Combined uncertainty estimation for the determination of the dissolved iron amount content in seawater using flow injection with chemiluminescence detection.

This work assesses the components contributing to the combined uncertainty budget associated with the measurement of the Fe amount content by flow injection chemiluminescence (FI-CL) in <0.2 m filtered and acidified seawater samples. Amounts of loaded standard solutions and samples were determined gravimetrically by differential weighing. Up to 5% variations in the loaded masses were observed during measurements, in contradiction to the usual assumptions made when operating under constant loading time conditions. Hence signal intensities (V) were normalised to the loaded mass and plots of average normalised intensities (in V kg) vs. values of the Fe amount content (in nmol kg) added to a "low level" iron seawater matrix were used to produce the calibration graphs. The measurement procedure implemented and the uncertainty estimation process developed were validated from the agreement obtained with consensus values for three SAFe and GEOTRACES reference materials (D2, GS, and GD). Relative expanded uncertainties for peak height and peak area based results were estimated to be around 12% and 10% (coverage factor  = 2), respectively. The most important contributory factors were the uncertainty on the sensitivity coefficient (i.e., calibration slope) and the within-sequence-stability (i.e., the signal stability over several hours of operation; here 32 h). For GD, using peak height measurements, these factors contributed respectively 69.7% and 21.6% while the short-term repeatability accounted for only 7.9%. Therefore, an uncertainty estimation based on the intensity repeatability alone, as is often done in FI-CL studies, is not a realistic estimation of the overall uncertainty of the procedure.