SIGNIFICANCE OF THE ALKENONE RECORD

We assess next how to interpret alkenone results from Hole 1002C in light of oxygen isotope data and other constraints. Factors such as global changes in sea level, which regulate the depth of the sill connection between the Cariaco Basin and the open ocean; changes in local climate, including winds and freshwater runoff; and changes in regional surface temperatures and salinities all may affect the isotope and alkenone records to varying degrees. We do not expect salinity variations to influence the U^k´₃₇ temperature estimates (see Sonzogni et al., 1997), but they may indirectly modulate alkenone concentrations in the sediments by affecting patterns of inflow and outflow across the sills of the Cariaco Basin, and hence the nutrient and oxygen budgets of the basin. As Lin et al. (1997) point out, in the modern "estuarine" hydrography, the Cariaco Basin imports relatively nutrient-rich water at its sill depth of 146 m and exports nutrient-poor surface water into the open Caribbean (see also Richards, 1975). This circulation pattern promotes both high production of haptophyte algae in the water column and high preservation of alkenones in the sediments.

The alkenone concentration and paleotemperature estimates do not look very similar (see Fig. 4, Fig. 5). We assume the alkenone concentrations reflect haptophyte productivity, subject to possible influences of variable preservation and changing dilution by other sediment components (Villanueva et al., 1998). We have not converted our concentration estimates to fluxes because the order-of-magnitude changes in total C₃₇ ketones that define the major Milankovitch-scale pattern (Fig. 4) are too large to explain by dilution/concentration mechanisms. The changes in ketone abundance must primarily reflect alterations in the production or preservation of these compounds over time. Furthermore, we are wary of introducing artifacts into the time series because of the potential error in correlating a planktonic isotope record with clear local influence with the global oxygen isotope time scale. As suggested above in our discussion of core-top alkenone paleotemperature analyses, we consider the U^k´₃₇ record to largely record changes in mean annual SST over the past 160 k.y. in the Cariaco Basin.

The lack of coupling between alkenone temperatures and amounts rules out simple interpretations of the data (e.g., that high upwelling rates result in cold temperatures). Instead, each data set has its own logic. Global sea level appears to have played a decisive role in controlling the time variations in alkenone concentration. As Figure 6 demonstrates, the alkenone abundances largely correlate with variations in total organic carbon. A model in which the sill depth sets the tendency of the Cariaco Basin to import nutrients and to establish bottom-water anoxia seems to explain most of the variations documented here. In addition, secondary effects (such as the intensity of upwelling-favorable winds) may have played a role in determining how productive and anoxic the Cariaco Basin became during sea-level highstands, setting variations like those seen in marine isotope Substages 5A and 5C above Holocene levels of alkenone concentration. One feature of note is the low variation in long-chain ketone abundance during marine isotope Stages 2, 4, and 6 relative to the remainder of the record. Alkenone concentrations remain uniformly low during these intervals, despite the frequent presence of "faint to well-developed" laminations (Shipboard Scientific Party, 1997). This may suggest very low variability in the exchange of waters across the shallow sills during glacial maxima. Significant fluctuations in alkenone concentration during MIS 3 and 4 may record intermittently greater connections to the open ocean during these intervals of somewhat higher sea level.

Two observations of the paleotemperature record over the full glacial cycle analyzed here seem particularly significant: the absence of cooling during the maximum extent of glaciation and the higher than modern temperatures inferred for Substage 5E. We first defend why the alkenone reconstructions of SST for the Cariaco Basin provide plausible results and then consider the implications for such an interpretation for isotopic data derived from G. ruber. To identify possible problems in interpreting the alkenone paleotemperature results, we must ask ourselves what factors might make at least some of the estimates unrepresentative. The most likely of these would be seasonal variations in alkenone synthesis and changes in the assemblage of alkenone-synthesizing haptophyte species. We believe that there is little basis for either objection in this setting.

To what degree could alkenone temperatures at Site 1002 be strongly biased by upwelling signals? This problem breaks down into two related questions: (1) Could variations in coastal upwelling be significant enough to produce SSTs in the Cariaco Basin unrepresentative of the open ocean? and (2) Could alkenone paleotemperatures be biased to record only upwelling conditions? First, in order to determine quantitatively by how much seasonal upwelling reduces the average temperature of Cariaco Basin, we can compare the magnitude of the upwelling effect in the basin to the adjacent open Caribbean Ocean. The two closest grid points from the World Ocean Atlas (Levitus, 1994) give identical values of 26.6°C, compared to our estimate of 25.1°C for Cariaco SST. Coastal upwelling therefore reduces annual SST in the Cariaco Basin by ~1.5°C relative to the open ocean; if upwelling were to shut off, SST would rise by about this amount. Decreased coastal upwelling during the last glacial period might mask some regional cooling in the Caribbean during the last glacial period, but it could not by itself hide the 4°-5°C tropical cooling hypothesized by some (e.g., Guilderson et al., 1994).

There is also little basis to believe that alkenone temperatures are tightly correlated with upwelling temperatures, as we demonstrated above. Furthermore, alkenone temperatures do not deviate significantly from mean annual SST in other seasonal upwelling environments (Herbert et al., 1998). Our best present ecological understanding suggests that alkenone synthesis is fairly constant in tropical and mid-latitude settings, in part because haptophyte production may be suppressed by competition with diatoms during periods of peak export production (Holligan et al., 1993). We therefore conclude that although variations in upwelling intensity may provide a mechanism for changing SST and alkenone paleotemperatures in the Cariaco Basin on the order of 1°-2°C, they do not promote a strong bias toward the U^k´₃₇ signal.

Because alkenones appear to be produced within the mixed layer and with little seasonal bias in the late Holocene, past shifts in the depth of production could promote a bias toward U^k´₃₇ temperatures. This bias should be one-sided: the depth of production cannot shift upward relative to present-day conditions, but it could shift downward into the thermocline under different hydrographic conditions in the past. Examples of such subsurface production of alkenones and their cold bias have been documented for the North Pacific by Prahl et al. (1993). Note that this caveat does not affect our "warm" temperatures (like those observed during the LGM), but it could enter into some of the colder temperatures determined during intervals of MIS 3-5 (Fig. 4, Fig. 5).

Although U^k´₃₇ temperature offsets may occur between different strains or species of haptophyte algae (Volkman et al., 1995), core-top studies over large regions support the use of a single calibration function similar to the Prahl et al. (1988) culture study. This is true even in regions in which species such as G. oceanica form an important part of the nannofossil assemblage (Herbert et al., 1998; Müller et al., 1998).

Sea level and connections between the Cariaco Basin and the open ocean may have played a significant—and perhaps counterintuitive—role in determining glacial-interglacial changes in SST inferred from alkenone results. In this regard, we caution against interpreting our record as necessarily representative of regional patterns in SST. Consider how SST might be determined during sea-level lowstands. With a sill depth of ~26 m, the Cariaco Basin could not import waters colder than mixed-layer conditions in the Caribbean Sea (Lin et al., 1997). Temperatures below sill depth would be determined either by the temperature of this water or by production of deep water within the Cariaco Basin itself, perhaps influenced by seasonal or interannual salinity variations. In this situation, upwelling would have little, if any, efficiency in depressing SST because the thermocline would almost certainly be weak compared to the modern condition. More generally, a weakened thermocline would mean that all processes that act to mix the upper layers of Cariaco Basin would have had little impact on SST during sea-level lowstands.

Our explanation, then, pictures a basin during glacial maxima (which may have been much cooler than the Cariaco Basin) largely isolated from the open ocean. With relatively high insolation at its latitude of 10°N and few mechanisms to reduce SST, the Cariaco Basin may have been relatively buffered to glacial-interglacial variations in temperature. Swings to low SST occurred instead during intervals of somewhat higher sea level (such as during MIS 3-5C) and particularly during glacial terminations (see Fig. 5). At times of rising sea level, the deeper sill connections may have provided the cooler deep waters necessary to significantly lower SST compared to modern temperatures.

Such an interpretation presumes that colder SST characterized the open ocean outside the gates of the Cariaco Basin. There is no alkenone data at present to support this idea, and faunal studies suggest little if any cooling in the Caribbean during the LGM (CLIMAP, 1984). Cooling of 4°-5°C is supported by Sr/Ca values in corals (Guilderson et al., 1994) and by isotopic results from Andean ice cores (Thompson et al., 1995). The alkenone results presented here may or may not represent a regional or local response of surface temperatures to glacial-interglacial climate change.

If our alkenone temperature estimates are correct, they imply substantial evaporation/precipitation/runoff contributions to the G. ruber oxygen isotopic record of Peterson et al. (Chap. 4, this volume). The glacial-interglacial isotopic amplitude of ~2.2 requires a significant combination of temperature and/or salinity effects compared with the global ice-volume effect of 0.9-1.2 (Fairbanks and Matthews, 1978; Schrag et al., 1996). Very short-lived excursions of 1 or more during MIS 3-5C also require nonglobal influences on the surface isotopic record from the Cariaco Basin. We caution against point-by-point comparison of the alkenone and isotopic time series because samples are generally offset by tens of centimeters and because both data sets display high-frequency variations. Nevertheless, the alkenone paleotemperatures imply that ~1 of the glacial enrichment in oxygen isotopic composition of G. ruber must be caused by changes in the local runoff and evaporation/precipitation balance. This would require an increase in salinity at the LGM of ~2 using standard salinity ¹⁸O correlations (Craig and Gordon, 1965) and/or a reduction of local runoff whose effect on modern surface ¹⁸O values in the Cariaco Basin is not quantified.

This interpretation of the Cariaco ¹⁸O record is plausible but is not without problems. Lin et al. (1997) set an excellent standard for balancing the case for a favored model and considering objections to this model in order to explain variations in planktonic ¹⁸O. They documented the large amplitude of the G. ruber ¹⁸O signal from the LGM to the Holocene in the Cariaco Basin. In addition, the authors demonstrated that the foraminifers G. bulloides (an upwelling species) and N. dutertrei (representing deeper water conditions) display glacial-interglacial changes of only ~1.3, or nearly the expected ice-volume effect alone. Lin et al. (1997) interpreted the large enrichment in G. ruber isotopic values at the LGM to represent a 4°-5°C cooling of surface waters. They noted at the same time that the absence of isotopic enrichment in the other foraminiferal species raises problems in this interpretation. First, one would expect that bottom waters in the Cariaco Basin would actually have warmed during the LGM because the basin could not be ventilated by waters much below the mixed layer. This warming ought to partially offset the global ice-volume signal in the G. bulloides and N. dutertrei records. Second, if all of the G. ruber isotopic variance comes from cooling during the last glacial period, one obtains a reconstruction of a thermally homogeneous upper water column (for example, one can superimpose a 4°-5°C cooling at the surface on the profile shown in Fig. 2).

Our model reconciles isotopic results with alkenone paleotemperature estimates by requiring much higher salinities in the Cariaco Basin during sea-level lowstands than are found today. A large fraction of the salinity signal in this region today apparently originates from advection of Orinoco and Amazon discharges (Dessier and Donguy, 1994), which are modulated by the passage of the Intertropical Convergence Zone (ITCZ) (Hastenrath, 1990). Higher salinities during the glacial period would result from the greater physical isolation of Cariaco waters from the open ocean, from regional reductions in rainfall as ITCZ failed to advance as far northward as it does today, and from reduced freshwater input into the Cariaco Basin from local tributaries and from the major South American river systems. This model would require large changes in surface salinity but permit a combination of temperature and salinity variations at the depth and seasons represented by the G. bulloides and N. dutertrei records of Lin et al. (1997). In the latter examples, the isotopic effects of warmer deep waters and higher salinities may have canceled each other out, leaving the global ice-volume signal as the residual.

Several independent lines of evidence support this interpretation. Faunal studies fail to indicate surface cooling in Cariaco Basin sediments or in the adjacent Caribbean (CLIMAP, 1984; Lin et al., 1997). Planktonic foraminiferal faunas during the LGM are dominated by the tropical species G. ruber (Overpeck et al., 1989). Unusual benthic assemblages in the Cariaco Basin are similar to those found in abnormally high-salinity environments (Peterson et al., Chap. 4, this volume). Studies of the terrestrial paleoenvironment point to greater aridity in the region (Leyden, 1984, 1985; Margraf, 1989).

Our alkenone unsaturation time series also suggests that SST warmed during marine isotope Substage 5E to ~2°C warmer than present temperatures (Fig. 4). The unsaturation values lie near the end of the analytical range of the U^k´₃₇ index and past the highest growth temperature used by Prahl et al. (1988) in their culture calibration of the U^k´₃₇ index. Warmer than Holocene values for Substage 5E at Site 1002 are consistent with other alkenone determinations at open ocean locations, as follows: 1.5°C warmer in the northwest Arabian Sea (Emeis et al., 1995), 1°C warmer in the equatorial Atlantic, 1.5°C warmer at 11°S in the South Atlantic, 1.5°C warmer off the Congo Fan, 1.5°C warmer along the Angola Margin, 4°C warmer at the Walvis Ridge (Schneider et al., 1995), 3.5°C warmer in the tropical North Atlantic (Eglinton et al., 1992), 4°C warmer in Santa Barbara Basin (Herbert et al., 1995), and 3°C warmer in the North Pacific at ODP Site 1020 (S. Kreitz, unpubl. data).

In the end, studies like those reported here will have to be interpreted in light of similar records from the open Caribbean Ocean. The oceanography of the Cariaco Basin and its sensitivity to sea-level variations are unusual enough to make us cautious in inferring more than local significance to our alkenone time series. The presence of laminated sediments and high accumulation rates in the Cariaco Basin thus presents something of a paradox. Do they represent an unusual opportunity to observe the sensitivity of a tropical climate to global changes, or a system so highly amplified as to be unrepresentative of the open tropical ocean? Studies of oxygen-isotopic amplitude during the LGM from the tropical Atlantic are mixed: Curry and Oppo (1997) documented large anomalies from the Ceara Rise, whereas Stott and Tang (1996) found shifts consistent with only minimal cooling of surface waters. Milankovitch-scale sampling for ¹⁸O and alkenone paleotemperatures on pelagic cores outside the Cariaco Basin are needed to determine whether the general glacial-interglacial features presented here are representative of the region (in which case the high-frequency isotopic and U^k´₃₇ variations documented by Peterson et al. [Chap. 4, this volume] take on added significance) or whether the paleoceanography of the Cariaco Basin is largely a story of contrasts between the basin and the open ocean.