Planktonic foraminiferal depth habitat and δ18O calibrations : plankton tow results from the Atlantic sector of the Southern Ocean

Planktonic foraminiferal depth habitat and D 18 O calibrations: Plankton tow results from the Atlantic sector of the Southern Ocean P. Graham Mortyn 1 and Christopher D. Charles Scripps Institution of Oceanography/University of California, San Diego, La Jolla, California, USA Received 23 February 2001; revised 30 July 2002; accepted 18 September 2002; published 21 May 2003. [1] Plankton tows conducted in the Atlantic sector of the Southern Ocean allow analysis of the influence of water column structure on planktonic foraminiferal abundance and d 18 O composition. Foraminiferal abundance varies by several orders of magnitude across a large gradient in sea surface temperature and other hydrographic features, demonstrating high sensitivity of foraminiferal populations to regional differences in water properties. The depth of maximum abundance for key species such as Globigerina bulloides and Neogloboquadrina pachyderma is not constant from station to station. The pattern suggests that their abundance and shell chemistry are tied to density horizons or other conditions (such as food availability) that become more sharply defined with depth in the northern subantarctic. The consistent observation of Globorotalia inflata and Globoratalia truncatulinoides as relatively deep-dwelling species confirms their utility as indicators of upper thermocline properties. In d 18 O all species are observed to be isotopically lighter than predicted from water properties, but the species-specific offset is fairly uniform at all stations. These observations define the utility of multispecies d 18 O for reconstructing temperature and density stratification from past surface oceans. INDEX TERMS:4267 Oceanography: General: Paleoceanography; 4572 Oceanography: Physical: Upper ocean processes; 4870 Oceanography: Biological and Chemical: Stable isotopes; 3030 Marine Geology and Geophysics: Micropaleontology; KEYWORDS:stratification, d 18 O Citation: Mortyn, P. G., and C. D. Charles, Planktonic foraminiferal depth habitat and d 18 O calibrations: Plankton tow results from the Atlantic sector of the Southern Ocean, Paleoceanography,18(2), 1037, doi:10.1029/2001PA000637, 2003. 1. Introduction [2] A recurring problem in paleoceanography is to make inferences about the ocean in four dimensions from a onedimensional sediment core record. Of course, this problem is greatly simplified if we assume that the chemistry of sedimentary populations of planktonic foraminifera is homogeneous and is acquired at one time and place in the water column. This is an implicit assumption in most studies, despite evidence that this does not hold for most species [Lohmann, 1995]. Vertically stratified plankton tows are among the more direct means of determining how these simplifying assumptions can be related to precise physical processes. The advantage of a plankton tow system, such as the multiple opening closing net and environmental sampling system (MOCNESS) [Wiebe et al., 1976, 1985], is that it collects material on a depth-discrete basis at a fixed sampling time and location. This approach allows for realtime observation of the depth distribution of foraminiferal abundances and isotopic compositions for comparison to simultaneously collected data such as sea surface temperature (SST), salinity, and fluorescence (a proxy for chlorophyll a content) [Schreiber et al., 1998]. [3] A variety of studies have documented the vertical distribution of planktonic foraminifera in the tropics [e.g., Fairbanks et al., 1980, 1982; Ravelo et al., 1990; Ravelo and Fairbanks, 1992; Watkins et al., 1996, 1998], yet there is an equal need to calibrate the paleoceanographic proxies derived from planktonic foraminifera in higher-latitude regions. This is especially true given that foraminifera represent one of the few means for retrospective monitoring of the processes of deep-water formation and heat exchange between the high-latitude ocean and the atmosphere. Core-top calibration approaches reveal a few basic properties of modern high-latitude foraminiferal assemblages, but generally, they cannot discriminate the mechanisms through which sedimentary signatures are acquired. For example, Globigerina bulloides, one of the principal species studied at middle-to-high latitudes, follows the predicted d 18 O of calcite for North Atlantic surface waters only over a narrow range of about 1%(from 1 to 2%); outside of this range, G. bulloides core-top d 18 O does not reflect mean surface conditions [Bard et al., 1989]. The implication is that either seasonal growth or calcification at depth confounds the signal. By contrast, similar data from another planktonic species, Neogloboquadrina pachyderma, suggest a very different pattern in South Atlantic core tops: the d 18 O of this species follows the predicted d 18 O of calcite (and therefore SST variability) over a wide range of latitudes and hydrographic conditions [Charles and Fairbanks, 1990]. This pattern for N. pachyderma d 18 O is in marked contrast to that for the North Atlantic region [Kohfeld et al., 1996; Bauch et al., 1997]. Sorting PALEOCEANOGRAPHY, VOL. 18, NO. 2, 1037, doi:10.1029/2001PA000637, 2003 1 Now at Department of Earth and Environmental Sciences, California State University, Fresno, California, USA. Copyright 2003 by the American Geophysical Union. 0883-8305/03/2001PA000637 15 -1 out such differences between two widely used planktonic species at high latitudes, including the differences between Northern and Southern Hemisphere patterns, is an essential prerequisite for refining reconstructions of past surface ocean conditions. [4] Here we present a description of modern middle-tohigh—latitude planktonic foraminifera caught in MOCNESS tows taken in the South Atlantic sector of the Southern Ocean over a 1-month period during the austral summer of 1996. Specifically, we concentrate on the paleoceanographically important species that span a variety of depth habitats and latitudes. Previous work in the same study region made use of underlying core-top samples [Niebler et al., 1999] to decipher water column stratification. Our work here complements this and other work with core tops in the Southern Ocean [Labeyrie et al., 1996] because it allows a direct correlation between foraminiferal geochemistry and surface ocean properties. Our analysis does not consider all aspects of foraminiferal ecology, but we demonstrate that the unique intersection of depth habitat and South Atlantic water column structure might explain some of the main patterns of foraminiferal abundance and chemistry in the modern sediments (and by extension the paleoceanographic record). 2. Methods and Materials [5] During January–March 1996 we collected MOCNESS plankton tow samples at or near piston coring stations that later became coring stations for Ocean Drilling Program (ODP) leg 177. The primary goals of this operation, on cruise TNO57 aboard the R/V Thomas Thompson, were twofold: (1) to document the vertical distribution of microfossil-producing plankton (foraminifera) and, by comparison with water column properties, to understand the ecological controls on their distribution and (2) to understand how and where in the water column the foraminifera acquire their chemical and isotopic signatures. [6] Samples were collected from six stations spanning a latitudinal and longitudinal range between 41and 53S and 5and 12E, respectively (Figure 1). These subantarctic South Atlantic station locations were dictated by sediment coring interest, because they are areas with high sedimentation rates and important sedimentary sequences. However, these stations also happen to span surface waters that today exhibit, among other properties, an SST range of greater than 12C in only 12of latitude, making for one of the largest SST gradients in the global ocean [Levitus and Boyer, 1994]. The MOCNESS stations are therefore appropriately located in areas of interest for microfossil calibration. [7] Of concern in any plankton tow study is the question of how representative such a ‘‘snapshot’’ view may be of seasonal or longer-term conditions. We assume that our results, collected between 4 February 1996 and 5 March 1996, are representative of the summer season of biologic production in the Southern Ocean. Abelmann and Gersonde [1991] summarized biosiliceous particle flux (radiolarians and diatoms) for time series sediment trap deployments in the Southern Ocean between 1983 and 1990. Their results demonstrated that the flux during the austral summers accounted for 70–95% of the total annual flux, suggesting that our snapshot is as representative of sedimentary conditions as is possible for a 1-month period. Because of ship time constraints, our sampling strategy could not explicitly resolve possible population migrations over diurnal or lunar-modulated [e.g., Bijma et al., 1990] reproductive cycles. [8] Vertically stratified plankton tow samples (150-mm mesh net within a 333-mm net used for structural support) were collected using a MOCNESS device described in detail by Wiebe et al. [1976, 1985]. The MOCNESS system was also equipped with sensors to measure in situ properties from the water column, including temperature, salinity (and s t ), and fluorescence. The temperature and salinity data are presented in Figure 2, while the s t data are shown with fluorescence in Figure 3. The individual nets were sequentially opened and closed over discrete depth intervals between 0 and 800 m, and temperature, salinity, and fluorescence were measured in situ with attached conductivity-temperature-depth (CTD) and fluorometer probes. The nine sampled intervals and their vertical spacing varied from station to station and were chosen using hydrographic information obtained from a separate CTD/Rosette cast just prior to MOCNESS deployment. MOCNESS samples were stored in a 100% ethyl alcohol (ethanol) solution to prevent dissolution of foraminiferal shells and to minimize artifacts to isotopic measurements that can arise with other preservatives such as buffered formalin [Ganssen, 1981]. Splits of these samples were density-separated in a hypersaline solution according to a modification of a method described by Be´ [1959]. [9] Depending on station-specific foraminiferal abundances, sample splits were counted with the aim of totaling 300 specimens per sample. In some cases the total abundance was low enough that the total counts came from the whole sample at less than 300 specimens. The abundance of key species from these sites are presented (Figure 4) as number of shells per unit volume of seawater, a scale made possible because the volume of filtered seawater was measured with a flowmeter attached to the frame of the MOCNESS system. The flowmeter data should be considered only an approximate representation of the effective flow through the nets (especially if biomass concentrations are high enough to cause net clogging, lower flow estimation, and hence overestimation of foraminiferal standing stock). For most stations, however, the relationship between the flowmeter data and effective flow should be relatively uniform for all the nets on a given cast. Fluorescence (mg/m 3 ) is also presented with the absolute foraminiferal abundance data (Figure 4), after it was corrected by uniformly subtracting fluorescence contents below 500 m (which we assume to represent the background) for a given station. Planktonic foraminifera captured with the MOCNESS plankton tow system are not necessarily alive; however, their isotopic composition (discussed below) helps discern live from dead populations. 15 -2 MORTYN AND CHARLES: SURFACE STRATIFICATION AND d 18 O CALIBRATIONS [10] In order to minimize possible size-related differences, d 18 O analyses were carried out on the 150–250-mm fraction of G. bulloides, N. pachyderma, Globorotalia truncatulinoides, Globorotalia inflata, and Orbulina universa, where sufficient quantities were available. All samples were roasted under vacuum at 375C for 1 hour prior to isotopic analysis. Measurements were made at the Scripps Institution of Oceanography (SIO) using a Carousel-48 automatic carbonate preparation device coupled to a Finnigan MAT 252 mass spectrometer. The d 18 O precision (1s) of the NBS-19 standard was better than 0.09%for 190 standards run along with the samples over an 18month period during which the samples were analyzed. We will not deal with the d 13 C for the foraminifera (though the data exist) because measurements of seawater d 13 C (total dissolved CO 2 ) were probably compromised by sample storage effects. [11] In order to compare equilibrium calcite d 18 O(d 18 Oec) values against measured foraminiferal values we estimated the d 18 O of calcite precipitated in isotopic equilibrium with seawater. We first predicted the d 18 O composition of the seawater using our own salinity measurements obtained from the MOCNESS downcasts and a South Atlantic salinity-d 18 O (water) relationship from Geochemical Ocean Sections Study (GEOSECS) [1987] data shown by Charles and Fairbanks [1990]. The applicability of this relationship is assessed by representative d 18 O (water) (D. Hodell, personal communication, 1996) and salinity analyses from our own downcasts of the CTD (Figure 5). We then used the predicted d 18 O (water) and the MOCNESS temperature data to estimate the d 18 Oec, according to a standard mean ocean water (SMOW) to Peedee belemnite (PDB) unit conversion [Hut, 1987] and a carbonate-water isotopic temperature scale from O’Neil et al. [1969]. These predicted d 18 Oec Figure 1. Map showing our plankton tow station locations between Africa and Antarctica. The dashed line corresponds to the mean position of the modern Polar Frontal Zone (PFZ), which is the northern terminus of the subsurface T min layer bounded by the 2C isotherm in the 100–300-m layer in the summer and the vertical 2C isotherm in the winter [Belkin and Gordon, 1996]. MORTYN AND CHARLES: SURFACE STRATIFICATION AND d 18 O CALIBRATIONS 15 -3 data are then compared against d 18 O measurements from foraminifera collected in the water column in Figure 6, while also shown in latitudinal section form in Figure 2. [12] Among the various d 18 O paleotemperature equations proposed in the literature [e.g., Bemis et al., 1998, and references therein] we chose the inorganic calcite equation of O’Neil et al. [1969] as our reference point because it was the only one calibrated to temperatures as low as 0C. However, we recognize that this definition of equilibrium calcite is arbitrary and perhaps not even directly applicable to living planktonic foraminifera. Over the temperature range of interest here (0–15C) the absolute offset between measured foraminiferal d 18 O and ‘‘predicted equilibrium’’ varies significantly if other equations are extrapolated to the low temperature range [e.g., Kim and O’Neil, 1997; Erez and Luz, 1982]. Because there is as yet no agreement on how best to define equilibrium for biologically precipitated calcite, we cannot interpret the average apparent disequilibrium. On the other hand, the choice of paleotemperature equation does not affect the relative interspecies and intraspecies offsets from equilibrium, an important point that we discuss below. Figure 2. Latitudinal sections of temperature, salinity, and the predicted d 18 O of equilibrium calcite. The temperature and salinity data are downcast measurements collected with the CTD system, while the predicted d 18 O (calcite) data are calculated using d 18 O (standard mean ocean water (SMOW)) = 0.5(salinity) - 17%[Charles and Fairbanks, 1990], the Hut [1987] SMOW to Peedee belemnite (PDB) correction factor of 0.27%, and the O’Neil et al. [1969] paleotemperature equation (see Figure 6 caption for further details). 15 -4 MORTYN AND CHARLES: SURFACE STRATIFICATION AND d 18 O CALIBRATIONS [13] In order to compare the foraminiferal d 18 O from the water column to that from material at the modern sedimentwater interface, we also measured ‘‘core-top’’ samples beneath three of the six stations where material permitted. Core-top material is not presented from TNO57-22, TNO57-9, and TNO57-11 since we have reason to doubt the presence of modern or even late Holocene sediments at these locations. The representative data from the other stations are shown along the lines beneath individual panels of Figure 6. These samples were provided by either multicore or trigger core from each individual station for the depth interval of either 0–1 cm or 0–2 cm into the sediment. Table 1 presents a summary of the core samples used for the core-top data. Species-specific foraminiferal samples were generally picked from the 150–250-mm size fraction, although in one case (TNO57-21) it was necessary to Figure 3. In situ fluorescence and s t profiles based on downcast data collected with the MOCNESS system. Panels representing data from individual stations are arranged from north to south. Fluorescence profiles are shown along the left side of each panel, while s t profiles are shown along the right side of each panel. Fluorescence is reported in concentration units as it is a proxy for chlorophyll acontent [Schreiber et al., 1998]. Note that the data are limited to 500-m depth at the southernmost stations, TNO57-16 and TNO57-13. MORTYN AND CHARLES: SURFACE STRATIFICATION AND d 18 O CALIBRATIONS 15 -5 15 -6 MORTYN AND CHARLES: SURFACE STRATIFICATION AND d 18 O CALIBRATIONS supplement this with material from the >250-mmsize fraction. 3. Results 3.1. Oceanographic Setting [14] Figure 2 shows that the surface temperature ranges from about 1 to 13C across the 12latitudinal span of our study area. The density structure becomes much better defined to the north, with a relatively shallow thermocline and a clear surface mixed layer. The salinity section (Figure 2) shows generally more salty surface waters in the north, while also illustrating southern middepth anomalies that are likely the result of interleaving layers of water with different origin. Thus the local structure affecting the foraminiferal population is certainly partly a product of eddy mixing, which must be a perennial process associated with the Antarctic Circumpolar Current [Savchenko et al., 1978; Gille et al., 2000]. [15] Phytoplankton abundance, as in many other areas of the world’s oceans, is sensitive to the hydrographic properties associated with the pycnocline. Figure 3 shows that fluorescence maxima all occur within the mixed layer, but usually at relatively deep depths near the top of the pycnocline. The origin of the so-called ‘‘deep chlorophyll maximum’’ in other regions has been a source of long standing debate, and the possible mechanisms for its occurrence include the intersection of high nutrients and high light, the effect of density gradients on settling, or preferential grazing patterns. It is also possible that fluorescence peaks are partly the result of phytoplankton photoadaptation, involving changes in the carbon to chlorophyll or carbon to fluorescence ratios. Without ancillary evidence our data cannot test these various possibilities. Accordingly, we only note the relationship between fluorescence and foraminifera on a descriptive (nonmechanistic) level. 3.2. Planktonic Foraminiferal Abundance [16] Several important observations can be drawn from the abundance data presented in Figure 4. One is that the overall abundances vary considerably from site to site, by as much as 4–5 orders of magnitude between TNO57-16 and TNO57-13. The TNO57-9 site shows the highest surface foraminiferal abundance, about 235/m 3 of seawater, in apparent response to a phytoplankton bloom suggested by the coincident peak in fluorescence (>0.7 mg/m 3 ). The TNO57-16 site shows the highest middepth (75–100 m water depth) abundance, about 2484/m 3 of seawater, also in association with a fluorescence peak of about 1.4 mg/m 3 . The highest deep (>100 m water depth) abundances are shown at TNO57-16, TNO57-9, and TNO57-11. [17] The association of total foraminiferal abundance peaks with fluorescence maxima is not always straightforward. For instance, surface peaks in fluorescence at sites TNO57-21 and TNO57-11 are not coincident with abundance peaks at those stations. Also, fluorescence peak depths vary from site to site. For example, the peaks occur well below the surface at the TNO57-16 and TNO57-13 sites, while they generally occur closer to the surface at the other stations. Despite these differences, however, there is reasonable correspondence between foraminiferal abundance and fluorescence maxima. In fact, TNO57-22 shows a double fluorescence peak in the upper 100 m, with high foraminiferal abundances closely associated with each. [18] It is possible to separate the abundance patterns by species. Despite the fact that G. bulloides is encountered over a wide depth range the maximum of this species is consistently associated with a fluorescence peak (the only exception being TNO57-21), suggesting that G. bulloides closely tracks phytoplankton blooms. The next most abundant species is N. pachyderma (left-coiling), which shows a wide range of occurrence for abundance maxima (anywhere from the surface at TNO57-13 to the 75–175 m depth at TNO57-9. N. pachyderma (right-coiling) is observed mostly at middepths, with one exception (TNO57-22). Turborotalita quinqueloba also generally shows its highest abundances below 100 m, although the surface maximum at TNO57-22 is also an exception. G. inflata is consistently most abundant at subsurface depths (between 50 and 300 m), while the maximum abundances of G. truncatulinoides are always deeper than 100 m. The maxima of this species are even deeper at TNO57-21 and TNO57-9, roughly 500 and 200 m, respectively. O. universa is the least abundant planktonic species in these tows, showing measurable middle-to-deep quantities at only TNO57-21 and TNO57-22. [19] In addition to this close association with a probable food source in phytoplankton (suggested by the fluorescence peaks), maximum foraminiferal abundances generally occur within the mixed layer and just above the pycnocline (the only exception being TNO57-21). In all cases, G. bulloides maxima coincide with overall foraminiferal maxima because it is everywhere the most dominant species. 3.3. Oxygen Isotopic Signatures and Water Column D 18 O Gradients [20] Oxygen isotopic data (d 18 Oin%PDB) from the water column and core-top samples are presented in Figure 6. In the water column at all stations, predicted d 18 O curves are more positive in d 18 O relative to foraminiferal measurements, implying that all species are well outside our Figure 4. (opposite) Water column abundances of various species of planktonic foraminifera presented from each of our six stations, arranged as in Figure 3, along with in situ fluorescence data from the MOCNESS system. In each stacked histogram figure the varying column thickness corresponds to the towed depth interval in the water column, while the foraminiferal species absolute and relative abundances are shown along the bottom xaxis according to the legend color scheme. The fluorescence data are shown in each panel along the upper xaxis in concentration units of mg/m 3 (as in Figure 3). MORTYN AND CHARLES: SURFACE STRATIFICATION AND d 18 O CALIBRATIONS 15 -7 calculations of isotopic equilibrium (using the O’Neil et al. [1969] equation) in this region. However, at the depth of maximum abundance for any given species, the speciesspecific offset from expected equilibrium is fairly constant. We specifically calculated this d 18 O departure for G. bulloides, N. pachyderma, G. inflata, and G. truncatulinoides, using all six stations and all appropriate depths (where peak abundances coincided with measurable d 18 O). This collective analysis yielded an average d 18 O departure from equilibrium of 1.0 ± 0.4%, with a high degree of similarity for each species (G. bulloides = 0.95, N. pachyderma = 1.00, G. inflata = 0.94, and G. truncatulinoides = 1.10). 15 -8 MORTYN AND CHARLES: SURFACE STRATIFICATION AND d 18 O CALIBRATIONS [21] Figure 6 illustrates shallow-to-deep d 18 O gradients in different foraminiferal species that mimic the profiles of the predicted d 18 O-calcite data. Considering G. bulloides and N. pachyderma as shallow dwellers (at northern and southern sites, respectively) and G. inflata and G. truncatulinoides as deep dwellers (at all sites), some relevant gradient observations could be made. The shallower species show a near-constant offset from the predicted curve of about 1%above 100 m, while the deeper species show a similar offset below that depth. Data from TNO57-22, TNO57-9, and TNO57-11 are sufficient to illustrate this pattern, while those from the other stations are not available. At stations where G. bulloides d 18 O is invariant with depth we assume the deeper signal simply reflects the descending populations that calcified at shallower depths. If we connect the shallow G. bulloides data or the shallow N. pachyderma data to the G. inflata or G. truncatulinoides data below 100 m at TNO57-22 and TNO57-9, a reasonable similarity to the predicted equilibrium curve is observed. Thus, by combining species the foraminiferal d 18 O tracks the basic structure of the thermocline with a relatively uniform 1%d 18 O offset. Similarly, at TNO5711, analogous observations can be made with the gradient between shallow G. bulloides d 18 O data and deeper G. truncatulinoides d 18 O data. [22] The core-top foraminiferal d 18 Odataareshown along the bottom lines for each representative panel in Figure 6. The paucity of enough individual shells of a given species often prevented analysis, but where sufficient material was available, some comparisons could be made. In all cases the core-top foraminiferal d 18 O was more positive relative to that from the water column, by as much as 1.2–1.4%for G. inflata at TNO57-21, 1.25%for G. bulloides at TNO57-16, and 0.75%for N. pachyderma at TNO57-16. 4. Discussion [23] The following discussion is organized around the questions of (1) whether unique inferences of depth habitat can be established for the four principal species found in the high latitude sediments and (2) what influence these ecological tendencies have on the d 18 O compositions of foraminifera and therefore the sedimentary record of high-latitude climate change. A number of analogous studies from different regions of the global ocean offer useful comparisons [e.g., Ortiz et al., 1995; Kohfeld et al., 1996], and general similarities in the ecological relationships of major species do exist among the various high-latitude regions of the ocean. However, we emphasize that the intersection of these ecological tendencies with the unique hydrographic conditions in the South Atlantic sector of the Southern Ocean creates unique impacts on the sedimentary record. Thus results from other regions, for example, the high-latitude North Atlantic, cannot necessarily be extrapolated to the Southern Ocean. [24] One major question is whether there exists a truly surface-dwelling planktonic foraminiferal species in the South Atlantic. Our abundance data show high foraminiferal occurrence to be associated with peaks in fluorescence concentration at relatively deep portions of the surface mixed layer (Figure 4). This suggests that the foraminifera captured in our plankton tows are closely associated with a phytoplankton food source that is controlled by the density structure of the water column, which is not surprising since the top of the pycnocline likely sees high diffusion of underlying nutrients [Anderson et al., 1972]. Thus there is no reason to suspect that any of the most common subpolar species are restricted to the surface mixed layer. [25] From the sediment trap work of Thunell and Reynolds [1984], Reynolds and Thunell [1985], and Sautter and Thunell [1989] it is evident that G. bulloides is abundant in a wide range of thermal environments (at least 5–20C) and increases in abundance during periods of high phytoplankton productivity due to upwelling or bloom conditions [Sautter and Thunell, 1991]. Other MOCNESS studies in upwelling environments have found peak G. bulloides abundances to be associated with a chlorophyll maximum and relatively high biomass contents [Fairbanks et al., 1982; Ortiz et al., 1995]. Our results, although with much higher abundance counts than encountered in these previous MOCNESS studies, show a similarly strong correspondence of G. bulloides to fluorescence maxima. This observation is consistent with the idea that this species is tightly coupled with phytoplankton blooms but is inconsistent with the idea that this species is strictly surface dwelling. These considerations also raise the possibility that the shallower populations of G. bulloides caught in our surface net tows may simply have been mixed upward from deeper portions of the mixed layer. [26]N. pachyderma typically dominates the foraminiferal assemblage in other high-latitude regions of the ocean, and Figure 5. (opposite) Comparison of the predicted equilibrium d 18 O (in %PDB) of calcite estimated with the same method but with two different starting points. The first involves the prediction of the d 18 O of seawater based on salinity (light circles), while the second involves actual measurements of the d 18 O of seawater (dark squares, D. Hodell, unpublished data, 1996) where data are available from four of the six stations. The results of this comparison justify the use of a South Atlantic salinity versus d 18 O (water) relationship derived from GEOSECS [1987] data (see text for details). With either starting point the d 18 O (water) information is necessary to take the prediction to the next step. The d 18 O (water) prediction is converted to PDB units [Hut, 1987] and then used to estimate the d 18 O of equilibrium calcite according to the paleotemperature equation of O’Neil et al. [1969]. Panel arrangement is consistent with previous figures, and locations for which this comparison was not possible (TNO57-11 and TNO57-16) are not shown. MORTYN AND CHARLES: SURFACE STRATIFICATION AND d 18 O CALIBRATIONS 15 -9