Regional 14CO2 Offsets in the Troposphere: Magnitude, Mechanisms, and Consequences

See allHide authors and affiliations

Science  21 Dec 2001:
Vol. 294, Issue 5551, pp. 2529-2532
DOI: 10.1126/science.1066114


Radiocarbon dating methods typically assume that there are no significant tropospheric 14CO2 gradients within the low- to mid-latitude zone of the Northern Hemisphere. Comparison of tree ring 14C data from southern Germany and Anatolia supports this assumption in general but also documents episodes of significant short-term regional14CO2 offsets. We suggest that the offset is caused by an enhanced seasonal 14CO2 cycle, with seasonally peaked flux of stratospheric 14C into the troposphere during periods of low solar magnetic activity, coinciding with substantial atmospheric cooling. Short-term episodes of regional14CO2 offsets are important to palaeoclimate studies and to high-resolution archaeological dating.

The basic assumption about the atmospheric distribution of 14C is that, although sources of 14C-depleted C (such as outgassing of CO2 from the ocean mixed layer) or 14C-enriched C (such as intrusion of newly produced 14C from the stratosphere) are restricted to certain areas and thus could leave a regional 14CO2 imprint, rapid atmospheric mixing produces an efficient dispersion of14CO2 gradients. Hence, over even short time spans, on the order of 1 month, there are approximately uniform hemispheric levels of 14C. In consequence, a cornerstone of14C dating is the assumption of a spatially uniform14CO2 source level during carbon uptake by plants, allowing us to provide and employ a single universal14C data set for the calibration of the 14C time scale (1, 2).

Support for the assumption that any regional14CO2 differences are small and may be ignored (that is, they are close to the detection limits of the current radiocarbon technique) comes from carbon cycle modeling. Braziunaset al. (3) used a coupled atmosphere-ocean general circulation model (GCM) to calculate the preindustrial global atmospheric 14CO2 distribution. For the low- and mid-latitudes of either hemisphere, they found zonal mean meridional 14C activity gradients of only about 1 per mil (‰), which is equivalent to a difference of 8 14C years in measurement determination. In view of achievable measurement precision of no better than 2 to 3‰ even at high-precision14C laboratories, this work shows that the low- to mid-latitudes of each hemisphere are uniform in14CO2 concentration for practical purposes. Only at high latitudes does the 14CO2 gradient increase to up to 6‰ (equivalent to an apparent 14C age difference of 48 years) (2–4).

The ability to directly detect 14CO2gradients depends on measurement precision. The most extensive work has been carried out during the construction of the high-precision14C calibration data sets, when absolutely dated tree ring chronologies from various mid-latitude regions of Europe and North America were compared to study the magnitude of regional14C offsets at several, up to millennia-long, time intervals of the Holocene. The results of such studies have been inconsistent about whether measurable 14C gradients exist (5–8).

The majority of the data document mid-latitude mean regional differences well below 10 14C years (albeit only if the corrections of the Belfast laboratory are to be applied), but several shorter intervals show regional differences between zero and up to several decades. The latter are significant if correct. Some workers have suggested that changes in oceanic and atmospheric circulation patterns may be responsible for such offsets (2,6).

It is important to resolve this disagreement, because, if substantial regional offsets in 14C levels are real within the core low- to mid-latitude zone of the Northern Hemisphere, the calibration of 14C ages into calendar years would require regional corrections, in the same way as is done for marine14C ages (ΔR values). This issue is particularly relevant for building an accurate time frame for the prehistory of the eastern Mediterranean, because there small but highly important discrepancies between radiocarbon dates and dates derived from the interpretation of historical and protohistorical evidence have led to several decades of controversy, such as the debate concerning the age of the Thera volcano eruption (9).

We therefore designed a three-stage study [the Eastern Mediterranean Radiocarbon Intercomparison Project (EMRCP)] to establish the extent, if any, of a regional 14C gradient between a key region and a tree species used to build the international standard high-precision calibration curve (1, 10) [German oak (GeO) from southern Germany] and the eastern Mediterranean region, using wood from Anatolia. 14C measurements were performed at the Heidelberg Radiocarbon Laboratory. First, we determined the overall dating precision of the Heidelberg laboratory by replicating the 14C calibration data set, based on GeO, in the 17th and 16th centuries B.C. Second, we determined the 14C ages of decades of wood that grew at exactly the same time from absolutely dated wood samples from Anatolia and southern Germany grown in the 15th to 17th centuries A.D. Third, using the information gained in the first two stages, we then explored the anchoring of a floating Anatolian tree ring chronology in the 2nd millennium B.C. to the 14C data sets from the absolutely dated GeO and Irish oak (IrO) chronologies.

For the first stage of EMRCP, we selected 22 decades of GeO in the interval 1490–1710 B.C., from the middle Main river, southern Germany. This section had been previously measured in the Seattle14C laboratory (11). For stage 2 of EMRCP, we selected decadal samples from two absolutely dated tree ring chronologies: Turkish pine (TuP) (Pinus nigra) fromÇatacık in western Anatolia and GeO (Quercus robur) from historical buildings in southern Germany. These sequences cover the interval 1420–1649 A.D.; that is, a span of 23 decades (12).

An independent assessment of laboratory errors is possible through our replication of the GeO data comprising part of the INTCAL98 calibration data set (13). In Fig. 1, our measurements are compared to those of the Seattle 14C laboratory. The mean difference between the Heidelberg-Seattle decadal data is 2.3 14C years. The observed versus expected standard deviation of the differences is 29.0 versus 22.5 14C years. The increase in variance is dominated by the decade with midpoint at 1675 B.C., with an observed difference of 80 14C years (3.8σ variation based on quoted measurement errors). When this decade is omitted, the observed versus expected standard deviation drops to 23.8 versus 22.514C years. We repeated the Heidelberg measurements of the decades with midpoints at 1655–1675 B.C. (note the additional GeO data in Fig. 1), with results identical to the previous ones. The bidecadal14C age of IrO (1, 7) at 1670 B.C. is 3344 ± 21 14C years, which does not help to resolve the discrepancy satisfactorily. Therefore, we decided to omit the decade with midpoint at 1675 B.C. from the comparison. The total increase in the variance of the pairs is explained by an additional, unknown error of 7.8 14C years (18.314C years if the decade with midpoint at 1675 BC is included), caused by either one of the two laboratories or by both of them. From this comparison, we conclude that a conservative upper limit of an additional, unknown laboratory error for the Heidelberg facility is 8 14C years, included in all data presented here.

Figure 1

Comparison of Heidelberg and Seattle (13) measurements on similarly sourced GeO decadal samples.

A comparison of 14C ages of pairs of same-age decades from GeO and TuP wood samples is shown in Fig. 2. The mean absolute difference of the 23 pairs of GeO-TuP is 1.4 14C years, indicating a regional offset between Central Europe and Anatolia that is negligible. However, a clear trend is visible in the data, with all TuP ages being older than GeO in the interval A.D. 1440–1540 (mean difference between TuP and GeO, 17 14C years) but mostly younger in the intervals before and after, A.D. 1420–1440 and A.D. 1550–1640 (mean difference between TuP and GeO, –14 14C years for A.D. 1550–1640). The TuP > GeO episode occurs at a time of strongly rising atmospheric 14C levels, which leads to the rapid decrease of 14C ages as compared to calendar ages; this episode also corresponds with the Spörer minimum of solar activity (A.D. 1416–1534) (14) and the associated cooler climate episode in the mid-15th century A.D. in the Northern Hemisphere that has been detected in many studies (15,16).

Figure 2

14C ages of decadal samples of GeO and TuP, grown contemporaneously.

We observe the same pattern when we try to anchor the floating Aegean Bronze Age juniper chronology (ABJ) (17) to our GeO data (Fig. 1) and the 14C calibration curve (1). Using 52 ABJ decadal 14C age determinations made on samples of Juniperus foetidissima and J. excelsa, we can obtain a match (18) only when we accept that the ABJ 14C data in the early 8th century B.C. are offset by ∼30 14C years with respect to the14C data based on Central European wood (Fig. 3 and inset). Again, we note that the14C levels differ significantly, and for a short period consistently, only for intervals of strongly changing atmospheric14C, coincident with widespread cooling in the Northern Hemisphere (19, 20).

Figure 3

Wiggle-match of 52 14C decadal age determinations of the floating Aegean Bronze Age juniper chronology to the internationally recommended INTCAL98 calibration data set (1): See (31) for details and discussion. In the interval from the mid-9th to mid-8th centuries B.C. (inset), a period of rising atmospheric 14C levels (strongly declining14C ages), a significant 14C age difference exists between Central Europe and Anatolia.

What could cause such a difference in the atmospheric14C level between the two regions? The key to the answer may lie in the magnitude, timing, and location of injection of stratospheric 14C into the troposphere, combined with the mechanisms of carbon storage in tree rings. Based on the mean lifetime of 14C of 8267 years, in a steady state, every year 1/8‰ of the total 14C inventory is replenished by14C production in the stratosphere and troposphere. Ultimately, 14C reaches the deep ocean, which contains most of the global 14C inventory; but mixing within the stratosphere, latitude-dependent cross-tropopause exchange, carbon and14C exchange across the sea surface, and deep ocean ventilation—all of which operate on a wide range of time scales—attenuate and cause a phase shift of the 14C source. Because the atmosphere holds just ∼1.5% of the global14C inventory, it is a sensitive monitor of transient changes in any of the transport stages. In fact, from our present-day tropospheric 14C monitoring network (4) and from modeling bomb radiocarbon (21), we deduce a preindustrial/pre-bomb seasonal 14C component of ∼4‰ activity variation between low winter/spring and maximum summer activity. Thus, depending on the rate of carbon uptake during the growing season and the phase relation of the growing season between two regions, the comparison of annual tree rings may show part of this seasonal cycle: an earlier onset of warmth, followed by restricted water availability, largely constrain average tree growth in the Mediterranean to spring and early summer, whereas Central European oaks lay down a substantial contribution of the overall carbon storage in their annual rings later, in July and August.

For most of the time covered by our comparison, the net effect is too small to be detected at the level of measurement precision. However, the amplitude of the seasonal cycle is proportional to the14C influx from the stratosphere. During times of “deep” solar minima (22), the 14C production (23) and hence 14C flux were doubled as compared to the average of the 11-year solar cycle [figure 8. of (24)] at high latitudes. Thus, the seasonal cycle becomes notable, and we can explain the difference in magnitude of the regional 14C gradient between the A.D. and the B.C. event because the 14C increase in the 8th century B.C. was 50% higher than in the Spörer minimum episode.

The recording of the seasonal 14C signal by tree rings also may have been amplified by climate change. The two events are part of widespread hemispheric cooling episodes (15, 19, 20): the initial phases of the Little Ice Age after the Medieval Warm Period, and the wet and cold earlier 8th century B.C. For this type of event, a cyclonic circulation and higher frequency of cold polar air in the North Atlantic and northwestern Europe is assumed, with the eastern Mediterranean in antiphase (25). Thus, while Central European trees may have had their growing season shifted to later in the summer, because of the cold (continental-type) winters, the vegetation in the Mediterranean may have enjoyed favorable moisture and temperature conditions leading to early growth (26). Furthermore, during deep solar activity minima, the solar ultraviolet flux was considerably reduced (27), leading to changes in stratospheric heating and circulation (28–30). Thus, the partitioning of major mechanisms in stratosphere-troposphere 14C exchange, as well as tropospheric wave propagation, may have been notably different during these intervals, adding to the 14C seasonality.

Thus, although the general calibration of 14C ages is only marginally affected by our findings, and the standard calibration curves may be used with confidence in most cases, at certain times significant deviations occur. For example, the revised best estimate dates for the floating Bronze-Iron Age Aegean dendrochronology (31) now must be shifted to ages ∼22 years older—a matter of no little importance for archaeologists (9). Finally, it is clear that high-precision 14C analyses provide a valuable tool for studying decadal- to century-scale atmospheric dynamics in the past.

  • * To whom correspondence should be addressed. E-mail: bernd.kromer{at}


View Abstract

Navigate This Article