Comparison of Recent Coral Life and Death Assemblages to Pleistocene Reef Communities: Implications for Rapid Faunal Replacement on Recent Reefs

Marine ecologistsandpaleoecologists are increasingly recognizingthat the Pleistocene and Holocene fossilrecord of coral reefs is the exclusivedatabase fromwhichan assessmentof the long-term responsesof reef communities to envirorunental perturbations may be obtained. Theapparentpersistenceof coralcommunitiesin the faceof intensefluctuations in sealevelandsea surfacetemperatureduring glacial and interglacial stagesof Pleistocene time is in markedcontrastto dramaticfluctuations in reef communitystructuredocumentedby short-termmonitoringstudies. We comparedthe taxonomic structureof live and dead coral communitieson a modem patch reef currently undergoing a community transition to late Pleistocenefacies exposed in the CockburnTown fossil coral reef. Multidimensional scaling revealed that specific taxa and colony growth forms characterize life, death, and fossil assemblages. The recent decline of thickets of Acropora cervicornis is representedby their abundancein the death assemblage, whilePorites porites dominatesthe coral life assemblage. Althoughadditional study of Pleistocenereefal facies is required, the greater similarityof the death assemblage to the fossil assemblage suggests that the present Caribbeanwide decline of A.cervicornis is without a historicalprecedent


INTRODUCTION
Community ecologists and paleoecologists are becoming increasingly aware that the fossil record is an exclusive and crucialdatabase fromwhichto interpret longterm community patterns (Jackson 1992;Ricklefsand Schluter 1993;Jackson et al. 1996). Duringthe lastdecade, paleoecological studies in terrestrial (e.g. Davis 1986;Delcourt and Delcourt 1991;Davis et al. 1994;Reed 1994;Coope 1995) and marine(e.g. Jackson 1992; Buzas and Culver 1994;Allmon et al. 1996;Jackson et al. 1996;Pandolfi 1996) systems have demonstrated that the fossil record possesses a wealth of information applicable to current concerns for global environmental changeas wellas environmental perturbations on a localscale.
ThePleistocene fossilrecordofcoral reefsoverthelastmillion years is a particularly valuable data repository because of its generally spectacular preservation, and because, with few exceptions, the same coral taxa that inhabit modem shallow water reef environments are present in Pleistocene deposits. Although this is in part due to the younggeologic age of the interval, a great deal of qualitative (e.g. Mesollela 1967;Mesollela et al. 1970;White et al. 1984;White and Curran 1987;White 1989;Johnson et al. 1995;White and Curran 1995)and quanitative (Greenstein and Moffat 1996;Pandolfi 1996) data suggest spectacular preservation is common for reef coral assemblages accumulating during at least the last 600ka in boththe Indo-Pacific andCaribbean provinces. The Caribbean shallow water coral (and mollusk.) fauna has undergone littlespeciation or extinction sincefaunalturnover ended roughly a millionyears ago (Potts 1984;Allmon et al, 1993;Jackson etal. 1993;Buddetal. 1994;Jeckson 1994a,b) in spite of intensifying cycles in climate and sea level throughout the Pleistocene.
With the above as an underlying assumption, Pleistocene fossil coral reef deposits can potentially beused as a database CarbonQtes & Evaporites, v. 13,no. 1, 1998, p. 23-31. with whichto address a varietyof cardinal issuesfacingreef ecologists and marine resource managers, whose frustration over the temporally myopic view afforded by monitoring studies that rarely span a scientific career increasingly pervades the literature (e.g. Done 1992;Jackson 1992;Bale and Nieuwland 1995). Perhaps foremost among these issues, is an assessment of the response of coral reef communities to environmental perturbations. Important ecological influencesoncoralreefsmayoperateona variety of temporal and spatial scales (porter and Meier 1992), including decadal time scales (Bale and Nieuwland 1995;Done 1992), and the need for long term data sets has been recognized by a variety of workers (e.g. Likens 1987;D'Elia et aI. 1991;Jackson 1992). Jackson(1992) suggested thatthe fossilrecordrepresents the exclusive database from which responses of coral communities to globalchangemay be gauged For example, in Barbados, preliminary qualitative data fromthePleistocene raised reef terraces suggests that similar coral communities and zonation patterns have prevailed for the past 600 ka (Jackson 1992). In an overview of mollusk, reef coral, and planktic foraminiferal communities, Jackson (l994a) found little correlation between the magnitude of environmental changeand subsequent ecological and evolutionary response during Pleistocene time. In a very detailed study that examined geographic and tempera! changes in community composition, Pandolfi (1996) found limited community membership in Indo-Pacific reef corals from 125-30 lea. These studies give a very different picture of coral reef community structure and stability than that derived from traditional, small-scale ecological studies in modem environments.
Given the utilityof the Pleistocene fossil recordof coral reef communities for assessing long-term responses to environmental perturbations, quantitative comparisons of these assemblages to modem reef communities is essential. Here, wecomparethestructureof communities of liveanddeadreef corals presently found on Telephone Pole Reef to that preserved in late Pleistocene age (Sangamonian) facies exposednearCockbwn Townto determine theaccuracywith whichthe fossil record represents thetaxonomic structure of a once-living coralcommunity. Ourresultscontrastwith those obtained fromsimilarcompanion studies, andsuggestthatthe transition betweencoral communities currently occurring in Fernandez Bay (and indeed, throughout the Caribbean) may not have a historical precedent

Data analyses
Radiometric dates indicate that the Pleistocene reefs on San Salvador flourished between 119-131 ka (Chenet al. 1991).
In modem environments, we define the death assemblage as in situ dead coral material encountered along each transect (including dead portions of partiallydead colonies) and coral rubble accumulating adjacent to the reef framework. Dead coral coloniesencountered along the transect were identified to the species levelonly if we could recognize them without breaking them open or peeling off any algae or other overgrowth. Bulk rubble samples composed of dead coral weretaken at the 5, 15,25, and 35 m marksof each transect Thismethodology allowed us to adequately samplethe death assemblage as defmed above. Rubblesamples wereplacedin a 5 mm mesh bag constrained by a 10 L bucket Thus, coral species and growthformwererecorded for each specimen >5 mm in size that preserved morphology sufficient to permit identification. Taxonomic data obtained from the rubble samples were pooled with those obtained from dead corals encountered along each transect We constructed species sampling curves to investigate whether our methodology adequately accommodated the diversity present in the coral assemblages studied. Comparison of taxonomic composition was calculated using the Bray-Cwtis dissimilarity coefficient (Bray and Curtis 1957), which has been shown to be one of the most robust coefficients for the analysis of taxonomic composition data (Faithet al. 1987). Dissimilarity values range from 0 (for a pair of samples with identical taxonomic composition) to 1 (for a pair of samples with no taxa in common). Abundance data were ttansformed to their square roots prior to the analysis, to reduce the influence of occasional large abundance values for some taxa (Field et al. 1982). In addition, the transformed abundance values for each taxon were standardized by the maximum attained by that taxon. This standardization equalizes the potentialcontributions of taxa to the overall dissimilarity in composition. Without standardization by taxon, the Bray-Curtis values are dominated by thosetaxa thatattainhigh abundances (Faithet al. 1987). The resulting Bray-Curtis matrixwas subjected to anordination technique that provideda visual summary of the pattern of dissimilarity values among the samples. The technique employed wasglobalnon-metric multidimensional scaling, or GNMDS (Kruskal1964), whichhasbeenshownto be one of the most effective methods available for the ordination of taxonomic composition data (Minchin 1987). Each sample is represented as a point in a coordinate space with a given numberof dimensions. The distances between eachpairof pointsare,as faras possible, inrankorderwiththe corresponding dissimilarities in taxonomic composition; points that are close together on the resulting scatter plot represent transects with similar coral constituents. The degreeto whichthedistances on the scatterplotdepartfroma perfect rank order fit is measured by a quantity known as 24 Field Methods

STUDY SITES AND SAMPLING METHODS
We used the linear point intercept (LPI) method (Lucasand Seber1977) andconstructed transects onTelephone PoleReef and in theabandoned quarryareaof theCockburn Townfossil reef. In ordertoadequately estimatecoverforthewidestrange of coral growthforms and colonysizes, transects were 40 m long (Mundy 1991), each separated by 2O-5Om. At 20 em intervals along each transect, the transect intercept was observed. Fromthe modernreefenvironments, the following data wererecorded if the transect intercepted a coral:species, colonysize,colonyorientation, growth form,and whetherthe colony was alive or dead and whetherwhole or fragmented. Coloniesthat werepartiallydeadand largeenoughto provide multiple intercepts wererecorded as liveor deaddepending on what portion of the colony intercepted by the transect The same data (with the obvious exception of whether the coral colony was alive or dead) were collected from transects laid across the Pleistocene reef facies exposed in the quarry.
The regressive stratigraphic sequence exposed in the abandoned quarry in Cockbwn Town, San Salvador, Bahamas, includes a coral rubblestone facies, composed predominantly of Acropora cervicornis, and a coral stone facies that contains abundant in situ Acropora palmata, Montastrea annularis and Diploria strigosa. These facies were defined by White et al. (1984) and Curran and White (1985) and suggested to represent back reef and reef tract environments, respectively. This fossil reef provides an excellent opportunity for a comparative taxonomic study because of its proximity to analogous modem reef environments (Fig. 1). Telephone Pole Reef, located in Fernandez Bay,is a mid-shelfpatehreef that,in the past,was dominated by thickets of A. cervicornis set amidst larger heads of Montastrea annularis and other massive corals. Today, the degraded A. cervicornis thickets and rubble provide examples of the coralrubblestone faciesdescribed by White et al. (1984) and Curran and White (1985). The branching coralA. cervicornis has suffereda major decrease inabundance throughout the Caribbean regionduringthepast two decades due to a combination of factors (see below). On Telephone Pole Reef, the once abundant stands of A. cervicornis have been significantly replaced by Porites porites ).  Son Salvador Island, Bahamas. Telephone PoleReefis located in Fernandez Bay, whilethe Cockburn Townfossil reefis located onshore at the north end ofFernandez Bay, in Cockburn Town.   Results of ordination revealthatsamples fromlife,death,and fossilassemblages are clearlydistinctfromone another (Fig.  3). Moreover, the taxonomic composition of the death assemblage samples is moresimilarto samples fromthefossil assemblages thansamples fromthe life assemblages (Fig.3). Analyses of species richness indicate that a significant decrease (F(2.10)=121.31;p< 0.00(5) inspecies diversity occurs from life to death to fossilassemblages (Fig.4). community structure on Telephone Pole Reef thatis part of a Caribbean-wide phenomenon. Beginning at least as early as the 1980's, Acropora cervicornis has suffered an extreme decrease in abundance as a result of a confluence of factors including hurricanes (Woodley et ale 1981), spread of macroalgae consequent to sea urchinmassmortality (Lessios 1988), coral diseases andcoralbleaching (Brown and Ogden Thediscrimination between thelifeassemblage and thedeath 1993; Littler and Littler 1996;Miller 1996) and a variety'of and fossil assemblages by the ordination technique is human-induced effects

cervicornis, Porites astreoides, Montastrea annularis, and
Diploria strigosa). It is compelling that these"healthy"reef communities werereflected by the fossil assemblages in the FloridaKeys and Greatlnagua,whereasthe presentPoritesdominated community onTelephonePoleReefisnotreflected by thefossilassemblage exposed on SanSalvador. Moreover, it is sobering to consider the rapidity with which A cervicornis-dominated communities havebeenaltered. From this we canonly conclude that, although the A cervicornisdominated coral association persisted during Pleistocene climatic fluctuations, it isapparently vulnerable to thearrayof perturbations currently being inflictedon it A secondhypothesis is thatrapidchangesin coraldominance within a community commonly occur, but the fossil record does not have sufficient resolution to preserve these temporally short-term fluctuations in reef community structure. Short-term studies of living coral reefs have recorded fluctuations of dominant species at virtually all spatialscales; ranging frommeterquadrats (e.g, Hughes et al. 1987;Bak and Nieuwland 1995) through individual reefs (e.g. Porter et al. 1981;Woodley et al. 1981) to entire provinces (e.g. Lessios 1988). Moreover, short-term fluctuations may be a prerequisite for long-term stability (Chesson and Huntly 1989) andthusproducethetype oflongterm persistence of coral communities documented by Mesollela (1967), Jackson (1992), and Pandolfi (1996). We note here that several workers have outlinedsedimentologic (e.g. Whiteet at. 1984;Whiteand Curran 1987;Curran et al, 1989;White 1989;Whiteand Curran 1995)and taphonomic (e.g. Greenstein andMoffat1996)evidence forrapidburialof latePleistocene bank-barrier and lagoonal reef systems of the Bahama Archipelago during the post-Sangamonian regression. Thus live and dead corals were buried concurrently. Greenstein andMoffat(1996) demonstrated thatspecimens of A cervicornis preserved in the quarry near Cockburn Town were actually less degraded than dead specimens accumulating on Telephone Pole Reef. In this case, the taphonomic evidence doesnot supporta massmortality event of thisspecies during Sangamon time. In any event,it seems likely to us that rapid transitions similar to those observed presently intheCaribbean wouldbepreservedsomewhere had they occurred. The preliminary resultsreported by Aronson and Precht (1997) are encouraging in that they demonstrate that this hypothesis is eminently testable; and further microstratigraphic examination of Pleistocene coral-bearing strata should be undertaken. Foremost among these strata shouldbe theunitsdescribed here,as wellas thosefromwhich long-term stability of Caribbean and Indo-Pacific reef coral communities hasbeenreported (e.g. Mesollela 1967;Jackson 1992Jackson , 1994aPandolfi 1996).  Fig. SA). Additionally, the paucity of milleporids in either the deathor fossil assemblage relative to the life assemblage further segregates life assemblage samples from those obtained from the fossil and death assemblages (compare Figs.SA, 5B, and 5C). Susceptibility of thosehydrozoans to thevariety ofphysical,biological, andchemical processes that tend to destroy potential fossil material possibly implicates phylogenetic differences between hydrozoan skeletal microstructure and scleractinian skeletal microstructure as a source of difference in thepreservation potential of thesetaxa. The decrease in species richness from live to dead to fossil assemblage resultsfrom the absence of the three milleporids we distinguished in our surveys (Millepora complanata, M. squarrosa andM. alcicornis) aswellas taxa thatarerareinthe lifeassemblage (e.g.Diploria clivosa, Siderastrea radians. S. siderea, Motuastrea cavernosa, Meandrina meandrites and Dendrogyra cylindrus, Fig. 5).

Constancy and Change in Reef Community
The resultswe report here contrast with those obtained from companion studies comparing coral life and death assemblages occurring on modern Florida Keys reefs to Pleistocene reefsexposedin the KeyLargoLimestone andon Great Inagua, Bahamas (Greenstein et al. in press and Greenstein and Curran 1997,respectively). In thosestudies, life assemblages currently thriving in the Florida Keys reef tract were demonstrated to be more similar to Pleistocene fossil assemblages than to their contemporary death assemblage. The striking difference obtained here is thatthe deathand fossilassemblages are clearly most similarto each otherin terms of the coral taxa theycontain. There are three alternative hypotheses that explain theapparentfailure of the Pleistoceneassemblage exposedonSanSalvadortoaccurately reflectthe life assemblage currently in place offshore, while other Pleistocene strata we have studied apparently reflect modern reef coral life assemblages muchmore closely.
First, the demise of A cervicornis in the Bahamas and Caribbean and subsequent replacement by another coral species (on Telephone Pole Reef, P. porites) is without historical precedent In Belize, theonceabundant standsofA cervicornis have been replaced by Agaricia tenuifolia (Aronson and Precht 1997;Aronson and Plotnick, in press). Careful examination of cores taken through the reef sedimentary recordin Belizerevealed norecognizable signals (abruptchangesin coral taxa, or taphomonic evidence of an essentially monospecific death assemblage) of similar transitions, suggesting that the present drastic reduction of Acervicomis hasno precursorin therecentgeological past(at least 3,800 years; Aronson and Precht 1997). In the Florida Keys,we (Greenstein et al., in press;Greenstein and Curran 1997) purposely chose modern reefs for our surveys that A third alternative is that differences between the conformed to earlier~1980) descriptions (for example environmental setting of Telephone Pole Reef and that GREENSlEIN, HARRIS, ANDCURRAN interpretedfor the Cockburn Town fossilreef account for the differenceswe have obtained. Whiteet al, (1984)and Curran and White (1985) interpreted the main portion of the Cockburn Town fossil reef as a reef tract system, whereas Telephone Pole Reef is a patch reef. Thus, the physical environment may not have permitted a take over by Porites porites or Agaracia tenuifolia (e.g. Aronson and Precht 1997). However,Greensteinand Moffat(1996)examinedthe mode of preservation of both Acropora palmata and A. cervicornis collected from the Cockburn Town fossil reef.
Based on a varietyof taphonomic criteria,they suggestedthat the fossilcoralswere derivedprimarilyfroma life assemblage that had been rapidly entombed. Had a die-off of A.
cervicornis occurred during Sangamonian time, thickets of degradedA. cervicornis mightpotentially bepreservedduring such an event, and yield a much differenttaphonomic signal than that obtainedby Greensteinand Moffat (1996). A twostep analysis of additional Pleistocene facies exposed in the Bahamas is needed to address this issue. First, to avoid circular reasoning, environmental interpretations of fossil reefassemblages shouldbemadeusingcriteriaindependentof coral content (for example constituent particle analyses coupled with detailed field descriptions [pandolfi et al., in preparation]). Once a fossil assemblagehas been determined to representan environmentlike thatcurrently hosting a reef undergoing transition in coral dominance, a comparative taxonomicstudyshouldbeperformedon both the modemand ancient assemblages.

SUMMARY
A comparison of the results of systematiccensuses of live, dead, and fossil coral assemblagesoccurringin and adjacent to Fernandez Bay has yielded insight into the importanceof the Pleistocene fossil record of coral reefs as an instrument with which to determine whether the presently observed changes in patch reef community structure in the Caribbean region have a historicalprecedent We have determinedthat the rapid decline of Acropora cervicornis observed on a Bahamian patch reef (and observed around the Caribbean region) may bea unique perturbation that contrasts with the long-term persistence of this taxon during Pleistocene and Holocenetime.

CONCLUSIONS
Although the results of our species sampling curves dictate that our results must be interpreted cautiously, primarily owing to relativelysmall sample size, we offer the following conclusions: 1. The fossil assemblage exposed in the Cockburn Town quarryis less speciesrich thanits livingand dead counterparts in Fernandez Bay. This is the result of two factors: a) coral taxa thatare rare in the life assemblage(e.g.Diploria clivosa, Dendrogyra cylindrus.Meandrina meandrites, andMontastrea cavernosa) are increasinglyerasedduringthe transition to the 29 death and fossil assemblages; and b) the three species of hydrozoancommonon thelivingreef are rare in the deathand fossilassemblages.
2. Based on the assemblageof corals they contain, the death and fossil assemblages are more similar to one another than either is to the livingcoral assemblage. Thisresult is in stark contrast to similar comparisons reported from the Florida Keys and Bahamas. We submit that this is the result of comparinga reef currentlyundergoinga transitionto Porites porites-dominated coral assemblage from an Acropora cervicornis-dominated assemblage.
3. Further study of Pleistocene reef facies is needed to assess whether the current Caribbean-wide reduction in abundance of Acropora cervicornis has an historicalprecedent ACKNOWLEDGMENTS Logisticalsupporton SanSalvadorIslandwasprovidedby the BahamianField Station. Additional financialsupport to BJG and HAC from Smith College is gratefully acknowledged. John Pandolfi (Smithsonian Tropical Research Institute) provided the Bray-Curtis analysis. We acknowledge the Schultz-Sherman Fairchild Foundation for support to LAH. Kathryn Jermann, Sara Rosenzweig, and Sarah Smalheer (Smith College), served as field assistants. We thank Rich Aronson and an anonymous reviewer for constructive comments thatsignificantly improved this manuscript