Anatomy of the Holocene inundation of an isolated carbonate platform: Bermuda North Lagoon, western Atlantic

The detailed Holocene inundation history of the Bermuda North Lagoon may be used as model for transgressive and highstand sequences in carbonate platforms. Sedimentation and facies development were controlled largely by sea‐level rise and antecedent topography. Four late Pleistocene to Holocene sequences may be identified in North Lagoon based on a combined analysis of 200 km shallow reflection seismics and 39 cores including 29 radiometric and U/Th‐ages. The sequences were deposited during sea‐level highstands and are separated by subaerial exposure horizons that formed during sea‐level lowstands. Sequence 1 (inferred MIS 7) consists of well‐cemented carbonate sands. Sequence 2 (MIS 5) is up to 20 m thick and consists of well‐sorted, inter‐reefal sands and reef sediments with mound‐like structures. Sequence 3 (inferred MIS 3) is up to ca 6 m thick and accumulated in topographic lows of the underlying sequences some 20 m below modern sea‐level. Sequence 4 (MIS 1, Holocene) includes lagoonal sediments up to 10 m thick, and reefs that accumulated on topographic highs of the MIS 5 sequences. Holocene sediments in topographic lows include peat, peaty sediment, freshwater mud, restricted marine carbonates, and open lagoonal carbonate sediments deposited in seagrass beds, shallow water, and deeper lagoon areas. Upward fining is an expression of deepening and the development of a reef‐protected lagoon environment. Holocene sedimentation on topographic highs usually lacks freshwater and transitional facies and starts with shallow marine mollusc shell accumulations overlain by carbonate sediments that show fining upward. Packstone (68%), wackestone (22%), grainstone (9%) and mudstone (1%) textures occur in cores, with Halimeda, molluscs, coralline algae and foraminifera being the most common constituent particles; coral fragments are rare. During the Holocene, an estimated volume of 1 km3 of carbonate sediments was deposited in North Lagoon. Average sedimentation rates are estimated to be 0.32 m/kyr.


| INTRODUCTION
There are only a limited number of core studies investigating Holocene lagoons of atolls and oceanic (Darwinian) barrier reefs. Marshall and Davies (1982) and Tudhope (1989) reconstructed the growth of small atoll-like reefs (One Tree, Davies reefs) in the southern and central Great Barrier Reef of Australia, respectively. Both lagoonal background sedimentation and lateral accretion, i.e., lagoonward progradation of sediment is crucial during Holocene lagoonal infilling according to these studies. Smithers, Woodroffe, McLean, and Wallensky (1992) also used cores to identify lagoonal infill of Cocos (Keeling) atoll in the eastern Indian Ocean. Cores were taken in marginal lagoon areas, and lateral sand apron progradation was also underlined as an important mechanism of lagoonal infill. Models of Holocene development of Glovers, Lighthouse and Turneffe atoll lagoons in Belize (Gischler, 2003) and Rasdhoo Atoll, Maldives  are based on the sedimentological and chronological analysis of vibrocores. Typical successions included basal late Pleistocene soil, early Holocene mangrove peat, and overlying deepening-upward successions of carbonate sands, silts, and muds. Apart from vertical accretion (aggradation) of lagoonal deposits, lateral, progradational filling also appears to be of great significance in these systems in general (Isaack & Gischler, 2017;Purdy & Gischler, 2005). Still, the great majority of large atolls and oceanic barrier reefs are unfilled buckets, i.e., carbonate platforms with unfilled accommodation space (Purdy & Gischler, 2005). Even flat-topped banks such as Great Bahama Bank are to a large part (>80%) characterized by unfilled accommodation space (Boss & Rasmussen, 1995;Harris, Purkis, Ellis, Swart, & Reijmer, 2015;Purkis & Harris, 2016;Purkis, Rowlands, & Kerr, 2015). Only smaller platform systems appear to be characterized by significant lagoonal infill and shoaling (Gischler, Storz, & Schmitt, 2014;Purdy & Gischler, 2005). These findings have important implications for analysing carbonate platforms in the fossil record, especially for sequence stratigraphy and cyclostratigraphy where sea-level sequences are supposed to be characterized by filled accommodation space, i.e., where cycle thickness is supposed to approximately equal sea-level variation. In addition to sedimentological projects, detailed foraminiferal studies with atoll lagoon cores were made in the Marshall Islands (Yamano, Kayanne, Matsuda, & Tsuji, 2002), offshore Belize (Schultz, Gischler, & Oschmann, 2010), and in the Maldives (Storz, Gischler, Parker, & Klostermann, 2014). Coarse-grained layers within fine background sediments of atoll and oceanic barrier reef lagoons were used to identify echinoid die-off events in Belize (Gischler, 2010), tsunami sedimentation in the Maldives (Klostermann, Gischler, Storz, & Hudson, 2014) and tempestite deposition in French Polynesia (Toomey, Donnelly, & Woodruff, 2013).
Only a few studies have combined coring and shallow seismics to reconstruct atoll-like and oceanic barrier reef development during the Holocene. They include Heron Reef, Australia (Smith, Frankel, & Jell, 1998), Mayotte, Indian Ocean (Zinke, Reijmer, Taviani, Dullo, & Thomassin, 2005;Zinke, Reijmer, & Thomassin, 2001, 2003Zinke, Reijmer, Thomassin, Dullo, et al., 2003), as well as Bora Bora (Isaack et al., 2016) and Tahaa (Toomey, Woodruff, Donnelly, Ashton, & Perron, 2016;Toomey et al., 2013), French Polynesia. All these examples of combined core-seismic approaches stem from the Indo-Pacific realm, where transgressive-regressive sea-level development dominated in the Holocene. Only in the mixed-carbonate-siliciclastic barrier reef example of Mayotte, was the complete palaeoecological and sedimentological succession presented in great detail (Zinke et al., 2005), especially during the early Holocene initial inundation. Clearly, more studies are needed especially from atolls and from reefs in the Atlantic to evaluate the response of isolated reef systems to sealevel and environmental change. In the western Atlantic, Holocene sea-level development is different from the Indo-Pacific, and characterized by transgressive curves in which a mid-late Holocene highstand is missing. To fill in this gap of knowledge, a combined seismic and core approach has been used in an atoll-like lagoon system from the western Atlantic. The use of both shallow cores, which include complete Holocene sediment successions, and reflection seismics allows sedimentation patterns in the late Quaternary to be reconstructed in three dimensions. (1977) interpreted the Pleistocene elevation below the marginal reef to be a result of formerly longshore current transport and storm redeposition of sand. Seaward of the marginal reef, there is an extensive reef front terrace up to 6 km wide at ca 20 m depth, which is locally bounded at the outer margin by a relict reef or sandy ridge (Stanley & Swift, 1967, 1968. The adjacent sloping fore-reef shows another terrace at 65 m depth. Rhodolith pavements have been found on the fore-reef slopes at depths of 40-50 m (Focke & Gebelein, 1978;Fricke & Meischner, 1985). At the southern margin of the Bermuda platform, seaward of the islands, abundant red algal ("boiler") reefs occur in the littoral zone (Ginsburg & Schroeder, 1973). These reefs are mushroom-shaped, and fire coral (Millepora sp.) and vermetids are found besides massive accumulations of coralline algae. The fore-reef slope is characterized by reef growth to ca 50-70 m depth and submarine terraces at 8 m, 20 m and 35 m depth (Fricke & Meischner, 1985;Meischner & Meischner, 1977). North Lagoon reaches depths of 20 m and contains numerous patch reefs, the ecology and sedimentology of which have been investigated by Garrett et al. (1971) and Jordan (1973). Lagoonal patch reefs have various shapes including pinnacles, knobs, linear and cellular (reticulate) types. Coral reefs are also found in the inshore waters such as at Castle Harbour (Dryer & Logan, 1978). Modern sediment studies have been conducted in the south-eastern part of the platform (Todd, 1939), among patch reefs (Garrett et al., 1971;Jordan, 1973), in inshore waters (Dryer & Logan, 1978;Neumann, 1963), and along the shore and in small embayments of the large islands (Sepkoski, 1971;Upchurch, 1970). Bosellini and Ginsburg (1971) coined the term rhodolith based on a study of red algal nodule formation in a bay of Bermuda. Based on the analysis of 200 samples, Upchurch (1970) investigated surface sediment distribution of the entire Bermuda platform. Two major facies may be distinguished including (i) reefal, coarse to fine sands with fragments of corals, coralline algae, foraminifera such as Homotrema rubrum, and gastropods as well as (ii) lagoonal, medium sand to mud with Halimeda and mollusc fragments.
The Bermuda volcanic pedestal and two other volcanic edifices, Argus (Plantagenet) and Challenger banks, which are located to the south-west and reach depths of 50 m, form the Bermuda Rise ( Figure 1). Three deep drill cores on the Bermuda islands penetrated the carbonate cap and reached the underlying volcanic deposits. The oldest borehole in the south-west of the Bermuda islands reached volcanic material ca 74 m below sea-level; the base of the overlying limestone was biostratigraphically dated to Eocene/lower Oligocene (Pirsson, 1914). The volcanic basement was subsequently K/Ar-dated to 34 Ma (Gees, 1969). Seismics also showed the base of the carbonate units at −75 m across the platform (Officer, Ewing, & Wuenschel, 1952). Core holes in the north-eastern part of the Bermuda islands reached the volcanic pedestal some 35 m below sea-level. The absolute (K/Ar) age of lamprophyric sheets was 33 Ma and older tholeiitic lavas some 110 Ma (Reynolds & Aumento, 1974). Seismics detected the limestone-volcanic boundary 30-90 m below the inshore water of Castle Harbor (Gees & Medioli, 1970a), while the results from the inshore water of Harrington Sound were inconclusive (Gees & Medioli, 1970b). The Bermuda platform is considered rather stable with 100-200 m of subsidence to have occurred during the past 25 Myr, and to only 0.6-1.2 m during the last 100 kyr (Vacher & Rowe, 1997).
The geology of the islands of Bermuda is well-known and has been repeatedly reviewed (Hearty, 2002;Land, MacKenzie, & Gould, 1967;Vacher & Rowe, 1997; and references therein). There are six major Pleistocene carbonate units, including coastal-terrestrial and coastal-marine deposits, separated by palaeosols. The carbonate units were deposited during MISs 27/35?, 11, 9, 7, 5, and 1 (Hearty, 2002;Rowe & Bristow, 2015; and references therein). Late Pleistocene deposits have been used to detail late Pleistocene sea-level fluctuations (Harmon et al., 1983;Meischner, Vollbrecht, & Wehmeyer, 1995;Hearty & Kindler, 1995;Muhs, Simmons, & Steinke, 2002). Hearty, Kindler, Cheng, and Edwards (1999), Kindler and Hearty (2000), and Olson and Hearty (2009) presented outcrop and U/Th-age data from Bermuda and the Bahamas suggesting that sea-level stood some 20-22 m above the present level during MIS 11. McMurtry et al. (2007) interpreted these supposedly intertidal deposits as a tsunamite, a hypothesis subsequently challenged by Hearty and Olson (2008). Raymo and Mitrovica (2012) also found the 20-22 m value inconsistent with MIS 11 sea-level and suggested a range of ca 6-13 m above present level based on glacial isostatic adjustment (GIA) models. Recently, Rowe, Wainer, Bristow, and Thomas (2014) estimated the MIS 7 highstand to some 6 m above present level based on dated coral fragments from Bermuda, however, redeposition may not be excluded. Sea-level data from Bermuda together with those from other locations were used by Hearty, Hollin, Neumann, O'Leary, and McCulloch (2007) to detail sea-level during MIS 5e. Hearty and Tormey (2017) suggested a 6-9 m higher-than-present sea-level including the effects of strong storm activity during the late MIS 5e based on the occurrence of even higher run-up deposits.
The analysis of peats and peaty sediments of inshore waters, ponds, and peat marshes resulted in the reconstruction of local, transgressive Holocene sea-level curves (Kuhn, 1984;Neumann, 1971;Vollbrecht, 1996).
The warm climate around Bermuda is due to the influence of the Gulf Stream, which allows tropical coral reef growth at this comparatively high latitude. Sea surface temperatures around Bermuda range from 19.5 to 28°C (Kuhn, 1984). The range within North Lagoon is 16 to 29°C (Kuhn, 1984). Oceanic sea surface salinities in the area are around 36.5‰; in North Lagoon, salinity ranges from 36.2‰ to 36.9‰ (Morris, Barnes, Brown, & Markham, 1977). The area is microtidal with a tidal range of 0.75 m on average (Morris et al., 1977). Strong tidal currents have been observed at channels through the northern reef (Garrett & Scoffin, 1977). Tidal currents cause a relatively quick exchange of lagoon waters, i.e., the residence time of waters in North Lagoon is estimated to 4.2 days (Morris et al., 1977). Bermuda is located in the westerlies belt, and winds mostly blow from the south-west although winterstorms usually approach from west to north-west. Average precipitation amounts to 146 cm/yr (Vacher & Rowe, 1997).

| METHODS
Seismic profiling, vibracoring and sediment sampling for this study took place in 1979 and in 1980 during field work, organized by Dieter Meischner (University of Göttingen, Germany). The second author took part in the field work and subsequently analysed seismic and core data for his dissertation (Kuhn, 1984).
Shallow seismics were recorded in August and September 1979 on a 11.4 m long boat (RV Micmac) using a boomer system. It consisted of a Uniboom model 230-1, an energy transformer EG&G model 234, a hydrophone EG&G model 265 and an EPC model 3200 graphic recorder. Fifteen traverses radiating away from the Bermuda islands, two transverse sections across North Lagoon, and two traverses parallel to the islands and the northern reef margin were recorded, respectively ( Figure 3). The total length of seismic sections amounts to ca 200 km. The reflection seismic data were not stored digitally. Best results were obtained with boat speeds ranging from 3 to 5 knots. In North Lagoon, 4 impulses per second and 16 sweeps were commonly used, which produced a vertical resolution to 0.2 m. In greater water depths exceeding 100 m at the platform margin, hardly any reflected signals could be recorded. Likewise, in shallow-water depths (<3 m), the quality of the recordings decreased. The penetration depth of the seismic system varied depending on the sediment/rock type; in the marginal parts of North Lagoon 15 m were not uncommon. The seismic recordings show the sediment thickness as a function of the sonic wave travelling time. Because of the distance between source and hydrophone (ca 10 m), the depth recording is not linear. The deviation increases with decreasing water depth.
To determine the depth of the sediment surface, a scale for the printouts was made for the most often used frequency (16 sweeps). Sonic speeds in water, sediments, and rocks were estimated. A sonic velocity of 1,540 m/sec was assumed for water at 36.5‰ and 26°C (Smith, 1974). For the unconsolidated Holocene sediments, a sonic velocity 10% higher than that in water was assumed. In hardened Pleistocene sands and clays, sonic velocities as high as 2,000 m/sec were used, while the sonic velocity in peat is rather low (ca 1,000 m/sec). These values are in accordance with those reported in earlier studies, e.g., by Enos (1977) and Choi and Ginsburg (1982) and with those in later studies based on petrophysical analyses of deep drill cores from the Bahamas (Anselmetti & Eberli, 1993;Anselmetti, Eberli, & Ding, 2000) and south Florida (Anselmetti, von Salis, Cunningham, & Eberli, 1997). Because of these assumptions, the calculated thicknesses of seismic units contain inaccuracies. However, even though no exact sonic velocities were measured, sediment thicknesses as measured in cores allow conclusions to be drawn from sonic speeds. Reflected seismic signals do not necessarily reflect changes in sediment type, but in physical parameters such as grain size, sorting, porosity, or cementation. The signals are controlled by impedance, which is a function of density and sonic speed. Unwanted signals had to be identified and excluded, and comprise multiple reflections, reverberations ("singing"), side echos and point reflections. The analysis of seismic sequences included the identification of discontinuity surfaces, i.e., surfaces characterized by strong reflectors as well as potential internal structures, determination of outer shapes and the spatial relationships of units. The identified sequences are supposed to match a depositional sequence in core. Interpretations of seismic sections were corrected for boat speed, sonic speed in various materials, and by taking into account water depth soundings and thickness measurements in cores.
Thirty-nine vibrocores were taken with a high-energy pneumatic vibrocorer (Meischner, Torunski, & Kuhn, 1981) utilizing 6 m long, 8.5 cm diameter aluminium from a floating raft ( Figure 3; Table 1) during September and October 1980. There was an average compaction of 25% during coring, which was corrected for in core logs. Also, in the fall of 1980, 18 surface sediment samples from North Lagoon were collected with a Van Veen-grab sampler ( Figure 3; Table 1). In addition, eight surface sediment samples were taken from three beaches along the north shore of the Bermuda islands ( Figure 3). Positioning during sampling was performed by measuring angles between landmarks and nautical signs with a sextant, and by sounding with a compass and an angle prisma overlaid on nautical charts of the Bermuda platform 1 : 6,200 to 1 : 60,000 were used. Measured GPS coordinates for sample stations and seismic lines are not available, because the GPS system was officially implemented only in 1995. At surface sediment sample stations, water depths were measured using an Elac echoscope LAZ 36; depths were checked with the scale on the grab sampler rope. At the core stations, water depths were measured with a chain leadline and an echo sounder (Fahrentholz model V30/150-S-T).
At the home laboratory, aluminium core pipes were opened lengthwise with a saw and the core cut with a knife and subsequently photographed, described, and analysed.
Absolute age data from carbonates and from peat (n = 23) were obtained promptly after core opening by radiocarbon dating (Dr. H. Erlenkeuser, University of Kiel, Germany). These ages were recently calibrated using Beta-Cal 3.18 provided by Beta Analytic Inc. (Miami) and relying on the INTCAL 13 and MARINE13 databases (Ramsey, 2009;Reimer et al., 2013). A Delta R-value of −129 ± 29 has been used for Bermuda (Druffel, 1997) in order to correct for correction of the reservoir effect. Uranium-thorium-ages of Pleistocene corals (n = 6) were provided promptly after core opening by Dr. P. Schwarcz (McMaster University, Hamilton, Ontario, Canada) and by Dr. R. Harmon (Scottish University Research and Reactor Centre, Glasgow, UK) using the alpha-counting method. Prior to age dating, X-ray diffraction was used to identify aragonitic coral samples.

| Seismics
Selected examples of seismic profiles are shown on the Figures S1-S3. Complete, interpreted seismic lines are found on the Figures 4-7. Four seismic sequences may be distinguished, which include three Pleistocene units and one Holocene unit.

| Sequence 1 (inferred MIS 7)
The lower boundary of the oldest sequence is not visible, and the reflection patterns within the sequence are usually diffuse (Figures 4-7). Only a few locations exhibit laterally continuous reflecting horizons. Some of these horizons show erosional truncation at the top of the sequence. The upper boundary of sequence 1 is marked by a clearly recognizable reflector. The depth of sequence 1 below North Lagoon is rather uniform. It dips from the islands (−20 m) towards the northern platform margin (−30 m). Below the reefs, the depth of sequence 1 is hard to follow; however, it may be seen in sandy areas in between the reefs (Figures 4, 5 and 7). In the south-western lagoon area, the depth is somewhat shallower (18-19 m below sea-level). In the central northern part of North Lagoon (Three Hill Shoals), sequence 1 forms a few topographic highs of low relief (max. 3 m) that in cases are overgrown by Holocene reefs ( Figure 5). No absolute age data were obtained from this sequence, however, an MIS 7 age was inferred based on the fact that MIS 5 deposits overlie this unit.  Figure 5) and in a 3-7 km wide area along the northern shore of the Bermuda islands (Spanish Point to Fort St. Catherine) (Figure 7). In these areas, Holocene deposits overlie sequence 1. In large areas of North Lagoon, however, sequence 2 underlies Holocene sediments ( Figures 5-7). The upper boundary of sequence 2 is represented by a strong seismic reflector. Reflection patterns are diffuse in sequence 2, but parallel layering and diverging reflectors may be seen. Some of these layers concordantly overlie sequence 1 or exhibit onlap. In some areas, mound and divergent fill structures occur. At the top, some layers show erosion and render the upper boundary an unconformity. Below the marginal reef, the upper boundary of sequence 2 forms a topographic high that encircles the entire lagoon ( Figures 5 and 6). In places, the top of the sequence reaches 10 m below sea-level where sequence thickness may reach ca 20 m (Figures 4-6). Both at the platform margin and at lagoonal shoals, the Holocene reefs have accreted on topographic highs of sequence 2. In the central parts of North Lagoon, the top of sequence 2 shows little relief (up to 2 m), in back reef areas the relief is higher (up to 6 m). In the northern and north-western part of North Lagoon, topographic highs and lows have elongated shapes. Uranium-series ages of corals from sequence 2 (cores NL 027, NL 037, NL 042) range from 130 to 87.5 kyr BP and indicate deposition during one or more of the sea-level highstands of MIS 5 (Table 2). Relatively large error ranges of ages and possible coral redeposition render precise designations to the individual highstands of MISs 5a, 5c, and 5e difficult.

| Sequence 3 (inferred MIS 3)
Sequence 3 occurs only at a few places in North Lagoon, usually below 15 m depth, where it usually fills in topographic depressions of underlying sequences (Figures 4-7). In central and coastal areas of North Lagoon, sequence 3 overlies sequence 1. The thickness of sequence 3 averages 2-3 m although thicknesses slightly exceeding 6 m are found in back reef areas of the northern lagoon where sequence 3 fills in topographic lows of sequence 2. Due to strong reflectors, this sequence exhibits distinct layering, especially in the western part of North Lagoon ( Figure 6, section J). Layers concordantly overlie the lower boundary or show onlap. In places, diverging layers may be observed. Diffuse reflection patterns are rare. In the northern lagoon, there are regions with strongly inclined reflectors that show offlap. Some reflectors exhibit erosive patterns at the upper sequence boundary. In general, the relief of this sequence is rather low. The sequence usually abates the relief of underlying sequences and forms a relatively flat surface. There is one absolute age date from this sequence of 39.26 kyr cal BP, and together with its stratigraphic position and elevation, deposition during MIS 3 is likely.

| Sequence 4 (MIS 1; Holocene)
The base of sequence 4 is located ca 15 m below sealevel in the central parts of North Lagoon ( Figure 8). It deepens towards the south-west, north and north-east where it reaches depths exceeding 20 m. The base of the sequence rises at the platform margin, where it may reach to 10 m below sea-level (top of sequence 2). At one location (North Rock) at the northern platform margin, Pleistocene limestone reaches several metres above modern sea-level. In places, the base of sequence 4 shows a small scale relief characterized by steep slopes. In the lagoon area, the base of sequence 4 is mostly characterized by a strong seismic reflector (Figures 4-7). The reflection is strongest when soils or peats are developed at the boundary, e.g., in topographic lows, as a consequence of strong impedance. In the central part of North Lagoon, the boundary does not show strong reflections. This is the F I G U R E 4 Interpreted seismic sections from north-eastern part of Bermuda platform. Note the strong relief of the seafloor due to reef development. Corrected for vessel speed and sonic velocity in water GISCHLER AND KUHN case when sediments below and above the boundary have similar textures or where cementation is rather weak. The upper boundary of sequence 4 is the sea floor, which is also characterized by a strong reflector. Some 85% of North Lagoon is represented by a more or less flat lagoon floor. Reflectors within this unit are parallel. The reflectors are slightly wavy in a 1 km wide zone inboard of the marginal reef where Callianassa mounds are very common. Over steep topographic depressions of the underlying sequence, Holocene sediment may reach thicknesses of 10 m. Usually, the thickness of lagoonal sediments of sequence 4 ranges from 1 to 4 m ( Figure 9). In the central part of North Lagoon thicknesses are less, as they are in between patch reefs in the north-eastern part of the lagoon. In other lagoonal inter-reef areas, sediment thickness reaches 3-4 m. The thickness of sequence 4 in inter-reef sand areas of the platform margin amounts to >4 m. Lagoonal shoals are usually located over topographic highs (up to 5 m relief) of underlying sequences ( Figure 5). It is not entirely clear whether patch reefs also accreted over topographic highs of the antecedent topography. Deeper water lagoon reefs were dominated by branched corals such as Oculina and Madracis, and branched Millepora occur in the south-western parts of North Lagoon where their tops occur at depths of 8-13 m. Based on the seismic data in this area, these reefs did not accrete on topographic highs of the underlying Pleistocene sequences ( Figure 6). Coastal fringing reefs, that are found along the northern shore of the Bermuda islands, form only thin veneers on the underlying Pleistocene limestones ( Figure 7). The thickness of sequence 4 at marginal reefs is hard to estimate, because the lower boundary is visible only at sandy depressions where it indicates a reef thickness of 9 m on average. At the reef front terrace, sequence 4 has an average thickness of ca 6 m (Figures 4 and 6). Seventeen absolute age dates from sequence 4 were obtained from peat, peaty sediment, freshwater sediment and a variety of carbonate sediments. Ages range from 11.0 to 2.5 kyr cal BP (Table 2).

| Sands
Pleistocene sands are very abundant and are usually composed of reef detritus, i.e., fragments of coralline algae Lithoclasts are common in core NL 021 ( Figure S6). Sediments are for the most part very well-sorted, unimodal and with an abundance maximum within the medium sand fraction (0.20-0.63 mm). Exceptions exist as, e.g., in core NL 033/2, which is characterized by very poor sorting (Table 3). Pleistocene sands are usually consolidated in contrast to Holocene sediments. The sands show varying degrees of cementation from weak, patchy, to complete cementation. Both marine and phreatic cements have been identified by Vollbrecht (1990). Age and time appears to be a decisive factor. Sequence 1 sands are usually strongly cemented, whereas cementation in sequence 2 is somewhat weaker. Cementation is weak in sequence 3. Bivalve molluscs such as Transenella sp., Linga pensylvanica, Glycimeris sp. and Chione cancellata appear to be characteristic of the Pleistocene deposits in North Lagoon, because they were not found in overlying Holocene deposits. A single absolute age date from a bulk sample of the sand facies in core NL 001/3 of 39.3 kyr cal BP would correlate with MIS 3 while another radiocarbon age from a mollusc shell in core NL 002 produced a date of >42.3 kyr cal BP. U-series dating of corals from cores NL 027, NL 037 and NL 042 produced ages correlating with MISs 5e, 5c and 5a. They range from 130 to 87.5 kyr BP (Figures S8, S11 and S13; Table 2).

| Peat
Pleistocene peat was found only rarely. It consists of unconsolidated plant remains with a relatively high water content. Pleistocene peat, ca 10-50 cm thick, was recovered in cores NL 001/3 and NL 001/5 ( Figure 12). According to radiocarbon dating, it formed some 28.1, 23.8, and 16.9 kyr cal BP (Table 2). Because sea-level was significantly lower than the peat depth in core, the reported ages  , S7 and S12, S13-S16). Their thickness is usually <50 cm with seismic profiles indicating that they preferentially formed in topographic lows. They are primarily composed of clay minerals, Fe and Al oxides, quartz, and phosphates and the carbonate content is low. In contrast to the reddish soils exposed on the Bermuda islands, the Pleistocene soils recovered from cores below sea-level are olive-green and according to X-ray diffraction contain pyrite. Soils also contain organic matter in the form of plant detritus. In the lower portion of the soil profile, carbonate rock fragments may be found indicative of dissolution. The soil in core NL 021 contains strongly weathered volcanic rock ( Figure S6). The uppermost soils at the top of the Pleistocene units probably formed in the early Holocene. Based on the lack of comprehensive absolute age data, the Pleistocene-Holocene boundary is placed at the top of the soil underlying peat and/or unconsolidated (Holocene) carbonate sediments.

| Sedimentology of Holocene deposits
Peat, peaty sediment, freshwater-to-brackish mud and a variety of restricted to fully marine carbonate sediments including reefal successions have been recovered (

| Peat
Holocene peat occurs in topographic lows and was found in cores NL 023, NL 031, NL 039, NL 040, NL 047-49 and NL 051 (Figures 13 and 14; Figures S6, S12, S14 and S16). It usually forms soft, thin layers (ca 5-20 cm thick) of undifferentiated, rarely fibrous plant remains. In the lower parts, vertical plant structures are visible although at the top, these become less visible and transition to peaty sediments occur. Peat age data range from 11.0 to 8.8 kyr cal BP, a peat layer from core NL 049 was dated to 4.4 kyr cal BP ( Figure 14). Today, peat is still forming in marshes located in topographic depressions of the Bermuda islands, i.e., within freshwater lenses of the groundwater. using the INTCAL 13 and MARINE13 databases (Ramsey, 2009;Reimer et al., 2013). For correction of the reservoir effect, a Delta R-value of −129 ± 29 has been used for Bermuda (Druffel, 1997)

| Restricted marine carbonate sediments
Sections with carbonate-rich sediments, which are characterized by much lower contents of organic matter as compared to the freshwater muds, occur in cores NL 023/2, NL 031/4, NL 040, NL 044, NL 047 and NL 048 (Figures 13 and 14; Figures S6 and S14). The sediments are fine-grained (clay, silt) and are mostly very poorly sorted. There are packstones (60%), wackestones (20%) and mudstones (20%). Mollusc shell fragments (9%) and Halimeda platelets (11%) are the most common constituent grains; however, fines (<63 μm) make up >80% (see data for sample in core NL 040 in Table 4). At the base, gastropod shells are common. They include Bulla striata, Modulus modulus, Vermicularia spirata, Batillaria minima, Cerithium lutosum and Cerithiopsis greeni. Bivalve shell abundance increases upcore with common occurrences of Codakia orbiculata, Psammotreta intastriata and Tagelus divisus. The gastropod Astraea phoebia and the bivalve Codakia orbicularis, common in seagrass beds, also occur in higher sections of this facies. Where erosive bases of the marine sediments exist, epifaunal bivalve molluscs that live attached to hard substrate or mangrove roots such as Isognomon sp., Pinctada imbricata and Brachidontes domingensis are found. The shells in the restricted carbonate sediments are usually well-preserved. Gastropods commonly occur with opercula in place, and some bivalves with both shells attached together (Psammotreta intastriata) were apparently preserved in situ. At the top, sediments are often better sorted and contain fragments of Homotrema rubrum, suggesting transport from adjacent higher energy, reefal environments. An age date of 9.1 kyr cal BP was obtained from core NL 031/4 ( Figure 13), while an age from core NL 040 was 8.4 kyr cal BP (Table 2). Modern equivalents to this facies occur in the nearshoremuddy substrate biotope (Upchurch, 1970), i.e., in lowenergy bays and ponds (Osgood, 1970;Walton, 1969).

| Fully marine carbonate sediments
These carbonate sediments include beach sands, shallowwater sands, sediments of seagrass beds, sands of the open lagoon floor, fine sediments of the deep lagoon floor, as well as back reef and reef sediments. Beach sediments are rare in the cores. Only at the base of cores NL 049 and NL 050 (Figure 14; Figure S16), which were collected near the shore, very well-sorted sands with bimodal grain-size distribution, typical for beach sands occur. All samples may be categorized as grainstone (Table 3). The sand from core NL 050 ( Figure S16) has comparatively high amounts of silt and clay, possibly as a result of bioturbation and mixing with overlying sediments. Sandy beaches also exist along the north and south shores of the modern Bermuda islands. Along the northern side facing North Lagoon, sandy beaches occur in open bays preferentially in coastal areas with low relief. Because a precondition for beach development is a relatively stable sea-level, abundant beaches have been forming only since the late Holocene some 4 kyr BP. The analysis of modern beach sediment samples showed two abundance grain-size maxima (1.80-0.63 mm; 0.40-0.16 mm); the sediments are well-sorted. The portion of silt and clay grain sizes is negligible.
Shallow-water sands commonly occur at the base of the Holocene marine carbonate successions. The sands are usually poorly sorted and contain small amounts of fines and poorly rounded clasts of Pleistocene sands as in cores NL 021/1, NL 027-030 and NL 041 ( Figures S6, S8, S9 and S13). Texturally, 70% of the samples are packstones and 30% are grainstones (Table 3). The clasts are interpreted as relics from previously existing rocky shores. Most common constituent grains include Halimeda platelets, fragments of coralline algae, foraminiferal tests and mollusc shell fragments (see data for core NL 034 in Table 4). In cores NL 031/4, NL 037 and NL 044, shallow-water sands form the transition between restricted marine carbonate sediments below and sediments of seagrass beds above (Figure 13; Figures S11 and S14). Due to the proximity of seagrass beds, redeposited shells of Codakia orbicularis and Astraea phoebia are commonly encountered in shallow-water sands.
Sediments of seagrass beds are usually up to 50 cm thick and occur in the lower portions of cores NL 021, NL 031/4, NL 033/2, NL 035, NL 040 and NL 048 (Figures 10c, 11c and 13; Figures S6, S10 and S11). Large mollusc shells are very common and include Codakia orbicularis (6-9 cm long) and Astraea phoebia. Mollusc shell fragments and Halimeda platelets are the most common constituent particles; the fines (<63 μm) reach 40%-50% (see data for two samples from core NL 040 in Table 4). Texturally, 80% of the samples may be characterized as packstone and 20% as wackestone (Table 3). The sediments are mostly poorly sorted and exhibit bimodal grain-size distribution with modes from 2.0 to 0.8 mm and 0.315-0.1 mm. The portion of the >2 mm fraction is high with the large sizes presumably deriving from carbonate producers living in the seagrass beds, while the smaller sizes accumulated due to the baffling effect of and F I G U R E 8 Pre-Holocene topography of the Bermuda platform based on seismic and core data. For calculation of sediment thickness, a sonic velocity 10% higher as compared to water (1,540 m/sec) was used. In addition, depth soundings and thickness measurements in cores were taken into account GISCHLER AND KUHN | 231 carbonate production by encrusters on sea grass (Nelsen & Ginsburg, 1986;Patriquin, 1972). The seagrass beds in the cores do not show the typical features of modern seagrass beds such as fining upward as described by Wanless (1981). Bioturbation and redeposition usually modified the deposits recovered in core at the bases and the tops. Two radiocarbon dates from seagrass sediments amount to 6.4 kyr cal BP (core NL 048; Figure 14) and 7.4 kyr cal BP (core NL 051; Figure S16) ( Table 2). Modern seagrass beds in the North Lagoon are usually found in water depth <10 m with Thalassia testudinum and Syringodium filiforme predominating; Diplanthera sp. occur in protected environments and soft bottoms.
Sands of the open lagoon floor were usually encountered overlying shallow-water sands and sediments of seagrass beds as in cores NL 001, NL 019, NL 020, NL 022, NL 027, NL 030, NL 036, NL 038 and NL 041-044 (Figure 12; Figures S5, S6, S8 and S9, S11-S14). They contain less fine and less coarse material and are better sorted than the seagrass-bed sediments with 80% of the samples identified as packstones and 20 % as grainstones (Table 3). The sediments are often bimodal with modes of 1.25-0.63 mm and 0.40-0.18 mm. They are characterized by strong redeposition. The bivalve (egg cockle) Laevicardium laevigatum is very common. Fragments of the burrowing sea urchin Melitta sp. occur in well-sorted sands. An age date from core NL 040 (Figure 13) was 6.7 kyr cal BP; an age from core NL 051 was 3.9 kyr cal BP (Table 2).
Fine sediments of the deep lagoon floor commonly overlie coarser-grained sediments of the open lagoon floor. Typical characteristics of this transition include a marked decrease in grain sizes >2 mm and an increase in the silt and clay grain-size fractions (<63 μm). Of these core samples, 58% may be characterized as packstone, 39% as wackestone and 3% as mudstone. The fine grain-size fractions of the sediment are presumably derived from the decomposition of codiacean algae such as Penicillus sp. and Halimeda sp. as well as from fine-grained reefal F I G U R E 9 Thickness of Holocene sediments on the Bermuda platform based on seismic and core data. For calculation of sediment thickness, a sonic velocity 10% higher as compared to water (1,540 m/sec) was used. In addition, depth soundings and thickness measurements in cores were taken into account debris. Fragments of mollusc shells and agglutinating worms become more abundant while the abundance of Halimeda platelets decreases (see composition data of cores NL 001 and NL 040 in Table 4). Commonly occurring molluscs include Gouldia cerina, Pitar fulminata and Codakia costata. Shells of the gastropod Finella sp. are common in the grain-size fraction 1.6-0.8 mm. These characteristics are well-developed in cores NL 001, NL 002,  Figures S5, S6, S9, S15 and S16). In the south-western part of North Lagoon, sediments are especially fine-grained (cores NL 033/2, NL 035, NL 037; NL 039; Figures S10-S12). Shells of the vermetid gastropod Vermicularia spirata are abundant. In modern Bermuda, V. spirata lives in shallow-water environments and also colonizes the somewhat deeper Oculina lagoon reefs. In general, both sediments from the core tops and from the lagoon floor may be characterized by coarser grain sizes, which is probably due to either storm redeposition or bioturbation. When applying Dunham-equivalent textures, 47% of the surface sediments of North Lagoon are packstones, 28% are grainstones and 25% are wackestones; i.e., 75% of the lagoonal surface sediments have grain-supported and 25% mud-supported textures (Table 5). Still, the amount of the fine grain-size fractions (<125 μm) of surface sediment and core top samples from North Lagoon (n = 36) shows a statistically significant correlation with water depth, i.e., depositional energy (r 2 = 0.48, p < 0.000).
Lagoon sediments adjacent to reefs (back reef sediments) largely consist of silt and fine sand, which derive from the marginal reef. The material is made up of Halimeda sp. platelets and fragments of Homotrema rubrum, coralline algae (Amphiroa sp.), and alcyonarian spicules. Shells of the gastropod Finella sp. are common. Texture of the redeposited reefal and autochthonous lagoonal sediments is characterized by a polymodal grain distribution. The sediment is well-sorted; grain-size fractions <20 μm and >200 μm are not abundant.
Sediments of the marginal reef are very coarse-grained (abundance maximum ca 1 mm) and comparatively wellsorted (Table 3). The grain size <200 μm is virtually lacking. The majority (70%-75%) of reef and inter-reef sediments from channels and sand pockets are packstone; grainstone abundance reaches 13%-25%. The sediments are composed of fragments of coralline algae, Homotrema rubrum, mollusc shells and corals (Upchurch, 1970). Halimeda platelets are especially common in protected sandy areas. Sediments of lagoonal patch reefs consist of Halimeda, mollusc, coralline algae, corals and Homotrema rubrum, and make up F I G U R E 1 2 Logs of cores NL 001/3 and NL 001/5, including information on correlation to seismic sequence. For legend see Figure 15 30%-50% of the reef framework (Garrett et al., 1971;Jordan, 1973). Cores from lagoonal reef areas such as NL 052 and NL 053 (Figures 11b and 15) contain abundant large fragments of massive corals and higher amounts of the finer grain-size fractions when compared to samples from inter-reef, sandy areas. A radiocarbon date of 2.5 kyr cal BP was obtained from core NL 053 (Table 2) while the reef accretion rate based on this core is estimated to be 1.5 m/kyr. Cementation appears to be very low or lacking. Core NL 033/2 (Figure 11a; Figure S10) was taken in a deeper water

F I G U R E 1 3 Logs of cores NL 031
and NL 040, including information on correlation to seismic sequence. For legend see Figure 15 Oculina-Millepora-Madracis reef structure in the southwestern part of North Lagoon. The tops of this reef type, where branched corals form the reef framework, occur in water depths of 8-13 m. In the core, large fragments of Oculina sp., Millepora alcicornis and Madracis detactis occur in carbonate silts and clays, together with minor amounts of sands. Sediments may be characterized as wackestone. Molluscs include Arca sp., Barbatia domingensis, Chama sp., Vermicularia spirata and Laevicardium laevigatum. Halimeda platelets are common in more sandy, coral-free sections.
Based on measured thicknesses and available age data, sedimentation rates in the Holocene lagoon area are estimated to be 0.32 m/kyr and to be 0.92 m/kyr in the reefal areas.

| Correlation between seismics and cores
The large majority of vibrocores were collected on seismic lines (Figure 3) with the positions of core stations along the interpreted lines indicated in Figures 4-7. Likewise, the designations of seismic sequences are included in the core logs (Figures 12-15; Figures S5-S16). The seismic recordings show the thickness of sedimentary sequences as a function of sonic wave travelling time, however, due to the fact that sonic velocities in sediments, rocks, and water were estimated, thickness had to be corrected by seismic-core correlation. Sequence thicknesses on seismic profiles were estimated by measuring between significant reflectors that are a consequence of impedance differences among various sediments, between sediments and rocks, and between sediments and water. The clearest reflectors were encountered at interfaces between consolidated rocks and unconsolidated sediments, between carbonate sediments and peat and clay, and at the sediment-water interface at the seafloor. Two examples have been selected to show correlations between cores and seismic data (Figure 16). In core NL 041, well-sorted, fine-to-medium grained, partially cemented sand of sequence 2 (Pleistocene) is superimposed by poorly sorted, mediumto-coarse-grained, unconsolidated sand of sequence 1 (Holocene) that contains some lithoclasts (Figure 16a). The boundary of the sequences is defined by a relatively strong reflector. Although the internal reflectors in sequence 2 are overall wavy, reflectors in sequence 1 are planar. In core NL 047, a clayey soil forms the top of sequence 2. It superimposes a well-sorted, weakly cemented, fine-to-medium grained sand. The sequence 2 soil is overlain by medium-to-fine-grained, unconsolidated sand of sequence 4 that contains plant and shell remains and shows a few intercalations of peaty sediment layers (Figure 16b). Two closely-spaced, strong reflectors mark the sharp lithological change interpreted as sequence boundary in this case. Internal reflectors are well visible within sequence 4. In the first example, the strong reflectors at the sequence boundary are largely caused by the impedance created between the consolidated and unconsolidated sands; in the second case impedance contrasts relate to the transitions from carbonate sediments to peaty F I G U R E 1 4 Logs of cores 48 and 49, including information on correlation to seismic sequence. For legend see Figure 15 sediment and to clay. In both examples, the sea floor shows a strong reflector at the sediment-water interface.

| Late Pleistocene development
Sea-level fluctuations and antecedent topography played a decisive role in the development of the Bermuda carbonate platform during the late Quaternary. The late Pleistocene and Holocene carbonate sequences in North Lagoon were deposited during highstands of sea-level. Sequence 1 was presumably formed during MIS 7 when relative mean sea-level in Bermuda stood comparatively high, 4.5-6.0 m above modern level according to Rowe et al. (2014). Timeequivalent deposits of sequence 1 on the Bermuda islands would include the littoral facies of the Belmont Formation whose designation to MIS 7 has been questioned by Hearty (2002). Sequence 2 was likely deposited during MIS 5 based on the available U-series data, which may be used as a geochronological guideline. The age data range from 87.5 to 130 ka, which would include MISs 5e, 5c and 5a. Time-equivalent units on the Bermuda islands include the Rocky Bay and Southampton formations, with sea-levels standing 5 m and 1 m above modern level, respectively Rowe & Bristow, 2015). The dated F I G U R E 1 5 Logs from reefal core NL 052, NL 053/1 and NL 053/2, including information on correlation to seismic sequence. Legend for T A B L E 3 Grain sizes in some cores. Very well-sorted (<0.35), well-sorted (0.35-0.5), moderately well-sorted (0.5-0.71), moderately sorted (0.71-1), poorly sorted (1-2), very poorly sorted (2-4), extremely poorly sorted (>4). Mean grain size and sorting after Folk and Ward (1957 corals stem from cores in reefal deposits; however, it is not entirely clear whether or not they are preserved in situ. They belong to the genera Diploria, Orbicella, and Porites, which occur over wide depth ranges. For these reasons, the U-series coral age data may not be used as an accurate Pleistocene sea-level indicator but as an approximation of geochronology only. Sequence 3 was likely formed during the lower highstand of MIS 3 based on the age date in core NL 001/3 some 20 m below modern sea-level. As this is just one bulk sediment age, it has to be handled with caution. Reported sea-level highs during MIS 3 show quite some variability and range from −60 m to −30 m below modern level (Pico, Creveling, & Mitrovica, 2017); however, Simms, DeWitt, Rodriguez, Lambeck, and Anderson (2009) have suggested a level of as high as −15 m below modern level in the Gulf of Mexico. There are no MIS 3equivalent deposits on the Bermuda islands. Based on Pleistocene sediment composition that clearly shows reefal influence, textures that are usually characterized by good sorting, and the occurrence of bivalve molluscs such as Chione cancellata, it is obvious that the late Pleistocene deposits of the North Lagoon area were formed in more open, less protected, and higher energy environments when compared to the Holocene platform interior sediments. C. cancellata is also very abundant on the shallow northern inner platform along the modern Belize barrier reef (Purdy, Pusey, & Wantland, 1975). Possibly, marginal reefs in Bermuda were not as well-developed and continuous as during the mid-to-late Holocene. A comparable development has been reported from the Florida Reef Tract where the platform margin had a ramp-like character during late Pleistocene sea-level highstands (MIS 11 to MIS 5e). A bank-barrier reef developed only in the latest Pleistocene of MISs 5c and 5a for reasons not entirely clear (Multer, Gischler, Lundberg, Simmons, & Shinn, 2002). Based on the elevation of the sequence 1, MIS 7 sea-level reconstructions, and the fact that subsidence was minimal, water depths of up to 25-35 m may be expected during deposition of the oldest late Pleistocene unit. The depositional environment was presumably more open and less protected but deeper than the modern. The higher elevation of sequence 2 in North Lagoon, the sealevel reconstruction for MIS 5e, and the low subsidence suggests shallower conditions, around 15 m water depth, during deposition of this Pleistocene unit. Again, the platform must have been more open than the modern setting. T A B L E 5 Grain size and sorting in surface sediments. Very well-sorted (<0.35), well-sorted (0.35-0.5), moderately well-sorted (0.5-0.71), moderately sorted (0.71-1), poorly sorted (1-2), very poorly sorted (2-4), extremely poorly sorted (>4). Mean grain size and sorting after Folk and Ward (1957 In case the age of sequence 3 were representative, palaeowater depth over North Lagoon during MIS 3 would have been even shallower than during MIS 5, i.e., not deeper than ca 5 m. Late Quaternary soils of Bermuda formed during lowstands of sea-level, preferentially in topographic lows. Carbonate dissolution influenced the limestone island topography as seen, e.g., in the occurrence of limestone fragments in the basal parts of soils that exhibit clear evidence of dissolution, and in the ragged, steep relief seen on small scales and in the tower-shaped topographic highs along some of the seismic traverses (e.g., Figures 5 and 7), which is indicative of karst. The general saucer-shaped geomorphology of the Pleistocene Bermuda platform also suggests a karst origin, according to Purdy (1974, fig. 42). However, alleged karst geomorphologies such as the principal saucer or bucket shape of coral reefs or reticulate reef distribution patterns, which may be seen in lagoonal patch reefs of Bermuda, may also be the result of biotic selforganization as recently discussed by Schlager and Purkis (2015).

| Holocene development
The Holocene deposits (sequence 4) allow a detailed view into the sedimentological expressions of sea-level rise since ca 11 kyr cal BP (oldest Holocene peat age date) over a karst-influenced carbonate platform. Bermuda is supposed to be a region well-suited to reconstructing sea-level fluctuations, because it is located in an area of minor glacial isostatic adjustment (Clark, Farrell, & Peltier, 1978) and affected by minimal subsidence only (Vacher & Rowe, 1997). In addition, Bermuda has been considered tectonically stable (Harmon et al., 1983;Hearty, 2002;Ludwig, Muhs, Simmons, Halley, & Shinn, 1996;Meischner et al., 1995). However, these views have recently been challenged based on GIA models suggesting postglacial crustal subsidence (Raymo & Mitrovica, 2012) and the fact that evidence for faulting, fracturing and seismic activity exists (Rowe et al., 2014;and references therein).
In Bermuda North Lagoon, Holocene sedimentation is strongly influenced by the morphology of the underlying Pleistocene deposits (Figure 17). Typically, lagoonal carbonate successions are expressions of continuous deepening of the platform interior. In the following, the supposed Holocene development of end members of a topographic low ( Figure 18a) and a topographic high of the underlying pre-Holocene topography are discussed (Figure 18b). In topographic lows of the pre-Holocene surface, Holocene facies and facies successions are quite diverse. In the lows of the karst-influenced surface, carbonate-free soil developed in the late Pleistocene. During sea-level rise in the early phase of the interglacial, groundwater was piling up and peat formed in topographic lows. During further sea-level rise, shallow freshwater lakes formed, in which peaty sediments accumulated. The salinity of lakes gradually rose and freshwater and brackish water carbonates were deposited. The occurrence of Paludestrina bermudensis, which dwells in modern peat marshes and marine ponds (Haas, 1952), is indicative of this phase. The water bodies in topographic lows were still rather small and possibly stratified so that oxygen minimum zones developed in bottom waters. Fine-grained, laminated sediment was deposited. With rising sea-level, more and more marine gastropods appeared in the sediments. The oxygenation of the bottom water continued to improve and marine bivalves colonized the bottom. These include, e.g., Codakia orbiculata, Psammotreta intastriata and Tagelus divisus. For comparison, the latter occurs in modern muds and muddy sands of back bays, swamps and estuaries of Florida Bay (Fraser, 1967). However, low diversities still suggest restricted conditions. During continued sea-level rise, the water bodies on the platform increased in size and former terrestrial bars developed into submarine shoals providing increasingly less protection. As a consequence, marine ponds developed into enclosed bays that became more and more open and subject to increasing depositional energy. The high diversity in mollusc species indicates fully marine conditions in this phase. Sands started to accumulate and seagrass beds developed with occurrences of Codakia orbicularis and Astraea phoebia. Both species are also common in seagrass beds of south Florida (Wanless, 1981) and Belize (Hauser, Oschmann, & Gischler, 2007). With an increase in water depth, the influence of the submarine bars was mitigated and the submarine relief diminished. In open lagoon areas, well-sorted sediments with few fines Halimeda sp.

Syringodium filiforme (ASCHERSON 1868)
Diplanthera sp. Vermicularia knorrii (DESHAYES, 1843) were deposited in which shells of Laevicardium laevigatum are abundant. In greater water depth, hydrodynamic energy was low and finer-grained sediments with shells of Gouldia cerina were deposited. In addition, the development of marginal and lagoonal reefs led to increased protection of the inner platform area. G. cerina is also very common in the low-energy, protected lagoon of the Turneffe Islands atoll of Belize (Hauser et al., 2007). On highs of the pre-Holocene surface, Holocene facies and facies successions are not as diverse as in the topographic lows (Figure 18b). Usually, well-cemented Pleistocene sands or remains of Pleistocene reefs form topographic highs. During inundation by the rising Holocene sea, rocky coastal substrates formed that were characterized by erosion and redeposition. Rocky shores, and, in rare places, beaches developed. The base of Holocene sediments is commonly represented by shallow-water sands with clasts of Pleistocene bedrock and large shells. Besides Codakia orbicularis and Laevicardium laevigatum, gastropod shells such as Astraea phoebia, Modulus modulus, Cerithium sp., Nassarius albus and Columbella mercatoria occur. Brachidontes domingensis, a bivalve of the waveexposed part of the lower tidal zone, is found in a number of cores. Seagrass beds formed in shallow water although sediments of seagrass beds are commonly redeposited, as seen in the transported and disarticulated shells of Codakia orbicularis. With increasing water depths, the influence of seagrass beds decreased. Well-sorted sediments formed, which contain fewer large constituent particles. During the continued rise of sea-level, the depositional environment became deeper and less agitated. Fine-grained sediments were deposited in the deepening platform interior, which became increasingly protected due to the development of marginal reefs.
During the Holocene development of the platform interior, the pre-Holocene relief has been continuously levelled out. Because of this levelling and the increase in water depth, facies units became larger and more homogenous (Figure 19). Between the two endmembers discussed above, i.e., sedimentation in a topographic low and on a topographic high, various transitions were found in sediment cores. Except for the sequence in core NL 031/4 (Figure 13) complete facies successions are hardly ever developed. Usually, successions are incomplete and interrupted by hiatuses where the missing sequences were either not deposited or eroded. Similar observations were made in other Holocene lagoons of atolls (Gischler, 2003; and isolated barrier reef systems (Isaack et al., 2016;Zinke, Reijmer, Thomassin, Dullo, et al., 2003;Zinke et al., 2005). Especially in the early Holocene deposits, there are usually lag times of up to 2 kyr between the first marine carbonate sediments and underlying mangrove peat deposits suggesting that a certain amount of time is necessary until the marine carbonate factory has been fully established (Isaack et al., 2016;Kim, Fouke, Quinn, Kerans, & Taylor, 2012;Schultz et al., 2010;Tipper, 1997). A principal problem in this context is the fact that complete transgressive lagoonal successions are best developed only in the deepest lagoon parts, where Holocene sediment thickness usually reaches or even exceeds 10 m (Isaack et al., 2016;Zinke et al., 2001). To recover such long cores with portable vibrocore systems from smaller vessels navigable in shallow reef lagoons is technically very challenging and therefore the exception rather than the rule.
Calculations between radiometric dates and age data and core tops (considering the latter as having modern ages), Holocene sedimentation rates (reefal and lagoonal areas) amount to 0.42 m/kyr ( Figure 20). Based on an average depth of 17.7 m below sea-level for the pre-Holocene surface, an average sediment thickness of 2.7 m, and available sedimentation time of some 8.5 kyr, lagoonal sedimentation rates may be estimated to 0.32 m/kyr. An estimated, total sediment volume of 2.4 km 3 of Holocene carbonate deposits accumulated on the Bermuda Platform based on an average thickness of 8.3 m (Kuhn, 1984). In the North Lagoon (290 km 2 ) ca 1 km 3 were deposited, and ca 1.4 km 3 accreted in the reef rim (170 km 2 ). The sedimentation rate of North Lagoon is in the same range as that of other atolls and carbonate platforms in the Atlantic (Bight of Abaco, Bahamas: 0.24 m/kyr, Rasmussen, 1989;northern Isaack et al., 2016). In the case of the Bora Bora, the higher-than-present relative sea-level in the mid-to-late Holocene in the south Pacific was presumably responsible for a larger variability of texture and composition detected in lagoonal sediments (Isaack et al., 2016).
Holocene reefs on the Bermuda platform developed preferentially on topographic highs, especially at the platform margin. Exceptions to this observation exist and include some of the platform interior lagoonal patch reefs dominated by Oculina and Madracis. The existing age data are not sufficient to decide whether keep-up or catch-up growth modes predominated. Eventually, vertical reef accretion rate presumably decreased during the mid-late Holocene as seen in many other Caribbean reefs (Gischler, 2015; and references therein). The decrease in the rate of Holocene sea-level rise caused this phenomenon and also led to an increased lateral (progradational) reef growth. In Bermuda, inter-reef areas of sand pockets and channels have been gradually filled in with sediment. However, larger sand aprons as indicators of lateral lagoonal infill from marginal reefs are lacking. An explanation might be the fact that the marginal reef of the Bermuda platform is not surface-breaking and discontinuous. In the Holocene, relative sea-level in the western Atlantic has reached modern level later than in the Indo-Pacific (Camoin & Webster, 2015). It has been argued that lateral transport of carbonate sand produced at the margin towards the platform interior lagoon has therefore set in later and sand aprons are not as well-developed and extensive in the western Atlantic (Isaack & Gischler, 2017). Apart from sea-level change, the declining mid-late Holocene climate, i.e., decreasing temperatures, after the early Holocene climate optimum (Marcott, Shakun, Clark, & Mix, 2013) possibly enhanced the decrease in vertical Holocene reef accretion rate. In order to detail Holocene reef development, architecture, and stratigraphy in Bermuda, additional rotary drill traverses across marginal and lagoonal reefs would be necessary.
Despite the large quantities of carbonate sediment produced and the background sedimentation rates that fall in the same order of magnitude as in other platform lagoons, the available accommodation space in the Bermuda North Lagoon is far from being filled during the Holocene sequence. The same was presumably true for the older, late Pleistocene sequences 1 and 2. With a total size of some 650 km 2 , the platform is likely too large already for accommodation space infill. Previous data compilations have shown that isolated platforms with filled or almost-filled accommodation space usually cover a few hundred square kilometres only Isaack & Gischler, 2017). In addition, the  Figures S3 and S14 (no. 4). Vertical scales show both two-way travel time in m/sec and depths in metres. For the legend of core logs see Figure 15 marginal reefs of the platform largely lack sand aprons, which play a significant role in platform infill as sedimentation rates may be up to three times higher as compared to background sedimentation (Isaack & Gischler, 2017;Purdy & Gischler, 2005; and references therein). The fact that many modern platforms, including Bermuda, are characterized by unfilled accommodation space is crucial and has to be taken into account when F I G U R E 1 7 Holocene facies distribution along idealized section across Bermuda platform. Note that Holocene reefs are tied to topographic highs of the Pleistocene whereas unconsolidated sediment accumulations are largely located over topographic lows of the underlying Pleistocene deposits F I G U R E 1 8 Model of Holocene sedimentation on topographic low (a) and on topographic high (b) based on observations in all cores investigated. Note that succession on topographic low includes a stronger subdivision into various facies as compared to succession developed on topographic high GISCHLER AND KUHN | 249 analysing carbonate platform sequences and cycles in the fossil record (Eberli, 2013;Purkis & Harris, 2016;Purkis et al., 2015), because sea-level sequences are supposed to approximately equal sea-level variation.

| CONCLUSIONS
Shallow reflection seismics and vibracoring were used to detail the late Quaternary development of the North F I G U R E 1 9 Conceptual diagram of facies development in space and time in reefal and lagoonal parts of the Bermuda platform. Arrows indicate continuous and discontinuous sediment transport pathways. Lagoonal development (right hand side) usually includes deepening. Reefal sediments successions suggests initial deepening and subsequent shallowing due to decrease in rate of sea-level rise and sediment infill in topographic depressions in the reef F I G U R E 2 0 Holocene and Pleistocene age data from cores on age-depth diagram. Bermuda sea-level curve after Neumann (1971), Kuhn (1984 and Vollbrecht (1996) Lagoon of the atoll-like Bermuda carbonate platform. Four seismic sequences were identified, which were deposited during highstands of sea-level (MIS 7, MIS 5, MIS 3, MIS 1). Lowstands of sea-level resulted in soil formation and limestone/carbonate dissolution. Pleistocene sequence 1 consists of consolidated carbonate sand and forms the basement of Three Hill Shoals. Sequence 2 shows considerable relief at the top and forms large parts of the Holocene sediments. Holocene reefs accreted on topographic highs of this sequence. Sequence 3 was deposited in lows of underlying sequences. Sequence 4 was deposited during the Holocene, and investigated in detail. Sea-level rise and antecedent topography largely influenced sedimentation during the Holocene. Endmembers of Holocene sedimentation in an antecedent topographical low and on a topographic high of the platform interior are contrasted. Sedimentation in a low resulted in differentiated facies development including deposition of peat, peaty sediment, freshwater mud, restricted marine and open marine (shallow and deep lagoon, beach, seagrass beds) sediments. Sedimentation on a high led to much simpler successions including basal coarse sediments with large shell fragments and upward fining from shallow, open marine to deep marine platform interior sediments. Strong facies-indicative molluscs include Codakia orbicularis and Astraea phoebia in seagrass beds as well as Gouldia cerina, Pitar fulminata and Finella sp. in protected or deeper lagoon settings. Lagoonal successions are characterized by fining upward. Holocene reefs accreted on topographic highs of the Pleistocene, i.e., former sand bars or reefs. In contrast to the Holocene, Pleistocene conditions in the Bermuda platform interior were probably more open and characterized by higher energy environments. Late Pleistocene and Holocene sequences are characterized by unfilled accommodation space.