Tsunami deposits in Martinique related to the 1755 Lisbon earthquake

In order to assess tsunami hazard in oceanic islands, one needs to enlarge the observational time window by finding more evidence of past events. To that end, evidence of allochthonous deposits provides estimates of tsunami inundation, recurrence time and magnitude. However, in tropical islands, erosion due to the highly rainy climate generally prevents deposits to stay in place and when they are, relating them to a tsunami is not straightforward, as they can result either from a strong hurricane or from a tsunami. One notable exception concerns deposits sealed by subsequent events. In this paper, we present 5 evidence of an anomalously thick two-layer tsunami deposit in an excavation in Martinique. Analysis of the archaeological remains indicate that it is related to the 1755 Lisbon tsunami. We explain the thickness of the deposit by a tsunami-induced bore in the mangrove drainage channels of Fort-de-France. Our results highlight the benefits of collaborative research involving geology and archaeology, indicate a way to improve our tsunami databases and further constrain the use of numerical modelling to predict paleo-tsunami deposit thickness. 10


Introduction
The Lesser Antilles volcanic arc has been formed since Eocene by the subduction of the North American plate beneath the Caribbean plate (Bouysse and Westercamp, 1990), with a low convergence rate (ca. 2 cm/yr, DeMets et al., 2000).However, this subduction has the potential to generate moderate to large thrust type earthquakes (Feuillet et al., 2011;Hough, 2013), potentially tsunamigenic (Hayes et al., 2013).During historical times (about 500 yrs), 2 large thrust earthquakes occurred in deposits.The stratigraphic cross section of the excavation (Fig. 2) indicates that in a first stage, the XVII century building walls and ground were surfaced, and then covered by white sand and mortar embankment.At this precise moment, a sandy layer sealed the whole area, including the new foundations, and entered the opened rooms.Field observations indicate that the construction was stopped for a while (days or weeks), and when the construction resumes, the sandy layer is removed only where the new foundations are built and the XVIII century building is constructed.It is finished during the second part of the 20 XVIII century (De Bexon, 1782).
The 8 cm-thick deposit presents clearly defined up and down limits, confined between light colored embankment materials and is composed of two different layers (Fig. 3).The bottom thin light colored layer (1 to 2 cm) of broken shells and white sand is deposited everywhere on the site and follows the irregularities of the erosive contact.The upper black layer is mainly composed of coarse sand of volcanic origin, but include numerous marine shells and rounded pebbles.A terrestrial origin, 25 as a result of a lahar flow or river flood, is excluded based on the fact that the mineralogical type of the deposit exhibits a serious lack of terrestrial carbon materials, showing at most an occasional small piece of wood.The physical aspect of the deposit, and the fact that Fort-de-France's Bay is very well protected against storms and their associated waves and surges, also excludes a stormy origin of these sandy layers.It is confirmed by the lack of large hurricanes in Martinique between 1713 and 1780 (Romer, 1932).The lower limit of the deposit is irregular and erosive, which corresponds to the erosive action 30 of an incident flow before any depositional process begins.The black layer globally follows the light grey layer, indicating a coeval deposition.Its upper limit is a bit diffusive, probably due to the latter manual spreading of the embankment materials observed in the excavation.In general, tsunami deposit comes from inflow and backflow origin.In our case, the geometry of the site, open eastward and closed westward, and the sand particles sub-horizontal lamination (Fig. 3c) shows that the water enters the site and the sediments deposit slowly, and that there is no contribution from upper backflow.In summary, these two superimposed layers of distinct origin can clearly be associated with a single tsunami event.

Origin of the Martinique tsunami deposit
The temporal origin of the tsunami ranges between 1726 and 1783.During this period, Martinique was stroke by two tsunamis, 5 in 1755 and 1767.The 1 st of November 1755 M⇠8.5-9.0Lisbon earthquake triggered the most important known transoceanic tsunami that travelled across the Northern Atlantic Ocean (e.g., Baptista et al., 2003).According to many coeval historical reports, cross-fitted with numerical modeling results, this tsunami is well-known to have reached Portugal, Spain, Morocco, Great Britain, Azores, Madeira, Newfoundland, Bermuda, the Lesser Antilles, Cuba and Brazil, with observed waves up to several meters.In Martinique, numerous historical records report 1 to 3-m height waves in all the coastal areas, including 10 Fort-de-France (Roger et al., 2011).Respectively, the Barbados earthquake of April 24 th , 1767 generated a local tsunami (O'Loughlin and Lander, 2003;Lander et al., 2003) and 3-feet waves were observed only on the eastern coast of Martinique (see a newspaper extract in http://tsunamis.brgm.fr).It is doubtful that the impact of this latter event, which occurred only 12 years after the notable 1755 tsunami, would not have been reported in Fort-de-France.We can thus relate with certainty the Fort-de-France deposits a unique event, the 1755 Lisbon tsunami.

15
However, two questions remained unanswered concerning on one hand the origin of the 2-layer sandy materials, and on the other hand, the thickness of the deposit (⇠8 cm) related to a ⇠1 m historically reported and modeled tsunami (Roger et al., 2011).The spatial distribution of the tsunami deposit thickness after the Tohoku-oki earthquake shows that a 10-cm deposit correspond to a 2.5-3 m tsunami (Takashimizu et al., 2012).Partially, the lack of backflow, which has an erosive effect on the deposited sand (Furusato and Tanaka, 2014), contributes to the thickening of our deposit.The thin lightly-colored layer at 20 the basement of the deposit can be attributed to the bottom of Fort-de-France's Bay, which presently exhibits the same kind of materials.It was transported by the tsunami waves straight to the town.The origin of the black material, and its thickness are more puzzling.Comparing this black sand to the different samples we collected afterwards from neighboring beaches and in the upper Madame river, we find a high similarity rate in terms of mineralogical composition and grain sizes (Fig. 4) only with the sand coming from a beach located just westward to the Madame river mouth (located on Fig. 4).In May 2014, a 25 preliminary archaeological finding was observed at site 2 located 250 m northeastward from the first one (see location on Fig. 4): among the 8 trenches that were dug and immediately re-filled, a sandy layer, similar to the one of site 1 was present in 6 of them (A. Jegouzo, pers. comm).During the XVIII century, Fort-de-France has been constructed on a mangrove and numerous channels were present so they could drain the water from the city (Fig. 4).As it has been observed in many places (Peters et al., 2007;Tanaka et al., 2012;Ely et al., 2014) and during the recent tsunamis, the upriver propagation of tsunami waves can be 30 amplified by a tsunami induced bore (Chanson and Lubin, 2013).In the case of the 1755 tsunami inundation of Fort-de-France, we propose that the Madame's river mouth ridge was removed by the tsunami-induced bore and that the upstream propagation Nat.Hazards Earth Syst.Sci.Discuss., https://doi.org/10.5194/nhess-2017-238Manuscript under review for journal Nat.Hazards Earth Syst.Sci. Discussion started: 7 August 2017 c Author(s) 2017.CC BY 4.0 License.subsequently flooded the center of Fort-de-France, using the draining channels indicated on contemporaneous historical maps (Fig. 4).This phenomenon explains both the 2-layer deposit and the abnormal thickness of the dark sand layer.

Discussion
In the specific context of tropical volcanic islands, coastal areas are constantly affected by erosion from the sea (waves and wind) and from the land (river overflows, landslides, bioturbation and anthropogenic activities).In tropical climate, most of the 5 year and especially during the cyclonic period, important orographic rainfalls increase the soil erosion and the river flooding.
This results in absent or weak tsunami deposits, which most probably will be destroyed afterward by the human activity.In Martinique, we found that most of the present-day coastal cities are built on former mangrove forests, which were completely banked up to meet the growing need for housing the increasing population in this small montinous island.In addition, hurricanes are rather frequent in comparison with tsunamis, and are also able to inundate low-lying areas and to produce similar deposits 10 (Shanmugam, 2012).For these reasons, not only do these processes make it difficult to find deposits of marine submersion in the Lesser Antilles, but when found, it is extremely difficult to determine their origin, as it has been shown in Anegada, British Virgin Islands (Atwater et al., 2012(Atwater et al., , 2017)).
Nevertheless, we have found in Fort-de-France conclusive evidence of 1755 Lisbon tsunami deposits.The 1755 Lisbon tsunami has been widely reported in the whole Caribbean region as previously mentioned.In Anegada, Atwater et al. (2017) 15 propose that an important sheet of sand and shells dates to the interval 1650-1800 can be related to the 1755 Lisbon tsunami.
In Fort-de-France, a 0.9 m level elevation in the Madame river has been observed and is correctly reproduced by numerical modeling (Roger et al., 2011) : our deposition site is located within the modeled inundation zone of the 1755 tsunami (Fig. 4).But our results bring a new restriction to use of the deposit thickness as a proxy for tsunami intensity estimation.Our thin lower layer corresponds to the normal up-stream sediment transportation due to the incoming tsunami waves bringing 20 sediment straight from the offshore sea-bottom.The thick upper deposit was produced subsequently by upriver propagation of the tsunami, as observed during the recent tsunamis of Chile (Fritz et al., 2011) and Japan (Mori et al., 2011), and also seen in the numerical modeling of the 1755 tsunami in Lisbon (Baptista et al., 1998).This propagation in a narrow channel is highly turbulent and able to carry a large amount of suspended sediment (see review in Chanson and Lubin, 2013).In addition, the specific geometry of the Cour d'Appel building prevent the outflow from eroding the inflow deposit as also observed in 25 Kuril islands (MacInnes et al., 2009).It is interesting because the observed deposit thickness is commonly used to calibrate the tsunami propagation model and estimate the intensity of paleo-tsunamis (e.g., Jaffe and Gelfenbaum, 2007;Jaffe et al., 2012).
Recently, in the context of near-field earthquakes, Goto et al. (2012) have also shown that the liquefaction produced by the strong ground motion is responsible for 4/5 of the tsunami deposits, whereas only 1/5 comes from beach erosion.In the context of far-field earthquakes, our study indicates that about 7/8 of the deposits can come from a river bore, and only 1/8 from the 30 beach or bay erosion.Thus, a careful analysis of the tsunami deposit thickness and origin must be a prerequisite before any numerical modeling based on the tsunami deposit thickness.Conversely, the observed thickening of the tsunami deposits for a ⇠1m tsunami in case a river crosses the city provides important constraints on hazard assessment in densely populated areas.
It was a fortuitous chance for us to find undisturbed tsunami deposits buried in a town that has been modified so many times during its short history.But it is likely that this can occur in other coastal American cities.The urban development of many presently large cities in the Americas began during the XVIII century, and it is plausible that the 1755 tsunami deposits can also be found in locations in eastern coastal American cities by investigating historical building sites that may have sealed sediments from this event.Further, as the conservation of tsunami deposits are rare in natural context, we can also hypothesize 5 that collaborative geological and archaeological studies of pre-Colombian sites could also reveal paleo-tsunami deposits and enable us to enlarge our observation window and improve our tsunami catalogues.As demonstrated for archaeoseismology (Nur, 2007), archaeological analysis appears to be a powerful tool to precisely date paleo-tsunamis within a range of dates (⇠several months to years) that is not reachable by classic dating methods, especially for the last centuries.

Conclusions 10
We provide the first conclusive observation of the 1755 Lisbon tsunami deposits in the Americas.The 6-8 cm observed thickness of the sedimentary layers is abnormally large to have been produced by a ⇠1 m tsunami.It can be explained by 2 successive floodings, one related to the direct tsunami front waves and responsible for about 1/8 of the deposit, and the other one produced by a tsunami-induced bore.This indicates that the tsunami deposit thickness used in tsunami modeling is a parameter that must be carefully checked in order to avoid overestimation of paleo-tsunamis and to correctly assess the tsunami hazard.
Tsunami hazard assessment is of primary importance for oceanic islands because of the highly populated coastal area linked to increasing tourism development, the concentration of infrastructure in low-lying areas and the expectation of sea-level rise associated with climate change.Our results indicate that the use of all available geological methodologies and collaborative studies with historians and archeologists might enable us to improve our historical tsunami catalogs in the future, thus helping the preparedness of tsunami hazard plans for coastal communities.

Figure 1 .Figure 2 .Figure 3 .
Figure 1.Location of the 1755 tsunami historical records (red circles) and of the deposits that could be related to the 1755 event with large uncertainties (green stars).The orange star represents the area of the Lisbon earthquake and the filled blue star the location of the deposits in Martinique.See Data Repository for precise locations and references.