Chapter 6. Determination of Dissolved Oxygen by the Winkler Procedure 
1.0 Scope and field of application
This procedure describes a method for the determination of dissolved oxygen in seawater, expressed as mmol kg -1 . The method is suitable for the assay of oceanic levels, e.g. 0.5 to 350 mmol kg -1 of oxygen in uncontaminated seawater and is based on the Carpenter (1965) modification of the traditional Winkler titration. As described it is somewhat specific to an automated titration system. A manual titration method is also described. There are currently alternative methods of assessing the endpoint (e.g., potentiometric) that give comparable precision, but these are not described here. This method is unsuitable for seawater containing hydrogen sulfide.

2.0 Definition
The dissolved oxygen concentration of seawater is defined as the number of micromoles of dioxygen gas (O2 ) per kilogram of seawater (mmol kg -1 ).

3.0 Principle of Analysis
The chemical determination of oxygen concentrations in seawater is based on the method first proposed by Winkler (1888) and modified by Strickland and Parsons (1968). The basis of the method is that the oxygen in the seawater sample is made to oxidize iodine ion to iodine quantitatively; the amount of iodine generated is determined by titration with a standard thiosulfate solution. The endpoint is determined either by the absorption of ultraviolet light by the tri-iodide ion in the automated method, or using a starch indicator as a visual indicator in the manual method. The amount of oxygen can then be computed from the titer: one mole of O2 reacts with four moles of thiosulfate.

More specifically, dissolved oxygen is chemically bound to Mn(II)OH in a strongly alkaline medium which results in a brown precipitate, manganic hydroxide (MnO(OH)2).

After complete fixation of oxygen and precipitation of the mixed manganese (II) and (III)hydroxides, the sample is acidified to a pH between 2.5 and 1.0. This causes the precipitated hydroxides to dissolve, liberating the Mn(III) ions. The Mn(III) ions oxidize previously added iodide ions to iodine. Iodine forms a complex with surplus iodide ions. The complex formation is desirable because of its low vapor pressure, yet it decomposes rapidly when iodine is removed from the system. The iodine is then titrated with thiosulfate; iodine is reduced to iodide and the thiosulfate is oxidized to tetrathionate. The stoichiometric equations for the reaction described above are:

The thiosulfate can change its composition and therefore must be standardized with a primary standard, typically potassium iodate. Standardization is based on the co-proportionation reaction of iodide with iodate, thereby forming iodine. As described above, the iodine binds with excess iodide, and the complex is titrated with thiosulfate. One mole of iodate produces three moles iodine, and amount consumed by six moles of
thiosulfate.
  4.0 Apparatus

4.1 Sampling apparatus

4.1.1 Sample flasks: custom made BOD flasks of 115 ml nominal capacity with ground glass stoppers. The precise volume of each stopper-flask pair is deter-mined gravimetrically by weighing with water. It is essential that each indi-vidual flask/stopper pair be marked to identify them and that they be kept together for subsequent use.

4.1.2 Pickling reagent dispensers: two dispensers capable of dispensing 1 ml ali-quots of the pickling reagents. The accuracy of these dispensers should be 1% (i.e. 10 ml).

4.1.3 Tygon Ò tubing: long enough to reach from spigot to the bottom of the sample bottle.

4.1.4 Thermometers: one thermometer is used to measure the water temperature at sampling to within 0.5°C. Two platinum resistance temperature sensors are used to monitor the temperatures of the titrating solutions in the laboratory.

4.2 Manual titration apparatus

4.2.1 Titration box: a three-sided box containing the titration apparatus. The walls should be painted white to aid in end point detection.

4.2.2 Dispenser: capable of delivering 1 ml aliquots of the sulfuric acid solution.

4.2.3 Burette: a piston burette capable of dispensing 1 ml and 10 ml of KIO 3 for blank determination and thiosulfate standardization. An alternate, precisely calibrated dispenser may be used for these steps.

4.2.4 Magnetic stirrer and stir bars.

4.2.5 Burette: a piston burette with a one milliliter capacity and anti diffusion tip for dispensing thiosulfate.

4.3 Automated titration apparatus

4.3.1 Metrohm 655 Dosimat burette: a piston burette capable of dispensing 1 to 10 ml of KIO3 for blank determination and standardization.

4.3.2 Metrohm 665 Dosimat Oxygen Auto-titrator. The apparatus used for this technique consists of a thiosulfate delivery system (the Dosimat) and a detector that measures UV transmission through the sample in a custom designed BOD bottle.

4.3.3 AST computer. The burette, endpoint detector and A/D convertor are controlled by an IBM compatible PC, in a system designed by R. Williams (SIO).

4.3.4 Dispenser: capable of delivering 1 ml aliquots of the sulfuric acid solution.

4.3.5 Magnetic stirrer and stir bars.

5.0 Reagents
5.1 Manganese (II) chloride (3M: reagent grade): Dissolve 600 g of MnCl2 *4H2O in 600 ml distilled water. After complete dissolution, make the solution up to a final volume of 1 liter with distilled water and then filtered into an amber plastic bottle for storage.

5.2 Sodium Iodide (4M: reagent grade) and sodium hydroxide (8M: reagent grade): Dissolve 600 g of NaI in 600 ml of distilled water. If the color of solution becomes yellowish brown, discard and repeat preparation with fresh reagent. While cooling themixture, add 320 g of NaOH to the solution, and make up the volume to 1 liter with distilled water. The solution is then filtered and stored in an amber glass bottle.

5.3 Sulfuric Acid (50% v/v): Slowly add 500 ml of reagent grade concentrated H2SO4 to500 ml of distilled water. Cool the mixture during addition of acid.

5.4 Starch Indicator (manual titration only): Place 1.0 g of soluble starch in a 100 ml beaker, and add a little distilled water to make a thick paste. Pour this paste into 1000 ml of boiling distilled water and stir for 1 minute. The indicator is freshly prepared for each cruise and stored in a refrigerator until use.

5.5 Sodium Thiosulfate (0.18 M: reagent grade): Dissolve 45 g of Na2S2O3 * 5H2O and 2.5 g of sodium borate, Na2B4O7 (reagent grade) for a preservative, in 1 liter of distilled water. This solution is stored in a refrigerator for titrator use.

5.6 Potassium Iodate Standard (0.00167M: analytical grade): Dry the reagent in a desiccator under vacuum. Weigh out exactly 0.3567 g of KIO3 and make up to 1.0 liter with distilled water. Commercially prepared standards can also be used. One ampule of Baker’s DILUT-IT KIO3 analytical concentrate solution is diluted 1:10 to create a 0.0167M stock solution. This solution is diluted 1:10 for titration use, 0.00167M. It is important to note the temperature of the solution so that a precise molarity can be calculated.

6.0 Sampling
6.1 Collection of water at sea, from the Niskin bottle or other sampler, must be done soon after opening the Niskin, preferably before any other samples have been drawn. This is necessary to minimize exchange of oxygen with the head space in the Niskin which typically results in contamination by atmospheric oxygen.

6.2 Sampling procedure:

6.2.1 Before the oxygen sample is drawn the spigot on the sampling bottle is opened while keeping the breather valve closed. If no water flows from the spigot it is unlikely that the bottle has leaked. If water does leak from the bottle it is likely that the Niskin has been contaminated with water from shallower depths. The sample therefore may be contaminated, and this should be noted on the cast sheet.

6.2.2 The oxygen samples are drawn into the individually numbered BOD bottles. It is imperative that the bottle and stopper are a matched pair. Two samples are drawn from each Niskin and the order of sampling is recorded.

6.2.3 When obtaining the water sample, great care is taken to avoid introducing air bubbles into the sample. A 30–50 cm length of Tygon  tubing is connected to the Niskin bottle spout. The end of the tube is elevated before the spout is opened to prevent the trapping of bubbles in the tube. With the water flowing, the tube is placed in the bottom of the horizontally held BOD bottle in order to rinse the sides of the flask and the stopper. The bottle is turned upright and the side of the bottle tapped to ensure that no air bubbles adhere to the bottle
walls. Four-five volumes of water are allowed to overflow from the bottle. The tube is then slowly withdrawn from the bottle while water is still flow-ing.

6.2.4 Immediately after obtaining the seawater sample, the following reagents are introduced into the filled BOD bottles by submerging the tip of a pipette or automatic dispenser well into the sample: 1 ml of manganous chloride, followed by 1 ml of sodium iodide-sodium hydroxide solution.

6.2.5 The stopper is carefully placed in the bottle ensuring that no bubbles are trapped inside. The bottle is vigorously shaken, then reshaken roughly 20 minutes later when the precipitate has settled to the bottom of the bottle.

6.2.6 After the second oxygen sample is drawn, the temperature of the water from each Niskin is measured and recorded.

6.2.7 Sample bottles are stored upright in a cool, dark location and the necks water sealed with saltwater. These samples are analysed after a period of at least 6-8 hours but within 24 hours. The samples are stable at this stage.

7.0 Titration Procedures
The basic steps in titrating oxygen samples differ little regardless of whether one uses the manual or the automated procedure. First the precise concentration of the thiosulfate must be determined. Next the blank, impurities in the reagents which participate in the series of oxidation-reduction reactions involved in the analysis, is calculated. Once the standard titer and blank have been determined, the samples can be titrated.
The fundamental differences between the manual and automated titration methods are the means of endpoint detection (visual versus a UV detector) and the method of thiosulfate delivery. The auto-titrator rapidly dispenses thiosulfate. As the changes in UV absorption are noted, the rate is slowed, and finally the continuous addition is stopped. The endpoint is approached by adding ever-smaller increments of thiosulfate until no further change in absorption is detected, indicating that the endpoint has been passed. Standardization, blank determination, and sample analysis are described generically below for both
methods, with specifics where warranted.

7.1 Standardization:

7.1.1 To one BOD bottle add approximately 15 ml of deionized water and a stir bar.

7.1.2 Carefully add 10 ml of standard potassium iodate (0.00167 M) from an “A” grade pipette or equivalent or the Metrohm 655 Dosimat. Swirl to mix.Immediately add 1 ml of the 50% sulfuric acid solution. Rinse down sides of flask, swirling to mix, thus ensuring an acidic solution before the addition of
reagents.

7.1.3 Add 1 ml of sodium iodide-sodium hydroxide reagent, swirl, then add 1 ml of manganese chloride reagent. Mix thoroughly after each addition. Once solution has been mixed, fill to the neck with deionized water.

7.1.4 Titrate the liberated iodine with thiosulfate immediately. In the manual method, use the 1 ml burette to titrate the standard with sodium thiosulfate (approximately 0.18 M) until the yellow color has almost disappeared. Add 1–2 ml of the starch indicator, which should turn the solution deep blue to purple in color. Titrate until this solution is just colorless and then record room temperature. This titration should be reproducible to within ± 0.03 ml, once the varying BOD bottle volumes have been accounted for.

7.1.5 The automated titrator system delivers 0.2 N thiosulfate to the acidified stan-dard solution and reads the change in UV light absorption in the solution. As the endpoint is approached, it delivers progressively smaller aliquots of thio-sulfate until no further change in absorption shows that the endpoint has been
reached.The endpoint is determined by a least squares linear fit using a group of data points just prior to the endpoint, where the slope of the titration curve is steep, and a group of points after the endpoint, where the slope of the curve is close to zero. The intersection of the two lines of best fit is taken as the endpoint. Reproducibility should be better than 0.01 ml l -1 .

7.1.6 The mean value should be found from at least three and preferably five replicate standards, and standards should be run at the beginning, end, and periodically throughout the time that samples are being titrated.

7.2 Blank determination:

7.2.1 Place approximately 15 ml of deionized water in a BOD bottle with a stir bar. Add 1 ml of the potassium iodate standard, mix thoroughly, then add 1 ml of 50% sulfuric acid, again mixing the solution thoroughly.

7.2.2 Before beginning the titration add the reagents in reverse order: 1 ml of sodium iodide-sodium hydroxide reagent, rinse, mix, then 1 ml of manga-nese chloride reagent. Fill the BOD bottle to just below the neck with deion-ized water. Titrate to the endpoint as described for the standardization procedure.

7.2.3 Pipette a second 1 ml of the standard into the same solution and again titrate to the end point.

7.2.4 The difference between the first and second titration is the reagent blank. Either positive or negative blanks may be found.

7.3 Sample analysis:

7.3.1 After the precipitate has settled (at least 6-8 hours for the automated method), carefully remove the sealing water taking care to minimize distur-bance of the precipitate. Wipe the top of the flask to remove any remaining moisture and carefully remove the stopper.

7.3.2 Immediately add 1 ml of 50% sulfuric acid. Carefully slide a stir bar down the edge of the bottle so as not to disturb the precipitate.

7.3.3 Titrate as described in the standardization procedure.

8.0 Calculation and expression of results
The calculation of oxygen concentration (mmol l-1 ) from this analysis follows in principle the procedure outlined by Carpenter (1965).

R  = Sample titration (ml) R Std = Volume used to titrate standard (ml)
R b/k = Blank as measured above (ml) M IO3 =Molarity of standard KIO3 (mol/l)
VIO3 = Volume of KIO3 standard (ml) E = 5,598 ml O2 /equivalent
Vb = Volume of sample bottle (ml) DOreg = oxygen added in reagents
Vreg = Volume of reagents (2 ml)

8.1 The additional correction for DOreg of 0.0017 ml oxygen added in 1 ml manganese chloride and 1 ml of alkaline iodide has been suggested by Murray, Riley and Wilson (1968).
O 2 ml/l ( )

8.2 Conversion to mmol/kg: To make an accurate conversion to mmoles/kg, two correc-tions are needed: (1) to correct for the actual amount of thiosulfate delivered by the burette (which is temperature dependent); and (2) to correct for the volume of the sample at its drawing temperature. Both calculations are undertaken automatically in many versions of software driven titration. Two pieces of information are required:
(a) the temperature of the sample (and bottle) at the time of fixing; the reasonable assumption being that the two are the same; (b) the temperature of the thiosulfate at the time of dispensing. Some versions of the automatic titration may also call for in situ temperature, as well as salinity, which allow for the calculation of oxygen solu-bility and thus the percentage saturation and AOU.

9.0 Quality assurance
9.1 Quality Control: For best results, oxygen samples should be collected in duplicate from all sample bottles. This allows for a real measure of the precision of the analy-sis on every profile. A mean squared difference (equivalent to a standard deviation of repeated sampling) is the measure of precision for these profiles. As this replication takes into account all sources of variability (e.g. sampling, storage, analysis) it gives
a slightly larger imprecision than indicated by the analytical precision of the titration (e.g. repeated measures of standards in the lab). In addition, periodic precision tests are done by collection and analysis of 5–10 samples from the same Niskin bottle. This precision should be better than 0.01 ml l -1. Field precision can vary from 0.005 to 0.03 depending on the sea conditions and the performance of the auto-titrator.
Samples are reduced to oxygen concentrations prior to the next cruise to identify degradation of the precision, before too many additional profiles have been collected.

9.2 Quality assessment: No absolute standard exists for oxygen analysis. Standards are made by gravimetric and volumetric measurements of reagent grade chemicals. Commercially prepared standards such as DILUT-IT can be used for comparison with the freshly made up standard in the lab. Standard solutions are relatively stable and provide an early warning of errors by changes in their titer. Profiles of oxygen are examined visually and numerically. At any depth where the replicates differ by 0.04 ml/l or greater, the samples are carefully scrutinized. The profile is compared with the historical profiles for consistency, particularly in the deep water. These profiles are also compared with the CTD oxygen sensor. Although CTD oxygen sensors are very imprecise and inaccurate, they provide a continuous record. Deviations from the general shape of the profile by a single oxygen sample is evidence of inaccuracy in the wet oxygen measurement.

10.0 References
Carpenter, J.H. (1965). The Chesapeake Bay Institute. Technique for the Winkler oxygen method. Limnol. Oceanogr., 10, 141–143.
Grasshoff, K. Ehrhardt, M, and K. Kremling (1983). Methods of Seawater Analysis.
Grasshoff, Ehrhardt and Kremling, eds. Verlag Chemie GmbH. 419 pp.
Murray J.N., Riley, J.P. and Wilson, T.R.S. (1968). The solubility of oxygen in Winkler reagents used for the determination of dissolved oxygen. Deep-Sea Res., 15, 237–238.
Strickland, J.D.H., and Parsons, T.R. (1968). Determination of dissolved oxygen. in A Practical Handbook of Seawater Analysis. Fisheries Research Board of Canada,Bulletin, 167, 71–75.
Williams, P.J.leB., and Jenkinson, N.W. (1982). A transportable microprocessor-controlled precise Winkler titration suitable for field station and shipboard use. Limnol. Oceanogr., 27 (3), 576–584.
Winkler, L.W. (1888). Die Bestimmung des in Wasser gelösten Sauerstoffen. Berichte der Deutschen Chemischen Gesellschaft, 21: 2843–2855.