RADIOCARBON CALIBRATION AND COMPARISON TO 50 KYR BP WITH PAIRED 14 C AND 230 TH DATING OF CORALS FROM VANUATU AND PAPUA NEW GUINEA

Transcription

1 RADIOCARBON, Vol 4, Nr 3, 4, p by the Arizona Board of Regents on behalf of the University of Arizona RADIOCARBON CALIBRATION AND COMPARISON TO 5 KYR BP WITH PAIRED C AND 23 TH DATING OF CORALS FROM VANUATU AND PAPUA NEW GUINEA K B Cutler 1,2 S C Gray 3 G S Burr 4 R L Edwards 1 F W Taylor 5 G Cabioch J W Beck 4 H Cheng 1 J Moore 4 ABSTRACT. We calibrated portions of the radiocarbon time scale with combined 23 Th, 231 Pa, C measurements of corals collected from Espiritu Santo, Vanuatu and the Huon Peninsula, Papua New Guinea. The new data map C variations ranging from the current limit of the tree-ring calibration [11,9 calendar years before present (cal BP), Kromer and Spurk 199, now updated to,4 cal BP, see Kromer et al., this issue], to the C-dating limit of 5, cal BP, with detailed structure between to 1 cal kyr BP and 19 to 24 cal kyr BP. Samples older than 25, cal BP were analyzed with high-precision 231 Pa dating methods (Pickett et al. 1994; Edwards et al. 1997) as a rigorous second check on the accuracy of the 23 Th ages. These are the first coral calibration data to receive this additional check, adding confidence to the age data forming the older portion of the calibration. Our results, in general, show that the offset between calibrated and C ages generally increases with age until about 2, cal BP, when the recorded C age is nearly yr too young. The gap between ages before this time is less; at 5, cal BP, the recorded C age is 4 yr too young. Two major C-age plateaus result from a 13 drop in C between 15 cal kyr BP and a 7 drop in C between cal kyr BP. In addition, a large atmospheric C excursion to values over occurs at 2 cal kyr BP. Between and cal kyr BP, a component of atmospheric C anti-correlates with Greenland ice δ 1 O, indicating that some portion of the variability in atmospheric C is related to climate change, most likely through climate-related changes in the carbon cycle. Furthermore, the 2-kyr excursion occurs at about the time of significant climate shifts. Taken as a whole, our data indicate that in addition to a terrestrial magnetic field, factors related to climate change have affected the history of atmospheric C. INTRODUCTION The radiocarbon dating method has been widely applied since its development in the 195s (Libby 1955). Its time range from about 5, yr ago to the present, applicability to a wide variety of materials, and high-precision age determinations make it the predominant means of age control for the end of the Quaternary period. The C age equation, however, is based on the assumption of constant atmospheric C/ C. Changes in atmospheric C/ C ( C) over time are well documented and are attributed to changes in the earth s magnetic field, variability in solar activity, and shifts in the rates of exchange within the global carbon cycle (Suess 197). The first two controls modulate the C production rate, while the third governs the distribution of C among the earth s carbon reservoirs after it is produced. For these reasons, C ages require calibration with an independent chronometer. Over the last half-century, intensive efforts have been devoted to calibrating the C time scale. An accurate and precise C calibration allows the true age of a sample to be established from the C age, which is essential for comparing C-dated records with records dated by other means, and for establishing rates of change. The calibration also furnishes the history of atmospheric C. This history contains important clues about past changes in the global climate system linked to shifts in carbon cycling. Tree-ring counts provide a continuous and high-resolution (bidecadal) calibration from present day back to 11,55 ± cal BP (calendar years before present, BP = AD 195) (Kromer and Spurk 1 Minnesota Isotope Laboratory, Department of Geology and Geophysics, University of Minnesota, 3 Pillsbury Drive, SE, Minneapolis, Minnesota 55455, USA. 2 Corresponding author. drscutler@earthlink.net. 3 Marine and Environmental Studies, University of San Diego, Alcala Park, San Diego, California 921, USA. 4 NSF-Arizona AMS Laboratory, University of Arizona, Physics Department, Tucson, Arizona 5721, USA. 5 Institute for Geophysics, University of Texas, 44 Spicewood Springs Road, Bld., Austin, Texas , USA. Institut de Recherche pour le Développement (IRD), BPA5, Nouméa, New Caledonia. 17

2 1 K B Cutler et al. 199). (This has now been extended to,4 cal BP; see Kromer et al., this issue.) However, extending the dendrocalibration back in time has become difficult due to the scarcity of appropriate fossil tree specimens. As a result, a number of studies have explored the suitability of other chronometers in calibrating older portions of the C time scale. With the development of high-precision 23 Th dating techniques in the late 19s (Edwards et al. 197a,b, 19; Edwards 19), fossil corals have become a prominent calibration tool. The use of 23 Th for high-resolution C calibrations was first proposed and tested by Edwards et al. (19). Despite this possibility, obtaining sequences of corals for the period needed in the calibration was a significant challenge because these corals are submerged well below present sea level. In 199, Fairbanks was able to recover corals that grew since the Last Glacial Maximum (LGM) through underwater drilling off the coast of Barbados. A field expedition to Papua New Guinea in 19 headed by Bloom and Chappell was also successful in obtaining the last half of the deglacial sequence by drilling into uplifted shorelines. These methods for recovering late-quaternary coral samples were further developed in subsequent studies, and, through the use of high-precision, paired 23 Th- C dating of the corals, individual points (Bard et al. 199, 1993, 199, 199; Edwards et al. 1993) or high-resolution sequences (Burr et al. 199) have been added to the calibration curve with as little as 3 error in the 23 Th age. Each calibration point in the former case is established from a single coral colony. In the latter case, a sequence of points are established through combined C dating and counting of annual growth bands in a single coral head; the results, averaged over yr to achieve decadal resolution, are then located on the calibration curve using the coral s 23 Th age range. To address the possibility of diagenetically-altered 23 Th ages, procedures were developed early on (Edwards et al. 1993; Gallup et al. 1994) for identifying altered samples through the comparison of the coral s initial 234 U/ 23 U ratio (δ 234 U i ) with the modern marine δ 234 U value. If we consider the cumulative coral calibration data, analyzed with high-precision 23 Th methods (Bard et al. 199, 1993, 199, 199; Edwards et al. 1993; Burr et al. 199) and screened based on a comparison between its measured uranium isotopic value and the known modern marine value (see Methods section), the following picture emerges. The interval between 11.9 and cal kyr BP consists of about points and a high-resolution sequence between 11.7 and.4 cal kyr BP. The - kyr interval between and 24 cal kyr BP, however, is represented by only points. Beyond 24 cal kyr BP, there is a solitary point at 3.2 cal kyr BP. In general, the gap between the calibrated and C ages increases with age: at cal kyr BP, the C age is about 17 yr too young, and at 3 cal kyr BP, the C age is about 4 yr too young. The low resolution in the older time range is due to the inherent difficulty in retrieving well-preserved samples of the appropriate age. In addition to coral calibration methods, other techniques have been employed to extend the calibration beyond the range of tree rings. One method developed in the 197s (Tauber 197) uses seasonally laminated sediments as an annual clock for C-dated materials contained within the laminae, such as plant macrofossils and foraminifera. Recent use of this method on laminated lake sediments (Kitagawa and van der Plicht 199; Goslar et al. 1995, ; Hajdas et al. 1993, 1995) and marine sediments (Hughen et al. 199, ) has produced continuous calibration curves, some of which extend to 45, yr, near the limit of the C time scale. Other methods that show promise include 23 Th dating of stalagmites (Vogel and Kronfeld 199; Genty et al. 1999; Beck et al. 1), 234 U- 23 Th isochron dating of aragonite and lignite (Schramm et al. ; Geyh and Schluchter 199), application of other dating methods (Stuiver 197), and correlation of C-dated deep-sea records with the GISP2 Greenland ice-core chronology (Voelker et al. 199; van Kreveld et al. ;

3 C Comparison to 5 kyr BP with Paired Dating of Corals 19 Bard et al. 4; Hughen et al. 4). Stuiver et al. (199) created a curve that can be used for C calibration (IntCal9) which is based on records derived from tree rings ( to 11.9 cal kyr BP; Kromer and Spurk 199), laminated marine sediments (11.9 to.5 cal kyr BP; Hughen et al. 199), and corals (11.9 to 24 cal kyr BP; Bard et al. 199, 1993, 199, 199; Edwards et al. 1993; Burr et al. 199) and extends to 24 cal kyr BP. The combined records provide detailed structure between cal kyr BP and present. However, prior to cal kyr BP, questions related to resolution or accuracy of the record remain. In this study, we analyze over 7 corals using high-precision C and 23 Th dating techniques to improve the resolution of portions of the calibration and to extend it beyond its current limit. To do this, we followed the strategy of drilling into uplifted shorelines to obtain deglacial and earlier corals and collected drill core samples at sites in Vanuatu and Papua New Guinea. This strategy allowed us to recover some unique samples as well as sequences of samples of higher resolution than earlier studies. Because older corals are more likely to have undergone diagenesis, we applied parameters, in addition to δ 234 U i values, to detect alteration in corals more than 25 kyr old. High-resolution 231 Pa dating methods (Pickett et al. 1994; Edwards et al. 1997) provided a rigorous second check on the accuracy of 23 Th ages and add confidence to those portions of the calibration established from samples that record concordant 23 Th and 231 Pa ages. This is the first coral calibration study to employ such methods. FIELD AREA AND SAMPLING METHODS The field area consists of sites (Figures 1 and 2). Tasmaloum and Urelapa are located at the southern tip of Espiritu Santo, Vanuatu and have uplift rates of roughly 4. m/kyr and 3.4 m/kyr, respectively. Kwambu, Kanzarua, Kanomi, and Gagidu Point are located on the Huon Peninsula in northeastern Papua New Guinea and have uplift rates of 1.9 m/kyr (Stein et al. 1993), 2. m/kyr (Ota et al. 1993), 2. m/kyr and 2.1 m/kyr, respectively (Cutler et al. 3). 5ºS Papua New Guinea 7ºE Huon Peninsula 157ºE Solomon Islands 17ºE km 15ºS ESPIRITU SANTO 15ºS Australia Vanuatu Espiritu Santo Urelapa Tasmaloum Aore Malo 15º3'S Figure 1 Maps of field areas. Boxes show areas depicted in Figure 2. 1º3'E 17ºE Coral samples were drilled from Tasmaloum, Urelapa, Kwambu, Kanomi, and Gagidu Point during drilling operations that took place between A total of drill cores, 17 from Vanuatu and 3 from Papua New Guinea, are used in the study (Figures 2 and 3, Appendix). All cores were retrieved onshore and all were drilled vertically except Vanuatu cores 9H, 9J, 9L,,, which were drilled at angles of to 3 (Figure 3, Appendix) to recover offshore samples. Rapidly uplifting coastlines were selected, as in previous studies (Chappell and Polach 1991; Edwards et al. 1993; Burr et al. 199), to reduce the drill depth required to reach older corals that grew when sea level

4 113 K B Cutler et al. a HUON PENINSULA 1.9 m/kyr North Kwambu Kanomi Kanzarua 2. m/kyr 2. m/kyr Huon Gulf Gagidu Point 5km 2.1 m/kyr b TASMALOUM North A-B, 7 9A-L 4. m/kyr Ocean 1km c URELAPA 3.4 m/kyr (inclined 21º) 11 North angle of inclined drilling Ocean Figure 2 Plan views of sampling sites. In (a) drilling sites are at Kwambu, Kanzarua, and Gagidu Point. There is also surface collection in Kanzarua and Kwambu. Drilling sites in (b) and (c) are indicated by solid symbols. Uplift rates are indicated in (a), (b), and (c). 2.5 m 13 (inclined 3º)

5 C Comparison to 5 kyr BP with Paired Dating of Corals 1131 was well below its present level. Because of this approach, the depth required to reach Last Glacial Maximum (LGM) corals that grew about 1 m below present sea level (Fairbanks et al. 199) is reduced by 4 to m for coastlines uplifting at 2 to 5 m/kyr, respectively. Core depths range from 15 to 95 m and LGM material was recovered from 9 cores. Of the 74 samples analyzed in this study, were obtained from drill cores. The remaining are from surface exposures at Kwambu and Kanzarua, from 9 to 75 m above present sea level. ANALYTICAL METHODS As an initial screen for altered or contaminated samples, all corals were examined for visible evidence of diagenesis or recrystallization and were subjected to X-ray diffraction analyses, which indicate the presence of calcite (in excess of 1%) formed during diagenesis. Those that appeared pristine and contained no detectable calcite were included in the study. C Technique All samples were subjected to a selective dissolution procedure as an additional measure to eliminate contaminated material prior to C isotopic analyses. Diagenetic processes can add significant amounts of carbon to corals with high C/ C ratios and potentially alter the C ages. Selective dissolution dissolves away the outer portion (~5%) of the skeletal structure, thereby removing such contaminants (Burr et al. 1992). The present C/ C ratio of each sample was established using accelerator mass spectrometry (AMS) at the National Science Foundation Accelerator Facility for Radioisotope Analysis at the University of Arizona, Tucson, USA. The data reported here are given as C ages and C values. C is a measure of the deviation of the atmospheric C/ C content at a given time from the pre-industrial atmospheric value in 15 in parts per thousand; positive values indicate an excess in C relative to 15, and negative values indicate a relative deficit. C is calculated from present C/ C and 23 Th age and is corrected for decay of C over time. The analytical methods and calculation of C ages and C values are detailed in Burr et al. (199). The reported uncertainty in C includes uncertainties in the C and 23 Th ages (Burr et al. 199). Reported C ages were also corrected for the reservoir age of surface water at the sample site. A reservoir correction of 5 yr was used for Tasmaloum samples (Burr et al. 199) and 4 yr for Urelapa and Papua New Guinea (Edwards et al. 1993) samples. 23 Th and 231 Pa Techniques Coral 23 Th and 231 Pa ages were measured at the University of Minnesota using thermal ionization mass spectrometry (TIMS). 23 Th methods are modifications of those of Edwards et al. (197a,b, 19), as described by Cheng et al. (). 231 Pa methods are those of Shen et al. (3) and Edwards et al. (1997), which are modifications of those of Pickett et al. (1994). 23 Th ages were screened for alteration using the initial 234 U/ 23 U (δ 234 U i ) value (Edwards et al. 1993) and, for samples older than 25 cal kyr BP, 23 Th- 231 Pa age concordancy. During coral growth, the coral δ 234 U value is indistinguishable from the marine value (Cheng et al. ). Because the coral δ 234 U value is sensitive to isotopic exchange of uranium after skeletal growth, comparison of this value with the modern marine δ 234 U value provides an indicator of coral diagenesis. Here, the coral δ 234 U value is considered indistinguishable from the modern marine value when the sample δ 234 U i values are within ± of the modern marine value and the 2-σ analytical errors are. The previously reported marine δ 234 U value was 9.3 (Galewskyr et al.

6 1132 K B Cutler et al. Height relative to sea level (m) 4 - a 9A-L (3.~4.7 m) 7 (13.2 m) kyr B A (29. m) (29.4 m) kyr kyr -4 1 Distance from shoreline (m) Height relative to sea level (m) b South 9H,J,L reef flat 21,57 yr (-29 m) 9B-F 9A,5 yr (-24 m),4 yr (- m) 9D,E 24,5 yr (-24 m) 9C?, yr (-13 m) North reef unit Unconformity m Figure 3 Cross-sections of Tasmaloum drill sites: (a) schematic cross-section perpendicular to shoreline at sites and 7. For comparison, the drill holes at site 9 have been projected onto it. Contour lines, based on sample ages and depths (Table 1), are at 2-kyr age intervals and are interpolated between cores. Vertical exaggeration is 2 ; (b) schematic cross-section perpendicular to shoreline at site 9. The top ages in each of these cores was roughly 4 cal BP. Ages of the oldest corals above the unconformity are indicated. Below the unconformity, there are samples that are, yr or older (Taylor et al., forthcoming). 199). For this work, we use the Cheng et al. () revised 234 U and 23 Th half-lives for 23 Th-age and δ 234 U determinations and adjusted marine value of 5. ± 1.7 for comparison to the coral δ 234 U value. As a second check for alteration, we dated corals older than 25, yr with high-precision 231 Pa dating methods. The different isotopic systematics in 231 Pa and 23 Th dating allow 231 Pa ages to act as

7 C Comparison to 5 kyr BP with Paired Dating of Corals 1133 an independent check on the 23 Th age (Cheng et al. 199). Samples that record concordant 23 Th- 231 Pa ages are significantly less likely to be altered. Ages are considered concordant when the 231 Pa ages are within ± yr of the 23 Th age and the 2-σ analytical errors are yr (Cutler et al. 3). RESULTS Seventy-four corals were analyzed using paired 23 Th- C dating, from Vanuatu and from Papua New Guinea. Six of these samples were from surface exposures; the remainder were from drill cores. The ages of the corals range from 9 to 5 cal kyr BP; 11 of the samples record ages between and 25 cal kyr BP, and 9 record ages older than 25 cal kyr BP. Eleven samples, including all nine with ages older than 25 cal kyr BP in addition to two others, were also dated with 231 Pa methods. The results, including isotopic composition, C, 23 Th, and 231 Pa ages, and C values, are presented in Table 1. Table 1 also gives coral species, site location, and present elevation relative to sea level. Screening for Sample Diagenesis The δ 234 U i results shown in Figure 4 reveal that the majority of analyses fall within of the modern marine value (hatched box). We interpret the clustering of δ 234 U i values as strong evidence that the modern marine δ 234 U value has been constant within bounds of for the past several tens of thousands of years. Also, based on oceanic residence time arguments, there is reason to believe that marine δ 234 U should be constant within a percent or so over time scales of 4 yr (Edwards et al. 3; Edwards 19; Richter and Turekian 1993; Hamelin et al. 1991; Henderson et al. 2). For these reasons, we conclude that the points that lie outside of the ± range have likely been altered, and we eliminate their ages from further consideration in this study. The 23 Th- 231 Pa isotopic results shown in Figure 5 indicate that out of the Pa analyses (representing 11 samples), 4 analyses (2 samples) have ages that are discordant by more than yr. These samples are eliminated from further consideration in this study. The remaining data lie on or very close to concordia, evidence that those samples did not experience significant diagenetic shifts in the relevant isotope ratios. The concordia plot reveals the increased probability of diagenesis in older corals collected from surface exposures. Plots of C age versus 23 Th or calendar age between and kyr are presented in Figure. Samples from 3 Tasmaloum drill cores, A, B, and 7, are shown as inverted triangles; samples from the other cores are shown as solid circles. The 29 samples from the other cores agree within error with the IntCal9 spline, with only 1 sample slightly removed. A number of samples from cores A, B, and 7, however, lie outside this spline in a time interval when the spline is well constrained with high-resolution data (Kromer and Spurk 199; Hughen et al. 199; Bard et al. 199, 1993, 199, 199; Edwards et al. 1993; Burr et al. 199). This suggests that samples from the A, B, and 7 cores were more susceptible to diagenesis than samples from the other cores. Figure 4 shows that several of these same samples have unusually low δ 234 U i values relative to the bulk of samples in the same time range, adding support to this conclusion. To understand why Tasmaloum cores A, B, and 7 may have been more susceptible to alteration, the location of these cores was examined (Figures 2b and 3a). The cores lie more than a kilometer away from the other Tasmaloum site where cores 9A L were collected. Moreover, they were drilled significantly further inland and from a significantly higher elevation (Figure 3a, Appendix) than any of the other 17 cores, with the exception of 9C. These observations suggest that corals at this site

8 1134 K B Cutler et al. Table 1 U, Th, Pa, and C analytical results. Sample a Sp b (m) (BP) Ana i (ug/g) (pg/g) [ 23 Th/ 23 U] ( ) (BP) age h (BP) ( ) Ev c 23 U 232 Th 23 Th age e Avg 23 Th C agecorr C j δ 234 U d m δ 234 U f i h ( ) δ 234 U g c Tasmaloum samples A b, ? 13 (1) 2.94 (2) 323 (3).91 (3) 5.3 (1.1) 927 (32) 9. (1.1) P 927 (32) 15 () 4 (11) A a, (A)? 13 (1) 3.13 (2) 79 (2).9 (2) 1.5 (1.2) 915 (27) 5.2 (1.2) P 9131 (19) 15 (7) 5 94 () (B) 3.22 (3) 71 (4).9 (3) 1.5 (1.2) 93 (2) 5.2 (1.2) P A (A) P (1) 2.91 (1) 475 (53).9 () 1.9 (1.2) 95 (7) 5. (1.2) P 91 (B) 2.97 (2) 21 (1).9 (7) 1.2 (1.2) 91 (72) 4.9 (1.2) P (C) 2.9 (2) 3 ().922 (2) 1.7 (1.3) 917 (23) 5.4 (1.3) P B b<13? 17 (1) 3.51 (2) 93 (22).74 (3) 4.2 (1.1) 774 (31) 7.4 (1.1) P 774 B.5 1? (1) 2.15 (2) 32 (29).59 (5) 4.7 (1.1) 55 (51).2 (1.1) P 55 (21) 54 () 5 4 (11) (31) 97 (5) (51) 71 () 75 (9) 2 9 () B (A) A 3 (2) 3.91 (2) 2 (213). () (1.1),174 () (1.1) F,973 (7),25 (1) 3 53 (17) (B) 3.9 (3) 93 ().2 (3) 9.5 (2.1),1 (3) (2.2) F (C) 3.93 (4) 3 ().9 () 13.5 (1.4),973 (7). (1.4) P B (A) A 11 (2) 3.74 (2) 1 (15).1 (4) 13. (1.2),193 (43) 3.7 (1.2) P,299 (2) 11,7 () 4 29 (13) (B) 3. (3) 5 (3).17 (5) 13. (1.3),221 (54) 2. (1.3) P (C) (1) 43 ().3 (5) 13.9 (1.2),539 (53) 3.9 (1.2) P 7 b>,<<15? 1 (1) 2.54 (2) 1 (15).34 (3) (1.1),441 (31) 3.1 (1.2) P,441 (31) 97 (1) 3 1 (1) 7 a>,<<15 (A)? 1 (1) 2. (2) 59 (2).4 (3) 13.9 (1.2),441 (31) 3.1 (1.2) P,419 (21) 97 () 3 52 (13) (B) 2.51 (1) 7 (17).35 (3) 137. (1.1),399 (29) 1.1 (1.1) P Fd 2 (1) 2.54 (2) 1 ().1151 (5) 13.7 (1.2) 11,1 (52) 3.3 (1.2) P 11, (a) A 7 (1) (2) 225 (4).1151 (5) 139. (1.2) 11, (52) 4.2 (1.2) P 11, (52) 93 () (17) (52),3 (7) 4 () (b) (A) A (4) 3.7 (3) ().133 (7) 13.4 (1.2) 13,94 (73) 3.9 (1.2) P 13,74 (39), (1) (17) (B) 3.73 (3) 3 (2).131 (4) 13. (1.3) 13, (4) 4.3 (1.3) P 9A + (A) D (1) (2) 74 ().49 (3) 1.1 (1.3),2 (32).2 (1.3) P,4 (19),5 () (13) (B) 2.59 (2) 72 (244).4 (4).9 (1.),22 (4). (1.7) P (C) 2.25 (1) 77 (9).4 () 139. (1.3),35 (3) 4.7 (1.3) P (D) 2.1 (3) 7 (7).43 (3) 139. (1.),57 (3) 4. (1.7) P (E) 2.5 (2) (7). (5) 1.1 (1.2), (53).3 (1.2) P 9B (c) P 9 (1) (3) 29 ().1 (4) 2. (1.),247 (44) 7. (1.) P,247 (44),4 (95) (15) 9B (a) (A) A 15 (1) 3.79 (2) 37 (2).29 () 13. (1.2),4 (9) 4.4 (1.2) P,55 (44),43 () () 9B (B) 3.21 (2) 99 (1).24 (5) (1.2),537 (57) 5. (1.2) P 9C A 2 (1) (3) 2 (1).1 (3) 1.2 (1.3) 11,45 (31) 5.7 (1.3) P 11,45 (31) 93 () () 9C A 9 (1) 3.5 (1) 13 (4).71 () (1.2), (3) 4.4 (1.2) P, 9C (A) A 13 (1) 3.4 (1) 4 (15).1379 (4) (1.2),45 (42) 5. (1.2) P,57 (B) 3.2 (2) 5 (7).131 (5) (1.2),7 (5) 4.7 (1.2) P 9D P (1) (1) 7 (4).1137 (3).1 (1.3) 11,449 (37) 4.7 (1.3) P 11,449 (3) 11,1 (1) (34) 2 15 (),27 (7) 19 (11) (37), () 4 9 (13) 9D A 9 (1) 3.3 (5) 2 ().13 (5).1 (1.2),151 (54) 5. (1.2) P,151 (54),27 (7) 211 (13) 9E (A)? 15 (1) 2.3 (1) 3 ().52 (4) 13. (1.1),7 (47) 3.9 (1.1) P,7 (42),4 (9) 4 29 (1) (B) 3.2 (3) 9 (21).53 () (1.3),7 () 3. (1.4) P 9E (B) A 1 (1) 3.2 (2) 9 (). () 1. (1.4),259 (7). (1.5) P,259 (7),5 (1) 3 11 (19) 9E A 1 (1) (2) 3 (9).9 (5) (1.2) 15,27 (55) 3.3 (1.3) P 15,27 (55) 13, (9) (1) 9E (A) A 19 (1) 3.45 (3) 957 (1).1935 (5) (1.2),32 (5) 1.3 (1.3) P,35 (53) 17,2 (1) 2 37 (29) (B) 2.29 (4) 311 (2).194 (11) (1.2),52 (132) 1.9 (1.3) P 9E A 24 (1) (3) 41 (1).229 () (1.) 24,577 (1).5 (1.9) P 24,577 (1) 19, () 3 (4)

9 C Comparison to 5 kyr BP with Paired Dating of Corals 1135 Table 1 U, Th, Pa, and C analytical results. (Continued) Sample a Sp b (m) (BP) Ana i (ug/g) (pg/g) [ 23 Th/ 23 U] ( ) (BP) age h (BP) ( ) Ev c 23 U 232 Th 23 Th age e Avg 23 Th C agecorr C j δ 234 U d m δ 234 U f i h ( ) δ 234 U g c Tasmaloum samples 9F 3 P 11 (1) 2.3 (3) 1773 ().4 (4). (1.1),74 (4). (1.2) P,74 (4),7 () 222 (31) 9F 5 A 15 (1) (4) 1 (5).131 (5) 13. (1.),42 (53) 3. (1.1) P,42 (53),41 (1) 1 15 (27) 9F 139 A 17 (1) 3.32 (4) 35 (7).57 (9) 13.5 (1.2), (2) 4.5 (1.2) P, (2),1 (19) 1 24 (33) 9F 17 Fd 24 (1) (2) 3215 ().1957 () (1.1),559 (75) 2.1 (1.1) P,559 (75) 17,93 (3) 1 29 (53) 9H 92 P 13 (1) (3) 3551 ().1 (7) (1.1),157 (79) 4. (1.2) P,157 (79),29 (1) 2 (29) 9H 113 (a) A 1 (1) 3.27 (3) 199 (9).1392 (5) (1.1),19 (55) 3.1 (1.2) P,221 (39),27 () 1 2 (22) (b) (5) 249 (15).4 () (1.2),29 (72) 3. (1.2) P (c) 3.33 (5) 1332 (1). () (1.3),233 () 4. (1.3) P 9H 23 (A) Fd 29 (1) 3.5 (5) 43 (3).25 (7) 131. (1.2) 21,4 (1) 139. (1.2) P 21,5 (5) 1,49 (3) (52) (B) 3.97 (5) 17 (7).52 (7) 13. (1.2) 21,739 (3) 13.9 (1.3) P 9J 4 (A) As 1 (1) 3.95 (4) 91 (15).1371 (7) 13.7 (1.3) 13,92 (73) 4.3 (1.3) P 13,959 (41),5 (1) 1 7 (25) (B) 3. (3) 4 (11).137 (4) (1.2) 13,973 (5) 5. (1.3) P 9J 1 A 1 (1) 3.93 (4) 1 (7). () 13.2 (1.1),35 () 3.9 (1.1) P,35 (),244 (2) (34) 9J 7? 1 (1) 1.11 (2) 3 (3).1379 (5) (1.3),59 (57). (1.3) P,59 (57),42 (7) 2 1 () 9J 155 P 27 (1) (3) (7).174 () (1.1) 19,23 (). (1.2) P 19,23 () 17,1 (1) 2 27 (3) 9L 9 P 15 (1) 2.75 (2) 253 ().13 (15).2 (1.1) 13,27 (1) 5. (1.2) P 13,27 (1) 11,5 () (32) Urelapa samples 42 P 5 (1) (4) 73 (27).1 (4).1 (1.1) 75 (4) 2. (1.1) P 75 (4) 17 (13) 1 (1) 4 P (1) 3.11 (4) 75 (19).97 (3).3 (1.1) 33 (35) 3. (1.1) P 33 (35) (95) 2 1 () P 1 (1) 3.2 (4) 5 (1).99 (3) 13.4 (1.1) 973 (34) 2.4 (1.2) P 973 (34) 3 (1) 1 3 (22) P 19 (1) 2.19 (3) 22 (23).47 (4) 139. (1.4),457 (42) 4. (1.4) P,457 (42) 917 () (15) G 24 (1) 2.34 (3) (23).13 (5) 2. (1.) 11,221 (52) 7.4 (1.1) P 11,221 (52) 9732 () (1) 152 L 25 (1) (2) 532 (4).1137 (5) 1.2 (1.1) 11,37 (5) 5. (1.2) P 11,37 (5) 9777 (1) (19) 1 P 27 (1) (3) 43 (2).117 (4) 2. (1.1) 11,9 (44). (1.1) P 11,9 (44),92 (9) (15) 194 A 43 (1) 3.2 (5) 49 ().1554 () (1.2) 15,945 (7) 4.3 (1.2) P 15,945 (7) 13,332 (17) 1 39 (3) 9 Fd 4 (1) 2.3 (3) 4224 (19).4 () (1.2) 21, (1) 2.7 (1.3) P 21, (1) 1,372 (3) 1 3 (54) 2 P 49 (1) (2) 53 (4).7 () (1.1) 21,97 (7) 2.5 (1.2) P 21,97 (7) 1,1 (1) (3) 217 Fd 5 (1) (3) 9 (57).24 (9) (1.1) 22,2 (1) 3.4 (1.1) P 22,2 (1) 1,72 (24) (47) 221 P 51 (1) (5) 445 (49).2137 () (1.4) 22,79 (7).4 (1.5) P 22,79 (7) 19,542 (4) 2 39 (9) 11 7 P 25 (1) 2.7 (3) 272 (21).15 (4). (1.2) 11, (4) (1.2) P 11, (4) 957 (13) () 11 4 P 35 (1) 2.54 (3) 7 (1).5 (4) 2.4 (1.), (49) 7. (1.) P, (49), (32) (9) P 22 (1) 2.77 (3) 133 (9).17 (3).2 (1.1) 11,2 (37) 4. (1.1) P 11,2 (37) 9727 () 1 17 (21) 13 4 P 13 (1) 2.4 (2) 1 (2).33 (3) 5.4 (1.1) 15 (2). (1.2) P 15 (2) 532 () 1 77 () 13 P 21 (1) 2.57 (2) (4).2 (3) 2.3 (1.4) 719 (33) 5. (1.4) P 719 (33) 797 (1) 1 74 (15) 13 74/75 (a) P 24 (1) (2) 13 (5).974 (3) 1. (1.) 979 (29) 5.5 (1.) P 979 (29) 522 (13) 1 11 (1) (b) (3) 159 ().975 (3) 2.2 (1.3) 91 (32).2 (1.3) P 13 5 A 3 (1) (5) 449 (25).1157 () 13.7 (1.2) 11,32 (7) 3.4 (1.2) P 11,32 (7) 992 (1) (2) 13 P 34 (1) 2.7 (4) 11 (23).53 ().9 (1.1),3 (). (1.2) P, Fd 52 (1) 2.31 (3) 23 (21).3 () 9. (1.1) 22,7 (9) (1.2) P 22,7 54 P 24 (1) 2.49 (3) 173 ().31 (4) (1.7),29 (44) 3.3 (1.7) P,29 (),977 (24) (9) (3) 1,2 (3) (55) (44) 93 (13) () 7 D 5 (1) 2.3 (3) 33 (17).213 () (1.3) 22,9 (139) 2. (1.4) P 22,9 (139) 19,2 (2) 2 35 (53)

10 113 K B Cutler et al. Table 1 U, Th, Pa, and C analytical results. (Continued) Sample a Sp b Ev c (m) 23 U (ug/g) 232 Th (pg/g) [ 23 Th/ 23 U] δ 234 U d m ( ) 23 Th age e (BP) δ 234 U f i ( ) δ 234 U c g Avg 23 Th age h (BP) C agecorr h (BP) Ana i C j ( ) [ 231 Pa/ 235 U] 231 Pa age k (BP) 231 Pac g Papua New Guinea samples GT 3. A 33 (1) (3) 2 (9).33 () 1.3 (1.2),449 (4).4 (1.2) P,449 (4),4 (2) (35) GT 4.95 A 35 (1) 2.39 (3) 9292 (3).51 (7) 1. (1.2),542 (13).9 (1.3) P,542 (13),53 (5) (27).23 (5), (3) P KNM 4.7? 41 (1) 2.4 (3) 249 (5).31 (5) 139. (1.2),11 (52) 5.7 (1.2) P,11 (52),39 (24) (3) KNM 5. (B)? 45 (1) (3) 5 (11).22 () 13. (1.2) 23,2 (9) 139. (1.2) P 23,44 () 19,4 (2) (5).41 (3) 24, () P (C) 3.27 (3) (7).22 (7) (1.3) 23,2 (9) 1.4 (1.3) P.47 (5) 24,7 (4) P KNM 5.7 (A) P 53 (1) (3) 1 (7).234 (7) 13.2 (1.1) 25,134 (9).3 (1.2) P 25,292 (),13 (1) (2).42 (9) 2, () P (B) (3) 9 (5).237 () (1.3) 25,42 (99) 2. (1.3) P.413 (15) 25, (1) P KNM T-2 * (a)? 22 (2) (3) 193 (1).324 () 5.7 (1.2) 3,79 (1) (1.4) P 3,79 (12) 32,7 (4) 3 42 (1).53 (15) 3,3 (1) P (b) 2.99 (3) 231 (2).3242 (31) 7.2 (1.1) 3,73 (423) 1.1 (1.3) P.525 (9) 35, (9) P KWA 41.24? 3 (1) 2.3 (13) (1).7 (9) 1.4 (7.2),93 ().7 (7.4) P,93 (),925 (1) 1 22 (31) KWA.2? 55 (1) 3.9 (5) 739 (1).25 () (1.) 27, (91) 2.4 (1.) P 27, (91),39 (192) 2 13 (5).43 () 27, (5) P KWA 4.3 (B)? 5 (1) 3. (4) 397 (94).219 (13).7 (1.2) 2,45 (19) (1.3) P 2,.3 (2) 21,5 (2) 2 97 (79).475 () 3,5 (7) P (C) () 4197 (51).229 () 135. (1.3) 2,54 (177) 7.2 (1.4) P.433 () 2, (5) P KWA I-1* (A) P 9 (2) 2.92 (3) 1 ().3921 (3) 1.1 (1.3) 4,43 (445) 137. (1.4) F 4,44.4 (11) 34, () 27 (29).15 () 45,1 (947) P (B) (4) 1 (24).392 (13) (1.2) 4,4 (197) 13.2 (1.3) F. (5) 44,71 (595) P KWA K-1* (A) P 3 (2) (4) 51 (5).4 (13) 119. (1.2) 49,7 (3) 137. (1.3) F 4,49.4 (3) 34, () 1 4 (22).5 () 43,92 (1) F (B) 3.23 (4) 173 (5).3945 (13) 11.9 (1.1) 47,17 () (1.2) F.552 (9) 37,941 (95) F KWA N-1 * (A) P 9 (2) 2.7 (3) 3 (5).4173 (2) 3.2 (1.1) 5,232 (45) 2. (1.2) P 5,3 (221) 4, (2) 4 49 (492).45 (7) 49, () P (B) (4) 2 (9).42 (17) 1.1 (1.2) 5,79 (23) 139. (1.4) P.59 () 5, (1) P KAN D-4* P 75 (2) 3.23 (4) 41 ().49 (24).7 (1.2) 4,75 (32) 7.7 (1.4) P 4,75 (32) 3,9 (19) 1 29 (95).25 () 4,299 (3) F KAN C-2* F 45 (2) (3) 497 (19).454 (21) 1.1 (1.1) 4,55 (31) 13.9 (1.2) P 4,55 (31) 43, (2) 3 43 (545).45 (9) 31,32 () F a Numbers in left column identify Vanuatu drill cores; letters identify New Guinea collection sites: KNM = Kanomi, KWA = Kwambu, KAN = Kanzarua, GP = Gagidu Point. Replicates denoted with (A), (B),... are fragments of the same sample and (a), (b),... are aliquots of the same sample split. * identifies samples collected from surface exposures. b Genus: A = Acropora sp., As = Astreopora, D = Diploastrea, Fd = Faviidae, G = Goniopora gr. Lobata, L = Leptastrea, and P = Porites sp. c Present elevation relative to sea level. d δ 234 Um is the measured δ 234 U value. e23 Th ages calculated from the equation ( 23 Th/ 23 U) 1 = exp( λ23 T) + (δ 234 U m /)[λ 23 /(λ 23 λ 234 )][1 exp(λ 234 λ 23 )T], where T is the age, λ 23 = yr 1, and λ 234 = 2.23 yr 1 (Cheng et al. ). f δ 234 U i = δ 234 U m exp(λ 23 T). g δ 234 Uc = δ 234 U i screening criteria and 231 Pa c = 231 Pa screening criteria defined in text; P indicates the sample passed and F indicates the sample failed the screening criteria. h Ages from replicate samples are weight averaged; ages in bold are from samples that passed all screening criteria (see text). C ages are reservoir corrected: 4 yr for New Guinea and Urelapa samples, 5 yr for Tasmaloum samples. i Number of C measurements on separate subsamples. j C = [F RC exp(λt 1)], where t is the 23 Th age, λ = 1/27 yr 1, and F RC is the reservoir-corrected fraction of modern carbon (Burr et al. 199). k231 Pa ages are calculated from the equation T = ln[1 ( 231 Pa/ 235 U)]/λ231, where λ 231 = yr 1 (Robert et al. 199).

11 C Comparison to 5 kyr BP with Paired Dating of Corals δ 234 U i ( ) Age (kyr BP) Figure 4 δ 234 U i versus age diagram. Inverted triangles are data from cores A, B, and 7; solid circles are data from other Vanuatu cores; and open circles are data from Papua New Guinea. Plots include replicate analyses. Error bars are 2 σ and, if not visible, are smaller than the symbol. The envelope with hatch marks indicates boundaries of δ 234 U i screening criteria: ±. The open envelope delineates earlier coral δ 234 U i results from Papua New Guinea (Edwards et al. 1993) and provides the marine reference value for that period (see text). may have been more susceptible to alteration. This may have resulted because the environment surrounding this particular locality was anomalously conducive to diagenesis, or more likely, because the greater distance both horizontally and vertically from the sea allowed greater interaction with meteoric water over longer periods of time. Based on these arguments, we eliminated samples from cores A, B, and 7 from further consideration in this study. Our continued discussion focuses on the 17 other drill cores that show little evidence of diagenetic alteration. Calibration Results The screened calibration data, derived from corals which record marine δ 234 U values that give concordant 231 Pa and 23 Th ages (for samples analyzed with 231 Pa) and are acquired from drill cores with little evidence of alteration, are shown in Figures 7 and. Replicate analyses were combined using a weighted average. The results show that the coral data that overlap in the range of the tree-ring calibration (represented by the IntCal9 envelope) are in close agreement (Figure a). Thirteen calibration points extend beyond the tree-ring calibration to cal kyr BP. The close agreement between these points and the IntCal9 spline confirms the data upon which this portion of the spline is based (Hughen et al. 199; Bard et al. 199, 1993, 199, 199; Edwards et al. 1993; Burr et al. 199).

12 113 K B Cutler et al [ 231 Pa/ 235 U] [ 23 Th/ 234 U] Figure Pa- 23 Th concordia diagram. [ 231 Pa/ 235 U] [ 23 Th/ 234 U] concordia diagram of samples analyzed for 231 Pa. All samples are from Papua New Guinea; error bars are 2 σ and, if not visible, are smaller than the symbol. The curved line, concordia, shows equal 231 Pa- 23 Th age values. Crosses on concordia indicate age in thousands of years (kyr). In addition, 23 calibration points fall between and 24 cal kyr BP, an interval in which, prior to this work, there were few data points. The new points provide detailed structure for the intervals between and 1 cal kyr BP and between 19 and 24 cal kyr BP. Two of these samples, with ages of.5 and 23. cal kyr BP, were also analyzed for 231 Pa and produced concordant 23 Th- 231 Pa ages. Finally, samples range in age from 24 cal kyr BP to the limit of C dating, 5 cal kyr BP. The 5 samples older than 25 cal kyr BP are 23 Th- 231 Pa age concordant and were C dated multiple times (Table 1). The rigorous analysis of these samples lends significant credence to the extension of the calibration presented here. The results, in general, show that the offset between calibrated and C ages generally increases with age until about 2 cal kyr BP. At cal kyr BP, the C age is about 1 yr too young, and at 2 cal kyr BP, the C age is about yr too young. The 2 points prior to 2 cal kyr BP indicate less of a gap between the calibrated and C age. At 37 cal kyr BP, the C age is about 4 yr too young, and at 5 cal kyr BP, the C age is about 4 yr too young. This older trend is supported by the solitary point of Bard et al. (199a) that shows an age gap of 4 yr at 3 cal kyr BP.

13 C Comparison to 5 kyr BP with Paired Dating of Corals 1139 C age (kyr BP) a kyr 1kyr 2kyr 3 kyr offset b Calendar age (kyr BP) Figure Screening for drill core diagenesis. Calibration results between and cal kyr BP. Curved lines are the IntCal9 calibration envelope; inverted triangles are data from cores A, B, and 7; and solid circles are data from all other cores. In this age range, only 1 sample was analyzed for 231 Pa. This sample gives concordant 23 Th- 231 Pa ages and is marked with a box at.5 cal kyr BP. Error bars are 2 σ and, if not visible, are smaller than the symbol. Diagonal line is the one-to-one reference line of equal age; parallel dashed lines show -yr offset lines. All data between cal kyr BP are shown in (a) and data set after removal of data from drill cores showing evidence of diagenesis are shown in (b) (see text). Significantly, the new data reveals 2 major C age plateaus that prevent precise C dating during these times. The first age plateau lasts for yr, between.9 and 13.9 cal kyr BP, and gives a nearly constant C age of.4 ±.2 kyr BP (Figure 9) associated with a C drop (Figure ). The second, longer age plateau extends for 3 yr, between 25 and 22 cal kyr BP, and gives a uniform C age of 19. ±.3 kyr BP (Figure 11) associated with a C drop (Figure ). For the most part, the Lake Suigetsu varved sediment record of Kitagawa and van der Plicht (; Figure, small triangles) compares well with the coral record. However, it disagrees between 25 and 2 cal kyr BP, the timing of the larger C age plateau. This discrepancy between the 2 records is plausibly due to missing varves or the assumption of constant sedimentation rate for portions of the record. The younger portion of the Bahamas speleothem record of Beck et al. (1; Figure, grey envelope) also agrees well with the coral record. However, because the speleothem record has a gap

14 1 K B Cutler et al. 5 C age (kyr BP) 4 3 kyr kyr offset between 2 and 2 cal kyr BP, a comparison cannot be made during the timing of the larger C plateau. There is a discrepancy between the 2 records of more than 15 yr at the 37-cal kyr BP coral point. This discrepancy could be due to the assumption of a constant dead carbon fraction in the speleothem record. DISCUSSION Figure 7 Coral calibration results between and 5 cal kyr BP. Solid circles are screened coral data from this study. 23 Th- 231 Pa age-concordant samples are marked with a box. Open circles represent existing coral calibration data (Bard et al. 199, 1993, 199 in this time range) for corals that record marine δ 234 U values according to δ 234 U i criteria (see text). Parallel dashed lines indicate -yr offset lines from solid line of equal age. Error bars are 2 σ and, if not visible, are smaller than the symbol. C Age Plateau: 15 cal kyr BP Calendar age (kyr BP) Figure 9 shows the tree-ring and coral calibration data between and 1 cal kyr BP and the corresponding atmospheric C record. The new findings are in close agreement with the earlier data and support the previously identified C plateau between.5 and 11.5 cal kyr BP (Edwards et al. 1993; Goslar et al. 1995; Hughen et al. 199). In addition, the new results reveal a second plateau between 15 and cal kyr BP. This plateau, resulting from a 13 drop in atmospheric C, compresses the C time scale such that all materials between.9 and 13.9 kyr old give nearly the same C age:.4 ±.2 kyr BP. This has repercussions for C dating: the C time scale does not allow for precise C age determinations over this interval.

15 C Comparison to 5 kyr BP with Paired Dating of Corals 11 a b C age (kyr BP) kyr 2 kyr offset kyr 4 kyr offset 1 3 c d 5 C age (kyr BP) kyr kyr offset kyr kyr offset Calendar age (kyr BP) Calendar age (kyr BP) Figure Details of coral calibration results. Calibration results between and cal kyr BP (a), 11 and 17 cal kyr BP (b), 13 and 32 cal kyr BP (c), and 24 and 52 cal kyr BP (d). Symbols are the same as in Figure 7. In addition, curved lines are the IntCal9 calibration envelop, small triangles represent the Lake Suigetsu varved sediment record of Kitagawa and van der Plicht (), and the gray shaded envelope represents the Bahamas speleothem record of Beck et al. (1). Parallel dashed lines indicate -yr offset lines from the solid line of equal age. Error bars are 2 σ. An enlargement of the 15 -cal kyr BP plateau is shown in Figure 9c. In addition to the new results, the plot includes the IntCal9 envelope and the previous coral and marine varve data that were used to establish the IntCal9 spline in this interval. While the new data are in good agreement with the original data, the IntCal9 envelope does not see the plateau and crosses diagonally through it. This points to the clear need for higher-resolution records. Several studies propose that the drop in atmospheric C associated with a later C plateau, between.5 and 11.5 cal kyr BP, was caused by changes in the carbon cycle likely associated with the Younger Dryas climate event (Edwards et al. 1993; Goslar et al. 1995; Hughen et al. 199;

16 12 K B Cutler et al. Calendar age (kyr BP) a C age (kyr BP) C( ) C age (kyr BP) b c kyr 3 kyr offset 2kyr 1kyr 15 Calendar age (kyr BP) Figure 9 C age plateau between 15 cal kyr BP. Solid circles are data from this study, open circles are from previous coral records (Bard et al. 199, 1993, 199, 199; Edwards et al. 1993; Burr et al. 199; data were screened with the δ 234 U i criteria), and the jagged line is the tree-ring calibration (Kromer and Spurk 199). Parallel dashed lines indicate -yr offset lines from solid diagonal line of equal age. Rapid drops in the atmospheric C record between and 1 cal kyr BP (a), and corresponding C age plateaus in the calibration (b), are marked with boxes. The data support the previously identified plateau between 11.5 and.5 cal kyr BP and reveal a second plateau between and 15 cal kyr BP. An enlargement of the 15-cal kyr BP plateau (d) includes the Cariaco Basin varved sediment record (grey 2-σ envelope, Hughen et al. ) and the IntCal9 calibration (black envelope). All error bars are 2 σ.

17 C Comparison to 5 kyr BP with Paired Dating of Corals 13 Calendar age (kyr BP) a C( ) b -43 Cresidual( ) 5 Tree rings/ corals GISP2-39 δ 1 O( ) Calendar age (kyr BP) Figure Atmospheric C and the GISP2 δ 1 O record, 1 cal kyr BP. (a) Crosses are a combined atmospheric C data set based on tree rings after 11.9 cal kyr BP and corals (this study and previous records, see Figure 9) prior to 11.9 cal kyr BP. Jagged solid line is 3-point running average of this data set; dashed line represents C values expected from changes in geomagnetic field intensity (Guyodo and Valet 199). (b) Solid line is the 3-point running average from (a) with the geomagnetic trend subtracted out (see text). Gray line is the GISP2 oxygen isotope record (Grootes et al. 1997). Boxes mark the times of the C age plateaus shown in Figure 9.

18 14 K B Cutler et al. C ( ) a 5 1 b 3 4 Calendar age (kyr BP) C ( ) 4 35 c C age (kyr BP) 3 25 kyr kyr offset Calendar age (kyr BP) Figure 11 C excursion at 2 cal kyr BP. Line connects tree-ring data, coral data from this study (solid circles), and coral data from previous records (open circles; Bard et al. 199, 1993, 199, 199). All errors are 2 σ. Boxes around symbols indicate samples with concordant 23 Th- 231 Pa ages. Dotted line represents C values expected from changes in geomagnetic field intensity. The rectangle identifies timing of plateau. (a) Atmospheric C record between and 45 cal kyr BP showing long-term geomagnetic contribution to C values. (b) Excursion and subsequent drop in atmospheric C causing C age plateau in calibration (c).

19 C Comparison to 5 kyr BP with Paired Dating of Corals 15 7 a -72 Cresidual ( ) 5 3 Tree rings/ corals δ 1 O ( ) -42 H1 H2 H3 GISP Calendar age (kyr BP) -44 b cooler GISP2 δ 1 O ( ) -4-3 H1 H2 H3 H4 H5-32 warmer Calendar age (kyr BP) Figure Atmospheric C and the GISP2 δ 1 O record, 37 cal kyr BP. (a) Dark line is the tree-ring and coral data, as in Figure 11; gray line is the GISP2 oxygen isotope record. H1,2,... indicate the timing of Heinrich events 1,2,...; (b) GISP2 record (top) and cooling trends (bottom). Cooling cycles are found to culminate in Heinrich events followed by abrupt warming, causing the saw-tooth pattern described by Bond et al. (1993).

20 1 K B Cutler et al. Stuiver et al. 1995; Oeschger et al. 19). One mechanism discussed is an increase in the ocean circulation rate. The atmospheric C/ C value is highly sensitive to small changes in the circulation rate because the ocean contains over 5 times more carbon than the atmosphere and has a significantly lower C/ C ratio due to C decay during its long residence time (Siegenthaler et al. 19; Keir 193). As such, an increase in the mixing rate will lower the atmospheric C/ C ratio to approach the oceanic value. Thus, an increase in ocean mixing during the Younger Dryas may at least be partially responsible for the atmospheric C drop recorded during this time interval, although recent studies do show evidence for slowing in ocean circulation during this period (McManus et al. 4). Alternatively, it is possible that a decrease in the C production rate, perhaps associated with changes in the solar magnetic field (Goslar et al. ), may have also contributed to this drop. Here, we examine whether there is a climatic link to the C drop recorded between 15 and cal kyr BP. First, we note that at least a portion of this interval corresponds to the timing of the first melt-water pulse, the most dramatic melting event during Termination I (Fairbanks 199; Bard et al. 199). Second, this interval corresponds to a period of generally increasing δ 1 O in Greenland ice (cf. Figure b), resulting from significant temperature increases, and decreasing δ 1 O of Chinese speleothem calcite, resulting from increases in summer Monsoon intensity (Wang et al. 1). In addition to climatically induced carbon cycle changes, this feature could also be linked to a decrease in C production due to changes in geomagnetic field intensity or solar activity. We first investigate the possibility that C production changes arising from fluctuations in the terrestrial magnetic field were the cause of the C drop observed between 15 and cal kyr BP. To evaluate this, we converted the SINT stacked paleointensity record of Guyodo and Valet (199) into the corresponding C production curve using the calculations of Lal (19). Because atmospheric C oxidizes and becomes incorporated into the larger reservoirs of the global carbon cycle, we then used the C production data as input to a simple carbon cycle model (Houtermans and Suess 1973). The results are broadly consistent, but slightly higher than those of Beck et al. (1) based on the SINT stacked geomagnetic field record and an invariant modern balanced 13-box carbon cycle. The predicted C values resulting from changes in dipole shielding are shown as a dashed line in Figure a. Although terrestrial magnetic field changes predict a general diminution of C over the interval, we find no evidence of a shift in dipole field intensity that could have produced this large, rapid C drop observed in the data. We next explore evidence that climatic shifts and associated carbon cycle changes resulted in the 15- to -cal kyr BP drop in C. In Figure b, we subtract out the geomagnetic contribution to the C record, where the tree-ring and coral C data has been combined with a 3-point moving average trendline, and compare the residual with the inverted oxygen isotope record from the Greenland Ice Sheet Project (GISP2; Grootes et al. 1997). The figure reveals a strong correlation between climate and residual C changes: both the.5- to and the 15- to -cal kyr BP intervals of falling C values that create the age plateaus are associated with similar transitions in the GISP2 record. The records also hint that Greenland climate changes preceded atmospheric C changes by a few hundred years. This apparent lag of atmospheric C may well result from the significant residence time of carbon in the ocean. The observed correlation between climate and residual C could be caused by two phenomena, either a change in ocean circulation associated with a change in climate, which changes carbon cycling and affects C, or a change in solar activity. Although some solar involvement cannot be ruled out, a number of arguments suggest that the main cause of the C-climate relationship is carbon cycle changes related to changes in climate. If the

21 C Comparison to 5 kyr BP with Paired Dating of Corals 17 C changes were caused by solar changes, one would expect changes in Be to correlate with changes in C in this interval. This is indeed the case for much of the Holocene (Alley et al. 1995; Finkel and Nishiizumi 1997). However, the atmospheric/greenland ice Be flux is similar during the Holocene and Late Glacial (Alley et al. 1995; Finkel and Nishiizumi 1997; Hughen et al. ), whereas the Holocene C fluctuations are factors of several smaller than the Late Glacial C fluctuations considered here. Thus, unless there is large amplification of the C response to solar variations during the Late Glacial, or there is a climatic artifact in the ice-core Be data, it would appear that solar variation is not the main cause of the large Late Glacial C shifts observed here. The last remaining possibility, that shifts in ocean circulation caused the observed drop in C recorded between 15 and cal kyr BP, is difficult to evaluate directly. For example, we do not have a good understanding of paleo-ocean circulation geometry and how this may change on millennial and sub-millennial scales during deglaciation. Nevertheless, as was done for the Younger Dryas C drop, we can make a qualitative argument for the 15- to -cal kyr BP C feature. Periods of warming in Greenland observed during this interval in the GISP2 record were likely to have been accompanied by higher rates of North Atlantic Deep Water formation, and thus higher rates of ocean circulation. As discussed previously, greater ocean ventilation rates would result in a lowering in atmospheric C values (cf. Stuiver et al. 1995). Thus, whereas we cannot rule out solar-related production changes, there seems to be a viable mechanism for carbon-cycle modulation of C in this interval. C Excursion at 2 cal kyr BP Figure 11 shows the new extension of the calibration between and 3 cal kyr BP and the corresponding atmospheric record. Notably, 2 points with concordant 23 Th- 231 Pa ages at face value suggest a large C excursion at 2 cal kyr BP. The 1 offset corresponds to a C age that is. kyr younger than the calendar age. The period prior to this peak is sparsely populated with atmospheric values hovering around 5, while the interval after this peak shows a clear line of descent between 7 and 4, in agreement with the single calibration point of Bard et al. (199a). Including this earlier data point, the drop in atmospheric values is composed of calibration points based on corals from 5 different cores drilled in Papua New Guinea, Vanuatu, and Barbados. The drop in atmospheric C creates a 3-yr age plateau in the calibration with a C age of 19. ±.3 kyr. If confirmed, this extended C plateau would severely complicate the use of C dating methods to establish a chronology of events leading up to the Last Glacial Maximum. Several aspects of the calibration data forming the C plateau lend some support for its existence. First, the corals produce a self-consistent atmospheric history despite having originated from such diverse environments; secondly, two of the samples record concordant 23 Th- 231 Pa ages; and finally, the samples were preserved below sea level for much of their history, making them less likely to be altered. However, 2 recent studies did not show a large C excursion in this time range (Hughen et al. 4; Bard et al. 4). These are both based on stratigraphic correlations to the GISP2 chronology and, therefore, both have inherent uncertainties associated with the accuracy of the correlations. Nevertheless, both record intermediate values for C in this time interval. If valid, the origin of the atmospheric C excursion at 2 cal kyr BP is puzzling. One possibility is diagenetic inaccuracy that was not detected despite use of 231 Pa. Thus, it is imperative that future efforts test whether this peak can be replicated. If diagenetic inaccuracy is not the cause, there are several possible causes. One possibility is a change in cosmogenic nuclide production rates. Be records from polar ice reveal a similar spike that is attributed to a dramatic shift in cosmogenic nuclide production rates of undetermined origin (Raisbeck et al. 197; Yiou et al. 1997). However,