The distribution of radiocarbon in the glacial ocean

. Accelerator mass spectrometric radiocar-bon measurements on benthic foraminifera shells, picked from samples on which concordant ages were obtained on the shells of two species of planktonic foraminifera, reveal that the age of deep water in the equatorial Atlantic during glacial time was 675+80 years (compared to today's age of 350 years) and that the age of deep water in the South China Sea was 1670-Z_105 years (compared to today's value of 1600 years). These results demonstrate that the 1.3 to 1.5 times higher radiocarbon content of carbon in glacial surface waters of the Caribbean Sea reconstructed by Bard et al. [ 1990] was primarily the result of a higher global inventory of radiocarbon rather than a decrease in rate of mixing between surface and deep waters of the ocean. The results are also consistent with the


INTRODUCTION
During peak glacial time 20,000 to 14,000 years ago, the Earth was a very different place.Not only was it colder and more ice covered, but it was dustier and poorer in greenhouse gases.About 14,000 years ago a change occurred which created conditions more akin to those of today.Broecker and Denton [ 1989] postulate that this change was brought about by a reorganization of the entire ocean-atmosphere system.If so, then it is important to learn as much as possible about the manner in which the system operated during glacial time.Fortunately, a wealth of information regarding the patterns and rates of large-scale circulation in the ocean is available in deep-sea sediments.
To date the most definitive information in this regard comes from cadmium to calcium ratios in shells of benthic foraminifera [Boyle and Keigwin, 1987;Boyle, 1988a].These results reveal that during glacial time the pattern of circulation was much different than today's.The nutrient constituent maxima which in today's ocean lie at intermediate depth were shifted toward the bottom.The strong contrast between the nutrient content of deep water in the Atlantic and Pacific observed in today's ocean was smaller during glacial time.
Although not a recorder of any specific change in ocean operation, the lower CO2 content of the glacial atmosphere can only be explained through a major change in the interaction between the sea's mixing A property of importance to the understanding of today's rate of deep-sea ventilation, i.e., the 14C/C difference between surface and deep water, can now be reconstructed for glacial time.Radiocarbon measurements by the accelerator mass spectrometric (AMS) method on shells of foraminifera handpicked from deep-sea sediments offer the possibility that this reconstruction can be accomplished for the last 20 thousand years or so [Broecker et al., 1984].The idea is that the ratio of the 14C/C for benthic foraminifera to that for planktonic foraminifera coex- In this paper we present new radiocarbon results which broadly support the conclusions reached in previous papers and which overcome the uncertainties associated with the previous results from the tropical Pacific.

RESULTS
We use as a criterion for validity of each benthicplanktonic age difference measuremere that agreemere exist (at the 20 level) between the ages obtained on two separate species of planktonic foraminifera.The logic behind this strategy is that neither the averaging of discordant planktonic results nor the arbitrary selection of one result over the other is likely to produce a reliable estimate of the benthic-planktonic age difference.core shows that the 13C/C ratio was higher during glacial than during Holocene time (see Figure 5).As can be seen, the age differences for glacial age samples are slighfiy smaller than today's.
The relationship between the Holocene-glacial 1513C change recorded in benthics and the planktonic-benthic age difference for glacial time is shown in Figure 6.

DISTRIBUTION OF RADIOCARBON IN • GLACIAL OCEAN
An attempt at reconstructing the distribution of radiocarbon in the glacial ocean is shown in Figure 7.

Based on measurements on corals recovered off
Barbados by Fairbanks [1989], Bard et al. [1990] show that during peak glacial time a discrepancy of about 3200 years exists between the radiocarbon age and the radiothorium age scale.For example, a coral

IMPLICATIONS FOR GLACIAL CIRCULATION
In order to gain some insight as to how the deepwater radiocarbon ages depend on mixing rates, we employ our Pandora "something like the real ocean geochemical model" [Broecker andPeng, 1986, and1987].For this exercise we have converted the interbox transfer fluxes used previously into seven circulation loops (see Figure 8).In this way we can alter the magnitude of any one of the loops without having to concern ourselves with conserving water, for in this loop scheme the water fluxes are automatically balanced.In addition to the seven loops shown in Figure 8   One further complication must be considered.The •80/160 records for the Greenland ice cores suggest that millennium-long events punctuated the climate of the northern Atlantic region during much of glacial time [Dansgaard et al., 1982].A likely cause for these oscillations is the turning "on" and "off" of deepwater production in the Atlantic [Broecker et al., 1985[Broecker et al., , 1990b].If such alternations were occurring during the time period covered by our measurements (i.e., 13,000 to 19,000 years ago) then because of bioturbation we would obtain the average of the age of Atlantic deep water during the "on" and "off" parts

CONCLUSIONS
The results presented here suggest that the ratio of the radiocarbon age of deep water in the Pacific Ocean to the radiocarbon age of deep water in the Atlantic Ocean was smaller during glacial time than today.Taken together with the nutrient distribution reconstruction for glacial time, this result suggests that the mixing zone between North Atlantic Deep Water and waters in the Antarctic moved well northward from its present position at 30 ø to 40øS in the Atlantic.
isting in deep-sea sediments does not change with time during the glacial period.Hence radiocarbon measurements on coexisting benthic and planktonic foraminifera shells provide a measure of the 14C/C difference between surface water and water at the depth from which the core was taken.For convenience this difference is expressed as an age (i.e., the time required for the surface water 14C/C ratio to decay to that for deep water).In today's tropical Pacific this age difference is about 1600 years, while in the tropical Atlantic it is about 350 years.Using this strategy Andree et al. [1986] showed that over the course of Holocene time the age difference for waters at 2 km depth in the western Pacific was the same as today's to within the measurement uncertainty (i.e., -200 years).Based on measurements on a core from the Ceara Rise, Broecker et al. [1988a] suggested that the age of deep water in the western tropical Atlantic was roughly twice as great during glacial time than it is today.Shackleton et al. [1988] and Broecker et al. [1988a] obtained results suggesting that the age of deep water in the tropical Pacific was somewhat greater during glacial time (-2000 years) than it is today (-1600 years).These latter results are however far from conclusive.The Shackleton et al. [1988] results come from an area where upwelling currently influences the 14C/C ratio in the photic zone.Because of this, the planktonic results do not necessarily reflect conditions typical of the surface ocean.Further, they present only two results from peak stage 2 time (i.e., 20,000 to 14,000 years ago).Both suggest a lower age difference than those before 20,000 years ago.. Two of the three cores reported by Broecker et al. [1988a] come from areas of such low accumulation rate that the bioturbation-abundance couple [see Broecker et al., 1984] likely introduces biases in the age difference.The third core analyzed by Broecker et al. [1988a] showed a systematic (and unexplained) discordance between the ages obtained on two planktonic species, G. sacculifera and P. obliauiloculata.Broecker [ 1989] pointed out that the smaller surface to deep 14C/C ratio difference for the glacial Atlantic could be explained either by a slowdown of the Atlantic's conveyor or by an inversion of the Atlanfic's circulation.However, with an inverted glacial circulation he was not able to account for Boyle's [1988a] observation that although the interocean nutrient content difference was smaller during glacial time, waters in the deep Atlantic still had a smaller nutrient content than those in the deep Pacific.

Fig. 1 .
Fig. 1.Carbon and oxygen isotope composition of benthic foraminifera shells as a function of depth in western equatorial Ariantic core KNR110-50GGC [Curry et al., 1988].Also shown are the levels at which planktonic-benthic age differences were determined.At the left the age of the horizon is given (planktonic mean -400 years).At the right are the differences between the benthic age and the mean planktonic age.

Fig. 2 .Fig. 3 .Fig. 4 .Fig. 5 .Fig. 6 .
Fig. 2. Carbon and oxygen isotope composition of benthic foraminifera shells as a function of depth in western equatorial Atlantic core KNR110-66GGC [Curry et al., 1988].Also shown are the levels at which planktonic-benthic age differences were determined.At the left the age of the horizon is given (planktonic mean -400 years).At the right are the differences between the benthic age and the mean planktonic age.

Fig. 7 .
Fig. 7. Reconsmaction of the radiocarbon content of various parts of the ocean-atmosphere system for glacial time.The estimate for tropical surface water is based on the 23øTh-lnC age comparisons on corals by Bard et al. [ 1990].The difference in 14C/C ratio for tropical surface water and the overlying atmosphere takes into account the lower CO2 content of the glacial atmosphere.The deep Ariantic value is for the site of the Ceara Rise cores (i.e., equatorial western basin).The deep Pacific value is for the South China Sea.The numbers within the diagram are the ratios of the 14C/C in the water of interest to that in the glacial atmosphere.cline.This scheme proves to be an effective means of reproducing the observed properties of t•xtay's surface ocean.In Table 4 we list the flux½s adopted in order to replicate the Holocene ocean and the resulting distribution of properties.• Our fu'st attempt to duplicate the observed glacial distribution of radiocarbon and nutrients involves a slowing down of the three loops (1, 2 and 3), which constitute the flow of NADW (see Figure 9).In order to maintain a nearly uniform 14C/C difference between the surface Pacific-Indian box and the deep Pacific-Indian box, decreases in the strength of loop 1 are matched by corresponding increases in the strength of loop 6 (see Table 5).By cutting the strength of theNADW loops we bring about an increase in the surface to deep •4C/C difference for the Atlantic Ocean (i.e., an increase in the deepwater age).By contrast, the phosphate content of deep water in the Atlantic does not change significantly with the NADW flux.To understand this, one must recall that the residence time of phosphorus in the ocean is many times longer than the oceanic mixing time.Hence, no significant gain or loss of this nutrient can occur during a single pass through the Atlantic.Since, at least in our model, only NADW is exported, the phosphate content of NADW must be equal to the average in the water entering the Atlantic.As in our model the mix of incoming water is nearly independent of the flux of water through the Atlantic, very little change in the phosphate content of Atlantic deep water occurs as the rate of NADW production changes.Thus, this simple exercise reveals an important truth: radiocarbon and nutrient distributions in the sea are not so tightly tied one to the other as one Fig. 9. Response of Pandora to changes in the flux of the combined NADW loops (i.e., 1, 2 and 3) from 30 Sv for the Holocene standard to values as low as 6 Sv.Note that in order to keep the age of deep water in the Indian-Pacific nearly constant, reductions in the strength of loop 1 are matched by increases in the strength of loop 6 (see Table5).The dashed line labeled "G" marks the flux at which the age of deep water in the Atlantic is reduced by a factor of 2.

Fig. 11 .
Fig. 10.Plot of alaC against 813C for deep water (> 2 kin) from throughout the world ocean.The 13C and •4C data for the Indian and Atlantic Oceans are from Ostlund et al. [1987].Those for the Pacific are fi'om [Kroopnick et al., 1970; Ostlund and Niskin, 1970].

TABLE 3 . Sramtory of Age Estimates for Glacial Deep Water (i.e., Benthic-Planktonic Age Differences) Water Mean depth, Number Benth-Plank Location Core Latitude Longitude km of Pairs Age, Years
*Sill depth of Caribbean.