Atmospheric Radiocarbon Calibration to 45,000 yr B.P.: Late Glacial Fluctuations and Cosmogenic Isotope Production

See allHide authors and affiliations

Science  20 Feb 1998:
Vol. 279, Issue 5354, pp. 1187-1190
DOI: 10.1126/science.279.5354.1187


More than 250 carbon-14 accelerator mass spectrometry dates of terrestrial macrofossils from annually laminated sediments from Lake Suigetsu (Japan) provide a first atmospheric calibration for almost the total range of the radiocarbon method (45,000 years before the present). The results confirm the (recently revised) floating German pine chronology and are consistent with data from European and marine varved sediments, and combined uranium-thorium and carbon-14 dating of corals up to the Last Glacial Maximum. The data during the Glacial show large fluctuations in the atmospheric carbon-14 content, related to changes in global environment and in cosmogenic isotope production.

The atmospheric14C content (expressed in Δ14C) (1) is sensitive to geomagnetic field strength and solar fluctuations (also through magnetic effects) as well as rearrangements in equilibrium between the major C reservoirs (atmosphere, ocean, and biosphere). Detailed calibration of the radiocarbon time scale into the glacial period is critical for accurate dating and a better understanding of changes in the Earth system.

Radiocarbon calibration can be performed by 14C dating of samples that can also be dated by an independent, preferably absolute dating method. The ideal samples for this purpose are tree rings, which can be dated by dendrochronology. Dendro-calibrations with (for the most part) 20-year tree-ring resolution have been obtained for almost the complete Holocene, back to about 7900 B.C. for the absolute chronology and to about 9400 B.C. including a matched floating tree-ring curve (2).

Beyond the range of tree rings, calibration has been problematic. Radiocarbon dates of terrestrial macrofossils from annually laminated sediments can potentially provide a high-resolution record of atmospheric 14C changes. However, varve chronologies have been revised several times (3). At present, calibration data from glacial varves provide a consistent data set back to about 11,000 B.C. (4-6). In addition, a marine calibration curve for the Late Glacial period is obtained by combined14C and U-series dates of corals (7,8) and 14C measurements on foraminifera from varved marine sediments (9). Because these data are for marine materials, they have to be corrected for the apparent14C age of the surface oceans, known as the reservoir effect.

Here we present a high-resolution atmospheric radiocarbon calibration from annually laminated sediments for the total range of the radiocarbon dating method [<45,000 cal yr B.P. (10)]. The sediments were taken from Lake Suigetsu (35°35′N, 135°53′E) near the coast of the Sea of Japan (11). The lake is 10 km around the perimeter and covers an area of 4.3 km2. It is a typical kettle-type lake with a nearly constant depth at the center,34 m deep. A 75-m-long continuous core (Lab code, SG) and four short piston cores were taken from the center of the lake in 1991 and 1993. The sediments are laminated in nearly the entire core sections and are dominated by dark-colored clay with white layers resulting from spring-season diatom growth. The seasonal changes in the depositions are preserved in the clay as thin laminations or varves. The sedimentation or annual varve thickness is relatively uniform, typically 1.2 mm/year during the Holocene and 0.61 mm/year during the Glacial. The bottom age of the SG core is estimated to be older than 100,000 years, close to the beginning of the last interglacial period.

To reconstruct the calendar time scale, we counted varves, based on gray-scale image analyses of digital pictures, in a 10.43- to 30.45-m-deep section, producing a 29,100-year-long floating chronology. Because we estimated the varve chronology of older than20,000 yr B.P. (19-m depth of SG core) by counting in a single core section, the error of the varve counting increases with depth, and the accumulated error at 40,000 cal yr B.P. would be less than 2000 years, assuming no break in the sediment (12).

The 14C/12C and13C/12C ratios of more than 250 terrestrial macrofossils (leaves, twigs, and insect wings) in the sediments were measured by accelerator mass spectrometry (AMS) at the Groningen AMS facility (13), after proper sample pretreatment (14).

The floating varve chronology was connected to the old part of the absolute tree-ring chronology (2, 15) by 14C wiggle matching (16), resulting in an absolute calendar age covering the time span from 8830 to 37,930 cal yr B.P. (17). The age beyond 37,930 cal yr B.P. is obtained by assuming a constant sedimentation in the Glacial.

The combined AMS 14C and varve ages provide an extension of dendro-calibration range (Fig.1). The features in the data overlapping the absolute tree-ring record agree very well, and our varve chronology also supports the recent revision of the floating German pine chronology, which was shifted by 160 years to older ages (6,15). Beyond the range of the dendro-calibration (11,700 cal yr B.P.), there is also general agreement with the European sediments (5, 6) as well as with marine calibrations (7-9).

Figure 1

(A) Radiocarbon calibration up to 45,000 yr B.P. reconstructed from annually laminated sediments of Lake Suigetsu, Japan. The small circles with 1σ error represent the 14C ages against varve ages. For the oldest eight points (>38,000 years, filled circles), we assumed a constant sedimentation during the Glacial period.The green symbols correspond to the tree-ring calibration (2, 15), and the large red symbols represent calibration by combined 14C and U-Th dating of corals from Papua New Guinea (squares) (8), Mururoa (circles), and Barbados (triangles) (7). The line indicates that radiocarbon age equals calibrated age. (B) Agreement of the 29,000-year floating varve chronology with absolute and recently revised floating German pine chronology, and radiocarbon calibration during the deglaciation.

The detailed record in atmospheric Δ14C during the deglaciation shows millennium-scale fluctuations superimposed on a long-term increasing trend into the past resulting from a decreasing geomagnetic intensity as reconstructed from geomagnetic records (18) (Fig 2). The long-term increase is also consistent with a model-generated estimate of atmospheric Δ14C increase (19), on the basis of 10Be flux records (20). Abrupt Δ14C decreases correspond to radiocarbon plateaus in the calibration curve. Near (a few centuries after) the onset of the Younger Dryas (YD), the Δ14C value decreases by 80 per mil from 10,800 to 9800 yr B.P. (12,500 to 10,000 cal yr B.P.); the decrease thus extends into the Preboreal (the earliest Holocene). This radiocarbon plateau occurs in both marine (8) and terrestrial (5) records and is referred to as the YD plateau. Our calibration shows that the YD plateau consists of two subplateaus at 10,000 and about 10,400 yr B.P.; the older one is characterized by a time of slow increase of the radiocarbon age. A similar decrease in Δ14C of 100 per mil (including magnetic effect) is observed from 12,600 to 12,100 yr B.P. (or 15,000 to 13,800 cal yr B.P.), which starts within the Oldest Dryas (OD) cold period and extends until nearly the end of the Allerød/Bølling warm period. This plateau can be related to the radiocarbon plateau recorded in (nonvarved) sediment cores from Switzerland (21). Our data strongly indicate a plateau around the OD cold period. It appears that the two radiocarbon plateaus in the YD and OD cold periods started a few centuries after the warm-to-cold transition. Furthermore, we also observe the millennium-scale fluctuations of about 100 per mil in Δ14C before the OD; maxima at 16,000, 17,500, and 19,000 cal yr B.P.; and minima at 15,500, 17,000, 18,000, and 19,500 cal yr B.P.

Figure 2

Atmospheric 14C changes during the deglaciation (<20,000 cal yr B.P.), normalized to the present value and given in Δ14C in per mil, reconstructed from annually laminated sediments of Lake Suigetsu, Japan. The symbols are as in Fig. 1. The features for the part overlapping the tree-ring record are excellent, including the wiggles in the radiocarbon calibration curve.

When compared with changes in the atmospheric CO2concentration measured directly in polar ice cores (22) and reconstructed from South American peat (23), minima at 13,500 and 15,500 cal yr B.P. seem to respond to sharp increases in atmospheric CO2 concentration. These coupled signals suggest that CO2 degassed from the ocean, especially from the intermediate-deep ocean, induced the change of atmospheric Δ14C because the ocean contains more than 90% of the global 14C inventory and the oceanic C is depleted in14C. For a recent discussion concerning the connection of atmospheric 14C and paleo-oceanographic parameters, we refer to (24). Paleo-oceanographic observations suggest that millennium-scale oscillations of the ocean thermohaline circulation (THC) occurred during the deglaciation (25). The THC oscillation can be linked to atmospheric Δ14C as follows: Δ14C increases when formation of North Atlantic Deep Water (NADW) is reduced abruptly and THC decreases, after climatic cooling such as the OD and YD cold periods; in contrast, Δ14C decreases when THC northward heat advection and deep water production resume. Our atmospheric Δ14C record implies that repeated cycles of such a process occurred, and a possibility that the NADW weakened for three or four periods during the last deglaciation is superimposed on the general trend of increasing ocean ventilation (26).

From the Last Glacial Maximum to 31,000 cal yr B.P., the long-term trend of Δ14C agrees well with reconstruction of cosmogenic isotope production rate deduced by the 10Be deposition reconstruction (19) and geomagnetic field intensity reconstruction (18) (Fig.3). For this time span, we observe two pronounced peaks in Δ14C at 23,000 and 31,000 cal yr B.P. The apparent Δ14C increases correspond to an increase in the concentration of another cosmogenic isotope, 10Be, at 23,000 and about 35,000 cal yr B.P., respectively, observed in ice cores from the Antarctic (27) and Greenland (28) as well as in marine sediments (29, 30). Furthermore, a 14C anomaly at these times has been observed previously in speleothems, dated by both 14C and U series (31). The time gap between the 14C and10Be enhancements can be explained by errors in both varve and ice-core chronologies, as well as by the different geochemical behavior of these isotopes; 14C is present in gaseous form (CO2) and gradually diffuses in the Earth system, whereas10Be is a solid attached to aerosol particles and deposited with precipitation (32).

Figure 3

Atmospheric Δ14C changes beyond the Last Glacial Maximum. The symbols are as in Fig. 1. The absolute (calendar) ages older than ∼20,000 yr B.P. should be considered as minimum ages. The red dashed line shows an instance with accumulated error of 2000 years at 40,000 cal yr B.P. in the varve chronology [see text and (12)]. The two peaks at 23,000 and 31,000 yr B.P. are superimposed on the long-term Δ14C change, probably caused by geomagnetic field change.

The broad increase in cosmogenic isotopes (both 10Be and 14C) at 23,000 yr B.P. can be explained as the increase in production rate by geomagnetic effects (28). The sharp14C peak we observed at 31,000 yr B.P. is roughly 300 per mil in Δ14C after removing the long-term trend. The 10Be increases by a factor of 2 in ice cores during a period of 2000 years (27). This factor of 2 increase corresponds to a 14C increase by a factor of 1.3 or 300 per mil (33), which is exactly what we observe in our data. These sharp enhancements in 10Be and14C at the same time are too large to be explained by rearrangements of the C reservoirs on the Earth.

Increased cosmogenic isotope production caused by a nearby supernova explosion has been suggested as a cause for the drastically increased10Be levels at this time (29, 34). Another possible explanation is a magnetic excursion with a sharp change in inclination of the geomagnetic field and the implied concomitant decrease in the geomagnetic field strength. Such events are observed as the Mono Lake and Laschamp excursions, dated at 28,000 and 33,000 yr B.P. (uncalibrated), respectively (35). The sharp Δ14C increase from Lake Suigetsu corresponds chronologically to the Mono Lake excursion. However, all of these explanations remain hypothetical.

The atmospheric radiocarbon calibration curve covering the past 45,000 years provides the basis for developing a better understanding of the past global C cycles and cosmogenic isotope production. Our high-resolution calibration curve is consistent with other proxies until 31,000 cal yr B.P. Beyond 31,000 cal yr B.P, much work is still needed to obtain a better understanding of the atmospheric Δ14C signal. Here, our calibration deviates from paleomagnetic records (18, 20) and from recent combined U/Th and 14C dating of speleothems (36). These data suggest that 14C dates at this time are 5000 years too young. This discrepancy can be caused either by speleothem dating problems (such as unknown initial14C age and possible detrital Th contamination) or missing varves in the older section of Lake Suigetsu. New 10Be data from the Arctic GRIP and GISP2 cores show the large spike at ∼41,000 cal yr BP (37), which is inconsistent with both our Δ14C maximum and the Antarctic 10Be record (27). This finding would indicate either a problem in one of chronologies, or that the 10Be and 14C peaks do not have the same cause. Future work on the Lake Suigetsu core can help to solve the remaining questions.


View Abstract

Stay Connected to Science

Navigate This Article