Bahamian coral reefs yield evidence of a brief sea-level lowstand during the last interglacial

The growth of large, bank-barrier coral reefs on the Bahamian islands of Great Inagua and San Salvador during the last interglacial was interrupted by at least one major cycle of sea regression and transgression. The fall of sea level resulted in the development of a wave-cut platform that abraded early Sangamon corals in parts of the Devil's Point reef on Great Inagua, and produced erosional breaks in the reefal sequences elsewhere in the Devil's Point reef and in the Cockburn Town reef on San Salvador. Minor red caliche and plant trace fossils formed on earlier interglacial reefal rocks during the low stand. The erosional surfaces subsequently were bored by sponges and bivalves, encrusted by serpulids, and recolonized by corals of younger interglacial age during the ensuing sea-level rise. These later reefal deposits form the base of a shallowing-upward sequence that developed during the rapid fall of sea level that marked the onset of Wisconsinan glacial conditions. Petrographic studies reveal a diagenetic sequence that supports this sea-level history. Preservation of pristine coralline aragonite, coupled with advances in U/Th age dating, allow these events in the history of the reefs to be placed in a precise chronology. We use these data to show that there was a time window of 1,500 years or less during which the regression/transgression cycle occurred and that rates of sea-level change must have been very rapid. We compare our results with the GRIP ice-core data, and show that the history of the Bahamian coral reefs indicates an episode of climate variability during the last interglacial greater than any reported in what is widely believed to be the more stable climate of the Holocene interglacial.


INTRODUCTION
Bank-barrier and patch coral reefs flourished on the Bahamian islands of San Salvador and Great Inagua (Fig. 1) duringthelastinterglacial-variously knownas substage 5eof the marine isotope scale, the Sangamon of North America, and theEemianof Europe. We haveconducted detailed field studiesof fossilreefs near Devil's Point on the west coastof Greatlnagua,and of the Cockburn Townand SuePointfossil reefs on the west side of San Salvador (Fig. 2) in order to determine their geologic history, particularly with respectto sea-level events. These three fossil reefs are the largest currently documented from the Bahamas. Petrographic studies have been usedto determine the diagenetic history of these reefs as it relates to the sea-level changes that have affected them.
Advances in the dating of aragonite using TIMS U/Th

Devil's Point FossilReef
A well-developed wave-eut platform withintheSangamonian sequence at Devil's Point (Fig. 3A) provides the most prominent field evidence for a fall and subsequent riseof sea level during the last interglacial. Corraded surfaces of large coral heads of Montastrea and Diploria fossilized in growth position (Fig.3B)suggestremoval of severaldecimeters of the coralheadtops. Otherfieldevidence showsthat the wave-eut platform developed during the Pleistocene, and that it is not a modern or Holocene feature. In places, the platform surface has rhizomorphs, the fossilized traces of terrestrial plants, directly on topof planed-offcorals (Fig.3C). Elsewhere, coral rubblestone, collapsed but essentially in place coral debris (Curran and White 1985;White and Curran 1987),overlies the wave-cut surface (Fig. 30). These deposits commonly extendvertically intoa shallowing-upward sequence of strata deposited during the transition from marine subtidal to nonmarine eoliandeposition resulting fromthe regression caused by the onset of Wisconsinan glaciation (White andCurran 1987, 1995)(see also later discussion of profileB-B' Figure  6). Laterally, the sea-level fall is represented by an irregular erosional surfaceuponwhicha well-preservedcoralpatchreef of Sangamon age is preserved (see later discussion of profile C-C' Devil's Point Figure7). . analyses, coupled with the preservation of pristine aragonite in fossil corals from these reefs, allows us to create a comprehensivechronology forthesequenceofeventsrevealed by thedetailed fieldand petrographic work. The methods and resultsof theagedatingand theoverallstratigraphic settingof the sampledcorals werepresented in Chenet al. (1991). The presence of a wave-cut surface within parts of the Devil's Point reefal sequence also was noted. Subsequent detailed field studieson Great Inagua and San Salvador revealed the widespread occurrence of an erosional surface within the reefalsequences. InthecaseoftheCockburn Townfossil reef, these later studies were facilitated by new exposures created during the construction of a marina.
Recent evidence from ice-eore data, pollen studies of lake sediments, and data from benthic foraminifera and stable isotopes indicates that the last interglacial included one or more episodes of extreme climate fluctuation. This is a controversial conclusion, as the last interglacial has hitherto been considered to have had a stable and equable climate, based in large part on comparison with the Holocene (Broecker 1994;Dowdeswell and White 1995). We explore this issue in the light of our detailed sea-level history withits precisely determined chronology.

Cockburn Town Fossil Reef
Although lacking a distinct wave-cut platform, many other features foundintheCockburn Townreef wereformedduring a sea-level lowstand that interrupted the growth of the reef. Corals are preserved in growth position on an undulating erosional surfacethat is interpreted as equivalentto the wavecut platform in the Devil's Point reef (Figs. 4A and 4B). Corals beneath the erosional surface were truncated, in some casesincluding the lithophagid boringswithinthecorals( Fig.  4C). Elsewhere the erosion surface itself was bored by sponges, and subsequently these borings were encrusted by corals (Fig.40).
Fissures, erosional channels, and small caves formedduring thesea-level lowstand, and theycut through bothin situcorals and associated lithified subtidal sediments. Some of the channels are wider in the lower part, thereby forming overhung cavities. These cavities and openings were subsequently filledby subtidal sediments (Figs. SA and 5B), which in some places have been removed by subsequent erosion to reveal cavity floors, walls, and roofs that have sponge and bivalve borings and serpulid encrustations (Fig.  SC). Red paleosols overly the fissures and the infilling subtidal calcarenites (Fig. 50). Following the physical stratigraphy of Carew and Mylroie (1995a) (Chenet al. 1991). This suggests that not all A well-preserved fossil patch reef is exposedalong the coast Bahamian fossil coralreefs were in existence throughout the northof theCockburnTown reef (White 1989). No wave-cut lastinterglacial and evidence for themid-Sangamon lowstand  fragments, which form the lower layer of a shallowingupward sequence. Such pristine preservation requires that corals be rapidly removed fromthe taphonomic environment (Greenstein andMoffat1996). Thefollowing coral ages were obtained from samples collected below the erosion surface: Profile B-B' (Figs. 2 and 6) is from the locality where the Acropora palmata, 1303±13 lea; Diploria strigosa. photographs showninFigure3 weretaken. Herethewave-cut 125.4±1.7 lea; Montastrea annuiaris, 128.4±1.2 and surfaceis underlain by in situcorals, and is overlain by coral 124.9±2.l lea. One sample of Montastrea annularis from rubblestones of collapsed but unabraded and pristine coral abovethe surfacehasan age of 123.8±1.5 lea. All coral ages warrs, CURRAN, ANDwn..SON ..t!' . and comprehensive documentation of relevant geochemical and analytical data are presentedin Chen et al. (1991). Lf':j A.~~<lorTW1anl me'e,..  from above theerosion surface hasaU-TIl ageof 123.8±1.7lea (Chen et aI. 1991, includes all analytical data). Thisgives a range of coral ages from below the erosion surface of 132.6±1,3 to 125.3±1.7 lea, and from above the surface a single ageof 123.8±1.7 lea.

PETROGRAPIllC EVIDENCE
Changing sea levels may expose nearshore reefs and associated subtidal grainstone facies to a sequence of diagenetic environments, each of which may leave a distinctive imprint that creates a record of these changes in   environment Figure 9A showsa grainstone with a sequence of earlier marine cements consisting of micritic high-Mg calcite followed by aragonite. Also present are later sparry calcite cements typical of meteoric, non-marine diagenetic environments in unburied Quaternary limestones. The grainstone shownin Figure9B has a similaroverallsequence but the amountof earlier marinecementis much less and the rockis cementedlargelyby non-marine calcsparwitha patchy distribution typicalof a vadosezone. Thecementsequence is reversed in the rock illustrated by Figure 9C, with the rock being largelycementedby an earliernon-marine isopachous calcsparwith a small amountof later,irregularly distributed, marine high-Mg calcite and subsequent aragonite. A more completely developed sequence of later marine cements is illustratedin Figure9D whereearlyisopachous to patchynonmarine calcsparis followed by isopachous marine high-Mg calciteand aragonite. Usingthewell-established principles of cementstratigraphy (Meyers1974), wecandeducea sequence of diagenetic environments from marine to non-marine and then a return to marine. This cementsequence indicates that an interval of sea-level lowstand occurred during the diagenetic history of these rocks and that sea level fell far enoughto exposereefaland associated rocksto thefreshwater phreatic and vadose environments.

Timing of the Sea-levelExcursion
A compelling body of field and petrographic evidenceshows that the development of coral reefs on Great lnagua and San Salvador islands during the Sangamon interglacial was interrupted by a fall of sea level that exposed the reefs and associated sedimentary facies to non-marine conditions. Sea levelprior to the fall wasat least4 m higherthanthe lowstand level. SUbsequently, sealevelroseto approximately +6m and the reefs flourished once again.
Corals from beneathand above the erosionsurfaceproduced 17 during the lowstand are exposed in close vertical proximity along profiles B-B' (Fig. 6) and C-C' (Fig. 7) at the Devil's Point reef and C-C' (Fig. 8) at the Cockburn Town reef. Uranium-thorium ages of corals from these localities are shownin Figure 10. In attempting to evaluateall of our data from a geological perspective we have tended to follow the common practice of focusing on the central tendency of the agedata (see forexampleChenetal. 1991;CarewandMylroie 1995a;Eisenhauer et al. 1996). Our interpretations are constrained by the knownrock record,and we areimpressed by the relationships between the clustering of dates in relationship to their stratigraphic source. Failure to follow this multifaceted approach leads to geologically absurd conclusions, for example placing younger rocks beneath an erosional disconformity and older ones above it Following these techniques the data from Figure 10, which represent situations wheredatedcoralsarein verticaljuxtaposition from beneathandabovetheerosion surface,indieateatimewindow for the regression-lowstand-transgression sequence in the range of 1.1 to 1.5lea.

Ratesof Sea-levelChange
Sea level prior to the regression to the lowstand was a minimum of 4 m above present sea level and the ensuing transgression raised sea level to 6 m above present sealevel (Curran and White 1985 Table1.Calculated averageratesof sea-level changeforthemid-Sangamon fallandriseof sealevel,assuming various durations for the lowstand and using two estimates (1100and 1500years) derived from data of Figure 10 for the length of the sea-level cycle. SangamonClimateInstabilityand Sea-level Changes coolingduring the last interglacial thatcoincide with colder periodsindicated by the GRIP ice-coredata (Thouveny et aI. 1994). Seidenkrantz et aI. (1995) presented data on the abundance of benthic foraminiferal speciesfoundin twocores of marineshelfsediments from Denmark. Theyinterpret their data as indicating two cooling events during the last interglacial, thattheycorrelate with colderintervals indicated in the GRIP ice-core data. Furthermore, these authors conclude that climatic change was rapid, on the scale of decades or centuries. Recently, Seidenkrantz and Knudsen (1997) presented a detailed analysis of benthic foraminifera fromoneof thesecoresthatsupports theirearlierconclusions. Thus, the known signatures of climatic instability duringthe last interglacial extend from Greenland to Europeand are, in fact, believedto be global (Broecker 1994).

Literature survey> Theevidence indicatesthatrapidchanges
of temperature of several degrees Celsius occurred on a decadal time scale during the last interglacial, producing a variety of signatures recorded in several parts of the globe (Broecker 1994). The question arises whether such temperature changes would also cause rapid changes in the volume of terrestrial ice and attendant rapid changes in sea level. Based on strarigrsphic studies of carbonate rocks on Oahu, Hawaii, Sherman et aI. (1993) reported two distinct sea-level highstands during the last interglacial. However, theiragedatinghadwideerrorbarsthatfell within thegeneral agerangeof thelastinterglacial, but werenotaccurate enough tosubdivide it Morerecentstudiesby MuhsandSzabo(1994) of uranium-series dating of the Waimanalo Formation on Oahudo not supportthe doublesea-level highstand, and they concluded that Oahu and similar tectonically active Pacific islands are unsuitable as reference points for determining last interglacial highstands.

Non-reefal Evidence for ClimaticInstability During the Sangamon
The Bahamas are generally regarded as lacking tectonic activity that would explain relative sea-level changes that occur in tandem over the whole archipelago (Carew and Mylroie 199580 b). Such changes must be due to absolute changes in the volume of sea water, most likely caused by changes in terrestrial ice volume. By comparison with the generally stable climates of the Holocene, it was widely assumed that the last interglacial was also a period of stable climate. This stabilitywas in sharp contrastto the 100lea of the last glacial period when relatively stable climate periods lastingseveral millenia were disrupted by abrupt changes to radically different climate states that occured within a few decades (Broecker 1994). This viewof stableclimates during the last interglacial was called into questionby ice-core data whichshowedthatabruptchangesof temperature occurred in Greenland duringthat time (GRIPmembers 1993;Johnsen et aI. 1995).
In a reviewof the Greenland ice-core data and the record of rapid climate changes, Dowdeswell and White (1995) comment that the terrestrial and marine records of the last interglacial give mixed signals. They concluded. somewhat cautiously, thatthe weightof evidence fromthe terrestrial and marine records suggests stable climate during the last interglacial. Not all of the evidence supports that conclusion, however. In a brief summary, Tzedakis et aI. (1994) report that pollen data from Europe supportthe view thlu climatic fluctuations occurred during the Eemian. More detailed pollen data from annually laminated lake sediments in Germany, and peats from France, showed that the last interglacial climatewas moreunstable than the Holocene, and that at times, winter temperatures reached levels similar to those that occurred during glacialperiods (Fieldet aI. 1994). Magnetic susceptibility, pollen, and organic carbon records from maar lakes that formed in explosive volcanic cratersin Precisely dated corals from a core through fossil reefaI the Massif Central of France show two periods of rapid deposits in the Houtman Abrolhos islandsoff the tectonically WI-llTE, CURRAN, AND WILSON

+6
The ice-core data show a major fall in temperature of One of the main features of the ice-core record for the last interglacial is rapid, and commonly significant, temperature fluctuations. However, somegeneraltrendscan bediscerned. From a low of approximately 1O"C below average Holocene temperatures around 142ka, a generalwarmingtrend brought temperatures aboveHolocene levelsbyapproximately 133lea. These warmer conditions continued as a general trend of temperatures approximately 4"Cabove Holocenevaluesuntil 126 ka, when a rapid cooling began. This period of higher temperatures coincidesquite well with the first stage of coral growth in the Bahamian reefs whichextendedfrom 132-125 lea. Rapid cooling events punctuated this early last interglacial warm period, although the most severe of these pre-datethe oldestcoralsyetfound in the Bahamian reefs that we havestudied. A short-lived cold spell,some timebetween 129 and 128 ka, brought temperatures approximately 2"C below Holocene values. We have found no record in the Bahamian reefs of this cold spell, but it is possible thatany effectsarebeyondthe resolution of field studiesin a complex reef facies. exposures of some Bahamian fossil coral reefsand associated facies, detailed field work. the preservation of pristinecoral aragonite, and breakthroughs in U/Ib age dating techniques creates an opportunity to compare evidence of sea-level changes in the Bahamas, with climate changes recorded elsewhere fromthe lastinterglacial period. Figure II is based on data presented in GRIP members (1993), and shows measured changesin CS 180 of the portionof the Greenland ice core that formed during the last interglacial, and the calculated temperature fluctuations basedon the isotopedata. Because these ice core data yield the best presently available information about the duration and timing of climate fluctuations duringtheEemian,we usethisdiagram todiscuss the results of our work on sea-level changes recorded in Bahamian fossil coral reefs, in the context of temperature fluctuations recorded in Greenland ice. Figure 11. Temperature data spanning the last interglacial based on oxygen isotope data from a Greenland ice core. Dashed line represents average Holocene temperature. Redrawnfrom data inGRIP members, 1993. Horizontal scale is non-linear. passive coast of Western Australia give some information about sea-level changes and elevations during the last interglacial (Eisenhauer et al. 1996). Sea level rose steadily from -4 m belowpresentaround 134ka, passedpresentlevel between 130 and 127 ka, and reached a maximum of at least 3.3 m above present at -124 lea. Sea level fell below present datumat -116 lea. Eisenhaueret al. (1996) foundno evidence of intra-interglacial fall of sea level, but they state that the resolution of their data would limit the duration of any such undetected regressions to 1.0 ka, or less.
Based on the island geology of Bermuda and The Bahamas, Hearty and Kindler (1995) developed a chronology for sealevel highstands for the past 1.2Ma. For the last interglacial, they proposed two highstands separated by a regression that lasted from approximately 128 to 123 ka based on protosols that lie between the marine deposits of the two highstands. They referred to Chen et al. (1991) in pointing out that coral reefs grew during the earlier oscillations. However, corals presentlyin situ above present sea level also are reported by Chen et al. (1991)from much of the time interval represented by the proposedlowstandof Heartyand Kindler(1995). In a more recent paper, Neumann and Hearty (1996) focused mainlyon evidencefor a rapid riseand subsequent fall of sea levelat theend of the last interglacial. However, theyinferred thatsea level for most of the interglacial remained near 2 m above present,interrupted only by a short-lived. sea-level fall of about 1.5 m at approximately 125 lea. The timing of this event appears to be constrained largely by data from Chen et al. (1991). To explainbioerodednotches at 6 m abovepresent sea level,Neumannand Hearty(1996) invoked a rapidriseof sea level from+ 2 to + 6 m at the end of the interglacial and a highstand that lasted only a few hundred years. This interpretation was questioned by Carew (1997) who stated that the postulated rate of notch formation is much too high and that the notches could not have been formed in a few hundred years, and by Mylroie (1997) who reinterpreted the notches as the eroded remnants of flank margin caves. Regardless, such caves also would need much longer than a few hundred years to form. Our work on the fossil reefs indicates that sea level was at or close to the + 6 m level for several thousand years from about 124 ka to approximately 119lea. This providessufficienttime for both notchand cave formation but obviatesthe need for invoking a veryrapid rise in sea level close to the end of the interglacial.
Interesting information is presented by Precht(1993)basedon studiesof reefalfaciesof the Falmouth Formation in Jamaica. According to Precht (1993), the Falmouth Formation comprises two shallowing-upward parasequences that represent reefal development during two separate sea-level highstands of substage 5e. Uranium-series dating of aragonitic corals suggests that the lower parasequence corresponds to a sea-level highstand at 134-127 lea, and the upper sequenceto a highstand at 124-119 lea.
The evidence from thisstudy.--The combination of excellent approximately 9"C from 126 to 125 lea. just prior to the dramatic fall in sea level recorded in the Bahamian reefs beginning at approximately 125 lea. The most reasonable conclusion from these data is that this major cooling led to greatly increased snowfall and the rapid increase in the accumulation of land-based ice.
Following 125lea. a slowwarming is indicated by theicecore data, but average Holocene temperature values were not reached untilapproximately 121 ka. Temperatures thenrose quickly to approximately 4 DC above Holocene levels, where they remained, with minorfluctuations, for a littlemorethan a millennium. In the fossil reefs discussed in this paper this warming trendcoincides witha rapidrise in sea leveland reestablishment of the coral reefs and the secondstageof their growth, which lastedfrom 124-119 lea. Although the general trends of warming climate, rising sea level,and regrowth of coralsarecoincident, therearediscrepancies in themagnitude of theseevents. The ice-core data show that temperatures in Greenland remained belowHolocene values for muchof this interval, whereas sea levelandaccompanying coralgrowthin the Bahamian reefs was higherthan presentlevels.
The reason for the discordance between rapidly rising sea level and slowly rising temperatures in Greenland is not known. However, somespeculations arepossible. Pollendata from Germany and France show that the average mean temperature of thecoldestmonth remained lowand relatively stableduring theIanerpart of thelastinterglacial beforerising rapidly during the last 1000 years (Field et al. 1994). This corresponds more closely to the ice-core data than to the Bahamian sea-level history. Summer temperatures remained relatively highduring this period,suggesting a moreextreme seasonality thanispresently experienced. Therateof increase in mean annual temperatures in the higher Iatitudes of the Northern Hemisphere may have lagged behind the average global increase, perhaps due to the lack of penetration by relatively warm oceanic currents. In this scenario, the low Northern Hemisphere temperatures are anomalous and the rapidly rising sea levels identified from the Bahamian reefs are the globalnorm. Hollin(1980) proposed that somepolar icemasses couldpassa criticallimitandsurgeintotheoceans, and thus cause rapid sea-level rise. It is possible that such eventsoccurred duringthe earlystages of the warming in the latter part of the last interglacial, and accelerated the rate of sea-level rise.
The ice-core data show that temperatures in Greenland fell approximately9"C in a5OQ-year spanbetween 119and 118ka to 5"C below average Holocene levels. This timing corresponds to the record from Bahamian coral reefs with the youngest mown coralbeing119.9±1.4 lea. Theexcellent state of preservation of Bahamian coralreefshasbeen attributed to rapidburial by the entombing sandsof a shallowing-upward sequence resulting from rapid sea-level drawdown beginning at about 119lea (White et al. 1984;Whiteand Curran 1995). This interpretation has been supported by the recent 20 taphonomic studies of the Cockburn Town fossil reef by Greenstein and Moffat (1996).
The Greenland ice-core data showone morewarmerepisode that occurred around 115 ka when temperatures reached approximately 4"C above the Holocene average for a short interval, beforefalling to theglaciallevelsthatpersisted until the Holocene. No corals of this age are known from the Bahamian fossil reefsthatwe havesmdied, norareany corals found encrusting the terminal shallowing-upward sequence sands.

CONCLUSIONS
I. Bahamian coral reefs on San Salvador and Great Inagua islands developed during two stages within the last interglacial separated by a short-lived sea-level lowstand.
2. The first stageof reef growthoccurredduring the interval 132-125 ka when Greenland ice-core data indicate that temperamres were generally about 4"C higher than the average for the Holocene. Sea level during this first reef development phasewas at least4 m abovepresent 3.According to ice-core data,temperatures fellapproximately 9"C during the interval 126-125 lea. This corresponds to a rapid sea-level fall that interrupted coral growthand led to a period of erosion in the earlier reef, and to an episode of freshwater diagenesis. Sea level fell duringthe interval 125-124 ka to about the current level, at rates probably significantly in excess of 10 mm per year. 4. After 124lea. sea levelrose rapidlyto a maximum of 6 m above present, and coral reef growth was renewed. This second phase of reef growth lasted from 124-119 lea. Although the ice-core data indicate a slow warming during this second interval of reef development, the extent of the warming seems insufficient to explain the magnitude and highrate of sea-level rise. The reason for this discrepancy is unknown. Warming inGreenland may havelagged theglobal rate due to its isolation from warming ocean currents, perhaps due to the kinds of longitudinal shifts in ocean circulation patterns described by Johnsen et al. (1995). 5. A rapidcooling of approximately 9"C beginning at 119ka coincides with a veryrapidfall in sea levelbeginning shortly after119lea. This fallwasduetoan earlyphaseof ice volume increase that marked the beginning of the Wisconsinan glacial stage. Falling sea level led to the rapid burial of the coral reefs in regressive facies sands and to their excellent preservation. 6. A brief return to temperatures up to 4"C higher than Holocene values isindicated by icecoredataat around115lea. Ifthisledto a sea-level highstand higherthanpresentwe have detected no impactin the Devil's Point and Cockburn Town reefs. The ice-core data show no other time between 115ka and the beginning of the Holocene when temperatures approached the Holocene average. 7. Our data support the concept that the last interglacial climate was subject to rapid and significant fluctuations in temperature that had dramatic effects on sealevel, which in turn affectedthe development of coralreefs.
8. Markedchangesof sea level appearto have separatedtwo intervals of several millenia duration when sea level was higher than presentand reef growthflourished.
9. The frequency of sharptemperature fluctuations shownby the ice-core data is greater than thatreflected in the observed effects in the coral reefs of sea-level changes. Whetherthis meanstherewereno effectson sealevelor thereefs,orthatthe effects are too subtle for present methods of analysis to discover, is unknown.
10.Basedon therecordof the lastinterglacial, therecan be no assurance that dramatictemperature fluctuations and changes in relativesealevelwill not occurduringtheremainder of the present interglacial.
11. As thechangesof sealevelduringthelastinterglacial most likelyare due to changesin the volume of land-based ice, it is wrong to think of interglacials as being free of ice sheets. Perhaps terminology misleads us and the terms glacials and interglacials mightbe better changedto greaterglacials (i.e., hyperglacials) and lesser glacials (i.e., hypoglacials) respectively.