During the planning phase of WOCE, the accelerator mass spectrometry (AMS) technique for measuring 14C was still relatively new in the United States. The general procedures had been worked out; however, no laboratory was prepared to handle the large number of samples expected from the WOCE program, nor had it been demonstrated that the AMS technique could deliver the required precision on a routine basis. The National Ocean Sciences AMS Facility (NOSAMS) at Woods Hole Oceanographic Institution (WHOI) was established in 1989 to serve this purpose. In planning the WOCE Pacific field work, it was recognized that sample collection would begin well before NOSAMS could deliver the high precision offered by conventional beta counting techniques. Therefore, both techniques were utilized.
On those WOCE legs that included both large-volume (LV) and small-volume (SV) sampling, the LV stations were spaced at an average interval of once every five degrees (~300 nautical miles). LV stations normally included two casts of nine Gerard barrels each, covering the water column from ~1000 m to the bottom. The upper kilometer of an LV station was covered by 16 SV samples taken from the CTD/rosette cast. One to three SV stations were placed between each LV station. At SV stations, only the upper thermocline region was sampled. Sixteen SV samples were taken at these stations.
Radiocarbon was extracted from the LV samples at sea as 14CO2, absorbed on excess NaOH and returned to shore in well-sealed glass bottles using a modification of the technique described by Fonselius and Östlund (1959). Once ashore the samples were sent to researchers at one of two laboratories for analysis: G. Östlund, Tritium Laboratory, University of Miami, Miami; or M. Stuiver, Quaternary Isotope Laboratory, University of Washington, Seattle. A short description of the measurement procedure and a cross-check between these two laboratories are available in Stuiver et al. (1974). Stuiver reports an error estimate for each analysis that ranges from 2.5 to 4.0 per mille (°/°°), while Östlund reports a uniform sample error of 4°/°°. In both cases the reported uncertainty is primarily counting error and does not include any error due to sample collection.
All SV 14C samples were collected from standard CTD/rosette casts into 500-mL glass bottles fitted with high-quality ground-glass stoppers. The samples were poisoned with HgCl2 immediately after collection and then returned to the United States for extraction and analysis at NOSAMS. Details of the extraction, counting, etc., are available from Key (1991), McNichol and Jones (1991), Gagnon and Jones (1993), and Cohen et al. (1994). The standard used for the 14C measurements is the National Bureau of Standards oxalic acid standard for radiocarbon dating. All results are reported as Δ14C, which is the deviation (in °/°° ) from unity of the activity ratio of sample to standard, isotope corrected to a sample Δ13C value of -25°/°°, where Δ13C was calculated from
|Δ13C = 1000 x [13C/12C]sample - [13C/12C]reference|
to correct for fractionation and dilution by anthropogenic CO2. For more information on standards and calculation methods, refer to the papers by Broecker and Olson (1961), Stuiver and Robinson (1974), and Stuiver (1980). As measurements were completed, the results were communicated from the analytical laboratory to the principal investigator responsible for the cruise via periodic data reports. R. Key gathered the Δ14C data; merged it with hydrographic data supplied by either the chief scientist or WHPO; added WOCE quality-control flags; and finally submitted the data to WHPO along with a final report for the WOCE Section P17C, Cruise TUNES, Leg 1 (Key 1996a, b). All of the LV samples collected in the Pacific Ocean will be processed by 1997, and the Pacific SV samples will be completed by 1998.
During the GEOSECS Program, the precision of the LV technique was established as 2-4°/°°. This precision is primarily a function of sample counting time and has remained constant throughout the succeeding large-scale ocean-survey programs. At the beginning of WOCE, the ultimate precision of the AMS technique and the degree of compatibility of the AMS and LV data (i.e., the absence of systematic errors in either data set) were unknown. NOSAMS is currently processing water samples with a mean "external" precision of 3.6°/°°. This degree of precision is indicative of the AMS target preparation and counting and does not include any uncertainty resulting from sample collection, storage, or stripping. A better estimate of the sample precision can be obtained by comparing the results from duplicate samples. A summary of all the true WOCE duplicates (i.e., two different sample bottles rather than two analyses from the same bottle) analyzed at NOSAMS through mid-1996 shows that the average standard deviation for the pairs was 4.6°/°°. The reason for the difference between this number and the external precision estimate (3.6°/°°) is currently unknown; however, it is attributed to either sample collection or sample processing prior to counting. Sample storage experiments at NOSAMS and other facilities have so far indicated that this is not a source of error. A reproducibility of 3°/°° is needed for the AMS technique to be equivalent to the average uncertainty for the LV technique.