A sensitive method for iron determination in seawater has been adapted on a submersible chemical analyser for in situ measurements. The technique is based on flow injection analysis (FIA) coupled with spectrophotometric detection. When direct injection of seawater was used, the detection limit was 1.6 nM, and the precision 7%, for a triplicate injection of a 4 nM standard. At low iron concentrations, on line preconcentration using a column filled with 8-hydroxyquinoline (8HQ) resin was used. The detection limit was 0.15 nM (time of preconcentration = 240 s), and the precision 6%, for a triplicate determination of a I nM standard, allowing the determination of Fe in most of the oceanic regimes, except the most depleted surface waters. The effect of temperature, pressure, salinity, copper, manganese, and iron speciation on the response of the analyser was investigated. The slope of the calibration curves followed a linear relation as a function of pressure (C-p = 2.8 x 10(-5)P + 3.4 x 10(-2) s nmol(-1), R-2 = 0.997, for Theta = 13 degrees C) and an exponential relation as a function of temperature (C-Theta = 0.009e(0.1030), R-2 = 0.832, for P = 3 bar). No statistical difference at 95% confidence level was observed for samples of different salinities (S = 0, 20, 35). Only very high concentration of copper (1000 x [Fe]) produced a detectable interference. The chemical analyser was deployed in the coastal environment of the Bay of Brest to investigate the effect of iron speciation on the response of the analyser. Direct injection was used and seawater samples were acidified on line for 80 s. Dissolved iron (DFe, filtered seawater (0.4 mu m), acidified and stored at pH 1.8) corresponded to 29 +/- 4% of Fe-a (unfiltered seawater, acidified in line at pH 1.8 for 80 s). Most of Fea (71 4%) was probably a fraction of total dissolvable iron (TDFe, unfiltered seawater, acidified and stored at pH 1.8).