A community-wide intercomparison exercise for the determination of dissolved iron in seawater

• Published in: Marine Chemistry
• Main Authors: Bowie, A.R., Achterberg, E.P., Croot, P.L., de Baar, H.J.W., Laan, P., Moffett, J.W., Ussher, S.J., Worsfold, P.J.
• Format: Article in Journal/Newspaper
• Language: unknown
• Published: 2006
• Subjects: South Atlantic Ocean
• Online Access: https://eprints.soton.ac.uk/38073/

The first large-scale international intercomparison of analytical methods for the determination of dissolved iron in seawater was carried out between October 2000 and December 2002. The exercise was conducted as a rigorously “blind” comparison of 7 analytical techniques by 24 international laboratories. The comparison was based on a large volume (700 L), filtered surface seawater sample collected from the South Atlantic Ocean (the “IRONAGES” sample), which was acidified, mixed and bottled at sea. Two 1-L sample bottles were sent to each participant. Integrity and blindness were achieved by having the experiment designed and carried out by a small team, and overseen by an independent data manager. Storage, homogeneity and time-series stability experiments conducted over 2.5 years showed that inter-bottle variability of the IRONAGES sample was good (< 7%), although there was a decrease in iron concentration in the bottles over time (0.8–0.5 nM) before a stable value was observed. This raises questions over the suitability of sample acidification and storage. For the complete dataset of 45 results (after excluding 3 outliers not passing the screening criteria), the mean concentration of dissolved iron in the IRONAGES sample was 0.59 ± 0.21 nM, representing a coefficient of variation (%CV) for analytical comparability (“community precision”) of 36% (1s), a significant improvement over earlier exercises. Within-run precision (5–10%), inter-run precision (15%) and inter-bottle homogeneity (< 7%) were much better than overall analytical comparability, implying the presence of: (1) random variability (inherent to all intercomparison exercises); (2) errors in quantification of the analytical blank; and (3) systematic inter-method variability, perhaps related to secondary sample treatment (e.g. measurement of different physicochemical fractions of iron present in seawater) in the community dataset. By grouping all results for the same method, analyses performed using flow injection-luminol chemiluminescence (with ...