High-frequency measurement of seawater chemistry: Flow-injection analysis of macronutrients

We adapted a commercially available flow-injection autoanalyzer (Lachat Quik-Chem 8000) to measure seawater nitrate concentrations at a rate of nearly 0.1 Hz and phosphate and silicate concentrations at a rate half that. Several minor improvements, including reduced sample-loop size, high sample flushing rate, modified carrier chemistry, and use of peak height rather than peak area as a proxy for nutrient concentration aided in the increase in sampling rate. The most significant improvement, however, was the construction of a copperized cadmium NO3 – reduction column that had a high surface area to volume ratio and a stable packing geometry. Preliminary results from a cruise in the Ross Sea in austral spring of 1997 are shown. Precision of all three analyses is better than 1%. Comparison of the nutrient concentrations determined by the rapid analysis method described here with traditional discrete analyses shows that nitrate and silicate determined by the two approaches are within a few percent of each other, but that the phosphate concentrations determined by the rapid analysis are as much as 10% lower than those determined by the discrete analyses. *Phone: (541) 737-8121. Fax: (541) 737-2064. E-mail: bhales@coas. oregonstate.edu Acknowledgments Thanks to the Antarctic Support Association and the captain and crew of the RVIB NB Palmer for support in the field. Thanks also to W. O. Smith and R. F. Anderson for their direction of the Joint Global Ocean Flux Survey Antarctic Environment and Southern Ocean Process Study program. The analysis of check samples and working standards by L. Gordon and J. Jennings is greatly appreciated. This article’s quality has been greatly improved by the thoughtful and thorough comments of four anonymous reviewers. B. Hales was supported in this work by a Department of Energy Global Change postdoctoral fellowship. Ship time was paid by NSF grant OPP-9350684. Limnol. Oceanogr.: Methods 2, 2004, 91–101 © 2004, by the American Society of Limnology and Oceanography, Inc. LIMNOLOGY and OCEANOGRAPHY: METHODS high spatial resolution. Until now, most chemical measurements have relied on slow wet-chemical analyses of coarsely spaced discrete samples subsequent to sampling. As a first step to changing this approach, we recently modified a SeaSoar towed undulating vehicle to carry a high-pressure, positivedisplacement pump that delivers seawater to a shipboard laboratory via a tube embedded in the towing cable (the Lamont Pumping SeaSoar; Hales and Takahashi 2002), following the pumping sampling system of Friederich and others (Friederich and Codispoti 1987; Codispoti et al. 1991). We demonstrated that the water-column location of samples taken from the seawater stream can be determined to within less than 0.5 m vertically, and that structure with vertical scale of about 1 m can be resolved in the sample stream. Taking advantage of the information present in a sample stream such as that supplied by the LPS requires an increase in the frequency at which nutrient concentrations can be measured: from one measurement every few min (≈ 0.01 Hz), which is typical of standard flow injection or continuous flow analyses, to one measurement every few seconds (≈ 0.1 Hz). This paper describes such an increase in sampling rate and presents preliminary results obtained with the system in the Ross Sea Polynya in November 1997. Analytical background—The heart of our analytical system was a Lachat QuikChem 8000 FIA autoanalyzer system (http:// www.lachatinstruments.com) configured to analyze nitrate, phosphate, and silicate concentrations in discrete samples that we had in our laboratory. Briefly, seawater samples are introduced to the system via a sample loop on an injection valve. The sample flows through the sample loop when the valve is in the load position; when the valve is switched to the inject position, the sample loop is flushed by a distilled water carrier stream that sweeps the sample into the chemistry manifold. There, combination with a suite of reagents results in a colored complex whose concentration is directly proportional to the nutrient of interest. Ultimately, the concentration of the colored complex is determined via its absorbance with an optical detection system and related to the nutrient concentration through calibration with standards of known concentration. The standard approach with this system is to inject large (1 to 2 mL) sample loops, allow full return to the carrier baseline, and numerically integrate between successive baselines to determine peak area. The maximum sampling frequency attainable is < 0.01 Hz (one sample every few minutes). The Lachat chemical analyses are all essentially FIA modifications of classical wet-chemical colorimetric analyses described in Grasshoff et al. (1983). Nitrate analysis follows a modification of this method by Johnson and Petty (1983); briefly, samples are injected into a carrier stream through an injection valve and then buffered to pH ≥ 8 with an imidazole buffer rather than the ammonium chloride used previously (Grasshoff 1983; Johnson and Petty 1983). Nitrate in the sample is reduced to nitrite in a column packed with copperized cadmium granules. An acidic solution combining sulfanilamide and N-(1naphthyl) ethylenediamine dihydrochloride is then added to the carrier/sample stream. Nitrite reacts with sulfanilamide to form a variety of diazonium ion complexes, which subsequently react with N-(1-naphthyl) ethylenediamine dihydrochloride to form colored compounds with an absorbance maximum at a wavelength of 520 nm. The absorption at this wavelength is directly proportional to the concentration of nitrate (and any nitrite that may have been originally present) in the seawater sample, up to concentrations of 40 mM. Phosphate analysis is an FIA implementation of the method of Koroleff (1983a). Briefly, phosphate is reacted at pH ≤ 1 with an acidic solution containing ammonium molybdate and potassium antimony tartrate to generate a phospho(antimony:molybdate) complex. This complex is then reduced with a solution of ascorbic acid to form a blue-colored complex that has an absorbance maximum at 880 nm. Absorption at this wavelength is directly proportional to the phosphate in the sample up to concentrations of 2 mM. Silicate analysis is an FIA implementation of the method described in Koroleff (1983b). First a solution of oxalic acid is combined with the sample stream to suppress any interference from phosphate. Then the sample is reacted with an acidic ammonium molybdate at a pH ≈ 3 to form a silicomolybdate complex. This is then reduced with a stannous chloride solution to form a blue-colored complex that has an absorbance maximum at 820 nm. Absorption at this wavelength is directly proportional to the silicate in the sample up to concentrations of 100 mM. Sampling frequency limitations—Two factors ultimately limit the frequency at which nutrient samples can be analyzed in a FIA system: (1) the ratio of sample volume to carrier flow rate or the sample flow resolution, τample, and (2) axial smearing of the sample as it transits through the reaction manifold, τsmear. The effect of these two can be illustrated by envisioning the idealized square-wave pattern that immediately follows discrete injections of a sample into a carrier stream (Fig. 1). Clearly, samples cannot be analyzed more frequently than once every τsample. For typical FIA flow rates and tube diameters, flow is laminar (Re ≈ 200; turbulent flow begins at Re ≥ 2100). This results in a parabolic velocity distribution of the fluids in the reaction manifold tubing with a maximum at the center of the tube that is twice the mean velocity (Fig. 2A; see also, e.g., Bird et al. 1960), and a zero velocity at the walls. Such a velocity profile pushes the sample at the center of the tube forward faster than that nearer the tube walls, resulting in a smearing of the original square-wave input signal. The longer the sample is subjected to such a velocity distribution, the more smeared it will become, and ultimately the amplitude itself will be decreased (Fig. 2b). The ratio of maximum flow to mean flow requires that the first part of the signal that reaches the detector will get there in about half the time it takes for the mean, which, in turn, means that τsmear will be approximately equal to the total time that a sample resides in the laminar flow environment between injection and detection. In other words, τsmear ≈ τflow where τflow, given by the total Hales et al. Rapid nutrient analysis