Large Variations of Surface Ocean Radiocarbon: Evidence of Circulation Changes in the Southwestern Pacific

Radiocarbon (A14C) and stable isotope (slSo and 513C) records are presented for biannual samples from a 323-year banded coral series collected from the southern Great Barrier Reef, Australia. The high-precision A14C record contains variations on an interannual timescale, that are particularly large between A.D. 1680 and 1730. By comparison with tree ring A14C records [$tuiver and Quay, 1980; M. Stuiver, personal communication, 1992), it is clear that these shifts were not caused by changes in the A14C of atmospheric COo.. Changes in vertical mixing and large scale advective changes involving source waters to the western Coral Sea region are likely processes that could account for these large A14C variations. Most low A14C values for the period A.D. 1635-1875 coincide with E1 Nifio/Southern Oscillation (ENSO) events as reported by Quinn et al. [1987] for the eastern tropical Pacific. However, ENSO does not explain all of the variations, especially during 1875-1920 when A14 C values remained high. Cross-spectral analysis of the early half of the A14C and 5180 records (A.D. 1635-1795) reveals that the 6-year period is coherent; this coherency is not present in the latter half (A.D. 1797-1957) of the isotope records. These data support the concept of century timescale changes in the nature of ENSO, as it is manifest in the southwestern Pacific. Our coral record shows no evidence of a Suess effect, the lowering of A14C from late 1800s through 1955 due mainly to COo. input from fossil fuel burning. This is coincident with the ctmnge we observe in the nature of ENSO and is further evidence that a long-term ctmnge in mixin(cid:127) of upper waters occurred in this region.


INTRODUCTION
Prior to nuclear weapons testing in the 1950s and early 1960s, radiocarbon was produced in the stratosphere solely by the reaction of cosmic ray secondaries (neutrons) and atomic nitrogen. Prebomb A14C of atmospheric COo., as reconstructed from dendrochronologicMly dated tree rings, varied by 10-20•.. on century timescales [Suess, 1958[Suess, , 1968Stuiver, 1961]. Stuiver et al., [1991] assert that the similarity (in timing and magnitude) of the records of tree ring A14C and ice core løBe [Beer et al., 1988] indicates that solar forcing appears to have been the primary cause of century timescMe variations.
A question that is yet unanswered is the role that the oceans have played in controlling the atmospheric AlPC record. Changes in average wind speed or rates of ocean mixing, if sufficiently large, could cause changes in the A 14 C of atmospheric COo.. Time histories of high-precision AlPC vMues for the surface waters of the ocean are the missing information needed to quantify the oceanic contribution toward shaping the observed variations in prebomb atmo-dates for samples prior to 1948 will be referred to as the latter year of the 2-year pair, i.e., 1946-1947 is referred to as 1947. The cut was made directly underneath the highdensity band, which makes the midpoint of each 2-year band about December of the first year. Radiocarbon measurements were performed on coral samples (~25 g) that had been acidified and converted to acetylene gas via a lithium carbide intermediate [Griffin and Druffel, 1985]. Samples were counted in six quartz gas proportional beta counters (M1 volumes 1.5 L) for six to seven 2-day periods at 900 mm of Hg and 22øC.
All A x4C measurements were corrected for isotope fractionation to a 5XaC (relative to PDB-1) of-25.0% o and for decay since the time of formation [Stuiver and Polach, 1977].
The 5xaC vMues for this correction were measured on CO2 produced from reburned acetylene gas. Errors on the measurements based on counting statistics were 4-2.0ø/00. Laboratory or system random errors are 25% of this value for modern samples; thus the reported error is 4-2.5ø/00. All of the biannual and annual corm samples were measured for 5XaC and 5XSO using previously described methods [Druffel, 1989]. In addition, selected 4 to 6 year portions of the core were ground off using a dremel tool into 0.10 to 0.15-year wide layers and measured for stable isotopes. Isotope ratios were measured on a V.G. Micromass 602E mass spectrometer.
The error determined from the standard deviation of replicate analyses of numerous corm samples was 4-0.09ø/00 for both 5XSO and

High-Precision Radiocarbon in Australian Corals
The high-precision Ax•C time history for the Abraham Reef corm is presented in Figure  AX•C shifts from very high to very low values from 1677 to 1687, 1691 to 1703, and 1727 to 1729. These 20-227oo shifts in Ax4C are much larger than those expected from anMyticM scatter, as the l•r error of the measurements (4-2.57oo) is a factor of 8-9 times smaller than the range of values. In fact, these large shifts in Ax•C are similar to the entire spatiM range of prebomb (A.D. 1955) A•C values in Pacific surface waters between the central ocean gyres (-35 to-45ø/øø) and the nutrient-rich tongue off the coast of Peru in the eastern tropicM Pacific (-65ø/oo) [Druffel, 1981]. Subsequent to 1729, the interannuM variability of the A x •C values is significantly less than that in the preceding period.
The temporal variations in the post-IS00 results are smMler and generally occur at a slower rate. The magnitudes of these variations are 4 to 7 times greater than the la error of the individual measurements, less than those observed for the late seventeenth and early eighteenth century data.  [Eddy, 1976] and a decrease in after A.D. 1900 (Figure la) due to fossil fuel dilution and solar variation, known as the Suess effect [Suess, 1953]. Similarly, the corm record shows on average higher A X•C vMues during the Maunder minimum. However, there is little suggestion of a Suess effect in the corm record after 1900. Overall, the decadM-averaged tree ring and biannuM coral A•C records are similar in shape prior to 1900, with larger trends, as expected, in the tree ring data.

Comparison of Coral and Tree Ring
To determine the deviation of the long term corm trend from that expected from the atmospheric tree ring record, we compare our results (boxcar averaged over 20-year increments) with the model-calculated A X•C record of the mixed layer (using bidecadal atmospheric A•C values and a box-      Figure 2).

Correlation Between Coral A•4 C and ENSO Events
It is important to note that the majority of these occur within an anomolous period of 1875-1920. This anomolous period also encompasses the first half of the Suess effect in the atmosphere (see Figure la). The seasonal 5XaC measurements range from -2.2 to +0.2•/0., with a mean value of-1.0•/0. (Figure 4b). The seasonal regularity of the 5x3C record is not as apparent as that in the 5XsO data, and is not expected owing to the preponderance of vital effects that plague coral 513C records [Swart, 1983].    Island (4øN, 159øW) [Druffel, 1987] indicate that the prebomb A14C signature for the SEC was low, -60 4-5•oo, due to divergence along the equator; seawater DIC data indicate a prebomb A•4C signature for the EAC was-38 4-5•oo [Rafter, 1968]. The average of the recent coral results (-49•too) falls in the middle of this range, as one would expect for a mixture of these two sources. To specifically identify the relative magnitude of vertical and advective changes in circulation that occurred in the past, future studies need to include A x•C and stable isotope records from corals collected along the western boundary of the tropical Pacific. Seasonal variability of AX•C needs to be examined, so that changes in the water mass surrounding the coral on short timescales can be assessed. Prebomb A x • C measurements, or estimates based on bomb x 4 C distributions, of thermocline waters from the surrounding regions are essential. Proxies of nutrient levels in these waters, for example, Cd/Ca ratios in corals from Abrahams Reef [Delaney and Linn, unpublished data], will also provide a method of deconvolving vertical and advective (horizontal) changes in circulation. The isotope data also point to an anomolous period from 1875 to 1920. There was no correlation between low A•C and ENSO events. This suggests that warmer, higher A•C waters predominated this region for 5 decades. This could have provided the necessary mask to disguise the Suess effect that was anticipated, had these waters been in steady state with surrounding water bodies.