Glacial ventilation rates for the deep Pacific Ocean

Transcription

1 PALEOCEANOGRAPHY, VOL. 19,, doi: /2003pa000974, 2004 Glacial ventilation rates for the deep Pacific Ocean Wallace S. Broecker and Elizabeth Clark Lamont-Doherty Earth Observatory, Columbia University, Palisades, New York, USA Irena Hajdas and Georges Bonani Accelerator Mass Spectrometry 14 C Laboratory, Institute for Particle Physics ETH Hoenggerberg, Zurich, Switzerland Received 13 October 2003; revised 19 December 2003; accepted 20 January 2004; published 2 April [1] A key constraint in attempts to reconstruct the patterns and rates of the ocean s thermohaline circulation during the last glacial period is the difference between the 14 C to C ratio in surface and deep water. While imperfect, it is our best index of past deep-sea ventilation rates. In this paper we review published ventilation rate estimates based on the measured radiocarbon age difference between coexisting benthic and planktic foraminifera from glacial-age Pacific sediments. We also present new results from a series of eastern equatorial Pacific sediment cores. The conclusion is that the scatter in these results is so large that the apparent 14 C age of glacial deep Pacific water could lie anywhere between double and half today s. Further, it is not clear what is responsible for the wide scatter in the radiocarbon results. INDEX TERMS: 1620 Global Change: Climate dynamics (3309); 1635 Global Change: Oceans (4203); 4532 Oceanography: Physical: General circulation; KEYWORDS: radiocarbon, paleoventilation, Pacific Ocean Citation: Broecker, W. S., E. Clark, I. Hajdas, and G. Bonani (2004), Glacial ventilation rates for the deep Pacific Ocean, Paleoceanography, 19,, doi: /2003pa Introduction [2] A number of attempts have been made to constrain the rate of renewal of glacial-age deep ocean waters. They are based on three measurement strategies: (1) 14 C age differences between coexisting benthic and planktic foraminifera, (2) 14 C and 230 Th measurements on age-matched surface and deep dwelling corals and (3) 14 C ages on benthic foraminfera associated with tephra horizons whose terrestrial correlatives have been radiocarbon dated. In this paper, the results of published measurements on samples from the deep Pacific Ocean are summarized and a set of new results from the eastern equatorial Pacific are presented. [3] The primary motivation behind these studies is the obvious one, namely a desire to better constrain the rate of thermohaline circulation during glacial time. The recent finding by Adkins et al. [2002] that during glacial time there existed in the deep Southern Ocean a reservoir whose salt content exceeded that in the ambient deep sea by about one gram per liter, suggests that thermohaline circulation was quite different during glacial time. This difference may well have contributed to the lowering of the atmosphere s CO 2 content. Recently published radiocarbon ages on foraminifera from calendar-dated Cariaco Basin sediments [Hughen et al., 2004] confirm previous studies [Bard et al., 1998; Kitagawa and Van der Plicht, 1998; Schramm et al., 2000; Voelker et al., 2000; Beck et al., 2001] which point to a bumpy decline in the 14 C to C ratio in the atmosphere and surface ocean which began about 40,000 years ago and has continued up to the present. While it is not clear what Copyright 2004 by the American Geophysical Union /04/2003PA caused this decline, its bumpiness may in part be the result of changes in the rate of deep-sea ventilation [Hughen et al., 1998]. [4] In this paper, we concentrate our attention on the situation for the deep Pacific Ocean. As it is the largest of Earth s active carbon reservoirs, this certainly is the place to start. As we show below, the situation regarding the reconstruction of its glacial ventilation rate is messy. Hence we have a long way to go before this goal can be accomplished. 2. Radiocarbon Age Differences Between Coexisting Benthic and Planktic Foraminifera Shells [5] In today s ocean, the difference in the 14 CtoCratioin the dissolved inorganic carbon between surface water and deep water provides a measure of the ventilation age of the deep water. By ventilation age is meant the ratio of reservoir volume to the injection rate of new deep water (i.e., water that has been in contact with the atmosphere long enough to replenish its depleted dissolved-oxygen content). The crude ages discussed in this paper differ from ventilation ages in that they are referenced to the 14 CtoC ratio in warm surface water rather than that in the appropriate polar surface water. The crude ages shown in Figure 1 are based on the assumption that the initial 14 C to C ratio in the waters descending into the deep sea is the same as that in the overlying warm ocean (D 14 C ffi 50%). However, prior to 14 C production by nuclear tests deep waters formed in the northern Atlantic had a D 14 C of about 70% [Broecker et al., 1995] and those formed in the Southern Ocean a D 14 C of about 140% [Schlosser et al., 1994]. As 1of12

2 Figure 1. The map in the upper panel shows the distribution of D 14 C values at a depth of 3 km in today s ocean as determined on samples collected as part of the GEOSECS expedition. The map in the lower panel shows the age difference calculated from the ratio of the 14 C/C ratio for 3 km water and that for overlying surface water. See color version of this figure in the HTML. the deep waters resident in the Pacific and Indian Oceans consist of a mixture of these two end-members [Peacock et al., 2000], their initial D 14 C was about 105% instead of 50%. Thus for today s Pacific Ocean, the ventilation age is approximately 500 years smaller than the crude age (1750 years) calculated based on the D 14 Cfor the overlying warm surface waters (see Figure 1). [6] This raises the question regarding temporal variations in the magnitude of the D 14 C offset between newly formed glacial-age deep water and that for low-latitude surface waters. In our estimation, this offset was likely larger during glacial time. The reason is related to the greater extent of sea ice in both polar regions. While during glacial time the 14 C to C ratios for all Earth s active carbon reservoirs were significantly higher than those for preindustrial time, as our interest lies in the difference between the 14 C to C ratio for surface and deep water, we need not be concerned with the absolute values. [7] The use of 14 C measurements on foraminifera shells as proxies for glacial deep sea ventilation rates suffers from several potential sources of bias. One has to do with the fact that some of the planktic species which have been employed for 14 C measurements calcify, in part, in the oceanic thermocline rather than in the surface mixed layer. As the ratio of 14 C to C decreases with water depth, the use of these species could lead to an underestimate of the crude age. Another potential source of bias has to do with the possibility that some of the benthics employed form their shells 2of12

3 Table 1. Error in Years Resulting From the Limited Number of Shells Picked From Cores With a Range of Accumulation Rates a Number of Accumulation Rate, cm/10 3 years Shells a After Andrée [1987]. In each case the bioturbation depth is taken to be 8 cm. within the sediment pores. As it would be expected that the 14 C to C ratio in the pore water SCO 2 would be less than that in the overlying bottom water, the incorporation of pore water SCO 2 would lead to an overestimate of the crude age. [8] Yet another source of bias is the coupling between bioturbation and shell abundance gradients. This problem is especially severe in situations where sedimentation rates are low and where the abundance of the planktic entity of choice changes in one direction and that of the benthic entity of choice in the other direction (see Broecker et al. [1984] for model calculations). While a correction for this bias could conceivably be made based on measured abundance gradients, in situations where the abundance trends are spiky, as they often are, the deconvolution will not yield a unique answer. Clearly, high accumulation-rate sediments are to be preferred. [9] Still another limitation exists. It has to do with the number of shells picked. The problem is that bioturbation mixes together shells with a range of 14 C ages. To overcome this, a large number of shells must be picked. Andrée [1987] has calculated the magnitude of the error introduced in this way as a function of the number of shells picked, the sedimentation rate and the bioturbation depth (see Table 1). Again, high accumulation rate sediments are to be preferred. [10] Finally, as pointed out by Adkins and Boyle [1997], account must be taken of the decline in the 14 C to C ratio in atmospheric CO 2 which occurred over the last 40,000 years. As the lag in the response to this decline is greater for deep waters than for surface waters, it leads to a reduction in the measured benthic-planktic age difference. 3. Absolute Deep Water 14 C to C Ratios Obtained From 230 Th-Dated Benthic Corals [11] Many of the problems associated with the benthicplanktic radiocarbon age difference approach can be eliminated by making 14 C measurements on solitary benthic corals whose precise calendar age has been determined by uranium series dating. As is the case for surface dwelling corals, those formed in the deep sea incorporate almost no measurable 230 Th. However, manganese-rich coatings which form on the benthic corals do contain 230 Th. Hence precleaning is an essential step in the analysis. With time, through the decay of 234 U, 230 Th (half-life 75,000 years) is produced within the coral. Hence the ratio of 230 Th to 234 U records the coral s calendar age. This age can be used to calculate the initial 14 C to C ratio in the coral. Taken together with initial 14 C to C ratios obtained for 230 Th-dated surface dwelling corals of the same calendar age, a crude age for deep water can be calculated. Clearly, the ages obtained in this way do not take into account the difference between the 14 C to C ratios for warm surface waters and that for newly formed deep water. Also as is the case for benthic-planktic pairs, this age must be corrected for the bias introduced by the decline in the 14 C to C ratio in atmospheric CO 2. [12] The major drawback of this approach is the difficulty of obtaining corals in the desired locations and age ranges. To date, they come largely from dredge hauls made on rocky outcrops. In order to identify glacial-age corals, a preliminary sorting based on either 14 Cor 230 Th measurements must be conducted. As most of the corals recovered are Holocene rather than glacial in age, this is a time consuming and expensive task. However, the handful of crude ages obtained in this way has proven to be highly significant. 4. Radiocarbon-Dated Tephra [13] One attempt has been made to couple radiocarbondated volcanic tephra with 14 C measurements on coexisting benthic and planktic shells [Sikes et al., 2000]. The advantage of this approach is that it offers an additional piece of information, namely, the 14 C to C ratio difference between surface ocean SCO 2 and atmospheric CO 2. However, the success of this approach depends on the reliability of correlation between tephra layers in deep-sea cores and their radiocarbon-dated terrestrial counterparts. 5. Previous Results [14] Summarized in Table 2 are the results of published measurements by these three methods on samples from the deep Pacific Ocean. The sites represented by these measurements are shown in Figure 2. The age differences are summarized in Figure 3. Clearly, the situation is unsatisfactory. The reconstructed crude ages range from 500 years to 4000 years. Some are far greater than today s; others are far smaller. A number of possible explanations for this large scatter come to mind. [15] 1. As the samples come from a range of locales, water depths, and ages (14,000 to 25,000 years), it is possible that some of the differences are real. [16] 2. Incorporation of secondary calcite precipitated from the sediment pore waters would lead to ages which are too young. [17] 3. Incorporation of reworked foraminifera shells would lead to ages which are too old. [18] 4. Bioturbation coupled with opposing abundance gradients could bias the benthic-planktic 14 C age differences in either direction. [19] It is worthwhile mentioning the single result on a benthic coral from the Drake Passage. As already mentioned, benthic corals have significant advantages over foraminifera. Goldstein et al. [2001] obtained such a coral from 1125 meters water depth in the Drake Passage. The present-day crude age of the dissolved inorganic carbon at this location and water depth is 830 years. Based on the measured 230 Th age of years for this coral, these 3of12

4 Table 2. Published Glacial-Age Benthic-Planktic Radiocarbon Age Differences for the Deep Pacific Ocean Depth in Core, cm Benthic Benthic Age, years Planktic Planktic Age, years D Age, years South China Sea V N, E 1953 M a 215 mixed G. sacc " " " P. obliq mixed G. sacc " " " P. obliq " " " P. obliq mixed G. sacc. + rub " " " P. obliq mixed G. sacc " " " P. obliq mixed G. sacc " " " P. obliq Ontong-Java Plateau V N, E 3120 M a 42 mixed G. sacc " " " P. obliq " " " Bulk CaCO mixed G. sacc " " " G. rub " " " N. dut mixed G. sacc " " " G. sacc " " " G. rub " " " P. obliq " " " P. obliq Eastern Pacific TT S, W 3225 M a 31 mixed P. obliq mixed G. sacc " " " P. obliq mixed G. sacc " " " P. obliq mixed G. sacc Eastern Equatorial Pacific TR163-31B 3.6 S, 84 W 3210 M b 86 Uvigerina N. dut " " " " " " Uvigerina N. dut " " " " " " Guaymas Basin JPC N W 818 M c 1400 B. sub mixed B. sub " B. sem " B. sub. + sem " Sea of Okhotsk NESGGC M c 70 Uvigerina pach(s) SEA OF OKHOTSK B M c 225 Uvigerina pach(s) Sea of Okhotsk NESGGC M c 230 Uvigerina pach(s) Sea of Okhotsk NESGGC M c 215 Uvigerina pach(s) Sea of Okhotsk NESGGC M c 170 Uvigerina pach(s) California Margin F8-90-G N W 1600 M d 155 mixed mixed California Margin F2-92-P N W 800 M d 298 mixed mixed of12

5 Table 2. (continued) Depth in Core, cm Benthic Benthic Age, years Planktic Planktic Age, years D Age, years North Pacific CH N E 978 M e 690 Uvigerina bull South Pacific H S E 2065 M f 245 mixed ash " infl " infl South Pacific U S E 2700 M f 128 mixed ash " bull mixed infl South Pacific U S E 1300 M f 80 Uvigerina ash Uvigerina bull Uvigerina infl Drake Passage 47396B 59.7 S 68.7 W 1125 M g coral calibrated 14 C age Th- 234 U age a Broecker et al. [1988]. b Shackleton et al. [1988]. c Keigwin [2002]. d Van Geen et al. [1996]. e Duplessy et al. [1989]. f Sikes et al. [2000]. g Goldstein et al. [2001]. authors converted its measured 14 C to C ratio to the initial ratio. They then compared this initial ratio to that obtained for 230 Th-dated Barbados corals of the same age [Bard et al., 1998]. The difference between these ratios yielded a crude age of 1020 years. This single result suggests that the age of deep Pacific SCO 2 during glacial time was not greatly different from today s. [20] In our estimation, the radiocarbon ages adopted by Sikes et al. [2000] for the marine tephra layers at 45 Sinthe eastern Pacific cannot be correct. Either the tephras are Figure 2. Map showing the location of the sites in the Pacific Ocean where 14 C measurements have been made on benthic-planktic pairs of glacial age (circles). The triangle in the Drake Passage marks the location of 230 Th-dated benthic coral [Goldstein et al., 2001]. The water depths (in km) at each site are also given. For completeness but not discussed in this paper are the locations of the Ceara Rise site where Broecker et al. [1990b] conducted 14 C measurements on benthic-planktic pairs and that of the Blake Ridge locale where Keigwin and Schlegel [2002] made such measurements. Shown as well are the Atlantic sites at which measurements on benthic corals were made by Adkins et al. [1998] (38 N) and Mangini et al. [1998] (2 N). See color version of this figure in the HTML. 5of12

6 Figure 3. In the left-hand panel are the measurements of the D 14 C (upper scale) for dissolved inorganic carbon (SC0 2 ) at three stations occupied during the GEOSECS survey. Also shown are the equivalent crude ages (lower scale) calculated assuming that the D 14 C for surface mixed layer SC0 2 is 50%.Inthe right-hand panel are age differences for coexisting glacial age benthic-planktic pairs (solid circles). The cross is for the 230 Th-dated coral. The hatched region represents the present-day range of ages in the Pacific Ocean as a function of water depth. miscorrelated with their terrestrial counterparts or the 14 C ages for terrestrial organics used to date the tephra are anomalously young. Another possibility is that the sediments contain reworked foraminifera. As the benthicplanktic age differences obtained for this core are clearly incompatible with the Drake Passage benthic coral result, we reject them as anomalous. As for the results at 27 S, the same criticisms apply. However, while at this site the benthic-planktic-based crude age is more reasonable, the 1230-year difference between the 14 C ages on G. inflata is unacceptably large. [21] The reader might ask why the published Atlantic Ocean results are not included in this paper. The reason is that such a summary has already been published [Broecker, 2002]. The conclusion is that the benthic-planktic age difference of 1100 years obtained by Keigwin and Schlegel [2002] makes more sense than that of 700 years obtained earlier by Broecker et al. [1990a]. It is interesting to note that the Keigwin and Schlegel [2002] results were obtained on samples taken at depths of prominent benthic abundance maxima. While it is unclear why such samples are better, the excellent agreement in the age differences obtained on them is highly encouraging. 6. Thermocline [22] Some planktic species calcify beneath the surface mixed layer. Because the D 14 CinSCO 2 decreases with water depth, if these species are employed in paleoventilation rate studies, their 14 C ages may be somewhat older than would be the case if they produced all their calcite in the surface mixed layer. Shown in Figure 4 are measured D 14 C values versus water depth for four GEOSECS stations in the equatorial ocean. As can be seen, the SCO 2 in the upper several hundred meters at the time of this survey was enriched in 14 C as the result of the invasion from the atmosphere of bomb-test 14 CO 2. Broecker et al. [1995] showed that the D 14 C for thermocline waters free of tritium is inversely proportional to the SiO 2 concentration and proposed the following relationship for the estimation of the prenuclear (i.e., natural) D 14 C to C ratio: D 14 C ¼ 70 SiO 2 where the units of SiO 2 are mmol/kg. Although this relationship works well in the oceanic thermocline, it underestimates the prenuclear D 14 C values for the surface mixed layer (yielding 70% for silica-free waters). Rubin and Key [2002] show that salinity-normalized and nitratecorrected alkalinity serves better as a proxy for natural radiocarbon than does silica. Of importance to the problem at hand, the alkalinity proxy yields D 14 C values closer to the 50% for warm surface water. A comparison between the two approaches to the reconstruction of the natural 14 C profile for the eastern equatorial Pacific is shown in Figure 5. As can be seen, while the two proxies are in agreement at water depths deeper than 300 meters, they diverge above this depth. As can be seen, even if the forams calcify as deep as 200 meters, the age bias is no more than a few hundred years. Hence this cannot be the explanation for the large scatter. [23] As summarized in Table 3, Broecker et al. [1988] made radiocarbon measurements on a number of P. obliquiloculata (sub-surface calcifier)-g. sacculifer (surface calcifier) pairs. For those 15 pairs younger than 9000 years, the range in the age difference is 20 to 950 years with an average of 520 years. It is puzzling that the two cores yield such different averages, i.e., 120 years for the 6of12

7 Figure 4. Plots of the measured D 14 C, tritium and dissolved silica concentration versus water depth at four equatorial GEOSECS stations. Following Broecker et al. [1995], the prebomb D 14 C is calculated from the relationship D 14 C= 70 SiO 2 %, where the units of SiO 2 are mmol/kg. Rubin and Key [2002] have demonstrated that salinity-normalized alkalinity provides a somewhat better stand-in for prebomb D 14 C than silica. As shown in Figure 5, the major differences between these two stand-ins occur in the upper 300 m of the water column. Thus, the solid lines based on the alkalinity stand-in provide a better estimate of the prenuclear D 14 C-depth profile. 5 pairs from V36-6 and 670 years for the 10 pairs from V35-5. In any case, many of the differences are larger than expected. Hence factors other than calcification depth likely bias these ages. 7. New Radiocarbon Measurements [24] In order to see if we could obtain more consistent benthic-planktic D 14 C age differences, we selected cores from four water depths close to the site of the core on which Shackleton et al. [1988] made their measurements. The locations, water depths and sample sizes are listed in Table 4. The samples were obtained from depths which based on existing radiocarbon measurements should yield planktic radiocarbon ages in the range 16,000 to 20,000 years. Even in these 50-gram samples, the number of benthics is marginal. While we had hoped to obtain radiocarbon dates on more than one benthic type, this was possible in only 4 of the 7 samples. In order to minimize the impact of shell calcification in waters below the mixed layer, G. sacculifer and G. ruber were our planktics of choice. However, as Shackleton et al. [1988] measured 7of12

8 Figure 5. A composite plot of the reconstructed profile with water depth of natural radiocarbon for two WOCE stations at 85.8 E in the equatorial Pacific (one at 3.1 N and the other at 1.1 S). As can be seen, the silicate stand-in of Broecker et al. [1995] yields lower D 14 C values than the alkalinity stand-in of Rubin and Key [2002]. This plot was provided by Stephany Rubin. N. dutertrei which is known to calcify beneath the surface mixed layer, we also made measurements on N. dutertrei, G. tumida, P. obliquiloculata and O. universa. In hindsight, these additional analyses proved valuable in that they clearly demonstrate that serious biases exist. The results of these measurements are listed in Table 5. In addition to the radiocarbon ages, we list the number of shells picked and their weight. Following Andrée [1987], in order to assure that the shells dated typify the ensemble of individual shell ages present in the sample, in each case several hundred shells were picked. Averages based on these measurements are summarized in Figure 6. For comparison, the average for a core from the same region published by Shackleton et al. [1988] is also shown. We hasten to add that as the spread in individual age differences is far beyond that expected from measurement errors, the spread must be the result of some source of bias. As we do not understand the source of this bias, there is no reason to believe that it can be removed by simply averaging. It is likely that the true age difference lies either above or below this average. [25] Despite the high sedimentation rate, the results on the core from 0.62 km depth are decidedly highly anomalous. Hence we decided not to make any further radiocarbon Table 3. Radiocarbon Age Differences for Thermocline Versus Surface Dwelling Planktic a Depth in Core, cm Thermocline Thermocline Age, years Surface Surface Age, years D Age, years Abundance Ratio, G. sac./p. obliq. South China Sea V N, E 2030 M 10 P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac " N. dut " " P. obliq G. sac P. obliq G. sac South China Sea V N, E 1953 M 65 P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac P. obliq G. sac. + rub P. obliq G. sac P. obliq G. sac a Broecker et al. [1988]. 8of12

9 Table 4. Information Regarding the Samples Chosen for This Study Core Latitude Longitude Water Depth, km Sample Depth, cm Sample Weight, g Weight Coarse Function, g Average Sedimentation Rate, cm/10 3 years VM S 89.7 W VM S 82.1 W RC S 85.8 W " " " " " VM S W " " " " " Table 5. New Benthic-Planktic 14 C Age Difference Measurements From the Eastern Equatorial Pacific Ocean Depth in Core, cm Benthic Number of Shells Weight, mg Benthic Age, years Planktic Number of Shells Weight, mg Planktic Age, years V21-30 Lat. 1.2 S Long W 0.62 km mixed ± 170 G. sac. + rub ± 170 " Uvig ± 190 D Age, years V19-27 Lat. 0.5 S Long W 1.37 km mixed ± 110 G. sac ± 110 " Cib ± 130 a G. ruber ± 100 " Uvig ± 150 P. obliq ± 100 " N. duter ± 100 " G. tumid ± 100 " O.univ ± 100 mean mean mixed ± 130 G. sac ± 130 " Cibs ± 140 G. ruber ± 120 " Uvig ± 130 P. obliq ± 120 " N. duter ± 110 " G. tumid ± 120 mean mean RC Lat. l.5 S Long W 2.57 km mixed ± 120 G. ruber ± 120 " Uvig ± 140 G. sac ± 110 " N. duter ± 110 " P. obliq ± 110 " G. tumid ± 110 mean mean mixed ± 150 G. ruber ± 140 " G. sac ± 130 " N. duter ± 120 " N. duter. b ± 110 " P. obliq ± 110 " G. tumid ± 130 " G. tumid. b ± 110 " O. univ ± 150 " O. univ. c ± 140 mean mean V21-40 Lat. 5.5 S Long W 3.18 km mixed ± 110 G. ruber ± 100 " G. sac ± 100 " N. duter ± 100 " P. obliq ± 100 " G. tumid ± 100 mean mean mixed ± 150 G. ruber ± 120 " G. sac ± 130 " N. duter ± 250 " N. duter. b ± 210 " P. obliq ± 140 " P. obliq. b ± 130 " G. tumid ± 130 " G. tumid. b ± 130 mean mean a Excluded from average. b Intense cleaning with hydrogen peroxide: weight loss about 50%. c Crushed and rinsed (not sonicated). 9of12

10 Figure 6. Summary (large solid circles) are averages of the differences between the average age for the benthic samples and the average age for the planktic samples for three of the cores measured as part of this study (see Table 5). Also shown (by the cross) is the average benthic-planktic age difference obtained by Shackleton et al. [1988] (see Table 3). Finally, the Goldstein et al. [2001] result based on the Drake Passage benthic coral (see Table 3) is shown (large square). The small dots represent the crude ages (based on an assumed surface water D 14 C value of 50%) calculated from 14 C measurements made on dissolved inorganic carbon from GEOSECS station 331 (4.38 S, W). While presented in this way the trend with water depth appears to be similar to today s, when the wide spread in the individual ages is taken into account this agreement may well be fortuitous. measurements on this core. Rather, we concentrated our effort on the three deeper cores. [26] The results on the core from 1.37 km depth are the best we have yet obtained. The spread in the planktic ages in both samples is only slightly larger than expected from the measurement errors. In the shallower of the two samples, the average age of the deeper-dwelling planktics (P. obliquiloculata, N. dutertrei, G. tumida, and O. universa) is 270 years younger than that for the surface-dwelling planktics (G. ruber and G. sacculifer). A similar average difference was obtained for the deeper sample. Except in one case, the agreement among the benthic ages in both samples was quite good. Interestingly, the interface dwelling Cibicidoides yielded the oldest age in both samples. The benthic-planktic age differences of 1015 and 1085 years are consistent with that of 1020 years obtained by Goldstein et al. [2001] on a benthic coral from a similar depth in the Drake Passage. In both cases, a correction for the trend in the D 14 C for atmospheric CO 2 must be considered. Fortunately, as shown by Hughen et al. [2004] for the range in calendar ages of importance here (i.e., between 25,000 and 17,000 calendar years ago) there is no significant trend. However, there appear to have been fluctuations about the mean with amplitudes up to ±70%. Corrections for these centennialduration fluctuations would require a far more precise knowledge of the details than we now have. [27] Two samples were analyzed from the core at 2.57 km water depth, one from cm and the other from cm. For the shallower sample, the age for benthic Uvigerina separate was 1350 years older than that for the mixed benthic separate. A similar difference (i.e., 1230 years) was obtained between the G. ruber and the P. obliquiloculata ages. However, as the ages for the other three planktics (G. sacculifer, N. dutertrei and G. tumida) fall in between these extremes, the situation is not quite so bad. The difference between the average age for the benthics and the average age for the planktics turns out to be 1705 years. However, if the age for the Uvigerina is put aside, the difference drops to 1030 years. On the other hand, if the age for the mixed benthics is put aside, the difference becomes 2380 years. So clearly, the uncertainty in the age difference is unacceptably large. [28] The situation for the deeper sample is even worse. Only enough benthics were found to permit one measurement. The spread for the planktics was a staggering 4680 years (i.e., 20,800 years for O. universa and years for P. obliquiloculata). In order to get a handle on what might cause this spread, we re-picked G. tumida and N. dutertrei and gave them an extreme cleaning (half the weight was removed by peroxide soaking plus each shell was broken in order to rid the material trapped in the chambers). The age of the peroxide-cleaned G. tumida came out 1810 years younger than that for original analysis. In contrast, the age of the peroxide-treated N. dutertrei was the same as that for the original analysis. If all 9 of the planktic analyses are averaged, the benthic-planktic age difference comes out to be 2800 years. Again, the spread in planktic ages is unacceptably large. [29] For the two samples from the core at 3.18 km water depth, the situation is only slightly better. In neither were enough benthics available to permit more than one measurement. The range for planktic results was 1480 years for the shallower sample and 3280 years for the deeper sample. In this case, the peroxide-cleaned G. tumida was much closer in age to the G. tumida with the usual cleaning (19020 for the peroxide-cleaned sample and for the routinely cleaned sample, a difference of only 240 years). But the age difference between the peroxide-cleaned and uncleaned N. dutertrei was 2490 years. The difference between the age for the mixed benthics and the average for the planktics for the two samples agreed reasonably well, 1725 years for the shallower sample and 2230 years for the deeper one. [30] A plot of the benthic-planktic age differences based on the average given in Table 5 are shown in Figure 6. Putting aside for the moment the very large spread in the individual results, the average differences match reasonably well the present-day trend with water depth. Further, the new results from 3.18 km water depth are consistent with the average age difference of the 8 benthic-planktic pairs analyzed by 10 of 12

11 Figure 7. Whole shell weights ( mm size fraction) for G. sacculifer, P. obliquiloculata, and N. dutertrei from three of the eastern equatorial Pacific cores. Each of the three records extends back to approximately 20,000 calendar years. Shackleton et al. [1988]. But as already mentioned, averaging discordant results is a dangerous exercise. [31] However, when the scatter in the results is taken into account, then a very large uncertainty must be attached to these age differences. Before any useful conclusions can be drawn, the origin of this scatter will have to be tracked down and eliminated. Four candidates exist. [32] 1. Measurement error and/or contamination during sample preparation. The excellent agreement obtained on the samples from V19-27 suggests that neither is the answer. [33] 2. Differential gradients with depth in the core in the abundance of the various entities. This, however, cannot be the explanation for the large difference between the ages for the peroxide-cleaned and normal G. tumida s from core RC [34] 3. In situ contamination of the shells by secondary carbonate. As pore water DIC should have 14 C to C ratios higher than that in the surrounding sediment CaCO 3,the deposition of secondary calcite should decrease the age of the shells. However, the difference between the peroxidecleaned and normal results suggests contamination with old rather than young carbon, making this scenario unlikely. [35] 4. Incorporation of reworked shells. The fact that several of the peroxide-cleaned shells yielded younger ages could be construed to support this explanation. The idea is that the reworked shells are more fragile because they have lost more weight to dissolution. Therefore they are more easily destroyed during the cleaning process. [36] As a test to see whether the shells in these three cores had experienced significantly different extents of dissolution, we determined shell weights on three species from each over the depth range representing the last 20,000 calendar years (see Figure 7). As can be seen, the core-to-core differences are quite small. Hence this exercise did not help us in constraining the source of the age difference spread. 8. A Hot Clue? [37] Puzzled by the results on the first two peroxidecleaned results, we conducted three additional comparisons bringing the total to five. The results are summarized in Table 6. For three of the five, the peroxide-cleaned foraminifera are 1900 ± 500 years younger than the foraminifera not subject to this cleaning. The other two pair of ages show no significant age difference. The possible explanations are as follows. [38] 1. The peroxide treatment preferentially breaks up reworked shells preweakened by dissolution in their original environment. [39] 2. In removing about 50% of the shell material with peroxide, we preferentially remove the primary calcite and preserve the younger secondary calcite. [40] 3. Contamination during the peroxide step. 11 of 12

12 Table 6. Comparison of Radiocarbon Ages Obtained on Samples Given, Our Normal Cleaning (i.e., No Peroxide), and Those Obtained on Samples Where About 50% of the Weight Was Removed Using Peroxide Core No Peroxide Peroxide D RC G. tumida ± ± ± 160 RC N. duter ± ± ± 160 V21-40 N. duter ± ± ± 400 V21-40 P. obliq ± ± ± 180 V21-40 G. tumida ± ± ± 170 [41] Were (1) the answer, the youngest ages would be the most nearly correct. Were (2) or (3) the answer, the oldest ages would be the most nearly correct. 9. Conclusions [42] The differences among the 14 C ages of coexisting planktic species reveal a major problem not only in connection with attempts to reconstruct past deep sea ventilation rates but also with any attempt to achieve highly accurate radiocarbon chronologies. It appears to us that the most likely explanation for these differences is that the sediment consists in part of reworked foraminifera shells. As the source of these reworked shells likely has a different species make up than that in the local sediment, the magnitude of the age bias created by their inclusion differs from species to species. As these interlopers are more subject to breakup, they are preferentially eliminated during cleaning. If this is the correct explanation, then the youngest radiocarbon ages are closest to the true age of the sediment. This creates a dilemma. Rapidly accumulating sediments from the oceans margins have major advantages over slow accumulating counterparts from the interior. However, as they are by their very nature from areas more prone to reworking, they may be plagued by their own set of biases. [43] Acknowledgments. This material is based upon work supported by the National Science Foundation under grant OCE Lamont- Doherty Earth Observatory contribution References Adkins, J. F., and E. A. Boyle (1997), Changing atmospheric D 14 C and the record of deep water paleoventilation ages, Paleoceanography, 12, Adkins, J. F., H. Cheng, E. A. Boyle, E. R. M. Druffel, and R. L. Edwards (1998), Deep-sea coral evidence for rapid change in ventilation of the deep North Atlantic 15,400 years ago, Science, 280, Adkins, J. F., K. McIntyre, and D. P. Schrag (2002), The salinity, temperature, and delta O-18 of the glacial deep ocean, Science, 298, Andrée, M. (1987), The impact of bioturbation on AMS 14 C dates on handpicked foraminifera: A statistical model, Radiocarbon, 29, Bard, E., M. Arnold, B. Hamelin, N. Tisnerat- Laborde, and G. Cabioch (1998), Radiocarbon calibration by means of mass spectrometric Th-230/U-234 and C-14 ages of corals: An updated database including samples from Barbados, Mururoa and Tahiti, Radiocarbon, 40, Beck,J.W.,D.A.Richards,R.L.Edwards, B. W. Silverman, P. L. Smart, D. J. Donahue, S. Hererra-Osterheld, G. S. Burr, L. Calsoyas, A. J. T. Jull, and D. Biddulph (2001), Extremely large variations of atmospheric 14 C concentration during the last glacial period, Science, 292, Broecker, W. S. (2002), Constraints on the glacial operation of the Atlantic Ocean s conveyor circulation, Isr. J. Chem., 42, Broecker, W. S., A. Mix, M. Andrée, and H. Oeschger (1984), Radiocarbon measurements on coexisting benthic and planktic foraminifera shells: Potential for reconstructing ocean ventilation times over the past 20,000 years, Nucl. Instrum. Methods Phys. Res. B, 5, Broecker, W. S., et al. (1988), Accelerator mass spectrometry radiocarbon measurements on marine carbonate samples from deep sea cores and sediment traps, Radiocarbon, 30, Broecker, W. S., T.-H. Peng, S. Trumbore, G. Bonani, and W. Wolfli (1990a), The distribution of radiocarbon in the glacial ocean, Global Biogeochem. Cycles, 4, Broecker, W. S., et al. (1990b), Accelerator mass spectrometric radiocarbon measurements on foraminifera shells from deep-sea cores, Radiocarbon, 32, Broecker, W. S., S. Sutherland, W. Smethie, T.-H. Peng, and G. Ostlund (1995), Oceanic radiocarbon: Separation of the natural and bomb components, Global Biogeochem. Cycles, 9, Duplessy, J.-C., et al. (1989), AMS 14 C study of transient events and of the ventilation rate of the Pacific intermediate water during the last deglaciation, Radiocarbon, 31, Goldstein,S.J.,D.W.Lea,S.Chakraborty, M. Kashgarian, and M. T. Murrell (2001), Uranium-series and radiocarbon geochronology of deep-sea corals: Implications for Southern Ocean ventilation rates and the oceanic carbon cycle, Earth Planet. Sci. Lett., 193, Hughen, K. A., J. T. Overpeck, S. J. Lehman, M. Kasgarian, J. Southon, L. C. Peterson, R. Alley, and D. M. Sigman (1998), Deglacial changes in ocean circulation from an extended radiocarbon calibration, Nature, 391, Hughen, K., S. Lehman, J. Southon, J. Overpeck, O. Marchal, C. Herring, and J. Turnbull (2004), 14 C activity and global carbon cycle changes over the past 50,000 years, Science, 303, Keigwin, L. D. (2002), Late Pleistocene-Holocene paleoceanography and ventilation of the Gulf of California, J. Oceanogr., 58, Keigwin, L. D., and M. A. Schlegel (2002), Ocean ventilation and sedimentation since the glacial maximum at 3 km in the western North Atlantic, Geochem. Geophys. Geosyst., 3(6), 1034, doi: /2001gc Kitagawa, H., and J. Van der Plicht (1998), Atmospheric radiocarbon calibration to 45,000 yr BP: Late Glacial fluctuations and cosmogenic isotope production, Science, 279, Mangini, A., M. Lomitschka, R. Eichstädter, N. Frank, and S. Vogler (1998), Coral provides way to age deep water, Nature, 392, Peacock, S., M. Visbeck, and W. Broecker (2000), Deep water formation rates inferred from global tracer distributions: An inverse approach, Inverse Methods in Biogeochemical Cycles, Geophys. Monogr. Ser., vol. 114, pp , AGU, Washington, D. C. Rubin, S. I., and R. M. Key (2002), Separating natural and bomb-produced radiocarbon in the ocean: The potential alkalinity method, Global Biogeochem. Cycles, 16(4), 1105, doi: / 2001GB Schlosser, P., B. Kromer, R. Weppernig, H. H. Loosli, R. Bayer, G. Nonani, and M. Suter (1994), The distribution of 14 Cand 39 Ar in the Weddell Sea, J. Geophys. Res., 99, 10,275 10,287. Schramm,A.,M.Stein,andS.L.Goldstein (2000), Calibration of the C-14 time scale to >40 ka by U-234-Th-230 dating of Lake Lisan sediments (last glacial Dead Sea), Earth Planet. Sci. Lett., 175, Shackleton, N. J., J.-C. Duplessy, M. Arnold, P. Maurice, M. A. Hall, and J. Cartlidge (1988), Radiocarbon age of last glacial Pacific deep water, Nature, 335, Sikes, E. L., C. R. Samson, T. P. Guilderson, and W. R. Howard (2000), Old radiocarbon ages in the southwest Pacific Ocean during the last glacial period and deglaciation, Nature, 405, VanGeen,A.,R.G.Fairbanks,P.Dartnell, M. McGann, J. V. Gardner, and M. Kashgarian (1996), Ventilation changes in the northeast Pacific during the last deglaciation, Paleoceanography, 11, Voelker, A. H. L., P. M. Grootes, M.-J. Nadeau, and M. Sarnthein (2000), Radiocarbon levels in the Iceland Sea from kyr and their link to the Earth s magnetic field intensity, Radiocarbon, 42, G. Bonani and I. Hajdas, AMS 14 C Laboratory, IPP ETH Hoenggerberg, CH-8093 Zurich, Switzerland. (hajdas@phys.ethz.ch) W. S. Broecker and E. Clark, Lamont-Doherty Earth Observatory, Columbia University, 61 Route 9W, PO Box 1000, Palisades, NY , USA. (broecker@ldeo.columbia. edu; eliza@ldeo.columbia.edu) 12 of 12