Extreme waves in the British Virgin Islands during the last centuries before 1500 CE

This paper wrestles with problems in earthquake and coastal geology. The seismological issue is whether tsunamis are generated along a plate-bounding trench in the northeast Caribbean. The coastal challenge is how to distinguish, in hurricane alley, between the bouldery traces of a tsunami and those of a storm.

Tsunamis in Caribbean written history were generated remotely in a few cases and locally in most (Lander et al., 2002). The Lisbon tsunami in the year 1755, of the common era (CE), produced the largest of the far-traveled waves. All but one of the local tsunamis arose from sources within the Caribbean plate. No documented tsunami is known to have arisen along the Puerto Rico Trench (Fig. 1A), although the trench coincides with several conjectural sources of Caribbean and North Atlantic tsunamis (ten Brink et al., 2014).

The largest of these conjectural sources is the subduction thrust that dips southward from the trench as a predominantly strike-slip boundary between the Caribbean plate and the North American plate (Fig. 1B). A tsunami in Hispaniola was generated during a 1946 earthquake on a western continuation of this fault (Lynch and Bodle, 1948; Dolan and Wald, 1998). All historical tsunamis generated elsewhere in the northeast Caribbean, however, can be associated with faults within the Caribbean plate (ten Brink et al., 2011). Interplate coupling inferred from satellite geodesy is negligible along the Puerto Rico Trench (ten Brink and López-Venegas, 2012). To the southeast, where plate convergence is more nearly orthogonal to the trench axis, the Lesser Antilles subduction thrust is thought to be partly coupled (Hayes et al., 2014; Symithe et al., 2015). The fault in this area is the inferred source of a large earthquake in 1946 that may have been accompanied by a few tens of centimeters of tectonic subsidence (Weil-Accardo et al., 2016). It is also the inferred source of a larger earthquake in 1843 (Feuillet et al., 2011; Roger et al., 2013), but accounts of the 1843 earthquake provide few hints of an attending tsunami and no evidence for tsunami damage (Bernard and Lambert, 1988; Shepherd, 1992; Bernard and Lambert, 1992).

Tsunami sources along the Puerto Rico Trench may also include normal faults of the outer rise. Their distinct but undated scarps furrow the incoming North America plate on the outer trench wall (ten Brink, 2005; Andrews et al., 2013) (Fig. 1C). Normal faulting in this tectonic setting can generate tsunamis efficiently by displacing tall water columns with high-angle dip slip (Craig et al., 2014). Along the Japan Trench, for example, normal faulting produced strong shaking and a disastrous tsunami in 1933 (Kanamori, 1971; Usami, 2003). In the northeast Caribbean, the outer rise was neglected as a regional earthquake source (Mueller et al., 2003) but modeled as a potential source of Puerto Rican tsunamis (Mercado Irizarry and Justiniano Sepulveda, 2003). However, like the subduction thrust along the Lesser Antilles, the outer rise along the Puerto Rico Trench is not known to have generated a regional tsunami in written history that begins with Columbus in 1492.

Other potential sources of tsunamis are marked by scarps of youthful reverse faults and submarine landslides south of the trench (Fig. 1C). A reverse fault that displaces the ocean floor has been mapped for a distance of 20 km in deep water northwest of Anegada, south of Main Ridge (Grindlay et al., 2005). Large submarine slides, farther west, have estimated recurrence intervals of ∼100,000 yr (ten Brink et al., 2006). A large example is directly north of San Juan, Puerto Rico.

This paper presents evidence for extreme waves of pre-Columbian age at Anegada, a low-lying island 120 km south of the Puerto Rico Trench (Figs. 1 and 2). We put this evidence in context by reviewing previous work on Anegada’s deposits from recent hurricanes, and on the island’s assorted evidence for southward inundation earlier in post-Columbian time. We differentiate some of this evidence from signs of southward flooding during the last centuries before Columbus. The main goals of the paper are to describe several kinds of pre-Columbian deposits, constrain their time of emplacement, and provide constraints for modeling that might eventually show whether these deposits represent a Puerto Rican Trench tsunami or an unusual storm.

It is rare that a prehistoric tsunami can be shown to be responsible for boulders on storm-prone shores. The case for a tsunami commonly hinges on evidence that storm waves are incapable of having emplaced clasts that are exceptionally large and/or inland (Frohlich et al., 2009; Ramalho et al., 2015). Site-specific effects of physiographic setting and clast provenance have confounded attempts to identify, in boulder deposits, a field signature of tsunami deposition that is both unique and global (Switzer and Burston, 2010; Terry et al., 2013, p. 65–66).

Anegada is exposed to several kinds of extreme waves, known or plausible. As reviewed herein (in the section headed “Exposure to Extreme Waves”) recent hurricanes have flooded parts of the island with storm surge, and exceptional storms might produce tsunami-like bores by analogy with 2013 Typhoon Haiyan in the Philippines (details in “Tsunami-Like Bores Analogous to Those from Typhoon Haiyan”). The island is subject to far-field tsunami waves like those associated with the Lisbon earthquake of 1755, as well as hypothetical tsunamis from historically latent sources along the Puerto Rico Trench.

This paper focuses less on comparisons with modern storm and tsunami deposits of distant shores, than on field-based comparison among extreme-wave effects on Anegada. Combining previous findings with new observations, we assign Anegada’s evidence for extreme waves to three generalized groups, listed here by increasing rank and age.

• 1. Deposits of modern storms (Fig. 3) consist mainly of microbial debris deposited around interior salt ponds and fans, along with localized fans of sand and mud deposited along the island’s south shore. Boulders moved onshore by these storms are probably limited to a coral rubble ridge within a few tens of meters of the island’s north shore.

• 2. An inland sheet of sand and shell is found mainly beneath low-lying salt ponds (Fig. 3). This deposit, encountered almost entirely below mean sea level, extends more than 1 km southward from the island’s north shore. This sheet was deposited quickly, sometime between 1650 and 1800 calibrated years in the common era. This event age, expressed herein as the range 1650–1800 cal yr CE, or simply 1650–1800, is consistent with deposition by far-traveled waves of the 1755 Lisbon tsunami. We will infer that the sheet of 1650–1800 sets an upper bound for extensive Anegadan deposits of a large tsunami of remote origin. That is, the sheet either represents the 1755 tsunami, or it was laid down by a tsunami or storm between 1650 and 1800 that exceeded the 1755 tsunami in its depositional effects at Anegada.

• 3. The deposits of highest rank include fields of scattered coral boulders and cobbles (Fig. 4). Evidence of this rank includes, in addition to the coral clasts, the shells of large marine mollusks, an extensive sheet of bioclastic sand, and fields and clusters of limestone boulders and cobbles. Some of this evidence was previously lumped with the sand and shell sheet of 1650–1800 cal yr CE. To explain the deposits of highest rank, we will invoke offshore erosion and onshore flooding during an extreme-wave event sometime between 1200 and 1480 cal yr CE (herein, 1200–1480 cal yr CE, or simply 1200–1480). We will infer that this event exceeded any other extreme waves at Anegada in the past few thousand years, and we attribute it to either a Puerto Rico Trench tsunami or to tsunami-like bores from a Haiyan-like storm.

For convenience, we use the term “extreme wave” across these three degrees of extremeness, and without regard for wave period. Modern storm overwash that deposited sand and gravel on the island’s shores fell short of floods from the sea that produced the inland deposits dated to 1650–1800 and 1200–1480. As interpreted herein, those prior floods may have been tsunamis, or they might represent storm-generated bores intermediate in period between storm and tsunami waves.

We stop short of comparing the field evidence with storm and tsunami models. A previous comparison of field evidence with storm and tsunami modeling included estimation of forces involved in transporting the largest clasts in two low-lying fields of limestone boulders at Anegada. The emplacement of these clasts was found to be better explained by tsunamis generated along the Puerto Rico Trench than by a category 5 hurricane or the Lisbon tsunami (Buckley et al., 2012; Watt et al., 2012). This initial finding needs reappraisal in light of the additional field observations presented herein, the accompanying bathymetry and topography from lidar (light detection and ranging) surveys, and the tsunami-like bores from 2013 Typhoon Haiyan.

The paper contains three sets of illustrations¹. One set provides overviews of regional setting (Fig. 1), local geography and physiography (Fig. 2), previous field evidence (Fig. 3), new field evidence (Fig. 4), dating (Fig. 5), and estimates of inundation distance and height along nine cross-island profiles (Fig. 6). The other two sets provide supporting details, grouped by theme and location. Figures A1–A12 summarize field evidence by clast type, plotted island-wide. Figures B1–B7 provide closer views of field evidence, using large-scale maps of parts of the island (index map, Fig. 2A) and including photographs of most of the dated clasts.

Also presented are lists of physiographic data sources (Table 1), radiocarbon ages (Table 2), details about the control points in Figure 6 (Table 3), and author contributions (Table 4). Further lists, containing most of the points plotted on the various maps, can be found in the Supplemental Material². Included are the locations and elevations of all plotted points of coral clasts, reef rock, beach rock, and a widespread unit termed the main sand sheet. Clast size and sand thickness are reported there, although these measurements were not made routinely in reconnaissance mapping. The Supplemental Material also contains details of ∼27 sets of limestone slabs that were systematically measured for clast size and orientation.

The lidar data in this report are available from the U.S. Geological Survey (Fredericks et al., 2016). The data are from two surveys, one terrestrial and the other shallow water, made in 2014 (Table 1). The resulting base maps are presented with a Universal Transverse Mercator grid, with northings and eastings in zone 20N.

Anegada is situated on the north edge of a shelf that is perched south of the Puerto Rico Trench. Along a bathymetric profile proceeding southward from an Atlantic abyssal plain 5–6 km below sea level, the outer rise descends across fault scarps on the outer trench wall, reaches the trench floor at water depths of 8 km, rises from there to the vicinity of the island’s north shore, and continues south of the island on a shelf a few tens of meters deep (Fig. 1B).

The north side of Anegada, facing the Atlantic Ocean, is fringed by a coral reef. The reef extends along most of the 17 km length of the island and continues another 15 km farther southeast. The seaward edge of the reef, at 10–20 m water depth, forms a smooth west to southeast arc. Breaking waves mark a discontinuous reef crest (Fig. 2A), as further delineated by shallow-water lidar data from 2014 (Table 1; Fig. 2B). The reef is separated from the island’s north shore by a reef flat, a few meters deep at most, that extends 100–1500 m perpendicular to shore. The fringing reef has probably existed throughout the past few thousand years, as judged from the greatest of the 27 radiocarbon ages obtained on coral clasts found inland (Table 2, sample 32). This age, 1960 ± 60 radiocarbon years before 1950 CE (¹⁴C yr B.P.), corresponds to the range 216–550 cal yr CE.

Coral taxa of the fringing reef and the reef flat, when surveyed in the 1970s by Dunne and Brown (1979), included the brain coral Diploria strigosa (Fig. A5), the boulder star coral Montastraea (now Orbicella) annularis (Fig. A6), the mustard hill coral Porites astreoides (Fig. A7), and the elkhorn coral Acropora palmata (Fig. A8). Despite losses of coral cover on Caribbean reefs since the 1970s (Gardner et al., 2003), we found living colonies of all these species near Anegada’s shore in the past few years. All these taxa are represented in the pre-Columbian deposits described in this paper.

Those pre-Columbian deposits also include two edible mollusks that today inhabit the reef flat. The conch Strombus gigas, still harvested today, was noted on the reef flat north of the island in the 1970s (Dunne and Brown, 1979), and is represented in pre-Columbian shell mounds at the East End (Schomburgk, 1832; Davis and Oldfield, 2003), remains a local food source. We found juveniles of the tiger lucine Codakia orbicularis living in sand of the reef flat of Windlass Bight in 2016 (Fig. A9). Near the Florida Keys, this bivalve is most abundant in shallows of the passages between islands, in similarly open marine conditions (Perkins and Turney, 1972). In the southern Caribbean, C. orbicularis was part of the diet of pre-Columbian peoples in Grenada (Giovas, 2013).

Synonyms at the genus level have been widely used for two of these species. Orbicella annularis was known as Montastraea annularis before molecular and morphological revision of reef-coral taxonomy (Budd et al., 2012). The alternatives to Strombus in S. gigas include Lobatus and Eustrombus (National Marine Fisheries Service, 2014; National Marine Fisheries Service, undated). The paper uses Orbicella annularis and S. gigas.

The tide range at Anegada is <0.5 m. The range at Tortola, 50 km to the southwest (Fig. 1C), is similarly small in official predictions discontinued after 2013 (http://www.ntslf.org/sites/ntslf/files/pdf/predictions/Tortola_2013_+0400.pdf). Tide ranges were measured at Anegada during a day in 2008 and across one week in 2009, in support of geological studies (Atwater et al., 2012; Watt et al., 2012).

Many of the landforms of Anegada, although partly obscured by trees and shrubs, are apparent in bare-earth topography computed from the terrestrial lidar survey in 2014 (Figs. 2B and 6). The entire island is <9 m above sea level. Pleistocene deposits in the east underlie benches above the 5 m contour. Salt ponds in the east are hemmed in between a narrow line of Holocene beach ridges and high ground underlain by Pleistocene deposits. Larger salt ponds occupy much of the island’s western interior at elevations at or slightly below mean sea level. The western ponds adjoin knolls of Pleistocene deposits and are mostly enclosed by Holocene beach ridges. Additional beach ridges extend westward beyond the western ponds.

The island’s Pleistocene deposits consist of shallow-water carbonate deposits of reefs, beaches, and channels. Reefs are abundantly evidenced by coral skeletons preserved in surficial limestone (Dunne and Brown, 1979). A beach is marked by imbricated slabs of beach rock in the quarry plotted in Figure 2B. Relict channels trending across Anegada, between its Atlantic and Caribbean sides, are evidenced by the topographic trough holding Bumber Well Pond and, most distinctly, by a meandering, branching topographic low south of Table Bay (Fig. 2B and profile F-F’ of Fig. 6). Landforms of the Pleistocene deposits also include the limestone knolls that form high ground and localized coarse-clast sources within and among the western salt ponds. Many such knolls can be seen in shaded relief in Figures B2B and B3B.

A finger coral specimen from the Pleistocene deposits has been dated to marine oxygen isotope stage 5e by uranium series methods. Well-preserved specimens of Porites furcata were collected from unconsolidated Pleistocene deposits in the wall of a quarry ∼3 m below the natural ground surface of the largest area of high ground southeast of Bumber Well Pond (Fig. 2B). The specimens evidently formed near their former sea level; they are conformably overlain by a unit containing imbricated slabs of beach rock. Some of these coral clasts retain pink encrustations of the reef-dwelling protozoan Homotrema rubrum. The age of one of the specimens was measured by William G. Thompson of Woods Hole Oceanographic Institution. Thompson’s reported age of 121,110 ± 166 yr (2 standard deviations) includes corrections for measured evidence of exchange of U and Th with the environment. Details can be found in Atwater et al. (2014, Fig. S2 therein).

Subaerial weathering since the last interglaciation has cemented the uppermost 1–2 m of the Pleistocene deposits. The resulting limestone pavement armors most of eastern Anegada and forms a hard substrate, at depths rarely more than 2 m below present sea level, that underlies Holocene beach ridges, salt flats, and salt ponds as far west as Red Pond (example beneath a salt flat, Fig. B3C). The limestone is pitted from dissolution and is widely pocked with sinkholes (Fig. B4B). The pavement is typically broken into slabs and blocks, and the breaks commonly coincide with roots of trees and shrubs.

Bioclastic sand forms numerous beach ridges at Anegada (Fig. 2B). The ridges are fewer in the east, where they line the shore, than in the west, where they have accreted outward from the interior and have built the island 3 km beyond the westernmost limestone knolls. The highest of the ridges in the west are between 5 and 8 m above sea level (Figs. B1B, B1D). The ridges are founded on gently dipping, locally pebbly layers deposited by waves. A cross-bedded, probably eolian cap is exposed on eroding parts of the north shore and in sand pits (including the pit labeled in Fig. B3B). Many of the sand grains are pink from Homotrema rubrum derived from the fringing reef and the reef flat (Pilarczyk and Reinhardt, 2012). Longshore drift, driven by prevailing easterly winds, probably accounts for the island’s westward accretion (Schomburgk, 1832).

Salt ponds and their microbial mats occupy low areas, near sea level, that make up much of the island’s western half. The hydrology and microbial biology of these western ponds were examined by Jarecki and Walkey (2006); they found that the most saline of the large western ponds attain dry-season salinities of 250 parts per thousand; that the large western ponds merge when engorged by rain; that the areas flooded intermittently form seasonal salt flats with patchy mangroves; and that at least one of the ponds is connected to the sea more by percolation than by a tiny inlet on the south shore (location noted in Fig. 2A).

We doubt that the western ponds had an additional inlet, on the north shore, in the early 1800s. Such an inlet was reported by Schomburgk (1832, p. 158), who surveyed Anegada in 1831; writing of the western salt ponds, he stated that “a junction with the sea” had existed on the island’s northern side until “the hurricane of 1819 stopped its passage.” The island map in Schomburgk (1832) suggests that this inlet was north of Flamingo Pond. The mapped inlet, however, is not apparent in lidar topography of this area, which instead shows a complex of breaches and beach ridges between Flamingo Pond and the sea (Figs. B1A, B1B). Perhaps rain from the 1819 hurricane ponded in this area.

The main western salt ponds are vestiges of a single, merged pond, with an inlet probably large enough for salinities to remain close to that of ocean water, that occupied much of the area of western Anegada for thousands of years. This bygone marine pond is evidenced by muddy deposits that contain cerithid gastropods and other mollusks (Reinhardt et al., 2012). It was replaced by salt ponds, to the complete exclusion of this molluscan assemblage, at a time indistinguishable from an overwash event in 1650–1800 cal yr CE (Atwater et al., 2012, p. 74, 78). This timing is evidenced by stratigraphic superposition: the sand and shell sheet of 1650–1800 separates marine pond deposits below from salt-pond deposits above (Bumber Well photos in Fig. A3).

The island was first called anegada, Spanish for flooded, in 1493, during Columbus’s second voyage (Morison, 1942, v. 2, p. 78, 98). The name does not necessarily refer to the bygone marine pond, which might have been difficult to see from ships that remained outside the fringing reef on the north or beyond patch reefs to the island’s south. However, stratigraphy and paleoecology suggest that a marine pond covered much of western Anegada in 1493 and that such a pond persisted until the overwash event of 1650–1800 (Atwater et al., 2012).

The bygone marine pond likely accounts for a shoreline notch cut into Pleistocene limestone at an elevation close to present mean sea level. The notch, probably cut in part by endolithic bacteria, is conspicuous on some of the margins of the western salt ponds (Fig. 2B). No higher shoreline notch is evident at Anegada (Atwater et al., 2012, p. 62). Accordingly, we infer that the catastrophic inundation of Anegada in 1200–1480 cal yr CE took place when the island’s relative sea level was ordinarily no higher than it is today.

Anegada’s documented history of post-Columbian settlement begins in 1776 or 1784 (Dookhan, 1975; Pickering, 1983; Faulkner, 2005; Atwater et al., 2012; supplementary text ESMT-1 therein). A Methodist church counted 82 Anegadian members in 1796 (Pickering, 1983). A nautical survey in 1811 tabulated the island’s human population as 197 and noted production of livestock and cotton (Blachford, 1813).

Schomburgk (1832) reported that some of the island’s trees had been logged, and that some areas were devoted to crops or grazing. Areas he mapped as farmed correspond today to areas laced with stone walls, many of them overgrown (Figs. B4F, B5C, and B5D). Most of the island’s permanent population, then as now, resided in the place known as The Settlement (location, Fig. 2A).

Physiographic setting predisposes Anegada’s north shore (its Atlantic side, facing the Puerto Rico Trench) more to flooding by a tsunami than to flooding by an ordinary hurricane. Frequency of occurrence of course exposes the island overwhelmingly to storms. But in case of either a tsunami or a storm, how would inundation from each be affected by the deep water and the fringing reef north of the island? In the possibilities considered herein, the focus is on the north shore because the most dramatic of the island’s extreme-wave evidence extends inland from there (Fig. 4). The potential causes of flooding from the north include tsunami-like bores comparable to those generated in the Philippines during 2013 Typhoon Haiyan.

The written history of human settlement at Anegada begins a few decades after the Caribbean’s most widely documented tsunami, the transatlantic waves of the 1755 Lisbon tsunami. As summarized in Figure 1A, these far-traveled waves were noted in the Lesser Antilles, Hispaniola, and Cuba, and its effects exceeded those of the Lisbon tsunami of 1761 (Lander et al., 2002; Barkan et al., 2009; Roger et al., 2010, 2011).

In the concluding section, “Inferred History of Extreme Waves,” we speculate that the 1755 tsunami accounts for geological evidence summarized in Figures 3 and A3. This evidence, as reported previously (Atwater et al., 2012; Pilarczyk and Reinhardt, 2012; Reinhardt et al., 2012), suggests that Atlantic waters poured southward into the interior of western Anegada in 1650–1800 cal yr CE.

The deep water north of Anegada should facilitate tsunami inundation across the island’s north shore. A hypothetical tsunami advancing southward from a source in the Puerto Rico Trench remains in deep water until it closely approaches the island (Fig. 1B). Traversing this deep water extracts less energy than would shoaling on the shelf to the island’s south. Simulated tsunamis generated by faulting in the Puerto Rico Trench produce extensive flooding of northern Anegada (Buckley et al., 2012).

The fringing reef to the island’s north fails to prevent this extensive flooding in the simulations of Buckley et al. (2012). In simulations for other places, onshore inundation was found to be reduced by generic reefs (Kunkel et al., 2006) or to vary with reef parameters (Roger et al., 2014), or to increase in onshore height and inundation distance after crossing a fringing reef in American Samoa (Gelfenbaum et al., 2011).

In contrast with its effect on an incident tsunami, the deep water north of Anegada should limit storm surge from set-up by wind. When setting up a storm surge, surface wind stress is resisted by the entire water column, a resistance greater in deep water than in shallow water. Computed storm surge is accordingly smaller in the deep water to Anegada’s north than on the shelf to the south (Caribbean Disaster Mitigation Project, 2002).

Storm surges observed at Anegada are consistent with these expectations. The main example documented is from 2010 Hurricane Earl (Atwater et al., 2014). This storm was at category 4 when it passed 30 km north of the island (Fig. 1A). Its surge reached elevations of 1 m along the south shore, as judged from wrack lines and eyewitness accounts in The Settlement and from wrack surveyed beside nearby spillover fans (Figs. 3 and A1; profiles D-D’ and E-E’, Fig. 6). In the north, wrack beside Windlass Bight marked a high-water line that was surveyed as approaching 2 m above sea level, but this line represents the combined effects of surge and swash (Fig. B3B; profile C-C’ of Fig. 6).

In the absence of tsunami-like bores, storm effects on Anegada’s north shore may be further limited by the presence of a fringing reef and the reef flat. In a simulated hurricane of category 5 at Anegada, the storm surge from atmospheric pressure and wind set-up is 1 m off the island’s north shore, and the storm waves break on the fringing reef and dissipate further across the reef flat (Buckley et al., 2012).

Eyewitness observations of the storm surge from 1960 Hurricane Donna are available for the south shore only. Donna was at category 3 or 4 when its eye passed over Anegada. Its reported winds blew counterclockwise, from the north as the storm approached, and from the southwest after the eye had passed. Smooth water of a storm surge entered The Settlement only after the eye had passed, according to eyewitnesses interviewed in 2008 and 2009. This surge reached 2.5 m above sea level at reference points later surveyed by differential global positioning system (Atwater et al., 2012), and it extended several hundred meters farther inland than did the surge from Hurricane Earl (Fig. A1). A Donna limit in The Settlement is 1.5 m above the sea-level datum of the terrestrial lidar survey (Fig. 6, profile E-E’).

Parts of the Pacific coast of Eastern Samar, in the Philippines, resemble the north shore of Anegada in that they are protected from wind-driven storm surge by the deep water of a nearby trench, and from ordinary storm waves by a fringing reef. This coast was abundantly flooded, however, during 2013 Typhoon Haiyan (Shimozono et al., 2015; Kennedy et al., 2016; Soria et al., 2016). The Eastern Samar town of Hernani was entered by three or four bores recorded on video (Roeber and Bricker, 2015). The flooding during each bore was on the order of minutes; the bores were intermediate in period between a storm wave (tens of seconds) and a tsunami wave (tens of minutes).

These tsunami-like effects of an unusually large tropical cyclone have been explored with numerical models (Shimozono et al., 2015; Roeber and Bricker, 2015; Kennedy et al., 2016). In simulations for Hernani, sets of large waves alternate with sets of smaller waves that produce tsunami-like bores by interacting with the fringing reef and the reef flat. More generically the phenomenon is known as surf beat, infragravity waves, or wave group run-up. In model runs with Philippine storms of various sizes, the fringing reef protects the shore from incident waves of lesser storms as much as half the intensity (central pressure and maximum wind speed) of Typhoon Haiyan. In additional model runs with various reef widths, wave effects greater than Haiyan’s were obtained for reef widths that produce resonance with surf beat (Roeber and Bricker, 2015). Such Haiyan-like effects have not yet been modeled for Anegada.

Previous studies at Anegada have documented geological evidence of recent storms and assorted geological evidence for greater overwash in 1650–1800 cal yr CE (Fig. 3). The following review, updated with new findings about this post-Columbian overwash, highlights discrepancies that are resolved in subsequent parts of the report by inference of greater overwash in pre-Columbian time (Fig. 4).

Geological effects of 2010 Hurricane Earl were examined at Anegada in February 2011, six months post-storm, with surveys of high-water lines and of predominantly fine-grained deposits (Atwater et al., 2014). The storm’s effects were considered further in 2013 and 2015 with studies of a coral rubble ridge (Spiske and Halley, 2014; Spiske, 2016).

The fine-grained deposits noted in 2011 consisted of widespread microbial debris and localized sand and mud. The microbial debris formed a desiccating layer, a few centimeters thick at most, on salt flats beside salt ponds from which the debris was likely derived. The layer probably consisted of microbial cells and their extracellular polymeric substances that storm waves had stirred up. The layer was drying as chips in 2011 (Figs. A1 and B2C), and these became progressively less evident in subsequent years.

The microbial debris affords comparisons, at the same place, between modern storm deposits and the island’s pre-Columbian fields of scattered coarse clasts. The microbial debris, derived from the island’s interior, contrasts starkly with boulders that the debris surrounded (Atwater et al., 2014, Fig. S6 therein). For example, microbial debris from Hurricane Earl coated salt flats from where stray coral boulders protrude (Figs. B2C and B6D).

The localized sand and mud accumulated on spillover fans that extend a few tens of meters inland from the island’s south shore west of The Settlement (Figs. 3 and A1). A new fan deposit, consisting of sand and lime mud 10 cm thick at most, was found on top of a black microbial mat. Pit walls revealed underlying sand sheets intercalated with earlier microbial layers. Leaves in the earlier mat layers were found to contain post-bomb levels of ¹⁴C; the leaves gave ages mostly younger than 1950 cal yr CE (Atwater et al., 2014, Figs. S4 and S5 therein).

The coral rubble ridge on Anegada’s north shore differs from the fields of scattered coral clasts in landscape position, internal structure, clast size, clast composition, and age. Here we summarize published descriptions (Spiske and Halley, 2014; Spiske, 2016), cite ridge clast ages obtained by Xu et al. (2015), and briefly compare the ridge with storm deposits on rocky coasts elsewhere.

Ridge at Soldier Point. A coral rubble ridge extends discontinuously along ∼1.5 km of Anegada’s north shore, at and near Soldier Point (map views, Figs. 3 and B4B; field photos, Figs. A1 and B4E). In fair weather it is separated from the water’s edge by a mostly bare platform of Pleistocene limestone as much as 10 m wide. The ridge has a maximum measured width of 15 m and a crest elevation as high as 3.5 m. The seaward side of the ridge slopes more steeply than its landward side; local slumps probably contribute to steepness on the seaward side. The ridge is outboard (north) of a higher limestone surface that crests nearly 4 m above sea level to the south (Fig. 6; Table 3, profile D-D’).

The internal deposits and structure of the ridge were examined in a trench (location in Fig. B4B). The deposits are as thick as 0.8 m, and consist mainly of coral rubble that is rounded, encrusted, and bored (brain coral examples photographed nearby; Fig. A3F). The dominant coral genus was Acropora. Also noted were minor conch shells, serpulid rock, beach rock, and karstified Pleistocene limestone. The clast diameters, in centimeters, averaged 16 cm by 11 cm by 4 cm, and the largest measured was 75 cm by 40 cm by 5 cm. Coarse clasts support the ridge, without interstitial sand, and show seaward-dipping imbrication.

The coral rubble locally continues a few tens of meters inland from the ridge as scattered clasts. The landward margin of the ridge usually coincides with ground-covering plants (Fig. 3) that obscure the rubble limit and probably impede rubble transport (Spiske and Halley, 2014). However, 180 m west of the trench site, coral rubble of cobble size extends as much as 50 m inland through a gap in vegetation. Here the remains of a stone wall approach the north shore (Fig. B4F), and the most inland of the rubble extends to these remains.

Storms routinely refresh the ridge. The ridge accretion was at first thought infrequent because the subaerially exposed rubble surfaces are gray (Spiske and Halley, 2014). This patina, ascribed to endolithic bacteria, was assumed to represent tens of years of subaerial exposure. Ridge-altering storms were therefore supposed greater than 2010 Hurricane Earl. However, the trench dug in 2013, and left open at that time, was found in 2015 to have been partly filled (Spiske, 2016). In addition, clasts at the trench-fill surface had already turned gray. These findings showed that the coral rubble ridge had been both refreshed and newly weathered between 2013 and 2015.

Recent deposition is further evidenced by boulder ages. Xu et al. (2015) sampled the coral rubble ridge at Soldier Point to assess Sr/Ca ratios in skeletons of the brain coral D. strigosa as indicators of sea-surface temperatures. Xu et al. (2015) cored three of the freshest-looking coral boulders, and dated the cores by means of uranium-series dating (two boulders), pattern matching of growth-band density (among all three boulders), and further matching with patterns of sea-surface temperatures observed elsewhere (three boulders). The outermost growth band of one of the boulders was found to represent 1976. In each of the other two boulders the outermost growth band was assigned to 2008.

Comparison with storm deposits on other coasts. The coral rubble ridge at Soldier Point has the internal structure and external form of storm deposits on other rocky shores (Spiske and Halley, 2014). The hundreds of waves in a storm commonly organize coarse clasts as the imbricated framework of a shore-parallel ridge, an effect neither known nor expected from tsunamis (Weiss, 2012; Terry et al., 2013, p. 64–65). Modern storms have refreshed bouldery coastal ridges in Ireland (Hall et al., 2006; Knight and Burningham, 2011; Cox et al., 2012), Iceland (Etienne and Paris, 2010), France (Fichaut and Suanez, 2011), Hawaii (Richmond et al., 2011), and the southeast Caribbean (Morton et al., 2008).

These coastal, shore-parallel ridges contrast with the inland fields of scattered or clustered boulders and cobbles that the comparatively few waves in a tsunami have produced (Paris et al., 2009; Etienne et al., 2011; Goto et al., 2011; Richmond et al., 2011; Yamada et al., 2014). However, inland fields of scattered coarse clasts have also resulted from storms in settings similar to Anegada’s north shore. In the Philippines, flooding from the sea during Typhoon Haiyan shifted isolated boulders a few tens of meters on a shore behind a fringing reef (May et al., 2015) and deposited limestone boulders 1 m in diameter as much as 150 m from the shore (Kennedy et al., 2016). Near Okinawa, storms and tsunamis have both moved boulders from a fringing reef onto a reef flat; the main difference is distance of transport, thought greater for a 1771 tsunami than for subsequent typhoons (Araoka et al., 2010; Goto et al., 2010a, 2010b). Inland fields of scattered coarse clasts have been documented, in addition, in northwestern France, where boulders quarried and scattered by giant historical waves extend a few hundred meters downslope in the lee of a coastal cliff top (Fichaut and Suanez, 2011).

Evidence for unusual flooding from the north was previously reported from Anegada’s western salt ponds and their margins. The findings were based on field work in 2008 and 2009, and on related laboratory work and modeling, and they were reported in a set of papers dated 2012. The field observations focused on breaches in the northern beach ridges and on two kinds of deposits to their south that were mapped chiefly at and below sea level: a sand and shell sheet, and inland fields of scattered limestone boulders and cobbles. The various lines of evidence, summarized in Figure 3, were presented in an introductory paper (Atwater et al., 2012), in paleontological reports on the sand and shell sheet (Reinhardt et al., 2012; Pilarczyk and Reinhardt, 2012), and in papers that describe the scattered limestone clasts (Watt et al., 2012) and model their emplacement (Buckley et al., 2012).

Nearly all this initial evidence was ascribed, by Occam’s razor, to a single storm or tsunami in 1650–1800 cal yr CE. This simplification leads to contradictions that are highlighted in this section, and that are resolved thereafter by evidence, found chiefly since 2009 and summarized in Figure 4, for truly catastrophic overwash in 1200–1480 cal yr CE.

Breaches transect beach ridges and overwash fans of Anegada’s north shore (Figs. 3, A2, B1A–B1D, B2A, B2B, B3A, and B3B). The deepest breaches hold salt ponds that likely began as plunge pools. Shallower breaches are marked by seasonally flooded salt flats, and still shallower breaches contain mangroves.

Evidence for multiple times of breaching. Two sets of breaches at Anegada are evident in bare-earth lidar topography south of Cow Wreck High Point (Fig. B1B). Their seaward limits are distinctly separated by geomorphic superposition of ridges on this accretionary part of the island. The inner, earlier set of breaches coincides with a muted beach-ridge topography that may represent decapitation of ridges by sheet-like overflow (seen also northwest of Red Pond; Fig. B2B). An inner set of breaches was previously mapped southeast of Cow Wreck High Point on the basis of air-photo interpretation (Atwater et al., 2012, Fig. 2B therein), but those inferred breaches are disallowed by intact ridges now seen in lidar (Figs. B1A, B1B).

Two times of breaching are less evident geomorphically in the narrower beach-ridge plain farther east beside Red Pond and Bumber Well Pond (Figs. B2B and B3B). In this area, breaches cut in pre-Columbian time might have been reused in post-Columbian time. Stratigraphic evidence for such multiple use, noticed in reconnaissance (Atwater et al., 2012, p. 75 therein), is beyond the scope of this paper.

Post-Columbian age of a younger time of breaching. The younger set of breaches south of Cow Wreck High Point probably dates to post-Columbian time, according to limiting radiocarbon ages obtained in the past few years.

Pieces of the staghorn coral A. cervicornis were mapped in an area south of the point, and ages were measured on four pieces that retained branches (Fig. B1). Most of the mapped pieces track the tall beach ridges in the northeast and their projected extension to the southwest. The pieces stand in beach-ridge sand as if exhumed from it, rather than deposited on top of it. We therefore infer that storms deposited the staghorn pieces during construction of the ridges.

Two of the dated pieces were collected from a beach ridge that is conspicuously breached to the northeast and to the southwest. The younger of these gave an age corresponding to 1479–1672 cal yr CE. The breaching likely occurred after this interval and may therefore correlate with deposition of the sand and shell sheet of 1650–1800 cal yr CE (Fig. 5). Sand deposits that may correlate with this sheet have been found along the low-altitude path inland from the dated breach (Fig. 6, profile A-A’; field photo, Fig. A3).

Examples of breaches on other shores. The various breaches at Anegada can be compared to breaches elsewhere that have been cut during storms or tsunamis. In some cases these breaches on other shores were used more than once.

During a hurricane in Texas, overtopping of a barrier island created dozens of breaches, particularly where the storm surge covered all but the highest dunes (Suter et al., 1982). The large breaches, upon emerging from constrictions, splayed landward as distributary channels of washover fans. Some of the breaches followed paths cut in prior storms. In a more recent example, 2012 Hurricane Sandy cut breaches tens of meters across a sandy barrier in New Jersey (Irish et al., 2013).

Tsunamis in 1960, 2004, and 2011 are among those known to have cut through sandy beach ridges. The 1960 Chilean tsunami poured through Chilean beach ridges (Atwater et al., 2013) and breached paddy-field berms in Japan (Kon’no et al., 1961). Some of the Chilean breaches had also been used by tsunamis in previous centuries. The 2004 Indian Ocean tsunami cut and/or widened breaches in Thailand, Sri Lanka, and Indonesia (Choowong et al., 2007; Goff et al., 2007, p. 11, 29, 31; Fagherazzi and Du, 2008; Jankaew et al., 2008, Fig. S1 therein; Liew et al., 2010; Kain et al., 2014). Along the Japan Trench, breaches in beach ridges have been correlated with a documented tsunami of 1454 (Sawai et al., 2015), and sand spits were breached by the catastrophic tsunami of 2011 (Tanaka et al., 2012).

A layer of sand and shells extends beneath and beside the western salt ponds as much as 1.5 km south from the breached ridges (Figs. 3 and A3). Its thickness is mostly in the range 1–25 cm. The sheet is preserved best beneath microbial mats of perennial parts of the salt ponds. These areas lack crab burrows that commonly homogenize deposits beneath the seasonally flooded salt flats.

Composition. The sand is chiefly bioclastic in the north and pelletal in the south. The bioclastic sand includes pink Homotrema rubrum grains that were ultimately derived from the reef or the reef flat, but that may have been sourced more immediately from breaching of the north-shore beach ridges (Pilarczyk and Reinhardt, 2012). The pelletal grains could have been suspended from deposits of the interior marine pond.

Also present, where the sheet extends beneath the length of Bumber Well Pond, are angular grains, granules, and pebbles of limestone (dark gray in lower close-up photo, Fig. A3). Similar limestone clasts today litter limestone pavements that flank the pond.

The shell assemblage in the sheet resembles that of the underlying marine pond deposits (Reinhardt et al., 2012). As in those deposits, the main taxa of the sheet are cerithid gastropods (Cerithium variabile, Cerithidea beattyi, Cerithium eburneum, Batillaria minima), small bivalves (Anomalocardia brasiliana, Anomalocardia cunimeris), and bubble shells (Bulla striata). Many of the bivalves are articulated. It is unlikely that any of these molluscan species derived from settings as exposed to the sea as the reef flat. Similar assemblages have been found, instead, on muddy flats on the landward side of the Florida Keys (Perkins and Turney, 1972).

Where overlain by microbial mat layers of salt ponds, the sand and shell sheet includes an uppermost layer of lime mud a few centimeters thick. The lime mud commonly contains mangrove leaves and crab claws, and its contact with the mats locally coincides with a white crust a few millimeters thick (Atwater et al., 2012, Fig. 8 therein).

Numerical and relative ages. Radiocarbon dating and negative historical evidence together bracket the sand and shell sheet in the range 1650–1800 cal yr CE. Emplacement not before 1650 cal yr CE, at two standard deviations, is shown by radiocarbon ages of mangrove leaves and a twig in the mud cap (Atwater et al., 2012, Fig. 8 therein). The most definitive of these samples were obtained from margins of Red Pond and Bumber Well Pond (Figs. 5A and A3). Emplacement after 1800 is unlikely because no extensive sand or shell deposits have resulted from Anegadan storms of the past 65 years or more, and because there are ample reasons to doubt Schomburgk’s (1832) secondhand report that a hurricane in 1819 closed a northern inlet to a western salt pond (details in “Traces of Recent Storms” and “Holocene Beach Ridges and Ponds,” respectively).

By relative dating, deposition of the sand and shell sheet coincided with demise of the marine pond with its conversion into salt ponds (reviewed in “Holocene Beach Ridges and Ponds”). Shelly marine pond deposits are overlain by the sand and shell sheet, which in turn is overlain by microbial mats (Fig. 3). The hypersaline waters of the salt ponds excluded the marine pond mollusks, which have yet to return to the western interior of Anegada (Jarecki, 2003).

Interpretations. The sand and shell sheet might represent tsunami overwash, hurricane overwash, waves generated inside the marine pond, and the opening of tidal channels. Reinhardt et al. (2012) found that among these possibilities, tsunami overwash best explains this sheet’s lateral extent and form. Citing provenance of the shells and preservation of articulated bivalves, they inferred that the overwash widely eroded and winnowed bottom sediments of the bygone marine pond.

Reinhardt et al. (2012) puzzled over this exclusively inland provenance of the shells in the sheet. They expected a tsunami that poured across Anegada from the north to have introduced shells from the reef flat, and made an analogy with a lagoon in Oman where tsunami overwash is evidenced by allochthonous shells derived from offshore (Donato et al., 2008). Many of the shells in the Omani case remain articulated, and many are sharply broken but otherwise fresh, in contrast to bored and disarticulated shells of the lagoon.

At Anegada, during the field work of 2008 and 2009, several coarse clasts derived from the reef flat, or from the reef, were found in the area of the sand and shell sheet, but were not given due weight. They include a boulder of the brain coral D. strigosa north of Bumber Well Pond (Figs. A5 and B3C; Atwater et al., 2012, supplementary figure ESMF-7 therein); a larger brain coral colony protruding from a salt flat west of Red Pond (Figs. B2C–B2E); and articulated and disarticulated valves of the tiger lucine C. orbicularis beside northeastern Red Pond (photo, Fig. A3).

Herein we show that both clasts of these kinds consistently predate the sand and shell sheet by several centuries. In contrast with Atwater et al. (2012; Figs. 5 and 6A therein), who assigned nearly all evidence to a single inferred flood from the sea in 1650–1800, we find that a previously unrecognized time of flooding dates to the last centuries before Columbus, and we infer that flooding in 1650–1800 merely overwrote its effects. Those effects likely included emplacement of limestone boulders and cobbles that were previously thought coeval with the sand and shell sheet of 1650–1800.

Distribution. Several fields of coarse limestone clasts were noted in 2008 and 2009 in western Anegada (Atwater et al., 2012, Figs. 2A, 5, and 7 therein). Two of these fields were mapped in detail (Watt et al., 2012) and their largest boulders were used to test possible causes of overwash (Buckley et al., 2012). The individual clasts of these two fields, north of Bumber Well Pond, are mapped in Figures 3 and B3B, and the generalized trends plotted in Figure B3B are reproduced in Figures 4 and A11. In addition, the northern field can be seen in profile view along line C-C’ of Figure 6.

Both these boulder fields north of Bumber Well Pond extend southward from inland outcrops of Pleistocene limestone from which most of the clasts were probably derived. One field extends 200 m southward from outcrops 300 m from the north shore, and the southern half of the other field extends 100 m southward from a limestone knoll that is 800 m from the north shore. The spacing between clasts was found to increase southward in both fields.

Time of emplacement. The emplacement of these and other coarse limestone clasts was previously dated, during field work in 2009, by assessing the stratigraphic position of the lowest parts of the clasts (Atwater et al., 2012, p. 75). If resting on sand, or at least above the top of shelly marine pond mud, a clast was interpreted in 2009 as having been transported during deposition of the sand and shell of 1650–1800 cal yr CE. Conversely, if surrounded by shelly marine pond mud, a limestone boulder or cobble was interpreted in 2009 as having been emplaced before the sand and shell sheet of 1650–1800. This relative dating presupposes an isochronous change from marine pond to hypersaline ponds immediately after the deposition of that sand and shell sheet (reviewed in “Holocene Beach Ridges and Ponds”).

Watt et al. (2012, Fig. 5 therein) accordingly checked the stratigraphic positions of the keels of 161 limestone clasts in the 2 fields north of Bumber Well Pond. Among these dispersed limestone boulders and cobbles, 77 were found within sand, 54 on top of sand, 31 on Pleistocene limestone, 3 at or above a contact between sand and shelly marine pond mud, and none beneath that contact.

The results of this methodical work do not necessarily show, however, when the limestone boulder fields north of Bumber Well Pond were first emplaced. In Atwater et al. (2012, p. 75), it was presupposed that all sand in contact with the coarse limestone clasts belongs to the sand and shell sheet of 1650–1800. Not considered was the possibility that coarse limestone clasts first accumulated with an earlier sand body, or that sand brought in by later overwash merely remobilized the smallest of the limestone clasts.

Inferred mechanism of emplacement. A tsunami probably emplaced both of the fields of scattered limestone boulders and cobbles north of Bumber Well Pond, according to Watt et al. (2012) and Buckley et al. (2012). Emplacement by storm was deemed unlikely on the basis of location far inland and elongation perpendicular to shore (Watt et al., 2012). A category 5 hurricane was found insufficient to produce the flows inferred by modeling transport of the largest of the boulders (Buckley et al., 2012), before 2013 Typhoon Haiyan made clear that an unusually large storm can produce tsunami-like bores in a setting like that of Anegada’s north shore.

Watt et al. (2012) made analogy with modern tsunami deposits in which coarse clasts were scattered inland. They contrasted the two fields with shore-hugging ridges elsewhere in the Caribbean that Morton et al. (2008) attributed to storms. Not considered, however, were examples of boulders that storms managed to scatter (comparisons in “Coral Rubble”).

Buckley et al. (2012) estimated flow velocities for transport of the largest of the boulders and compared the results with flow velocities computed from simulations of four potential sources of extreme waves at Anegada. The simulations were plotted along a physiographic profile without the bathymetric and topographic constraints now available from lidar. The extreme waves simulated were from a category 5 hurricane, a Lisbon tsunami generated during an earthquake of M 9.0, a Puerto Rico Trench tsunami generated on the subduction thrust during an earthquake of M 8.7, and a Puerto Rico Trench tsunami generated by normal faulting on the outer rise during an earthquake of M 8.0. Only the scenarios of tsunamis from the Puerto Rico Trench were found capable of transporting the boulders. However, this conclusion preceded the widespread recognition, from 2013 Typhoon Haiyan, that storms can produce prodigious tsunami-like bores.

The preferred scenario of Buckley et al. (2012) is a tsunami from normal faulting during an earthquake of M 8.0 on the outer wall of the Puerto Rico Trench (Fig. 1B). A steep fault dip yields waves of shorter period than does slip on a subduction thrust. The relatively short-period waves break on the fringing reef. The maximum water heights exceed 5 m across the reef flat and the north shore. The maximum simulated flow velocities are nearly 10 m/s across the reef and the reef flat, and above 5 m/s as much as 500 m inland from the north shore.

A sand and shell sheet that lacks large offshore clasts appears incongruent with inland fields of limestone boulders and cobbles. How did south-directed flows fail to bring in brain coral boulders and tiger lucine valves, while still dislodging and transporting limestone boulders from sources hundreds of meters inland? This puzzle was recognized by Reinhardt et al. (2012), who wondered why, if emplaced by a tsunami, the sand and shell sheet of 1650–1800 contains no molluscan taxa other than those of an interior marine pond.

These puzzling combinations of flow competence and clast provenance are reconciled herein by interpreting western Anegada as a geomorphic palimpsest, akin to a reused writing surface that retains traces of previous text. A sandy fan containing articulated reef-flat shells predates a sand and shell sheet that is inset into it. The later sheet is limited to the lowest flow paths, and its molluscan assemblage is limited to that of an interior pond. Catastrophic emplacement of inland limestone boulders and cobbles in pre-Columbian time preceded lesser flooding from the north. The lesser flooding, in 1650–1800, merely reworked the smaller limestone clasts.

Evidence for a pre-Columbian catastrophe, although mapped chiefly in eastern Anegada (Fig. 4), began to be recognized in 2011 with field work on the effects of Hurricane Earl on western Anegada. Samples were collected then from the brain coral boulders west of Red Pond (Fig. B2D) and north of Bumber Well Pond (Fig. B3C), and from an additional brain coral clast found upside-down in reconnaissance of an eastern salt pond (Figs. B6A, B6B, B6D). All three were expected to date, like deposits in Figure 3, to the post-Columbian decades between 1650 and 1800. However, the ¹⁴C ages instead suggested that all three coral colonies died in the last centuries before Columbus (Table 2; samples with 2011 prefix).

This confounding result spurred searches for additional coral clasts. Many were encountered during further reconnaissance in eastern Anegada in 2012. The coral rubble ridge at Soldier Point was noticed, as were inland fields of scattered coral boulders and cobbles inland from there. Inland coral clasts were sampled south of Soldier Point (Fig. B4) and to the southeast near Table Bay (Fig. B5) and Warner (Fig. B6). Nineteen ¹⁴C samples were submitted, and all gave pre-Columbian ages (Table 2, samples without prefix). Also noted, but not dated, were brain coral fragments incorporated in stone walls (e.g., Fig. B4F).

Dozens of additional coarse clasts were encountered inland in 2013, 2015, and 2016. Anomalous groups of limestone boulders were noted between Bumber Well Pond and Red Pond (Fig. 3D). More coral clasts were mapped near Soldier Point, Cooper Rock, and Table Bay. Radiocarbon ages were obtained on two of the most inland of the brain coral heads near Table Bay, to check for the possibility that age would vary with distance from shore.

The mapping in 2015 and 2016 took advantage of newly acquired lidar topography and expanded in the variety of clast types that the entire field parties sought. The prior mapping focused on the easily recognized clasts of brain coral. The mapping on lidar located D. strigosa (Fig. A5) among other coral clast types, some of which can superficially resemble clasts of the Pleistocene limestone: the boulder star coral Orbicella (formerly Montastraea) annularis (Fig. A6), the mustard hill coral P. astreoides (Fig. A7), and the elkhorn coral A. palmata (Fig. A8). The recent work also increased attention to the cobble-size valves of the lucine C. orbicularis, encountered chiefly in pits (Fig. A9), and to clasts of reef rock and beach rock (Fig. A10). The mapping in 2015 and 2016 included attempts to keep track of bioclastic sand associated with these various marine clasts, and to identify its approximate inland limit. This work was commonly impeded in eastern Anegada by spiny shrubs and trees, and attempts to determine where the sand is absent were further confounded by difficulties in identifying, in the field, trace amounts of the sand in brown soil that commonly coats the Pleistocene limestone.

Also sought since 2011 were stratigraphic tests of repeated overwash, that is, whether the traces of a pre-Columbian event are juxtaposed with evidence for overwash in 1650–1800. The main result presented in this paper is the stratigraphy correlated among pits along a cross section that passes through the coral boulder exposed west of Red Pond (Figs. B2F). In an additional example, a buried crust is interpreted as the subtle contact between the sand and shell sheet of 1650–1800 and underlying sand that contains articulated lucine valves (Fig. B3F).

The most catastrophic flooding of Anegada in recent millennia has been dated to the last centuries before Columbus by radiocarbon analyses of 27 coral clasts from central and eastern Anegada, 11 radiocarbon ages on mollusks, and 3 ages on plant fragments. All but five of the coral clasts dated are brain coral. Eight lucine valves and three conch shells were dated from sites near Red Pond and Bumber Well Pond. The dated plant fragments were a twig west of Red Pond and wood associated with a brain coral boulder in the Warner area (Figs. 5, 6, and B2–B6).

Still other radiocarbon ages were measured on staghorn coral pieces from near Cow Wreck High Point (Figs. 5, 6, and B1). These ages (as interpreted “Breaches in Beach Ridges”) provide evidence that some of the breaches in western Anegada were first cut after Columbus (after 1492) by extreme waves later than those inferred from most of the field evidence in Figure 4. This post-Columbian overwash is speculatively ascribed herein (“Inferred History of Extreme Waves”) to the 1755 Lisbon tsunami.

Radiocarbon measurements on marine shells and coral skeletons require adjustment for the ¹⁴C age of the water from which the calcium carbonate was precipitated. In the northeast Caribbean, this reservoir correction is complicated by multiyear changes in the mixing of water masses of differing ¹⁴C age (Kilbourne et al., 2007).

We used a local marine-reservoir adjustment, ΔR, of –50 ± 50 ¹⁴C yr (one standard deviation), where ΔR is the age offset between local and modeled global reservoir ages of near-surface ocean waters (Stuiver and Brazunias, 1993). The negative sign in ΔR means water more enriched in ¹⁴C than in the global model. Constraints on ΔR near Anegada were obtained by Kilbourne et al. (2007) from a coral colony offshore southwestern Puerto Rico; their results for the 1950 growth band correspond to ΔR of –27 ± 24 ¹⁴C yr, according to the marine-reservoir compilation of Stuiver et al. (2016).

Data obtained by Kilbourne et al. (2007) from earlier growth bands, back to the year 1751, suggest variable mixing of two different water masses: upwelled equatorial water that is relatively depleted in ¹⁴C, and subtropical surface water that is relatively enriched in ¹⁴C. The results further suggest that Caribbean waters in 1950 were toward the depleted end of the range. Using ΔR = –50 ± 50 ¹⁴C yr allows for a broad range of local marine-reservoir values that is centered on slight enrichment in ¹⁴C relative to the global marine-reservoir model.

Contamination of four of the brain coral samples was checked by analysis of calcite content. Because the coral skeletons are precipitated as aragonite, the calcite is diagenetic and, at Anegada, may contain old carbon derived from Pleistocene deposits. The measured values, with respect to total calcium carbonate, are 0.66% in sample 24 and 1.23% sample 22, both east of Warner (Fig. A7); 1.53% for sample 2011–1, the solitary boulder west of Red Pond (Fig. A1); and 1.92% for sample 14, south of Soldier Point.

According to Olsson (1974), contamination of a young sample by 1% carbon of infinite age increases the sample’s age by 80 ¹⁴C yr, and for 2% contamination the increase is 160 ¹⁴C yr. The contamination evidenced by calcite may have thus increased the age of sample 24 by as much as 50 ¹⁴C yr, and sample 14 by as much as 150 ¹⁴C yr.

Weathered exteriors were present on three of the four dated staghorn samples from the vicinity of Cow Wreck High Point. The weathering rinds were removed before these samples were submitted for dating (Fig. B1E).

Adjusted for marine-reservoir age but not for contamination, radiocarbon ages of the coral and shell samples were converted to ranges in calendar years by means of the Marine13 calibration data of Reimer et al. (2013) and version 7.1 calibration software of Stuiver et al. (2015). The radiocarbon ages of a twig west of Red Pond (Fig. A1F) and of wood east of Warner (Fig. B6D) were converted with the IntCal13 calibration of Reimer et al. (2013) and the same calibration software. Details by sample are listed in Table 2.

Allochthonous clasts ranging from sand to boulders extend inland from the north and northeast shore of Anegada (Fig. 4). Most of these deposits terminate on Pleistocene uplands within 1 km of the shore, but some approach the south shore by way of paths near sea level through Red Pond, Bumber Well Pond, and Budrock Pond. The most widespread of the deposits consists of sand that continuously coats much of low ground near the north shore but becomes difficult to identify as it yields inland to brown soil and bare limestone. This sand, termed the main sand unit because of its extent, is commonly associated with coral boulders and cobbles, and with anomalous limestone clasts.

The main sand sheet consists mainly of fine to medium carbonate sand. Most of the grains are subangular or subrounded. Some are rod-shaped and are speckled with silt-size particles (examples at right in Fig. A4), and perhaps originated as fecal pellets. Under a hand lens, pink grains of Homotrema rubrum are more evident where the sand is thick and well preserved than in inland soils where the grains appear bleached and chalky.

We mapped the main sand sheet as pervasive where sand more than 5 cm thick covers at least three-quarters of an area tens of meters on a side; as patchy where sand is thinner, less extensive, or both; and as scarce or absent where we were unable to identify it with confidence in the field. The patchy sand is commonly mixed with soil and confined to depressions in limestone pavement. The symbols for these three categories appear in Figures 4 and 5, A4–A12, and B2–B7, and as page names in the data tables of the Supplemental Material (^{footnote 2}).

The thickness of the pervasive sand, at the 20 sites where it was recorded, does not exceed 50 cm and is rarely more than half that amount (Supplemental Material). Internal structure is mostly obscured or obliterated by crab burrows (Fig. B3I), aside from three successive normally graded beds at one site (Fig. B3D), and a common overall upward fining of framework grains (Figs. B7C, B7H). Lime mud high in the sand contributes to this upward fining, but the mud may have been introduced from above by bioturbation of pond deposits (Fig. B2H).

Unlike the sand and shell sheet of 1650–1800, which was found chiefly beneath microbial mats of salt ponds at or below sea level, the main sand sheet extends as much as 4 m above sea level (Fig. 6, profiles B-B’ and I-I’). The main sand further differs from the sand and shell sheet by being associated with coral clasts and with conch and lucine shells (Figs. B3G, B3I, B7F, B7G), and in rarely containing shell assemblages dominated by cerithid gastropods. The two sand units differ distinctly in age if the main sand unit is coeval with the 24 associated coral clasts and 11 associated mollusks dated; all of these have given age ranges that extend no later than the year 1500 (Table 2; Fig. 5).

The main sand sheet notably extends onto limestone uplands southwest of Red Pond and in the Warner area (Fig. A4). Southwest of Red Pond it discontinuously coats a limestone hill, known locally as a cay, across a crest 4 m above sea level (Fig. 6; Table 3, profile B-B’). Likewise in the Warner area, pervasive sand extends across a limestone ridge 4 m above sea level (Fig. 6 and Table 3, profile I-I’). In additional Warner examples, rod-shaped grains 1–2 mm long, similar to those evident in the pervasive sand, were found by washing soil samples midway across the island, in an area where sand was judged scarce or absent in the field (photographs in Fig. A4).

Salt flats near the south shore, south of the outlets of Red Pond and Bumber Well Pond, are widely underlain by sand that likely correlates with the main sand sheet. Sand commonly 20–30 cm thick overlies muddy sand that contains abundant cerithids (Figs. A4 and B7E). Coral clasts embedded in the sand (Figs. A4 and A7) include an Orbicella boulder 100 cm by 85 cm by 80 cm (Fig. B7H). Also common are flakes of the microalgae Halimeda (Figs. A4, B7C, and B7E). The sand far exceeds, in its combination of thickness and inland extent, the deposits of storm overwash deposits west of The Settlement (reviewed in “Previous Work”).

This southern sand probably represents a fan built from the north during the time when coral clasts were being scattered widely on the island (Figs. 4–6). This interpretation is supported by the ages of the associated lucine valves (Table 3, profile C-C’) and by limiting maximum ages from coral clasts (Fig. B7B; Table 3, profile B-B’). The probable fan can be explained by the same southward flow of Atlantic water that likely accounts for the numerous breaches and the washover fan northeast of Red Pond, and for a delta built from there into the site of Red Pond (Fig. A2, air photo; Fig. B3B). The water level exceeded 4 m as it deposited patchy sand on a limestone cay south of the main part of Red Pond (Table 3, profile B-B’). It built the probable fan upon emerging and spreading out from constrictions at the south end of what is now Red Pond, and at the southwest end of the trough that now holds Bumber Well Pond.

Inland limits of the main sand sheet are rarely apparent where it tapers southward on limestone east of the eastern half of the island. The patches can be difficult to identify where the grains make up small fractions of brown silty soil. Sand at low elevation along the south shore, south of Warner, may have been derived from southern ponds and beaches, and it may contain salt crystals mistaken for bioclastic sand. Sand in the Warner area and south of Red Pond has been locally reworked by wind. Eolian reworking may have extended the apparent area of the sand after inundation destroyed much of the vegetation, or after land was cleared for agriculture and livestock in the eighteenth and nineteenth centuries. However, we have not noticed widespread sand dunes associated with the main sand except on salt flats south of Red Pond, where drifting sand is being trapped, chiefly by vegetation, southwest of the dated coral clasts and lucines that are plotted in Figure B7A. An eolian cap as much as 10 cm thick may surround a dated P. astreoides boulder in that area (Fig. B7C).

The Supplemental Material lists 225 coral boulders and cobbles that have been found inland of Anegada’s north shore. They are scattered across various kinds of terrain to elevations as great as 4 m. Their collective distribution is approximated, in large part, by their most abundant variety, D. strigosa (Figs. 4, A5, and A12); brain coral makes up nearly two-thirds of the coral clasts mapped. Most of the coral clasts are in fields that extend onto Pleistocene uplands of eastern Anegada. Other clasts, near Red Pond and Bumber Well Pond, are more nearly solitary and were found near sea level (Figs. A12, B2, and B3). Few of the clasts were found more than 0.5 km from the north shore, except in the fan deposits south of Red Pond and Bumber Well Pond. There, the shortest distance to the north shore is 2 km and the inferred transport distances are 3 km (Fig. 6; Table 3; profiles B-B’ and C-C’).

The coral boulders mapped in eastern Anegada extend from Soldier Point to the East End. At Soldier Point, a field of 30 mapped coral boulders and cobbles extends 470 m inland, in stark contrast to the shore-hugging form of the coral rubble ridge (Fig. B4B). This field rests on a pavement of Pleistocene limestone that crests broadly 50–100 m inland at elevations of 4–5 m, and that descends gently landward from there (Fig. 6, profile D-D’). Coral boulders found in eastern Anegada extend farthest inland south of Table Bay, to a distance of 635 m (Fig. 6, profile F-F’; Fig. B5B). The setting there is a relict Pleistocene channel that is aligned with a gap in the modern ridge crest (Fig. 2B). The southeasternmost field identified, east of Warner (Fig. 6, profile D-D’; Fig. B6B), extends 350 m inland and rises from a seasonally flooded salt flat, from which some of the clasts barely protrude (Figs. B6C, B6D).

Most of the D. strigosa clasts are in the range 0.5–1.0 m in diameter; a particularly large one, south of Cooper Rock, spans 2.7 m (Fig. 4). A solitary brain coral boulder 1.7 m in diameter protrudes from a salt flat west of Red Pond (Fig. B2C). Maximum clast size diminishes with distance from the north shore and with elevation, but the many small clasts were found in the north and on low ground (Fig. A12).

Many of the brain coral clasts, including the large one west of Red Pond, resemble entire living colonies in having smoothly convex outer surfaces parallel to annual growth bands (Fig. A5). Some of these surfaces are dimpled where the coral colony probably grew around a worm tube (Figs. B2D and B5G), a skeletal response described by Patton (1976, p. 6). Some of the largest colonies have fractured into multiple pieces, perhaps as a result of centuries of subaerial exposure. The individual pieces remain adjacent to each other and may easily be mentally reassembled to estimate the original size of the colony (Figs. B2D, B5F, and B6C, sample 26).

Irregular surfaces that truncate growth bands give evidence for erosion, mechanical, biological, or both. One such brain coral boulder surface is encrusted with serpulids and Homotrema (Fig. B4D). Others, of cobble size, resemble clasts of the modern coral rubble ridge (samples 19 and 20; Figs. B4C, B4E).

Approximately 33 of the listed coral clasts are boulders or cobbles of the boulder star coral Orbicella annularis (Fig. A6). They include boulders both north and south of Bumber Well Pond, and numerous boulders east of Warner. A dated O. annularis northwest of Table Bay (Fig. B5E, sample 22) was found near a cobble-size piece of Manicina areolata, also dated (Fig. B5E, sample 32). A second dated O. annularis boulder, south of Red Pond and Bumber Well Pond (Figs. B7H–B7K), was found at sea level 2 km from the north shore (Fig. 6; Table 3; profile B-B’).

Of the coral clasts tabulated and mapped, 25 are of the mustard hill coral P. astreoides (Fig. A7). Many more were doubtless overlooked, because the corallites of this species are tiny, and the original form is bumpy. The species predominates, for unknown reasons, on the probable fan south of the outlets of Red Pond and Bumber Well Pond. Closer study of some of those southern clasts might show that they contain additional materials and are better termed reef rock, like a reef rock clast mapped north of Bumber Well Pond (Fig. 10).

Mapped clasts of the elkhorn coral A. palmata extend along the northeast shore and are most numerous near Table Bay (Fig. A8). As with P. astreoides, the clasts are easily overlooked because of superficial resemblance to Pleistocene limestone.

Large molluscan shells were encountered in pits near the western salt ponds. The majority of these are sand-filled valves of the tiger lucine C. orbicularis (Fig. A9). We found them articulated at most of the lucine sites mapped north of Red Pond and Bumber Well Pond (Figs. B2H, B3B, B3D, and B3E), in a pit south of Red Pond (Figs. A9, B7E, B7F, and B7G), and in a pit in the East End. A disarticulated valve was found in sand beneath a limestone boulder at the distal end of a limestone boulder field that forms part of an incised delta in northeast Red Pond (delta, Fig. B3A; pit, Fig. A9).

Conch shells of S. gigas are widespread at the ground surface as well as present in the subsurface. We did not map surficial specimens because many contain holes likely made by people when extracting the muscle. The specimens found underground were entire or broken. An example of an entire specimen was found in a pit 1 m from the coral boulder west of Red Pond, within the depth range of the nearly horizontal bottom of the boulder (Figs. B2G, B2H). This conch shell was associated with articulated C. orbicularis. Another entire conch shell was found in a pit north of Bumber Well Pond (Fig. B3G).

Anomalous limestone clasts are far more extensive on Anegada than implied by the maps in this paper. The mapped examples are limited to fields of coarse clasts, most of them elongate, and to certain clusters of imbricated slabs (Fig. A11). Not plotted or listed are anomalous clasts that might possibly be explained by upturning from tree roots, or by effects of land clearing or wall building.

The elongate fields not mapped previously include several small ones that extend southward from probable sources near the south shore of Red Pond (example at upper left in Fig. A11; identified also along profile B-B’ of Fig. 6), and larger fields evident on air photos of the East End (small example at bottom of Fig. A11). Southward emplacement of the boulder fields near the south shore of Red Pond is congruent with the presence of sand across the crest of the limestone cay farther south (Fig. B2B and profile B-B’ of Fig. 6).

The imbricated clasts mapped are limited to those where numerous slabs dip in a consistent direction. The orientations were measured systematically south of Table Bay among brain coral clasts that include the large round colony shown in Figures A5 and B5F. The mean of the 27 measurements in this area suggests flow directed inland approximately perpendicular to the shore (Figs. B5C, B5D).

The main sand unit was likely derived from sources along and north of the island’s north shore. Between Soldier Point and Warner, a paucity of breaches suggests that the main sand unit was sourced from beach ridges that were leveled by catastrophic overwash and were rebuilt after that. The southern limit of the bioclastic sand may have been extended by wind that swept across overwash areas where plants had been removed or thinned. However, neither lidar nor ground surveys provide evidence for dune morphology in inland areas where the sand is thin and patchy.

As noted in “Fringing Reef and Reef Flat,” all the main species of scattered coral clasts were reported living on the fringing reef or the reef flat off northern Anegada in the 1970s (Dunne and Brown, 1979) and were found live on the reef flat in 2015 or 2016 (Figs. A5–A8). In addition, brain coral colonies that lived through 2008 were found onshore at Soldier Point in 2013 (Xu et al., 2015).

Many of the scattered coral clasts were probably entrained directly from the fringing reef or the reef flat. Clustered ages and exterior form suggest that a large fraction of the brain coral boulders were deposited alive (“Inferred Time of Emplacement”). Had the clasts been derived mainly from a coral rubble ridge like the one now at Soldier Point (Fig. A1), the clast surfaces would be more abundantly bored and encrusted, like the clasts in Figures B4D and B4E, and would not parallel annual growth bands as cleanly as in Figure B2.

The youngest of the dated lucine valves, along with the dated conch shells and large quantities of sand, were probably derived from the reef flat off the island’s north shore. C. orbicularis was living there in 2016 (Fig. A9), as was S. gigas. In the lucine sketch redrawn in Figure A9, the valves in life position underlie sediment one shell diameter thick, or ∼5 cm for the fossil valves pictured. The main sand unit may then contain sand from the uppermost 5–10 cm of the bottom sediments of the reef flat north of the island. This sediment source is particularly extensive north of the Codakia-rich sand deposits that border Red Pond and Bumber Well Pond. Those deposits are due south of the reef flat of Windlass Bight, where sandy shallows extend more than 1 km between the island’s north shore and the fringing reef (Fig. 2A).

Older lucine valves, as discussed in the following, may owe their age to having been recycled from reef-flat deposits beneath beach ridges of the north shore. In this attempted explanation for discordant ages, some of the specimens remained articulated as they were entrained in breaches and incorporated in the washover fan as much as 100 m to their south (Fig. B2B). In the example in Figure B3F, an articulated lucine dated 916–1174 cal yr CE was collected <50 m inland from a plunge pool.

Exposed pavement of Pleistocene limestone provides widespread sources for the associated clusters of limestone boulders between Soldier Point and Table Bay. Bare limestone knolls were likely swept clean by flows that built boulder fields in their lee (Red Pond example, Fig. A11).

Among the 27 radiocarbon ages measured on coral clasts, 23 are between 890 ± 25 and 1280 ± 20 ¹⁴C yr B.P., 18 are between 890 ± 25 and 1050 ± 15 ¹⁴C yr B.P., and none are younger (Table 2). A Manicina areolata cobble gave the outlier age, 1960 ± 40 ¹⁴C yr B.P. (Fig. B5E, sample 32). Age otherwise varies little with coral clast type and appears unrelated to distance inland (Fig. 5B).

The frequency distribution of the ages on coral clasts can be described most simply as unimodal and asymmetrical, with a long tail from ages older than 1050 ¹⁴C yr B.P. (Fig. 5C). This tail likely resulted from reworking of preexisting deposits (“Inferred Time of Emplacement”).

With the estimated marine-reservoir adjustment (ΔR) of –50 ± 50 ¹⁴C yr and the standard calibration methods noted here, 25 of the 27 coral clast ages correspond mainly or entirely to a calibrated age range, at two standard deviations, that is within the round-number interval 1000–1500 cal yr CE (Fig. 5D). A narrower cluster of 18 coral clast ages gives a range of 1200–1480 cal yr CE; the mode corresponds to the last few centuries of pre-Columbian time. The oldest clast, the Manicina areolata cobble, gave a radiocarbon age that corresponds, in calibrated years, to a time no earlier than 200 cal yr CE.

The conch S. gigas, with one sample from each of three different sites, gave ages more consistent with one another, and with the youngest coral clast ages, than did eight shells of the lucine C. orbicularis (Fig. 5B). Six of the eight lucine ages were measured on articulated valves. The scatter among the ages on articulated valves, between ca. 1000 ¹⁴C yr B.P. and 1600 ¹⁴C yr B.P. (Table 2), was unexpected because it is thought that an articulated bivalve in an extreme-wave deposit is usually transported while muscles of the living organism still hold the two valves together (Reinhardt et al., 2006; Switzer et al., 2011, p. 424). The discordance among lucine ages is greatest close to breaches in northern beach ridges (Fig. B3B). Perhaps recycled lucines remained articulated by being derived from nearby excavation of plunge pools, if the pools extended downward into reef-flat deposits beneath beach ridges.

Lucines from sand in a pit south of Red Pond, 2 km from the north shore, gave slightly discordant ages of 1040 ± 25 ¹⁴C yr B.P. (articulated) and 1150 ± 25 ¹⁴C yr B.P. (disarticulated but in a matching pair; Figs. B7E–B7G). These ages figure below in the inference that overwash from the north, after coursing through the sites of Red Pond and Bumber Well Pond, constructed a sandy fan near the south shore and introduced a large Orbicella boulder there as well (Table 3; profiles B-B’ and C-C’).

Radiocarbon ages of plant remains afford partial tests of the ¹⁴C ages of two of the coral boulders. The results are consistent with the estimate herein that the coral boulders were emplaced in 1200–1480 cal yr CE.

The ¹⁴C age of a twig may be consistent with the radiocarbon age of the solitary coral boulder west of Red Pond. The twig was encountered in one of the pits dug along a north-south cross section to explore the stratigraphic context of the coral boulder (Fig. B2F). The twig was deposited in sand 170 m north of the boulder. This sand projects beneath the nearly horizontal bottom of the boulder, an upright brain coral boulder associated with the problematic lucine discussed herein. The twig, as detritus, should predate deposition of the sand, as well as deposition of the coral boulder and the lucine. The twig age, corresponding to 1169–1269 cal yr CE, is older than the coral age, 1294–1458 cal yr CE, and it overlaps with the lucine age, 967–1213 cal yr CE.

Radiocarbon ages on a piece of wood may be consistent with the age of one of the dated coral boulders east of Warner. The boulder is a brain coral colony found upside down, protruding from a salt flat, and a sample from this boulder gave an age corresponding to the range 1283–1444 cal yr CE (Fig. B6C, sample 2011–3). Digging beside the boulder revealed a horizon of organic sand, interpreted as a buried soil, that contained a piece of wood (Fig. B6D). The organic sand appeared to project beneath the boulder, but the boulder was not excavated to establish whether its lowermost part rests on the organic horizon or is surrounded by it. Two ages were obtained from the wood, which contained close to 20 annual rings; the ages correspond to 1189–1271 cal yr CE and 1192–1278 cal yr. CE. By stratigraphic superposition, the wood should predate deposition of the coral boulder if the organic horizon extends beneath the coral boulder. The compared ages meet this expectation, notwithstanding the possibility that deposition of the boulder happened some time after the death of the coral colony.

Inland emplacement during a single pre-Columbian event most simply explains the nearly unimodal radiocarbon ages of the scattered coral clasts, the long tail on the side of earlier ages, the life-like form of many of the brain coral boulders, approximate concordance with all three conch ages, and concordance with two of the eight dated lucine valves (Fig. 5B). It can be supposed that a majority of the coral clasts were alive on arrival in 1200–1480 cal yr CE, the range in which two-thirds of the dated coral clasts cluster (Fig. 5C), and that the other clasts were reworked from preexisting deposits.

Emplacement during multiple events is less likely for two reasons. First, the clustered ages probably limit emplacement to a small slice of late Holocene time. All the coral ages predate 1500 cal yr CE, and 23 of the 27 coral ages at two standard deviations postdate 1000 cal yr CE, as do all three of the conch ages (Table 2). Second, it is difficult to explain this clustering by supposing that Anegada lacked a fringing reef until as recently as 216–550 cal yr CE, the oldest of the ages obtained (Table 2, sample 32). A reef this young would be an anomaly, compared with Caribbean localities where reefs have been present throughout the late Holocene. These localities include Martinique and the U.S. Virgin Islands, as judged from radiocarbon ages on A. palmata (Toscano and Macintyre, 2003, p. 263).

This narrow age range notwithstanding, the sites of the scattered coral clasts were probably close to their present elevation, with respect to sea level, not just at the time of deposition but also for thousands of years before. Regionally, relative sea level has risen approximately one meter per millennium in the past few thousand years (Toscano and Macintyre, 2003). At Anegada, the absence of stranded Holocene shoreline notches (“Holocene Beach Ridges and Ponds” discussion; Fig. 2B) implies negligible net emergence in the centuries since demise of the bygone marine pond. The apparent absence of higher notches tends to rule out net emergence across thousands of years as well.

The net uplift at Anegada has also been small on a longer time scale. As noted in the discussion “Pleistocene Deposits,” a Pleistocene coral sample collected a few meters above present sea level gave a uranium-series age close to 120,000 yr B.P. (120 ka). Subsequent net uplift is not required if the coral lived in the last centuries before it came to rest, if it was deposited no more than a few meters below ambient sea level, and if this sea level was close to the last interglacial maximum. Dated corals elsewhere show that global sea level exceeded present levels for much of the time between 125 ka and 115 ka, and that it peaked ca. 124 ka in the range 6–10 m above present levels (Kopp et al., 2009). Examples near Anegada include those reported from Florida (Muhs et al., 2011) and St. Croix (Toscano et al., 2012).

The scattered coral clasts probably came to rest about as far inland as they are today. The greatest late Holocene changes in shoreline position at Anegada are evidenced by the beach ridges farther west, near Flamingo Pond. Northeast Anegada provides little room for late Holocene shoreline change between the Pleistocene limestone and the fringing reef. Much the same holds for the area north of Red Pond, where limestone crops out locally near the north shore. The only coral rubble ridge recognized, whether in the field or on lidar topography, is the one that adjoins the modern shoreline at Soldier Wash.

The scattered coral clasts in northeast Anegada, from Soldier Point southeastward to the vicinity of Warner, were likely transported through trees and shrubs, or through the freshly mowed-down remains of those plants. The plant cover was probably thinner and shorter within a few hundred meters of shore than it was farther inland, as judged from modern vegetation encountered during mapping south of Soldier Point and Table Bay (Figs. A3B and A5B).

The few brain coral clasts encountered to the west, north of Bumber Well Pond and west of Red Pond, likely encountered little vegetation during transport unless they were saltated across the crests and swales of shrubby beach ridges. These coral clasts crossed over or through beach ridges of the north shore before halting in the bygone marine pond. A longer version of this journey likely emplaced the Orbicella and Porites clasts dated from south of Red Pond and Bumber Well Pond. They faced little impediment from vegetation, and they may have transited a freshly cleaned limestone floor of the trough that now holds Bumber Well Pond (along profile C-C’ of Fig. 6; Table 3).

Anegada’s geological evidence for extreme waves can be grouped most simply into the three categories illustrated in Figures 3B, 3C, and 4B. The groups differ in typical landscape position; in sedimentary architecture; in the size, composition, and likely sources of clasts; and in age. The three groups, named in the following subheadings, connote a hierarchy of hazards. At the lowest rank are hurricanes like those of recent decades. An intermediate hazard is represented by what was perhaps the Lisbon tsunami of 1755. Both these hazards are dwarfed by the catastrophe in 1200–1480 cal yr CE.

Extreme-wave deposits of the lowest rank were formed or at least refreshed by modern storms. The coarsest are organized into a single coral rubble ridge a few tens of meters wide that hugs the island’s north shore. Its limited inland extent is consistent with a physiographic setting that limits routine storm effects: storm waves break on a fringing reef, and storm surge sinks into deep water beyond the reef. Extensive deposits of modern storms at Anegada have been limited to microbial debris. This debris contrasts with the sand and shell sheet of 1650–1800 cal yr CE, and even more with effects of the overwash in 1200–1480 cal yr CE.

The south-directed overwash responsible for the sand and shell sheet poured over and through beach ridges north of the western salt ponds. A pink bioclastic facies was probably derived from the northern beach ridges, whether by reaming out of breaches that had been cut in 1200–1450, or by cutting of new breaches. The flows probably reworked parts of the lowest of the limestone boulder fields north of Bumber Well Pond.

The sand and shell sheet of 1650–1800 might represent the 1755 Lisbon tsunami. In a simulation for Anegada, far-traveled waves of the Lisbon tsunami rise and fall no more than 2 m and have periods of ∼20 min (Buckley et al., 2012). Such waves may have risen gradually against beach ridges of the north shore without transporting coarse clasts from the reef or the reef flat. The simulated wave period could be compared further with observations of the 1755 tsunami elsewhere in the Caribbean (Roger et al., 2010).

In any case, the mainly interior source of the sand and shell sheet of 1650–1800 provides local calibration of how a large tsunami of Lisbon source would register geologically at Anegada. Although the sheet represents Anegada flooding more extreme than any other since 1650, this flooding failed to introduce coarse clasts freshly derived from the fringing reef or the reef flat. The 1755 Lisbon tsunami thus failed to replicate the pre-Columbian emplacement of coral boulders and cobbles at Anegada. Those coral clasts were not emplaced by a prior tsunami of remote origin unless its far-traveled waves somehow managed to dwarf those of the 1755 Lisbon tsunami at Anegada.

This reasoning leaves either a tsunami of nearby origin or an unusual storm that produced tsunami-like bores as the cause of the 1200–1480 catastrophe.

Anegada’s greatest overwash in recent millennia occurred in the thirteenth, fourteenth, or fifteenth century, probably in a single event. Examples of the inundation heights and distances attained are plotted in blue in Figure 6 and elaborated in Table 3. These examples set minima that might be used to evaluate simulations of tsunamis generated along the Puerto Rico Trench and of tsunami-like bores from an unusual storm.

The pre-Columbian event is evidenced by a great variety of deposits: the main sand sheet, its associated coral clasts and large mollusks, and inland fields of limestone boulders and cobbles. Some of these deposits were built into a marine pond that would later, after 1650, become hypersaline; others extended hundreds of meters inland across limestone uplands. The scattered coral clasts and associated molluscan shells demonstrate entrainment from the fringing reef, the reef flat, or both. Bioclastic sand was also deposited inland from the northeast shore, where it may have been derived through removal of ancestors to the beach ridges shaded yellow along the eastern profiles in Figure 6. In the west, beach ridges were abundantly breached, and a sandy fan with coral clasts and lucine valves was built near the south shore, probably by flows that crossed the island from north to south that attained heights of 4 m or more midway along this transit. A time window of 1200–1480 is indicated by the nearly unimodal, consistently pre-Columbian ages of the brain coral clasts and conch shells, and by the youngest of the lucine ages. This time window likely applies to initial emplacement of the inland fields of scattered limestone boulders and cobbles. Had those fields been inherited from prior overwash thousands of years ago, coral clasts should have given such earlier ages as well.

This geological catastrophe was either a tsunami of nearby origin or a tsunami-like effect of an unusual storm. Its effects at Anegada exceeded those (if any) of the far-traveled waves of the 1755 Lisbon tsunami, or of any other hurricane or tsunami in the past 500 yr or more. A tsunami of nearby origin, generated during normal faulting or thrust faulting along the eastern Puerto Rico Trench, was previously invoked to explain the fields of limestone boulders north of Bumber Well Pond. If instead produced by an unusual storm, the overwash in 1200–1480 probably required bores like those generated in the Philippines by 2013 Typhoon Haiyan at a fringing reef that adjoins deep water (“Tsunami-Like Bores Analogous to Those from Typhoon Haiyan”).

In modeling these two alternatives, the burden of proof will rest more on an unusual storm than on a tsunami of nearby origin. A large enough tsunami of nearby origin, particularly from normal faulting along the Puerto Rico Trench, has been shown to be capable of catastrophic inundation at Anegada (Buckley et al., 2012). It remains to be determined, from modeling calibrated to Typhoon Haiyan, whether tsunami-like bores from an unusual storm can produce this much inundation as well, not just where a reef flat is conspicuous (Fig. 6, profiles A-A’, C-C’, I-I’), but also where it is narrow or interrupted (profiles D-D’, F-F’).

Inland deposits ranging from sand to boulders rank highest in a hierarchy of extreme-wave effects on a low-lying Caribbean island near the Puerto Rico Trench. These deposits were scattered hundreds of meters beyond the island’s storm-built rubble ridge. The deposits also exceed the local geological effects, if any, of a tsunami of remote origin. They include coral clasts that were probably emplaced during the last centuries before the arrival of Columbus. The catastrophe far surpassed any other at the island in recent millennia. If not an extraordinary storm, the catastrophe was likely a tsunami generated by thrust or normal faulting along the Puerto Rico Trench. In either case the inland fields of scattered coral clasts attest to a Caribbean and North Atlantic hazard unknown from either satellite geodesy or written history.

Permissions were arranged by the Department of Disaster Management of the Government of the British Virgin Islands. Residents of Anegada extended many courtesies to the field parties. Quantum Spatial and Wayne Wright made the terrestrial and shallow-water lidar surveys, respectively, with geodetic help from Alberto López-Venegas. Craig Weaver arranged for much of the financial support for the terrestrial survey. Brian Andrews and Harvey Greenberg merged the lidar results.

The paper was improved through a series of insightful reviews. A short version of this manuscript, prepared in 2015, was reviewed by Rónadh Cox, Bruce Jaffe, Jim O’Connor, Raphaël Paris, Patrick Wassmer, and an anonymous referee for Geology (Geological Society of America). A longer version was reviewed first by Andrew Kennedy and later, for Geosphere, by an anonymous referee and Adam Switzer. Two reviews were provided by Steve Hickman.