Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
 The nature of tropical climate variability ∼350 ka is addressed using δ18O and Sr/Ca records from a modern and a fossil coral from Vanuatu (southwestern tropical Pacific Ocean). Modern El Niño events at Vanuatu produce positive coral δ18O and Sr/Ca anomalies; similar anomalies observed in the fossil coral records suggest that El Niño was operative 350 kyr ago. Seasonal variations in coral Sr/Ca, a sea surface temperature (SST) proxy, have the same amplitude in both corals, whereas seasonal δ18O variations are smaller in the fossil coral than in the modern coral. This is consistent with displacement of the South Pacific Convergence Zone toward the southwest during the boreal summer ∼350 ka. Mathematical modeling results preclude warmer SST and higher SSS at Vanuatu during this time, but permit the surface ocean to be ∼2°C cooler and 0–2 psu fresher than today. Assessing the potential of variations in late Quaternary seawater Sr/Ca remains the largest obstacle to accurately reconstructing tropical SST using pristine fossil corals.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Evaluating modulations in ENSO during different background climate states requires records of El Niño events from times of different background climate. Instrumental records that could be used to reconstruct ENSO, such as sea surface temperature (SST) or SST anomalies (SSTA) in the eastern equatorial Pacific Ocean and the Southern Oscillation Index, do not extend much further back in time than the end of the 19th century, if that far. Instrumental SSTA records of ENSO are more complete both spatially and temporally post 1950. Instrumental sea surface salinity (SSS) records of ENSO are extremely rare and the ones that exist extend back only to the 1970s [e.g., Lagerloef and Delcroix, 1999; Gouriou and Delcroix, 2002].
 The mechanisms driving ENSO variability on geologic timescales are still enigmatic [Clement et al., 1999]. Coupled ocean-atmosphere general circulation models have been used to investigate some of the possible processes involved. Climate modeling indicates that the precessional cycle of insolation can interact with zonal asymmetry in Pacific SST and cause an increased frequency of warm (cool) events when there is an anomalous cooling (warming) in the late summer/fall [Clement et al., 1999]. Another similar model indicates that an increase in warm events during early glaciation may induce high-latitude ice buildup [Kukla et al., 2002]. The paleoclimate data that exist are insufficient to adequately verify model results, and more data are needed.
 Proxy climate records (e.g., tree rings, ice cores, and corals; Dunbar and Cole  for an overview) are required to investigate changes in ENSO in the preinstrumental period. Proxy records that overlap with, and extend beyond, the instrumental period offer a unique perspective on ENSO variability and can be used to address tropical climate variability on submillennial timescales. Proxy climate records also provide a data set from the tropical oceans that can be used to test model output and to provide constraints on climate model input.
 Corals are ideal for reconstructing seasonal to interannual variations in the tropical surface oceans because they are long-lived (many survive for centuries), their skeletons have annual density bands that provide chronological control and subannual resolution, they are widely distributed, and their skeletons are amenable to geochemical analyses. Two of the most widely used geochemical determinations of coral aragonite are stable oxygen isotope ratios and Sr/Ca ratios. Coral oxygen isotopic ratios (δ18O) reflect both temperature and the oxygen isotopic composition of the seawater (δ18Osw) at the time of carbonate precipitation [e.g., Epstein et al., 1953; Goreau, 1977; Weber and Woodhead, 1972; McConnaughey, 1989; Fairbanks et al., 1997]. δ18Osw in the modern ocean responds to salinity changes brought about by evaporation or precipitation and, in coastal zones, by fluvial input [Craig and Gordon, 1965; Bigg and Rohling, 2000]. Sr/Ca ratios in the modern ocean are largely invariant and the Sr/Ca ratios in corals vary in response to changes in surface ocean temperature [e.g., Beck et al., 1992]. Paired analyses of Sr/Ca and δ18O on a single coral sample can result in the unique solution of SST and SSS [McCulloch et al., 1996; Gagan et al., 1998; Ren et al., 2003].
 Here we present δ18O and Sr/Ca data from an exceptionally preserved fossil coral recovered from Bougainville Guyot (16°00.56′S, 166°40.34′E), located near Espiritu Santo Island in the Republic of Vanuatu (Figure 1). The fossil coral is ∼350 kyr old, an age that places the most likely time of coral growth during a time when sea level was lower than today; possibly during Marine Isotope Stage (MIS) 10. In this study, we compare geochemical variations in the fossil coral with those from a modern coral from near Espiritu Santo Island (15.7S, 167.2E; Figure 1) to address two fundamental questions:
 1. What was the background climate state of the tropical Pacific ∼350 kyr ago, in terms of sea surface temperature (SST) and sea surface salinity (SSS)?
 2. Is there evidence of ENSO activity at ∼350 ka?
2. Ocean-Atmospheric Interactions in the Western Pacific
 Ocean conditions in the western tropical Pacific both respond to and force atmospheric phenomena. The Western Pacific Warm Pool (WPWP) is the largest oceanic heat source in the world and strongly affects global climate on a variety of timescales, especially interannually (ENSO). Ocean-atmosphere interactions in the Vanuatu region of the South Pacific are dominated by spatial and temporal variability in the linkages between warm SST associated with the WPWP and atmospheric convection associated with the South Pacific Convergence Zone (SPCZ).
 The SPCZ is a zone of atmospheric convergence of the trade winds, and can be thought of as an extension of the Intertropical Convergence Zone (ITCZ) that stretches roughly from Papua New Guinea east to 120°W, 30°S [Vincent, 1994]. The spatial extent of the SPCZ changes in response to changing wind fields and SST on seasonal, interannual, and multidecadal timescales. The SPCZ is stronger in the austral summer than in the austral winter. Summertime warm SSTs facilitate upward convection in the atmosphere and northeasterly trade winds merge into the southeasterly trades along the SPCZ [Vincent, 1994]. The combination of warm SSTs and atmospheric convergence are associated with low atmospheric pressure and a persistent cloud band over the SPCZ. In the austral winter, the WPWP shifts to the north as Northern Hemisphere summer progresses. A shift in the wind field moves the focus of low-level tropospheric convergence to the ITCZ in the Northern Hemisphere during the austral winter [Vincent, 1994]. The locus of convection moves northward with the warm water and the winds, resulting in the SPCZs apparent northward movement into the ITCZ.
 The strength and position of the SPCZ also varies on multidecadal timescales according to the phase of the Interdecadal Pacific Oscillation (IPO; a Pacific-wide manifestation of the Pacific Decadal Oscillation; Folland et al. ). SPCZ movements in response to the phase of the IPO are quasi-independent of, but of the same magnitude as, SPCZ movements in response to ENSO [Folland et al., 2002]. In general, a positive phase of the IPO is related to a northeast shift in the SPCZ (the case from 1977 to the present), and a negative phase of the IPO is related to a southwest shift in the SPCZ (the situation from the mid-1940s to the mid-1970s). These types of movements become important when interpreting a climate proxy record because the proxy has been recording a Eulerian measurement of a moving system. In other words, the corals remain in the same location but surface ocean currents and atmospheric phenomena may change their position with time.
3. Climatic Setting of Vanuatu
 Vanuatu is well situated to study changes in tropical climate variability because it is positioned beneath the SPCZ, along the southern edge of the WPWP, and it is affected by spatial and intensity variations in both. Thus ENSO forcing is strong at Vanuatu and surface-ocean conditions there reflect the influence of important climate forcing.
 Movements of the SPCZ and its associated salinity front control small seasonal SSS changes (0.36 average annual range 1980–1992; Levitus and Boyer ) and larger interannual SSS changes (1.91 maximum range 1980–1992; Gouriou and Delcroix ) in this region. The SPCZ lies to the northwest of Vanuatu during the austral winter and in the austral summer it moves southeast over Vanuatu, bringing with it enormous amounts of rain contributing to regionally lower SSS at that time [Gouriou and Delcroix, 2002]. During ENSO warm phase (cool phase) years, the SPCZ weakens (strengthens) and Vanuatu is dry (wet). Salinity increases sharply during a warm phase event due to a combination of more saline waters advecting from south of the SPCZ and an increase in the local evaporation to precipitation ratio [Picaut et al., 1996, 1997, 2001; Vialard and Delecluse, 1998].
 The average SST for a 1° × 1° grid box centered near Espiritu Santo Island, Vanuatu (hereafter Santo) at 166.5E and 15.5S is 27.7°C (1980–1992; Levitus and Boyer ). This places Santo on the edge of the WPWP (as defined by 28°C mean annual SST contour), so it is affected by spatial and temporal changes in the warm pool. The average seasonal SST cycle near Santo, is 2.5°C [Levitus and Boyer, 1994], and the interannual SST variations are small relative to the seasonal cycle. SSTs generally decrease (increase) during ENSO warm phase (cool phase) years, as warm water from the warm pool advects toward the central Pacific.
 It is important to note that not all ENSO events manifest in exactly the same manner at Santo (Figure 2). There are some warm phase years where a small SSTA is associated with a strong sea surface salinity anomaly (SSSA; 1983), while other years have strong SSTA and weak SSSA (1987). Still other “events” last for multiple years, such as the mild El Niño of the early 1990s [Trenberth and Hoar, 1996, 1997]. The key to recognizing ENSO from temperature and salinity records in the Santo region is determining the sense and duration of the SSSA and SSTA.
 For the purpose of this study ENSO events must be defined in terms recognizable in a proxy record, faithful to the local modern instrument record, and consistent with conventional definitions. In a fashion similar to Trenberth , the 5-month running mean of surface anomalies are used to define an ENSO event. We define an ENSO warm (cool) phase event at Santo (Figure 2) to be indicated by a simultaneous SSTA decrease (increase) and SSSA increase (decrease), of a magnitude greater than 1.5 standard deviations of the running mean for more than 6 months of either SSSA or SSTA (δ18O or Sr/Ca anomalies in proxy situations).
 The definition above differs from that of Trenberth  in two ways. Surface salinity anomalies are included in the definition in addition to SSTA because the local response to El Niño at Santo can be either strong salinity anomalies or strong temperature anomalies, and is often both. Another difference is the use of a fractional standard deviation threshold as opposed to a specific temperature (0.4°C) as in the work of Trenberth . The fractional standard deviation threshold captures anomalies that are proportionally large compared to the variance in the specific record. When considering records of past SSTA and SSSA that may have a smaller or larger amount of variance, the standard deviation threshold may be a more robust indicator of significant change relative to the mean.
4.1. Physical Sampling of Corals
 A living Porites lutea head was cored in 1.5 m of water within Malo Channel, which is located between Malo Island and Espiritu Santo Island in the Republic of Vanuatu (Figure 1) in 1992. The site was chosen because of its exposure to open-ocean conditions, minimizing any local microclimate effects. A 5 mm-wide slab was removed from the core (henceforth referred to as MCB) and subjected to X-radiography, revealing density band couplets. The slab is composed of 8 individual pieces, which fit together well, thereby ensuring stratigraphic completeness. Samples of the coral slab were removed as powder using a computer controlled drilling stage and drill mount. Powder samples were generated by routing a 3 mm wide swath, ∼1 mm deep and collecting the powder every 1.33 mm down the slab.
 The fossil Porites sp. sample was recovered from the carbonate sequence of Bougainville Guyot during ODP Leg 134 at Site 831B in 1992 (831B-18R-1, 110–125; Taylor et al., 1994). The sample (henceforth referred to as BG831) is 15 cm long and was recovered from 238.3 m below the seafloor in 1066 m of water. Similarly to the modern coral, BG831 was cut into a slab 5 mm wide, X-radiographed to reveal density bands, and drilled for aragonite powder. Because the fossil coral grew more slowly than the modern coral, individual samples from BG831 were extracted every 0.67 mm.
 Representative samples from the modern and fossil coral were subjected to a variety of tests to determine the preservation state of the fossil coral. Initially BG831 was subjected to X-ray diffractometry to test for the presence of calcite. A thin section of the same coral was examined under a petrographic microscope and both MBC and BG831 were examined using scanning electron microscopy for evidence of secondary aragonite or calcite.
4.2. Age Determination of the Fossil Coral
 The age of the fossil coral sample was determined by a combination of radiometric (uranium series and Sr isotopes) and stratigraphic modeling techniques. The U-series determinations were performed at the University of Minnesota using thermal ionization mass spectrometry (TIMS) and the results have been previously published [Taylor et al., 1994]. The Sr isotope determinations were performed at the University of Michigan radiogenic chemistry laboratory and the results have been previously published [Quinn et al., 1994].
 Stratigraphic modeling was performed to calculate the elevation of the sample through time using the subsidence rate [Taylor et al., 1994] and present-day depth of sample. The intersection of the sample elevation with Quaternary sea-level curves (SPECMAP Stack; Imbrie et al. ; and sea level from Shackleton ) dictates the sample age. Coral growth position errors of ±10 m were used in all of the calculations.
4.3. Stable Isotopes and Elemental Ratios
 Stable isotopic determinations were made on coral powder samples at the University of South Florida, College of Marine Science using a ThermoFinnigan Delta Plus XL dual-inlet mass spectrometer with an attached Kiel carbonate preparation device. Carbon and oxygen isotope data have a precision of 0.04 and 0.08‰, respectively (1σ), as calculated from repeated measurements of the standard NBS-19.
 Sr/Ca determinations were made on coral powder samples at the University of South Florida, College of Marine Science using a PerkinElmer Optima 4300DV dual-view ICP-OES, following the drift-correction methods of Schrag . Each coral sample (∼250 mg) is dissolved in 10 mL of 2% (v/v) trace metal-grade nitric acid diluted with superdeionized water (resistance equals 18 MΩ). Gravimetric standards were made by combining Mg, Ca, Sr, U and Ba from SPEX ultrapure solution standards in proportions to match those of a typical coral. The accuracy of the Sr/Ca ratio of our gravimetric standards was confirmed via TIMS analysis at the University of Minnesota. Bulk coral powder from a modern Porites lutea sample from Santo was also dissolved in 2% (v/v) trace metal-grade nitric acid to make a “coral standard” solution. The coral standard solution was routinely analyzed as part of each run. Blanks of acid and deionized water are routinely monitored for contamination. Analytical precision (2σ) for Sr/Ca determinations used in this study is estimated to be 0.15% (0.013 mmol/mol), based on 86 determinations of the coral standard solution.
 Geochemical data were converted from the depth domain to the time domain by matching the maxima (minima) in Sr/Ca to minima (maxima) in SST data and linearly interpolating between, using the software package AnalySeries [Paillard et al., 1996]. In the modern coral, SST data as extracted from a 1° by 1° grid box (GISST 2.3b; Parker et al. ) was matched to the coral Sr/Ca. For BG831, the monthly climatological average SST served as a substitute for monthly SST data. Checking the age models visually against annual banding evident in the X-radiographs ensured accuracy of our depth to time conversions.
5.1. Age of BG831
 Sample BG831 has a 230Th age of 450 kyr ± 20, a 234U/238U age of 612 kyr ± 15 and an initial δ234U value of 96 ± 7. The lack of concordance between the 230Th and 234U/238U ages is probably caused by leakage of 234U related to alpha-recoil processes [Edwards et al., 1991]. Alpha-recoil processes would yield calculated ages that are older than true ages so that the calculated ages of sample BG831 should be considered maximum ages [Taylor et al., 1994]. The initial δ234U value of 96 ± 7 for BG831 is substantially lower than the initial δ234U values modern coral samples (∼150 ± 10; Edwards et al. ; Bar-Matthews et al. ; Henderson et al. ) and is indicative of open-system behavior with respect to uranium, but not necessarily other elements within the aragonite crystalline structure (i.e., uranium can exhibit open-system behavior and strontium can exhibit closed-system behavior; Bar-Matthews et al. ).
 Additional coral samples constrain the age of sample BG831 to be younger than ∼370 kyr. A coral sample (831B-18R-2, 44–46; 238.89 mbsf) from 59 cm deeper in the core than sample BG831 has an apparent age of 270 kyr, a minimum age of 180 kyr and a maximum age of 370 kyr based on Sr isotopes [Quinn et al., 1994]. A coral sample from 85.22 m deeper in the core than sample BG831 (831B-27R-CC, 12–14; 323.52 mbsf) has nearly concordant ages of 371 ± 9 kyr (230Th) and 393 ± 11 kyr (234U/238U) and an initial δ234U value of 159 ± 7. The age of sample BG831 can also be estimated using a stratigraphic forward model (see section 4.2). Tectonic subsidence rates can be calculated given the present elevation of a sample, the age of a sample, and the elevation of sea level at the time of coral growth [Imbrie et al., 1984; Shackleton, 2000]). A minimum estimate of subsidence rate is 3.41 m/kyr ([1389.52 − 10.05]/[393 + 11]); an average estimate of subsidence rate is 3.50 m/kyr ([1389.52 − 47.95]/[[362 + 404]/2]); and a maximum estimate of subsidence rate is 3.66 m/kyr ([1389.52 − 65.42]/[371 − 9]). This range of subsidence rates is used as input to a stratigraphic forward model to calculate the age of coral sample BG831; minimum likely age, 335 ka; mean likely age, 341 ka; maximum likely age, 374 ka (Figure 3).
5.2. Assessment of Diagenetic Alteration
 Stratigraphic forward modeling can be used to demonstrate that coral sample BG831 was quickly submerged below the range of sea-level fluctuations shortly after deposition (Figure 3). Hence, sample BG831 has not likely ever been subaerially exposed nor has it been subjected to the ravages of freshwater diagenesis. Any postdepositional alteration of sample BG831 is likely restricted to those processes attendant with residence in marine waters and pore fluids.
 Sample BG831 was examined for physical evidence of diagenetic alteration. X-ray diffraction analyses revealed only the presence of aragonite; no calcite peaks were identified, as expected from an originally aragonite sample that has never been subaerially exposed. Scanning electron microscopy (SEM) revealed no secondary cements (aragonite or calcite) in sample BG831. SEM images of a modern Porites coral sample are strikingly similar to those of the fossil coral sample BG831 (Figure 4). Examination of a petrographic thin section of sample BG831 documents similar preservation between modern and fossil coral; neither secondary cements nor recrystallization are discernable. A photomicrograph of the BG831 thin section is strikingly similar to a thin section photomicrograph of a well-preserved fossil coral from Papua New Guinea [McGregor and Gagan, 2003, Figure 3].
5.3. Geochemical Data
 The stable isotope ratios and Sr/Ca ratios used in this study were obtained for the entire 12 cm length of BG831 and the first 26 cm of MCB. The data from BG831 and MCB both show clear annual cycles in all of the measured geochemical variables. BG831 grew at an average rate of 9.8 mm/year and MCB grew at 20.0 mm/year. All significance tests and confidence intervals were calculated with same the number of degrees of freedom as years in the records. This assumes that there is one independent data point per year in these serially correlated records.
 Mean Sr/Ca ratios in the fossil and modern corals are significantly different with greater than 95% confidence (Figure 5). The mean Sr/Ca ratios for BG831 and MCB are 8.986 ± 0.018 mmol/mol and 8.881 ± 0.016 mmol/mol respectively (1σ). However, the mean annual cycles are very similar (Figure 6). The average annual Sr/Ca range is 0.13 ± 0.05 mmol/mol and 0.11 ± 0.04 mmol/mol (2σ) for BG831 and MCB, respectively. Calibration of the monthly-resolved modern coral Sr/Ca record 1980–1992 to GISST 2.3b SST over the same time period yields a slope of −0.050 ± 0.05 mmol/mol °C−1 and a good correlation (r = −0.78).
 BG831 has a mean oxygen isotope ratio of −4.25 ± 0.07‰ VPDB (2σ, Vienna Peedee belemnite), whereas the mean MCB value is −4.99 ± 0.13‰ VPDB (2σ, Figure 5). They are different at the >99.9% confidence level. The annual δ18O range averages 0.27 ± 0.10‰ in BG831, compared to a value of 0.45 ± 0.23‰ in MCB (Figure 6). Part of the difference between the two δ18O means must be due to the mean temperature difference indicated by the Sr/Ca data. Removing the Sr/Ca temperature signal from the oxygen isotope offset with the Sr/Ca-SST calibration above and the δ18O-SST calibration of Gagan et al.  leaves 0.38‰ to be explained by a combination of salinity and ice volume. One cannot determine from this residual δ18O value alone whether it results from a small positive change due to increased ice volume and a small positive change due to increased salinity, or a larger positive change due to increased ice volume and a negative change due to decreased salinity. A model is put forth in section 6.2 to examine these possibilities.
6.1. Geochemical Evidence for Fossil Coral Preservation
 The Sr/Ca, oxygen isotope, and carbon isotope data provide good evidence that the coral has not undergone significant diagenesis. Precipitation of secondary calcite and recrystallization of coral aragonite to calcite causes a sharp decrease in Sr/Ca values (toward 0.4–2 mmol/mol from 8 to 9 mmol/mol; Katz et al. ; McGregor and Gagan ). The large difference in Sr/Ca between inorganic calcite and coral aragonite implies that a 1% contamination of coral aragonite by inorganic calcite would cause significant irregularities (up to 1.8°C) in the seasonal cycle [Enmar et al., 2000; McGregor and Gagan, 2003]. Thus the retention of smooth annual cyclicity in Sr/Ca ratios in the fossil coral is consistent with preservation of a primary geochemical signal. Secondary aragonite has a slightly higher Sr/Ca ratio than coral aragonite [Enmar et al., 2000; Müller et al., 2001]. Contamination of a coral aragonite sample with 2% secondary aragonite, an amount unlikely to be missed by petrographic inspection, would cause a SST error of ∼−0.5°C, about equal to the error in the calculations to derive temperature (see section 6.2). Calcitization of coral aragonite in the meteoric environment changes the original oxygen and carbon isotope values of the coral aragonite [e.g., Pingitore, 1976; Martin et al., 1986] and disrupts the seasonal cycle as recorded by geochemical variation in the coral skeleton [e.g., Hughen et al., 1999; Tudhope et al., 2001]. Inorganic aragonite or high-magnesium calcite precipitation from marine waters would also increase the carbon isotope values by several per mil [James and Choquette, 1983; McConnaughey, 1989], but such an increase is not observed in the BG831 carbon isotope record. The balance of evidence causes us to reject the hypothesis that sample BG831 has been significantly altered by postdepositional processes.
6.2. Modeling the Pleistocene Mean Climate State
 Modern climate reconstructions essentially utilize the relationships between Sr/Ca, δ18O, salinity, and temperature to solve two equations for two unknowns. Sr/Ca is used to calculate temperature, which is then input into the δ18O temperature equation, theoretically yielding δ18Osw, a variable directly related to salinity. However, over geologic timescales these relationships may not be so straightforward.
 Unlike modern climate reconstructions, changes in seawater chemistry must be accounted for when interpreting Pleistocene coral records. Over glacial/interglacial timescales, the oxygen isotopic composition of seawater changes with ice volume due to the effects of preferential evaporation and sequestration of 16O in glacial ice. This effect is well documented and the global average is 1.0 ± 0.1 ‰ for 120 m (sea-level equivalent) of ice build-up [e.g., Schrag et al., 2002]. The Sr/Ca ratio of the ocean may also change as a function of ice volume because a low sea level exposes Sr-rich shelf carbonates to meteoric dissolution and recrystallization [Stoll and Schrag, 1998]. In contrast to δ18Osw, seawater Sr/Ca variations over glacial/interglacial timescales are not well known. Model results of Stoll and Schrag  indicate that seawater Sr/Ca may have increased by 1–3% during glacial periods, but the change in seawater Sr/Ca since the last glacial maximum was more likely 0.5–1.1% [Stoll et al., 1999]. Some data agree with this model output, but more data are needed to verify the model predictions of the Stoll and Schrag  and Stoll et al. .
 A simple mathematical, mass balance model has been developed to investigate the effects of changing ocean chemistry on fossil coral Sr/Ca and δ18O. The following two equations were simultaneously solved under variable seawater Sr/Ca conditions and variable ice volume inputs:
where ΔSr is the difference between the modern Sr/Ca and fossil Sr/Ca ratios (fossil minus modern), and ΔO is similarly the difference between modern and fossil δ18O. The variables ΔT, ΔI, and ΔS are changes in temperature, ice volume, and salinity respectively, from modern to ancient times. The partial derivative terms are empirically determined and the values used are listed in Table 1.
Table 1. Definition of Variables Used to Solve Equations (1) and (2) for Sea-Surface Temperature, Ice Volume, and Sea Surface Salinity
Change in coral δ18O (‰ VPDB) per 1°C change in SST.
Change in coral δ18O (‰ VPDB) per 1 m change in sea level.
Change in coral δ18O (‰ VPDB) per 1 psu change in salinity.
Change in coral Sr/Ca (mmol/mol) per 1°C change in SST.
Change in coral Sr/Ca (mmol/mol) per 120 m change in sea level.
 Two constraints on the equations are essential for limiting the domain and range of the equations. First, we know that the maximum change in ice volume over glacial/interglacial cycles is equivalent to about 120 m of sea level and that we are currently at a sea-level highstand (ice-volume low). Sea level can be further constrained from the age of the sample. Results of stratigraphic forward modeling suggest that BG831 grew while sea level was between 119 and 20 m below present sea level (Figure 3). Thus the ΔI term is constrained to between 20 and 119 m. Second, seawater Sr/Ca could not theoretically change more than 5% in any glacial/interglacial cycle and is more likely to be <3% [Stoll et al., 1999]. The mathematical model was run several times, each time the ∂Sr/∂I term was held at different values.
 The full range of possible model results of SST and SSS are illustrated in Figure 7. Straight lines in Figure 7 denote different values for the ∂Sr/∂I term in the model, the boundaries of which are limited by maximal and minimal ice volumes. All of the lines converge on a single point that represents seawater Δ18O and Sr/Ca values the same as today (i.e., no change in ice volume). Increasing the amount of Sr/Ca change on glacial-interglacial cycles causes an increase in the temperature uncertainty, whereas decreasing the amount of Sr/Ca change on glacial-interglacial cycles causes an increase in the uncertainty of salinity.
 The model output (Figure 7) must be interpreted in light of other information to determine a mostly likely paleoclimate solution from the full range of model possibilities. A light gray parallelogram in Figure 7 represents temperature and salinity solutions limited by the age model and sea-level considerations. Thermodynamic processes may limit mean annual SST to no more than about 30°C [Ramanathan and Collins, 1991; Cubukcu and Krishnamurti, 2002]. This is mentioned because it may be useful with other data in other locations although it is not a useful constraint in this case, since the modern mean annual SST at Santo is 27.7°C and the maximum SST increase possible is close to the upper limits of the model.
 In this case, we use a less commonly used coral temperature proxy, Mg/Ca ratios, to constrain our model. Coral Mg/Ca can be used because seawater Mg/Ca is not thought to vary on glacial/interglacial timescales [e.g., Hastings et al., 1998]. The difference between the modern and the fossil coral mean Mg/Ca indicates that temperatures were 2.4 ± 0.5 °C cooler when the fossil coral grew, using the in situ temperature calibration of Quinn and Sampson . A slightly darker shaded parallelogram represents the model results consistent with the Mg/Ca temperature and the error associated with the temperature calculations. The area of overlap between the Sr/Ca and δ18O defined temperature and salinity, and the Mg/Ca defined temperature, denoted by the crosshatch pattern, represents the best estimate of surface ocean conditions ∼350 ka. Data from fossil coral BG831 indicates that the surface ocean ∼350 ka was ∼2°C cooler. Salinity estimates are not as well constrained as SST estimates because of the age range of the fossil coral permits a wide range of sea-level possibilities (119 m to 20 m lower than the present). Our best estimate of SSS ∼ 350 ka is that it was most likely 0–2 psu fresher than today. The 0–2 psu salinity change corresponds to an ice volume ≥43 m lower than the present; a condition that is met for ∼73% of the age range of the fossil coral. Furthermore, the central age estimate of the fossil coral of 341 ka, the most probable true age of the coral, corresponds to an ice volume of −114 m with respect to modern.
 Lacking the additional SST constraints derived from coral Mg/Ca, the potential of changing seawater Sr/Ca on glacial/interglacial timescales contributes the most uncertainty in the SST and SSS determinations. As an example, if a paleo-sea-level elevation was known to be 120 m below present, then there would still be a SST and SSS uncertainty of 5.1°C and 3.7 psu, respectively, simply from not knowing if seawater Sr/Ca at 350 ka was the same as today or if it was greater than today's value by up to 3%. This demonstrates that accurate estimates of past seawater Sr/Ca values remain the largest obstacle to the accurate reconstruction of tropical SST in the past using Sr/Ca and δ18O in pristine fossil corals.
 Mg/Ca determinations on fossil coral samples may provide a means by which to investigate the magnitude of glacial/interglacial changes in seawater Sr/Ca. Coral Mg/Ca-SST estimates from BG831 imply that seawater Sr/Ca ∼350 ka was not considerably different from modern values. If true, this contrasts with previous seawater Sr/Ca estimates based on model results and foraminiferal data [e.g., Stoll and Schrag, 1998; Stoll et al., 1999]. More work is clearly needed to better define the role that fossil coral Mg/Ca determinations might play in constraining glacial/interglacial changes in seawater Sr/Ca.
 Propagating the error in equations (1) and (2) [Beers, 1957, equation (37); Schmidt, 1999], demonstrates that the model SST and SSS values are robust. The exact magnitude of the standard error is dependent on the ice volume (ΔI) and temperature (ΔT) derived from the model, but is close to 0.5°C and 0.5 psu in all cases. The largest sources of error are the lengths of the coral geochemical records, specifically the variance in the mean value of each record. The second largest source of error is the ∂Sr/∂T relationship. Essentially, the more sensitive the paleothermometer, the greater the possible error associated with it.
Equation (2) assumes that δ18Osw is only a function of global ice volume and local salinity (due to mixing with meteoric waters of unchanging isotopic compositions), but there are other possibilities. One could change the relationship of δ18Osw and SSS by changing the isotopic composition of meteoric waters either by changing the source location for local precipitation or by changing the temperature of that source location [Charles et al., 2001]. A model of the isotopic composition of precipitation during glacial times (forced with both CLIMAP SST and with cooler tropical SST fields) indicates that the δ18O of precipitation around Santo would increase slightly during glacial times [Charles et al., 2001]. Similarly, the δ18O of precipitation in the eastern tropical Pacific would also generally increase, and that might increase the isotopic composition of the South Equatorial Current, which bathes Santo today [Reverdin et al., 1994; Qu and Lindstrom, 2002]. The coral may record an isotopic enrichment due to a combination of increased isotopic composition of rainfall or increased isotopic value of the South Equatorial Current.
 How representative is a 13-year record of SST relative to multidecadal to century long SST records? Is it possible that the environmental changes evident in the fossil coral only represent a short-term shift rather than the mean state over a longer period? This issue is addressed in two ways. First, 13-year moving averages of monthly Sr/Ca and δ18O data from the entire modern Santo coral (1928–1992; data not shown) documents that the mean for any 13-year period is never outside of the 95% confidence level on the mean Sr/Ca and δ18O calculated from the full records (8.93 ± 0.03 mmol/mol and −4.85 ± 0.12 for Sr/Ca and δ18O respectively). Second, 13-year moving averages of monthly SST data for the 20th century were used to demonstrate that this proxy-based conclusion is also robust when instrumental data are used. The mean SST value over the same time period was 27.57 ± 0.57°C and similarly to the results from the proxy record, the minimum and maximum 13-year mean were well within the error on the total mean value (27.27°C and 27.79°C for minimum and maximum respectively). Thus a 13-year record is representative of multidecadal to century long SST records and the mean value is not likely to be strongly affected by higher frequency oscillations.
6.3. Paleoclimatic Implications of Fossil Coral Data and Model Results
6.3.1. Paleoclimate of the Santo Region
 The seasonal range in δ18O is smaller in the fossil record than in the modern record, whereas the seasonal range in Sr/Ca is the same in both fossil and modern corals (section 5.3). These differences in the seasonal cycles of δ18O and Sr/Ca in the fossil and modern corals may be used to investigate the source of seasonal variability at Santo.
 Seasonal-scale salinity changes in the Santo region today are due to the migration of the SPCZ and the SSS front associated with it. The smaller amplitude annual cycle of δ18O in BG831 relative to the modern must be due to seasonal hydrologic variations in ancient times because ice-volume effects are not possible on seasonal timescales, and the difference cannot be attributed to temperature because the seasonal Sr/Ca changes (i.e., SST changes) are similar in amplitude between modern and fossil records. There are two possible hydrologic scenarios that are consistent with the reduced amplitude δ18O cycle in the fossil coral relative to the modern coral: a decrease in rain during the rainy season or an increase in rain during the dry season. The latter scenario is more consistent with our modeling results, which suggest a lower mean salinity ∼350 ka with respect to today. One mechanism by which to increase rainfall during the austral winter is to limit the northward migration of the SPCZ during this season. This limited migration of the SPCZ at 350 ka contrasts with present-day austral winter migration of the SPCZ, which occurs in response to increasing Northern Hemisphere temperatures and convergence of the wind fields in a more northerly location. A more southerly position of the SPCZ at 350 ka is consistent with increased meridional thermal gradients in the Northern Hemisphere during glacial times [Trent-Staid and Prell, 2002].
 Could the fossil coral represent a single decadal-scale shift in the SPCZ as has occurred this century due to shifts in the Interdecadal Pacific Oscillation? The difference in δ18O between the modern and the fossil is much larger than δ18O variations recorded by the modern coral (1928–1992) which are thought to originate from modern shifts in rainfall linked to IPO movements. It is therefore unlikely that the fossil coral represents a similarly scaled event, but instead represents a larger spatial displacement of organized convection.
6.3.2. Paleoclimate of the Tropical Indo-Pacific
 The age estimate for fossil coral BG831 allows for the possibility that this coral grew when sea level was 20 to 119 m lower than today. The age estimate is centered close to the minimum sea level of MIS 10, so there is a higher likelihood that the coral grew during the glacial period than at the extreme edges of the age estimate, when sea levels were only 20 m lower than today. Additionally, 2–3°C of tropical cooling indicated by the coral geochemical data is more consistent with coral growth during a glacial period. We find it most reasonable therefore, to compare our results with studies of other glacial periods.
 The 2–3°C cooling at Santo, as indicated by the Mg/Ca ratios and supported by the Sr/Ca-δ18O model, is consistent with other paleoclimate studies in the Pacific. Foraminiferal Mg/Ca ratios indicate ∼3°C cooling in the western tropical Pacific during the last glacial maximum (LGM; MIS 2) and other glacial stages over the past 450 kyr (tropical South China Sea [Stott et al., 2002]; Ontong Java Plateau [Lea et al., 2000]; Sulu Sea [Oppo et al., 2003]). Further evidence for cooling of this area during the LGM comes from modern analog technique-based faunal analysis at ODP site 828A, just west of Espiritu Santo Island, which suggests at least 1–2°C cooling [Barrows et al., 2000]. There has been much discussion of how much the tropics change during glacial periods (see Trent-Staid and Prell  for a recent summary). Our data fall into the growing consensus that the tropical Pacific generally cooled by 2–3°C during glacial periods.
 Salinity decreases in the surface waters of the western equatorial Pacific (WEP) during glacial periods have been suggested by Lea et al.  based on estimates of glacial seawater δ18O that were lower than estimates of the global ocean value and those from the eastern equatorial Pacific (EEP). Our model results (Figure 7) are consistent with decreased surface salinity in the WEP during glacial periods. To have more than a slight salinity increase ∼350 ka, one must assume that fossil coral BG831 grew during an interglacial period. While this is possible, stratigraphic modeling results and the depressed SST indicate that this scenario for coral growth is not likely.
 Planktonic foraminiferal δ18O and Mg/Ca ratio data from the WPWP indicates a moderate salinity increase during the LGM along the eastern edge of the Indonesian archipelago [Stott et al., 2002]. At first, the data from Stott et al.  appear to directly conflict with the data from this study and that of Lea et al. , which indicate decreasing salinity in the WEP during glacial periods. Stott et al.  attempted to reconcile this conflict by pointing out that the Ontong Java Plateau (OJP) receives large increases in rainfall during ENSO warm phase events as the locus of atmospheric convection shifts eastward from Indonesia. If atmospheric convection also shifted eastward during glacial periods, seawater around Indonesia might become more saline, and seawater at other locations (such as the OJP) might become less saline. With this scenario, Santo would also experience increased salinity, as it does during an ENSO warm phase event, but the fossil coral data indicate conditions were actually less saline. A scenario that fits all of the data is that the northward, seasonal migrations of both the ITCZ and the SPCZ ∼350 ka was more limited relative to today (i.e., both convergence zones remained closer to the austral summer configuration throughout the year). This would have the effect of shifting the mean salinity fields to the south, freshening Santo and the Ontong Java Plateau while leaving Indonesia saltier relative to modern conditions.
 Fully coupled atmosphere ocean general circulation models (AOGCMs) have recently been used to simulate LGM climate phenomena [Bush and Philander, 1998, 1999; Hewitt et al., 2001; Kitoh et al., 2001; Kitoh and Murakami, 2002; Kim et al., 2003]. Seasonality in the convergence zones has not yet been studied, however some patterns can be identified. Tropical SSTs are lower in the models, as is average humidity. The hydrologic cycle is generally diminished, but each model seems to have some regions where tropical convergence increases along with the P-E balance. The model used by Kim et al.  displays stronger trade winds and an increase in Walker circulation coincident with a decrease in ITCZ activity. The SPCZ (the upward limb of Walker circulation) is shifted slightly westward and SSS decreases compared to modern day beneath the convergence zone. This scenario can explain decreased average SSS at Santo and increases in SSS in Indonesia, an area usually doused with rainwater from the ITCZ, although it does not explain the freshening of the Ontong Java Plateau at the LGM.
 Model evidence for increased trade winds is further supported by other paleo data. Factor analysis of planktonic foraminifera data indicate an increased thermocline tilt, and thus stronger trade wind activity during the LGM [Andreasen and Ravelo, 1997]. SST data from foraminiferal assemblage work similar to the CLIMAP project [CLIMAP Project Members, 1976, 1981] suggest an increased zonal thermal gradient across the Pacific Ocean, further supporting an increase in the trade winds [Trend-Staid and Prell, 2002]. Admittedly, the coupled models result in SSTs that are cooler than the foraminiferal data indicate, but otherwise the general patterns are similar.
 In summary, reconstructing the hydrologic response to glacial forcing is a complex issue. Much evidence exists for stronger trade winds that would tilt the tropical thermocline to the west and encourage stronger Walker circulation that could in turn decrease salinity in surface waters beneath the SPCZ. Additional “paleo” data are needed to constrain the east-west thermal and salinity gradients across the Pacific Ocean during glacial periods, and to help resolve the discrepancies in current model and climate proxy results.
6.4. Modern and Pleistocene Interannual Variability
 The response of coral δ18O and Sr/Ca to interannual SST and SSS forcing must be demonstrated in the modern records before interannual relationships in the fossil record can be considered. ENSO warm (cold) phase events are manifest in this region as cool (warm) and dry (wet) anomalies [Delcroix, 1998]. Corals respond to lower (higher) SST and higher (lower) SSS values attendant with an ENSO warm (cool) phase by precipitating skeletal material that has higher (lower) δ18O and Sr/Ca values.
 The δ18O and Sr/Ca anomaly plots from MCB are very similar to the SSSA and SSTA plots from Santo (Figure 2). Coral δ18O and Sr/Ca both increase during the 1983, 1987, and 1991 El Niño events, and they decrease in the 1989 La Niña event. Additionally, modern coral geochemistry provides details about the characteristics of each event that we know to be true from the instrumental record. The 1983 warm phase event was characterized in the region by a large salinity anomaly and a small temperature anomaly. Salinity in the western Pacific lagged behind the eastern Pacific warming (represented by gray bars in Figure 2) during this event. The timing and relative magnitude of temperature and salinity anomalies in the proxy data match the instrumental records well (Figure 2). The 1987 ENSO event was manifest in this region by large temperature anomalies and smaller salinity anomalies; a scenario well recorded by the proxy data (Figure 2). By using the definition of ENSO events described in section 3, one can easily identify modern ENSO events in the geochemical proxy records, and even further describe details about the temperature and salinity anomalies associated with these events.
 The Sr/Ca and δ18O time series in the fossil coral from Santo contain two geochemical excursions that are consistent with our definition of ENSO events as recorded in the geochemistry of a modern coral (see section 3; Figure 2). The first “paleo-ENSO” event is primarily manifest as a temperature anomaly that lasts for essentially 16 months. The positive Sr/Ca anomaly reaches the 1.5-σ threshold criterion in June of year 2, and continues through September of year 3, with a 4-month period in between where the anomaly remains positive, but drops below the threshold criterion of 1.5 σ. Meanwhile, the δ18O anomaly does not increase until September of year 2, but remains primarily positive through September of year 3. The maximum SSTA is −1.8°C.
 The second “paleo-ENSO” event lasts 17 months and strictly speaking should be split into two events by our definition. Both proxies have anomalies during year 6 and again in year 7. The event is primarily recorded by a δ18O anomaly, indicating that salinity deviations dominated this event. δ18O reaches the 1.5 σ threshold criterion in July of year 6, drops below the threshold criterion in December of the same year, and then bounces back again in June through November of the following year. The Sr/Ca anomalies are minimal over this time, but they do show minor positive anomalies associated with the larger δ18O positive anomalies.
 These two climate events, interpreted as “paleo-ENSO” events, appear to agree with the findings of Tudhope et al. , who reported decreased interannual variability in coral δ18O during glacial periods at Papua New Guinea. The absolute amplitudes of the δ18O and Sr/Ca anomalies in BG831 are smaller than δ18O and Sr/Ca anomalies observed in the modern coral. However, as a percentage of the average seasonal range, the anomalies are very similar to modern events. BG831 is too short of a record to divulge significant information about the frequency of “paleo-ENSO” events.
 ENSO variability may be controlled by the ∼20 kyr precessional cycle [Clement et al., 1999, 2000]. Seasonal variations in insolation are especially important in modulating ENSO in the geologic past [Clement et al., 1999; Kukla et al., 2002] and one would expect lower frequency of El Niño events at times when perihelion occurs in March and aphelion occurs in September of minimal seasonality in equatorial insolation (March minus September insolation is at a minimum). Such minima in the equatorial seasonality of equatorial insolation occurred at ∼350 ka and during the last glacial period around 30 and 55 ka [Laskar, 1990].
 We have compared geochemical variations in a fossil coral with those from a modern coral from Espiritu Santo, Vanuatu to address two fundamental questions:
 1. What was the background climate state of the tropical Pacific, in terms of temperature and salinity, during ∼350 ka?
 2. Is there evidence of ENSO activity at ∼350 ka? This study confirmed that modern Porites spp. corals from Santo record significant surface-ocean variability in the δ18O and Sr/Ca of their skeletons. Hydrologic and thermal details of specific ENSO events are in fact resolvable in the modern coral geochemical record.
 This study also confirmed that climatologically significant information can be preserved in the skeletal aragonite of fossil corals over long (105 years) timescales. Changes in seawater chemistry can have large affects on the climatological interpretation of fossil coral records and must be accounted for even if the geochemical signature of environmental variability is preserved in the fossil coral skeletons. A simple mathematical model using empirically derived relationships joining δ18O and Sr/Ca with seawater Sr/Ca variations, seawater δ18O variations, SST, and SSS, facilitates the interpretation of coral geochemical records.
 Specific conclusions include:
 1. Coral proxy data and mathematical modeling of Pleistocene mean SST and SSS results in temperature estimates ∼2°C cooler and salinity 0–2 psu fresher than present-day conditions. Our fossil coral data and modeling results preclude warmer SST and substantially higher SSS at Santo at ∼350 ka. Accurate estimates of past values of seawater Sr/Ca remain the largest obstacle to the accurate reconstruction of tropical SST in the past using Sr/Ca from pristine fossil corals.
 2. Seasonal SST ranges reconstructed ∼350 ka are very similar to modern seasonal SST ranges. In contrast, seasonal hydrologic variations ∼350 ka are reduced in amplitude relative to the modern. The reduced seasonal SSS variations and increased SSTs near Santo are interpreted as evidence that the SPCZ at ∼350 ka did not migrate as far northward during austral winter as it does today.
 3. ENSO or ENSO-like interannual variations are not unique to interglacial time periods, nor just the past 130 kyr, but are also present in the tropical southwestern Pacific between 335 and 374 ka.
 The authors thank Anthony Greco for assistance with SEM analyses, Thierry Corrège for his help with the initial method development for the ICP, and Ethan Goddard for all of his tireless efforts in the analytical laboratories at USF. The manuscript was enhanced through interactions with Benjamin Flower, David Hollander, and Gary Mitchum and through the comments of two anonymous reviewers.