The majority of samples processed at the National Ocean Sciences AMS Facility (NOSAMS) thus far were collected as part of the World Ocean Circulation Experiment (WOCE). Due to the long storage time (2–3 yr) required to analyze thousands of samples on the accelerator mass spectrometer (AMS), a test was designed and implemented to determine the effects, if any, of storage time on 14C concentration. We find no systematic offsets in AMS measurements made over a 5-yr period between a total of 16 replicate sets from surface and deep water collected at the same locality. Furthermore, the average δ14C value from the deepwater AMS replicates (-213.1%, std. dev. 7.3) agrees very closely with the conventional 14C results published for GEOSECS (-212.7%) from station 320 taken 20 yr earlier.
A total of 73 WOCE shipboard replicate sets (162 AMS measurements) were analyzed with a mean precision of 4.3%. AMS results from 20 more shipboard replicate sets (44 AMS measurements) submitted as CO2 from the Stable Isotope Laboratory (SIL) at the University of Washington were analyzed with a mean precision of 3.4%. These results suggest no significant difference between water stripping methods used in each preparation lab.
To assess reproducibility, we calculate a pooled estimate of σ for replicates called s, which we use as an approximation of σTOT for a given sample type. The s for WOCE seawater replicates is 4.9% and 5.8% for SIL gas replicates. These numbers demonstrate an overall reproducibility of seawater AMS results at NOSAMS that is in line with reported errors. We take the difference between total error s and machine error as the overall standard deviation of combined uncertainties associated with preparation of samples and with AMS. For seawater samples processed at NOSAMS, σSPL is calculated to be 2.4%, and for the SIL gas replicates it is 4.8%.
Reproducibility of samples prepared with an acid hydrolysis technique is demonstrated using 24 coral samples submitted in triplicate by Dr. R. G. Fairbanks of Lamont Doherty Earth Observatory. Seventy-two replicates were prepared and analyzed at NOSAMS with a mean reported precision of 1.2%. The pooled estimate s for the Fairbanks triplicates is 2.6%. We calculate a laboratory reproducibility uncertainty for coral hydrolysis samples of 2.2%.
In 1993, NOSAMS participated in the Third International Radiocarbon Intercomparison (TIRI) Study. We report here 60 AMS analyses of the six TIRI test materials, five of which are organic carbon samples, to validate sample-processing methods for organic carbon sample AMS analyses at NOSAMS.