "carbonates" 348 endmembers (represented by Ca and Sr) and more or less influenced by the "salt" endmember. 349 Moreover, facies F3 is hardly linked to the "organic matter" endmember represented mainly by the Br. 350 Thus, facies F3 can be highlighted by the Sr/Br ratio, which increases in facies F3. (Fig. 5). size contour plot shows that all these coarser layers have a similar grain size distribution with a mode 355 around 800 µm (Fig. 5). Moreover, the Sr/Br ratio increases simultaneously with the grain size in F3 but 356 presents a higher resolution than the grain size samples. This enable us to use Sr/Br as a high-resolution 357 grain size proxy, which is useful to precisely identify the coarse deposits boundaries in the study area.  probably deposited between 1955 and 1963 AD (Fig. 6). Deeper L#24 and L#23 (Fig. 5) were probably 373 deposited before 1955 AD but during the 20 th century. 374 375 4.4.2. 14 C ages and age-depth model 376 The oldest age in core SCR1-18-01 is 464 +/-74 cal AD at 105 cm depth (  Fig. 7). A correlation between Core SCR1-18-01 379 and Core SCR1-18-03 was performed, based on F3 layers and using Sr/Br ratio and 14 C ages (Fig. 7). 380 An age-depth model was constructed for core SCR1-18-01 using the 10 14 C ages (Table 2)   Br and S ("Organic Matter" endmember, Fig. 4B). This high organic matter content is probably related to autochthonous algal productivity as terrestrial organic matter inputs through run-off reaching the 397 lagoon from the inner part of this small karstic island is very limited. 398

399
The very small angular quartz grains found in the organic matrix are unlikely to originate from Scrub 400 Island, which only comprises carbonates. Both the grain size and morphology can be interpreted as the 401 result of fragmentation of larger quartz grains. Moreover, the angular quartz grains suggest that they 402 have been transported in suspension, without an erosive rolling/saltation phase (Coude-Gaussen, 1991). Fe and Si on the PCA (Fig. 4B) suggests that Fe in F1 most probably had the same aeolian origin 410 ("silicates" endmember, Fig. 4B). These two poles are the major components of facies F1 (Fig. 4C). and their abraded tests, it is likely that these taxa are allochthonous and that they were transported in the 422 lagoon during marine incursions (e.g., Pilarczyk et al., 2016). Most F3 layers exhibit a sharp basal contact and thin inland, with the number of these units also decreasing landward as attested from core 424 correlation (Fig. 7). Consequently, F3 layers are interpreted as high-energy deposits resulting from 425 overwash of the sandy barrier related to either hurricanes or tsunamis. Cerithideopsis costata shells associated with small bivalve shells. The large amount of shells explains 429 the high counts of Sr for the "carbonates" endmember. This shell accumulation is found at the same 430 period (1450-1635 AD) in both cores (Fig. 7), suggesting that it probably extended in the entire lagoon. completely discounted, although it is also possible that the frequency of high energy events is close to 475 the frequency of sand layers recorded on Scrub Island. 476

Tsunami deposits 486
The upper layer displaying specific criteria corresponds to layer 17 (L#17, Fig. 5). It is 4 cm 487 thick in core SCR1-18-01. L#17 is also recorded in Core SCR1-18-03 (Fig. 7). L#17 is composed of a 488 sandy layer with a mud layer at the top. The sandy part displays an upward small grain size decrease 489 from halite precipitation on sand grains exposed to both sea spray /sea water and strong evaporation. 500 Such conditions are commonly found on beach barriers (e.g., Ward, 1967) but also on land following 501 tsunami inundation (e.g., Chagué-Goff et al., 2012), even one year after the event despite dilution by 502 rainfall (Chagué-Goff et al., 2014), due to capillary action. Thus, abundance of salt coating could be an 503 indicator of beach barrier sands. On the contrary, the absence of salt coatings can be related to an absence of strong evaporation, a situation which can be found in submarine sand, or sand from the lower 505 shoreface (e.g., Ward, 1967). Thus, the absence of salt coating on grains in L#17 might indicate a deeper 506 sediment source, farther away from the shoreline. Even though post-depositional processes should also 507 be considered, the covering of sandy deposits by the fine grained facies 1 allows the preservation of 508 halite coatings on the sandy grains. 509 Finally, the 2 cm thick mud layer above the sand deposit contains a large amount of wood and 510 vascular plant remains. This concentration of plant remains shows an important terrestrial input in the 511 lagoon that can be interpreted as resulting from the backwash during the L#17 event. Alternatively, this 512 concentration in plants remains could also be due to the incorporation of vegetation eroded from the 513 sand barrier by the tsunami wave. As organic matter is less dense, it is deposited on top of the tsunamis 514 deposit out of suspension, as previously reported (e.g., Chagué-Goff et al., 2011). 515

516
According to the age model (Fig. 8), this event is dated at 1720 cal. AD (1652-1810 cal. AD), 517 which could correspond to the 1755 AD Lisbon tsunami (Chester, 2001). Although originating in the 518 Eastern Atlantic, this tsunami was able to reach Scrub Island lagoon, which is oriented towards the west. 519 Rico Trench (Fig. 1B), which is a possible source for a tsunami in the Caribbean (Atwater et al., 2017). 544 In Bonaire (south Caribbean) sublittoral sediments interrupting onshore sedimentary sequences on the 545 eastern shore (Boka Washikemba), provided evidence for a major wave impact shortly before 500 cal. 546 BP (1450 cal. AD), which most likely corresponds to a tsunami (Engel et al., 2010). In the same region, 547 the study of a sediment core from a saline lagoon on Cayo Sal Island, (western Venezuela) showed 548 probably tsunami-laid reworked peat (Weiss, 1979). This event is dated between 1180 and 1450 AD 549 (Weiss, 1979), which correspond to an age between 1046 and 1481 cal AD (Reimer et al., 2013). These 550 two islands are more than 800 km south from the Puerto Rico Trench, indicating that if all the study 551 sites cited above recorded the same tsunami, it most probably hit the whole Caribbean area. 552 It is also possible that coarse layers dated to the expected age of this tsunami in other study sites 553 have been misinterpreted as hurricane or storm events, such as in St-Martin 25 km from Scrub Island, 554 within a lagoon with the same orientation (Malaizé et al., 2011) or in Vieques, Puerto Rico with a single 555 high energy event during a period with fewer intense hurricane strikes (Donnelly and Woodruff, 2007) 556 or in Belize where the event was described as "much stronger than all other storms in the record" 557 (McCloskey and Keller, 2009). If all the above-cited sediment records are related to the same event, 558 their wide geographical extent would exclude the hypothesis of a hurricane record, as one hurricane 559 generally cannot cross both the north and south of the Caribbean. Finally, while most of these studies proposed an age for this event with a more than a 200-year uncertainty, the new age constrain provided 561 by our age model gives a much more accurate age with only a 100-year uncertainty. The identification 562 of new sedimentary records of this event would therefore provide an opportunity to further refine its 563 chronology and source. A record of this event might be preserved in others islands in the northern or 564 southern Caribbean and it should be investigated. 565 566

Hurricane chronicle and climate forcing 567
If layers L#14 and L#17 are related to tsunami events, the 23 remaining sandy layers (F3 deposits) 568 deposited in the studied lagoon over the last 1600 years, can probably be interpreted as resulting from 569 storms/hurricanes (Fig. 8). Among those 23 layers, the three most recent F3 deposits occurred since the 570 mid-19th century and can be associated with historical hurricanes. Intriguingly only three high-energy 571 events are recorded in the lagoon as sandy layers, while 13 hurricane trajectories (Table 1) passed less 572 than 30 km from Scrub Island since 1852 AD (coast.noaa.gov/hurricanes, 2020). Thus, we assume that 573 Scrub lagoon has only recorded the closest and most intense hurricanes. 574 575 According to 137 Cs activities, L#25 was deposited between 1955 (first 137 Cs fallouts) and 1963 576 AD (maximum nuclear weapon test in the Northern Hemisphere, Fig. 6). This layer can be tentatively 577 related to the only hurricane that passed within 30 km from Scrub between 1955 and 1963, the Category 578 3 Hurricane Donna, which took place in 1960 AD (Table 1, coast.noaa.gov/hurricanes, 2020). According 579 to the short-lived radionuclide measurements, L#24 occurred before 1955 AD but during the 20 th 580 century, as the sediment still presents 210 Pbex activities (Fig. 6), and based on our age model, it is dated 581 between 1935 and 1959 AD (Fig. 8). There were two hurricanes in the region during this period: 582 Category 3 Hurricane Dog (1950) and Category 1 Hurricane Alice (1955) ( Table 1). As Dog seems to 583 be both closer and stronger than Alice, it may be recorded by L#24. L#23 is dated between 1903 and 584 1957 AD with the best age being 1930 AD according to the age model (Fig. 8). L#23 can be attributed 585 to the unnamed category 3 hurricane that occurred in 1922 AD (Table 1). 586 Based on the instrumental data available, it appears that three conditions must be fulfilled for a 587 given hurricane to leave a sedimentary record in the studied lagoon, assuming enough sediment availability: (1) at least category 3 on the Saffir-Simpson scale, (2) a hurricane track less than 30 km 589 from Scrub Island and (3) a hurricane heading southward of Scrub and northwest of Anguilla (Fig. 2). 590 It is likely that the hurricane path must face the barrier to be recorded in the lagoon and thus must reach 591 northwest Anguilla, as Anguilla Island appears to act as a barrier for hurricanes passing south of the 592 island. This could explain why Category 5 Hurricane Irma and Category 3 Hurricane Lenny were not 593 recorded in the studied lagoon, as their path bypassed Anguilla from the south (Fig. 2). Moreover, 594 hurricanes passing north of the lagoon cannot be recorded. 595

596
The studied lagoon provides an opportunity to estimate long term hurricane activity in the region over 597 the last 1600 years. A 101-year window event frequency was made for Scrub Lagoon (Fig. 9D)  hurricanes that passed within 30 km from Scrub, two also hit the Bahamas, two headed towards Belize 634 but did not reach this area and most of them have made their way close to the American East coast but 635 never reached the Massachusetts region (Fig. 9). Nevertheless, even if this regional comparison provides 636 a relative coherent pattern, small differences of hurricane activity between each site could be also 637  (Fig. 9D, E), which is consistent with the Bahamas signal. However, other 646 periods high hurricane record in Scrub Island correspond to a more southern ITCZ position (720-835, 647 1080-1230 and 1625-1695 cal AD, Fig. 9, D, E) showing that for the Scrub Island signal, the 648 correspondence between the ITCZ position and the hurricane activity intensification is not obvious. 649 Thus, other climatic forcing parameters are expected to influence hurricane tracks. Therefore, it is 650 necessary to investigate other sites in the Caribbean in order to better understand climatic parameters 651 modulating the trajectory of hurricanes. Caribbean, its large extent supports a tsunamigenic source. The presence or absence of salt coating on 667 sand grains, indicating a barrier or lower shoreface origin for sand grains, respectively, could be used as 668 a new criterion to distinguish tsunami (reworking both barrier and shoreface sediments) from hurricane 669 (mostly reworking barrier sediments) deposits.
The 23 remaining sandy layers were interpreted as resulting from hurricanes. The chronology 671 established with short lived radionuclides suggests that the three most recent deposits were associated 672 with historical hurricanes and it appears that three conditions must be fulfilled for a given hurricane to 673 leave a record in the studied lagoon: (1) at least category 3 on the Saffir-Simpson scale, (2) a storm track 674 at less than 30 km from Scrub Island and (3) a storm track located southwestward with respect to the 675 studied lagoon. 676 The studied sedimentary sequence provides an opportunity to explore the past hurricane activity  This is a non-peer reviewed preprint submitted to EarthArXiv The manuscript is under review at Sedimentary Geology This is a non-peer reviewed preprint submitted to EarthArXiv The manuscript is under review at Sedimentary Geology This is a non-peer reviewed preprint submitted to EarthArXiv The manuscript is under review at Sedimentary Geology This is a non-peer reviewed preprint submitted to EarthArXiv The manuscript is under review at Sedimentary Geology This is a non-peer reviewed preprint submitted to EarthArXiv The manuscript is under review at Sedimentary Geology