- Email: [email protected]
An interlaboratory study of dissolved oxygen in water
½1net Reveurch Vol 15. pp 32t to 325 (~ Pergamon Press Lid t081 Printed ii1 Grcat ~ l U l l n
00,13-13M~1/030321-05502.00/0
AN INTERLABORATORY STUDY O F DISSOLVED OXYGEN IN WATER R. J. WILCOCK,C. D. S~VENSO~ and C. A. ROBERTS Chemistry Division, Department of Scientific and Industrial Research. Private Bag. Peton¢, Ne~ Zealand
(Recewed January 1980) AIwtract--Dissolved oxygen concentrations in water samples free from chemical reductants may be stabilised by the addition of mercuric chloride (40 $ m-3} and storage in gas-tight bottles. This preservation technique has been used in an interlaboratory study of dissolved oxygen analysis in New Zealand laboratories. At reference coneentrations of 1.20 and 5.86 g m - 3 there was a significant positive bias in results reported for both the Winkler method (0.24 and 0.22 gm -3 respectively) and the membrane electrode method (0.59 and 0.62 g m-3 respectively). Inadequate precautions to avoid sample aeration during handling and analysis probably caused the bias.
INTRODUCTION
For a number of years this laboratory has organised a nation-wide programme of collaborative water analysis; the C H E M A Q U A programme (Kingsford et al., }973; Stevenson et al., 1975 ; Timperley, 1978). along similar lines to those undertaken in other countries (Wales & Traversy, 1972; Ekedahl et al., 1975; Simpson, 1978). The programme is designed to evaluate analytical procedures, assess interlaboratory variability of results, and create an awareness among the participating organisations of the need for regular assessment of analytical performance. Previous rounds have been concerned with major ions (Kingsford et al., 1973), nutrient species (Stevenson et al., 1975) and trace metals (T~mperley, 1978), at levels likely to be of significance in natural, potable and waste waters. We are not aware of any interlaboratory study of dissolved oxygen analysis in water. Such studies appear to have been hindered by the lack of an effective preservation procedure to ensure that all laboratories received samples containing the same concentration of dissolved oxygen. We have found that concentrations of this constituent can be stabilised by addition of mercuric chloride, to eliminate biological activity, and storage in gas-tight glass containers. This paper reports the successful application of this preservation procedure in an interlaboratory study. Dissolved oxygen fDO) is an important physico-chemical parameter of water quality, and has special significance for aquatic organisms in natural waters. It is crucial therefore that its measurement he both rapid and reliable. Of the many ways of determining DO. by far the most commonly used are the Winkler method and its modifications (Standard Methods. 1976). and the various membrane-electrode systems. Because results by these different methods are frequently compared with each other it is often of great interest to know the extent to which this is justified, or if one method is clearly superior to the other. 321
This study involved 45 laboratories, ranging from large research-oriented organisations engaged in regular analysis of water samples, to small "one-person" laboratories, not occupied full-time with water analysis. Each laboratory was sent two 2.25 1. samples, one containing a low level of D O (less than 2 g i n -3) ( C H E M A Q U A 11) and the other approx. 5 g i n -3 ( C H E M A Q U A 12) comparable with the level recommended by some authorities as the minimum concentration required to maintain good fish populations in natural waters (U.S. EPA, 1976). All participants in the study were issued with instructions on how to siphon the test solutions out of the containers, to minimize sample aeration. METHODS
Sample stability trials Before sending samples to the participating laboratories, trials were carried out to ensure that samples remained unaltered during their despatch to the various destinations. All samples were preserved from microbial activity by the addition of mercuric chloride (40gin-3). and the main concern of the trials was minimizing the physical exchange of oxygen between the samples and their surroundings. One of the participant laboratories expressed concern that mercuric chloride would interfere with the Winkler method. This was checked by taking 41. aliquots of distilled water having different DO contents, splitting them into 2 I. sub-samples, and preserving one of these with mercuric chloride. All of the samples were analysed for DO b~ the az~de modification of the Winkler method (Standard Methods. 19761 and results for the preserved and unpreserved pairs compared. No difference was observed between the samples over the range 1.0--7.5 g m- ~ (reproducibility within _+ 0.02gm-~). The water used for the study contained no chcmicall:r interfering substances. All samples in the preservation and stability trials were analysed by the azide modification of the Winkler method. The stability trials consisted of filling various types of container with water of known DO, and analysing them daily for a period of about 2 weeks. Unfilled polyethylene containers (approx. vol. 2.25 I./ were unsatisfactory because they permitted diffusion of
322
R J. WtLCOCK
ox,,gen through the container walls. Samples m these containers, at an initial D O of 1.0 g m - ~. had a steady rate of increase in D O concentration of 0.35 g m - ~ d a y - t. for a period of 10 days. Glass reagent bottles, while not permitting oxygen exchange through the container walls, tended to leak when they were inverted or placed on their sides. Lining the threaded bottle tops with polyethylene sheet prevented water loss from the bottles, but did not restrict gaseous exchange, and increases in DO of up to 0.90 g m - 3 were recorded for samples over 24 h. Glass (Winchester type) reagent bottles fitted with rubbet O-rings proved to be both gas-tight and leak-proof. Samples with D O concentrations in the range 1.8-9.5 g m - 3. analysed over a period of 12 days. had coefficients of variation less than 1°;,. This method was consequently adopted for the main trial. Preparation ol sample.~ /or collaborattre exerct.se The local tapwater, of artesian origin, provided a convenient supply of low D O content water for C H E M A Q U A I I Aliquots were collected by running tap water into the bottom of each bottle and displacing 2 vol before retaining sample water. The constancy of the D O concentration throughout the filling process was followed by analysing the first bottle filled, and thereafter every fifth bottle. These samples had a D O content of 1.00 +_ 0.02 g m Samples for C H E M A Q U A 12 were produced b~ bubbling compressed air through tap water in a large plastic tank. The solution was effectively stirred by the bubble action, and D O was monitored continuously with a D O electrode system". Samples were collected by siphoning partially aerated water into clean Winchester bottles, and displacing about 2 vol. before retaining the water. DO analyses were carried out on the first bottle, and on ever.,, tenth sample collected: g~vtng a D O of 5.76 _+ 0.08 g m - ' After each bottle was filled and labelled (CHEMAQL-X II or C H E M A Q U A 121. 4 m l of sample was withdrav, n with a syringe and 2 ml of 5°o mercuric chloride solution added. The bottles were fitted with rubber O-rings and tops. and then inverted several times to ensure thorough mixing of the preservative, and equilibrium between the sample waters and the 2 ml air bubbles Icreated to compensate for possible thermal expansion). The small gas space above each sample led to a slight increase m the sample DO levels. The magnitude of this change was estimated by assuming ideal behaviour of the gas phase, a gas-tight seal (i.e. a closed system .tnd a final equilibrium position described by Henry's law. This yielded C. = C, - I P t V s M . I O ~ VtRT) I 4-1KHV=M.103 V~RT) where C~. C., are respectivel.~, the retrial and final DO concentratzons, in g m - 3. P~ is the initial partial pressure of oxygen m the a~r space (aim'f) KH is the Henry'l lay, constant, m arm g " ~• m V). V~ are the gas volume and liquid volume respectively,
of C H E M A Q U A 1! and C H E M A Q U A 12. for 10 days after the date of despatch of the mare body of samples to the participating laboratories. The adjusted values were 1.20 ~ _ 0 . 1 0 g m -3 for C H E M A Q U A II and 5.86 -,- 0 . 1 2 g m "
for C H E M A Q U A 12
These were defined as the reference concentrations. (N.B. The 2 ml air space in each sample bottle was calculated to allow for a thermal expansion caused by a rise in temperature of 10:C. For this system in equilibrium at 20~C, the oxygen distribution between the gas and liquid phases is only affected slightly by temperature variations. The maxim u m change in DO, for a ~ 10:C temperature change, was estimated to be __. 0.05 8 m - ~ for C H E M A Q U A 12. and considerably less for C H E M A Q U A 11.) RESULTS All 45 p a r t i c i p a n t o r g a n i s a t i o n s r e p o r t e d their results as h a v i n g been o b t a i n e d either by t h e W i n k l e r m e t h o d (and p r e d o m i n a n t l y by the azide m o d i f i c a tion), or by m e m b r a n e - e l e c t r o d e s . A few r e p o r t e d results o b t a i n e d by b o t h m e t h o d s . T h e results are listed in T a b l e s I a n d 2. a n d are p r e s e n t e d g r a p h i c a l l y in Figs l a n d 2. R e s u l t s were a v e r a g e d w h e n m o r e t h a n o n e e s t i m a t i o n was s u b m i t t e d for e a c h s a m p l e . a n d all d a t a were r e c o r d e d to the n u m b e r of signific a n t figures r e p o r t e d by e a c h l a b o r a t o r y . Statistical treannent o l d a t a T h e results were i n s p e c t e d for o u t l y i n g values, a c c o r d i n g to t h e A S T M p r o c e d u r e ( A S T M , 1971), before c a l c u l a t i n g m e a n s a n d s t a n d a r d d e v i a t i o n s for the nett d a t a sets (Tables 1 a n d 2). R e s u l t s o b t a i n e d by e l e c t r o d e a n d W i n k l e r m e t h o d s were c o m p a r e d for degree of s c a t t e r (variance) by the F-test. a n d for significant differences b e t w e e n their m e a n s by the t-test (Davies. 1947).
DISCLSSION T h e results for b o t h C H E M A Q U A I I a n d 12 are s h o w n in Figs 1 a n d 2. It is e v i d e n t f r o m this t h a t a l t h o u g h m a n y of the results are in g o o d a g r e e m e n t with the reference values ( 1.20 g m - 3 a n d 5.86 g m - 3). t h e r e is an overall positive bias for b o t h m e t h o d s , at
,:]) M e t e r OWlnkler Reference value
(l.)
M ~s the molecular weight of oxygen R. T are the gas constant (arm.tool- + K - ~) and absolute temperature, respectivelyUsing pubhshed KH values tWilhelm ,,t -i.. 1977). these corrections were 0.20 _ 0.01 g m - ~ for C H E M A Q U A II and 0 . 1 0 _ 0 . 0 3 g i n -~ for C H E M A Q U A t2. These mcreases were verified experimentall> by analysing samples
0 " Yellow Sprmgs Instrument C o m p a n y Model 57 Meter and YSI 5739 probe. + I a t m = 101.325 kPa.
;} 2 3 Dissolved oxygen (g/rn3) Fig. ! Results for C H E M A Q U & I I
1.6 1.4 1.90 2.30 1.70 1.3 I 2.2 1.45 1.3 1.7 1.9 2.75 1.78 1.8 5.6* 6.4 6.30 6.1 6.3 6.25 5.9 7.08f 6.4 6.0 6.0 6.54 6.15 5.44 5.g6 6.10 6.40 6.0 6.4 6.18 6.25 6.5 4.64f 6.35 5.9 6.25* 7.84"I" 5.9 6.0 6.2* 5.75 6.32 5.87 5.9 5.91 5,75
(g m " J)
Winkler method
* Averaged result. t Outlier.
Ol 02 03 04 05 06 07 08 II 12 14 I5 16 17 18 19 20 22 24 26 27 28 29 30 31 32 33 34 35 36 37 4O 41 42 43 45
Laboratory code number
* Averaged result. t Outlier.
Reference value = 1.20
Range = 1.3-2.75
M e a n = 1.79 Standard deviation = 0.39 Ic.v. 22%)
04 06 09 19 24 26 28 30 31 36 37 38 39 42
Electrode (g m - s)
Mean = 6.10 Standard deviation = 0.27 (c.v. 4%) Range = 5,44-6,54
1.2" 1.6 1.40 1.3 1.2 1.35 1.4 2.8Ol' 2.0 1.95 I.lO 1.75 1.05 1.05 1.40 2.30 1.70 1.3 1.3 l.l I 1.35 2.2 2.40 1.40 1.0 1.3 1.69' 1.2 1.2 1.3 1.15 1.8 1.32 1.0 1.16 1.30
OI 02 03 04 05 06 07 08 II 12 14 15 16 17 18 19 20 22 24 26 27 28 29 30 31 32 33 34 35 36 37 40 41 42 43 45
Laboratory code n u m b e r 6.4 6.35 6.55 5.95 6.60 6.15 6.7 6.25 6.1 7.15 6.6 6.8 6.15 6.27 7.2
Electrode (g m - J)
Reference value = 5.g6
Range = 5.95 7.2
Mean = 6.48 Slandard deviation = 0.37 (c.v. 6%1
04 06 09 19 24 26 28 30 31 36 37 38 39 40 42
Laboratory code number
Table 2. Dissolved oxygen results for C H E M A Q U A 12.
Mean = 1.44 Slandard deviation = 0.37 (c.v. 26%) Range = I.O-2.40
(g m - 3)
Laboralory code numbers
Winkler method
Table I. Dissolved oxygen results for C H E M A Q U A I I
o
e,,
m
e~
o
~n
m o" O
,,,i
>
324
R.J. WILCOCKet ui.
Meter OWinkler
Z.
5
¢6
7
8
Dissolved oxygen (g/m3) Fig. 2. Results for C H E M A Q U A 12.
the two DO levels. The "'offset" in the Winkler method values becomes less significant however, when the uncertainty of the reference values ti.e. + 0 . 1 0 g i n -J) is taken into account. There was no obvious correlation between laboratory results that differed widely from the mean values, and the delay between their despatch and subsequent analysis. It is apparent from Figs 1 and 2 that analyses carned out by membrane electrode generally gave higher results than those obtained by Winkler titration. This is further illustrated when the means for each method are compared, after outliers have been rejected (Tables 1 and 21. The mean value for the electrode results is higher than the Winkler mean by 0.35 g m - 3 for C H E M A Q U A t 1, and 0.37 g m - 3 for CHEMAQUA 12. Application of the r-test showed that these differences between the methods are significant and indicate a consistent bias for the electrode method of determining DO. A comparison by F-test of the variances for each method, showed they were not significantly different, at either DO level. Our experience in surveymg a large number of different DO meter systems has shown that in general they tend to give positive errors at low DO levels li.e. less than 2 g m-J). This is partly because of their generally poor response times to reach equilibrium at these levels, probably associated with slow depletion of the oxygen in the bulk internal electrolyte necessary to decrease the "blank" response to a negligible level. The stirring that is necessary to maintain flow across the membrane in order to overcome localised DO depletion by the electrode can also often introduce a positive error through sample aeration. The results for C H E M A Q U A 11 are in accord with these observations. The potential sources of error in the Winkler method are well documented IStandard Methods. 1976: Carpenter. 19651. and a precision of + 0 . 0 2 g i n -~ is realizable in most circumstances. As with the electrode method, the most probable source of positive error arises from sample aeration during the various manipulations. The determination of low DO levels is especially important in corrosion studies. and in studies of natural and waste waters having an appreciable oxygen demand. These results suggest
that the Winkler method ~s probably more reliable than the membrane electrode techniques, when the DO is less than 1-2 g m-3, and that care should be taken when comparing results at this level obtained by different workers. This is particularly important when different methods have been used. The disparity between the Winkler and electrode method means for C H E M A Q U A 12 was larger than expected, as there is usually very good agreement between these methods at moderate to high DO levels (Reynolds. 19691. Response times for membrane-.electrodes are typically of the order of 10-20s at 5-6 g m - 3 DO, and there is usually no difficulty in carrying out measurements in this range and achieving accurate results ( ~0.05 gin-3). It is possible that in some cases aeration errors were introduced when the samples were removed from their bottles, especially if the samples were poured rather than siphoned, into suitable containers. All participants were advised of the proper sample handling procedure to minimize aeration errors, and the large samples distributed (2.25 I.I should have provided adequate volume for flushing of analysis vessels. It is difficult to understand how such errors would account for the difference between the two sets of results, and it is more probable that an additional consistent error, such as aeration caused by over-vigorous stirring. was responsible for the higher electrode results. In spite of the difference between the means for C H E M A Q U A 12. the overall standard of results for both methods was quite satisfactory as judged by the total error criterion for analytical acceptability (McFarren et al.. 1970) which rates both methods "'excellent" at this DO level. However. the results cast some doubt on whether, in practice, it is possible to distinguish reliably between DO levels of 5 g m - ~ and 6 g m -3, as is required in some regulatory water classifications. CONCLUSIONS
The tnterlaboratory comparison of DO measurement, at two concentrations, showed that a reasonable level of performance was achieved in most cases. and that the between-laboratory variability was about 7-10 times the precision attainable by each method. Those results obtained by membrane-electrodes were on average significantly higher than those measured by the Winkler method. It was not possible to discern whether this positive bias in the electrode method was caused by an inherent failing of the method, or if analysts were consistently making the same kind of error. It is thought that the latter is the case, and that sample aeration either during the stirring or sample transfer stages was responsible. The variability among analysts" results suggests care should be taken when comparing results obtained by different workers, and that too much importance should not be attached to small differences in DO without close examination of the methods used.
An interlaboratory study of dissolved oxygen in water
The results of the survey suggest that at D O levels of 1-2 g m -a or less, the Winkler method is more reliable than most conventional membrane-electrodes. REFERENCES ASTM (1971) Recommended practice for dealing with oullying observations. ASTM Designation E178-68 Carpenter J. H. (1965) The accuracy of the Winkler method for dissolved oxygen analysis. Limnol. Oceano#. 10, 135-140. Davies O. L. (1947) Statistical Methods tn Research and Production, pp. 55-70. Oliver & Boyd, London. Ekedahl G.. Junker P+ & Rondell B. 0975) lnterlaboratory study of methods for chemical analysis of water. J. War. Pollut. Control Fed. 47. 858-866. Kingsford M.. Stevenson C. D. & Edgerley W. H L. (1973) Collaborative tests of water analysis (the CHEMAQUA programme~. I. Sodium, potassium, calcium, magnesium, chloride, sulphate, bicarbonate, carbonate and conductivii). N.Z. J/Sci. 16, 895-902. McFarren E. F.. Lishka R. J. & Parker J. H. (1970) Criterion for judging acceptability of analytical methods. Analyt. Chem. 42. 358-365. Reynold J. F. (1969) Comparison studies of Winkler vs
325
oxygen sensor. J. Wat. Pollut. Control Fed. 41. 2002-2009. Simpson E. A. (1978)The harmonisation of the monitoring of the quality of inland freshwater. J. Inst. War. Engng Sciem. 32. 45-56. Standard Methods for the Exanunation of Water and Wastewater (1976) 14th Edn. Americap Public Health Association, New York, NY. Stevenson C. D.. Kingsford M. & Edgerley W. H, L. (1975) Collaborative tests of water analysis (the CHEMAQUA programme) 2, Nitrate nitrogen, reactive dissolved phosphorus, and total dissolved phosphorus. N.Z. JI ScL 16, 895-902. Timperley M. H. (1978) Collaborative tests of water analysis (the CHEMAQUA programme) 3. Trace metals, N.Z. JI Sci. 21. 557-564. U.S. EPA (1976) Quality Criteria for Water. Office of Water and Hazardous Materials, Washington, DC. Wales R. W. & Traversy W. J. (19721 lnterlaboratory quality control study number 2. Total phosphate organic nitrogen, nitrate nitrogen, and organic carbon. Inland Waters Branch. Dept. of Envim. Report Series 119. Ottawa. Canada. Wilhelm E., Battino R. & Wilcock R. J. 0977) Low pressure solubility of gases in liquid water. Chem. Rer. 77, 219-262.
00,13-13M~1/030321-05502.00/0
AN INTERLABORATORY STUDY O F DISSOLVED OXYGEN IN WATER R. J. WILCOCK,C. D. S~VENSO~ and C. A. ROBERTS Chemistry Division, Department of Scientific and Industrial Research. Private Bag. Peton¢, Ne~ Zealand
(Recewed January 1980) AIwtract--Dissolved oxygen concentrations in water samples free from chemical reductants may be stabilised by the addition of mercuric chloride (40 $ m-3} and storage in gas-tight bottles. This preservation technique has been used in an interlaboratory study of dissolved oxygen analysis in New Zealand laboratories. At reference coneentrations of 1.20 and 5.86 g m - 3 there was a significant positive bias in results reported for both the Winkler method (0.24 and 0.22 gm -3 respectively) and the membrane electrode method (0.59 and 0.62 g m-3 respectively). Inadequate precautions to avoid sample aeration during handling and analysis probably caused the bias.
INTRODUCTION
For a number of years this laboratory has organised a nation-wide programme of collaborative water analysis; the C H E M A Q U A programme (Kingsford et al., }973; Stevenson et al., 1975 ; Timperley, 1978). along similar lines to those undertaken in other countries (Wales & Traversy, 1972; Ekedahl et al., 1975; Simpson, 1978). The programme is designed to evaluate analytical procedures, assess interlaboratory variability of results, and create an awareness among the participating organisations of the need for regular assessment of analytical performance. Previous rounds have been concerned with major ions (Kingsford et al., 1973), nutrient species (Stevenson et al., 1975) and trace metals (T~mperley, 1978), at levels likely to be of significance in natural, potable and waste waters. We are not aware of any interlaboratory study of dissolved oxygen analysis in water. Such studies appear to have been hindered by the lack of an effective preservation procedure to ensure that all laboratories received samples containing the same concentration of dissolved oxygen. We have found that concentrations of this constituent can be stabilised by addition of mercuric chloride, to eliminate biological activity, and storage in gas-tight glass containers. This paper reports the successful application of this preservation procedure in an interlaboratory study. Dissolved oxygen fDO) is an important physico-chemical parameter of water quality, and has special significance for aquatic organisms in natural waters. It is crucial therefore that its measurement he both rapid and reliable. Of the many ways of determining DO. by far the most commonly used are the Winkler method and its modifications (Standard Methods. 1976). and the various membrane-electrode systems. Because results by these different methods are frequently compared with each other it is often of great interest to know the extent to which this is justified, or if one method is clearly superior to the other. 321
This study involved 45 laboratories, ranging from large research-oriented organisations engaged in regular analysis of water samples, to small "one-person" laboratories, not occupied full-time with water analysis. Each laboratory was sent two 2.25 1. samples, one containing a low level of D O (less than 2 g i n -3) ( C H E M A Q U A 11) and the other approx. 5 g i n -3 ( C H E M A Q U A 12) comparable with the level recommended by some authorities as the minimum concentration required to maintain good fish populations in natural waters (U.S. EPA, 1976). All participants in the study were issued with instructions on how to siphon the test solutions out of the containers, to minimize sample aeration. METHODS
Sample stability trials Before sending samples to the participating laboratories, trials were carried out to ensure that samples remained unaltered during their despatch to the various destinations. All samples were preserved from microbial activity by the addition of mercuric chloride (40gin-3). and the main concern of the trials was minimizing the physical exchange of oxygen between the samples and their surroundings. One of the participant laboratories expressed concern that mercuric chloride would interfere with the Winkler method. This was checked by taking 41. aliquots of distilled water having different DO contents, splitting them into 2 I. sub-samples, and preserving one of these with mercuric chloride. All of the samples were analysed for DO b~ the az~de modification of the Winkler method (Standard Methods. 19761 and results for the preserved and unpreserved pairs compared. No difference was observed between the samples over the range 1.0--7.5 g m- ~ (reproducibility within _+ 0.02gm-~). The water used for the study contained no chcmicall:r interfering substances. All samples in the preservation and stability trials were analysed by the azide modification of the Winkler method. The stability trials consisted of filling various types of container with water of known DO, and analysing them daily for a period of about 2 weeks. Unfilled polyethylene containers (approx. vol. 2.25 I./ were unsatisfactory because they permitted diffusion of
322
R J. WtLCOCK
ox,,gen through the container walls. Samples m these containers, at an initial D O of 1.0 g m - ~. had a steady rate of increase in D O concentration of 0.35 g m - ~ d a y - t. for a period of 10 days. Glass reagent bottles, while not permitting oxygen exchange through the container walls, tended to leak when they were inverted or placed on their sides. Lining the threaded bottle tops with polyethylene sheet prevented water loss from the bottles, but did not restrict gaseous exchange, and increases in DO of up to 0.90 g m - 3 were recorded for samples over 24 h. Glass (Winchester type) reagent bottles fitted with rubbet O-rings proved to be both gas-tight and leak-proof. Samples with D O concentrations in the range 1.8-9.5 g m - 3. analysed over a period of 12 days. had coefficients of variation less than 1°;,. This method was consequently adopted for the main trial. Preparation ol sample.~ /or collaborattre exerct.se The local tapwater, of artesian origin, provided a convenient supply of low D O content water for C H E M A Q U A I I Aliquots were collected by running tap water into the bottom of each bottle and displacing 2 vol before retaining sample water. The constancy of the D O concentration throughout the filling process was followed by analysing the first bottle filled, and thereafter every fifth bottle. These samples had a D O content of 1.00 +_ 0.02 g m Samples for C H E M A Q U A 12 were produced b~ bubbling compressed air through tap water in a large plastic tank. The solution was effectively stirred by the bubble action, and D O was monitored continuously with a D O electrode system". Samples were collected by siphoning partially aerated water into clean Winchester bottles, and displacing about 2 vol. before retaining the water. DO analyses were carried out on the first bottle, and on ever.,, tenth sample collected: g~vtng a D O of 5.76 _+ 0.08 g m - ' After each bottle was filled and labelled (CHEMAQL-X II or C H E M A Q U A 121. 4 m l of sample was withdrav, n with a syringe and 2 ml of 5°o mercuric chloride solution added. The bottles were fitted with rubber O-rings and tops. and then inverted several times to ensure thorough mixing of the preservative, and equilibrium between the sample waters and the 2 ml air bubbles Icreated to compensate for possible thermal expansion). The small gas space above each sample led to a slight increase m the sample DO levels. The magnitude of this change was estimated by assuming ideal behaviour of the gas phase, a gas-tight seal (i.e. a closed system .tnd a final equilibrium position described by Henry's law. This yielded C. = C, - I P t V s M . I O ~ VtRT) I 4-1KHV=M.103 V~RT) where C~. C., are respectivel.~, the retrial and final DO concentratzons, in g m - 3. P~ is the initial partial pressure of oxygen m the a~r space (aim'f) KH is the Henry'l lay, constant, m arm g " ~• m V). V~ are the gas volume and liquid volume respectively,
of C H E M A Q U A 1! and C H E M A Q U A 12. for 10 days after the date of despatch of the mare body of samples to the participating laboratories. The adjusted values were 1.20 ~ _ 0 . 1 0 g m -3 for C H E M A Q U A II and 5.86 -,- 0 . 1 2 g m "
for C H E M A Q U A 12
These were defined as the reference concentrations. (N.B. The 2 ml air space in each sample bottle was calculated to allow for a thermal expansion caused by a rise in temperature of 10:C. For this system in equilibrium at 20~C, the oxygen distribution between the gas and liquid phases is only affected slightly by temperature variations. The maxim u m change in DO, for a ~ 10:C temperature change, was estimated to be __. 0.05 8 m - ~ for C H E M A Q U A 12. and considerably less for C H E M A Q U A 11.) RESULTS All 45 p a r t i c i p a n t o r g a n i s a t i o n s r e p o r t e d their results as h a v i n g been o b t a i n e d either by t h e W i n k l e r m e t h o d (and p r e d o m i n a n t l y by the azide m o d i f i c a tion), or by m e m b r a n e - e l e c t r o d e s . A few r e p o r t e d results o b t a i n e d by b o t h m e t h o d s . T h e results are listed in T a b l e s I a n d 2. a n d are p r e s e n t e d g r a p h i c a l l y in Figs l a n d 2. R e s u l t s were a v e r a g e d w h e n m o r e t h a n o n e e s t i m a t i o n was s u b m i t t e d for e a c h s a m p l e . a n d all d a t a were r e c o r d e d to the n u m b e r of signific a n t figures r e p o r t e d by e a c h l a b o r a t o r y . Statistical treannent o l d a t a T h e results were i n s p e c t e d for o u t l y i n g values, a c c o r d i n g to t h e A S T M p r o c e d u r e ( A S T M , 1971), before c a l c u l a t i n g m e a n s a n d s t a n d a r d d e v i a t i o n s for the nett d a t a sets (Tables 1 a n d 2). R e s u l t s o b t a i n e d by e l e c t r o d e a n d W i n k l e r m e t h o d s were c o m p a r e d for degree of s c a t t e r (variance) by the F-test. a n d for significant differences b e t w e e n their m e a n s by the t-test (Davies. 1947).
DISCLSSION T h e results for b o t h C H E M A Q U A I I a n d 12 are s h o w n in Figs 1 a n d 2. It is e v i d e n t f r o m this t h a t a l t h o u g h m a n y of the results are in g o o d a g r e e m e n t with the reference values ( 1.20 g m - 3 a n d 5.86 g m - 3). t h e r e is an overall positive bias for b o t h m e t h o d s , at
,:]) M e t e r OWlnkler Reference value
(l.)
M ~s the molecular weight of oxygen R. T are the gas constant (arm.tool- + K - ~) and absolute temperature, respectivelyUsing pubhshed KH values tWilhelm ,,t -i.. 1977). these corrections were 0.20 _ 0.01 g m - ~ for C H E M A Q U A II and 0 . 1 0 _ 0 . 0 3 g i n -~ for C H E M A Q U A t2. These mcreases were verified experimentall> by analysing samples
0 " Yellow Sprmgs Instrument C o m p a n y Model 57 Meter and YSI 5739 probe. + I a t m = 101.325 kPa.
;} 2 3 Dissolved oxygen (g/rn3) Fig. ! Results for C H E M A Q U & I I
1.6 1.4 1.90 2.30 1.70 1.3 I 2.2 1.45 1.3 1.7 1.9 2.75 1.78 1.8 5.6* 6.4 6.30 6.1 6.3 6.25 5.9 7.08f 6.4 6.0 6.0 6.54 6.15 5.44 5.g6 6.10 6.40 6.0 6.4 6.18 6.25 6.5 4.64f 6.35 5.9 6.25* 7.84"I" 5.9 6.0 6.2* 5.75 6.32 5.87 5.9 5.91 5,75
(g m " J)
Winkler method
* Averaged result. t Outlier.
Ol 02 03 04 05 06 07 08 II 12 14 I5 16 17 18 19 20 22 24 26 27 28 29 30 31 32 33 34 35 36 37 4O 41 42 43 45
Laboratory code number
* Averaged result. t Outlier.
Reference value = 1.20
Range = 1.3-2.75
M e a n = 1.79 Standard deviation = 0.39 Ic.v. 22%)
04 06 09 19 24 26 28 30 31 36 37 38 39 42
Electrode (g m - s)
Mean = 6.10 Standard deviation = 0.27 (c.v. 4%) Range = 5,44-6,54
1.2" 1.6 1.40 1.3 1.2 1.35 1.4 2.8Ol' 2.0 1.95 I.lO 1.75 1.05 1.05 1.40 2.30 1.70 1.3 1.3 l.l I 1.35 2.2 2.40 1.40 1.0 1.3 1.69' 1.2 1.2 1.3 1.15 1.8 1.32 1.0 1.16 1.30
OI 02 03 04 05 06 07 08 II 12 14 15 16 17 18 19 20 22 24 26 27 28 29 30 31 32 33 34 35 36 37 40 41 42 43 45
Laboratory code n u m b e r 6.4 6.35 6.55 5.95 6.60 6.15 6.7 6.25 6.1 7.15 6.6 6.8 6.15 6.27 7.2
Electrode (g m - J)
Reference value = 5.g6
Range = 5.95 7.2
Mean = 6.48 Slandard deviation = 0.37 (c.v. 6%1
04 06 09 19 24 26 28 30 31 36 37 38 39 40 42
Laboratory code number
Table 2. Dissolved oxygen results for C H E M A Q U A 12.
Mean = 1.44 Slandard deviation = 0.37 (c.v. 26%) Range = I.O-2.40
(g m - 3)
Laboralory code numbers
Winkler method
Table I. Dissolved oxygen results for C H E M A Q U A I I
o
e,,
m
e~
o
~n
m o" O
,,,i
>
324
R.J. WILCOCKet ui.
Meter OWinkler
Z.
5
¢6
7
8
Dissolved oxygen (g/m3) Fig. 2. Results for C H E M A Q U A 12.
the two DO levels. The "'offset" in the Winkler method values becomes less significant however, when the uncertainty of the reference values ti.e. + 0 . 1 0 g i n -J) is taken into account. There was no obvious correlation between laboratory results that differed widely from the mean values, and the delay between their despatch and subsequent analysis. It is apparent from Figs 1 and 2 that analyses carned out by membrane electrode generally gave higher results than those obtained by Winkler titration. This is further illustrated when the means for each method are compared, after outliers have been rejected (Tables 1 and 21. The mean value for the electrode results is higher than the Winkler mean by 0.35 g m - 3 for C H E M A Q U A t 1, and 0.37 g m - 3 for CHEMAQUA 12. Application of the r-test showed that these differences between the methods are significant and indicate a consistent bias for the electrode method of determining DO. A comparison by F-test of the variances for each method, showed they were not significantly different, at either DO level. Our experience in surveymg a large number of different DO meter systems has shown that in general they tend to give positive errors at low DO levels li.e. less than 2 g m-J). This is partly because of their generally poor response times to reach equilibrium at these levels, probably associated with slow depletion of the oxygen in the bulk internal electrolyte necessary to decrease the "blank" response to a negligible level. The stirring that is necessary to maintain flow across the membrane in order to overcome localised DO depletion by the electrode can also often introduce a positive error through sample aeration. The results for C H E M A Q U A 11 are in accord with these observations. The potential sources of error in the Winkler method are well documented IStandard Methods. 1976: Carpenter. 19651. and a precision of + 0 . 0 2 g i n -~ is realizable in most circumstances. As with the electrode method, the most probable source of positive error arises from sample aeration during the various manipulations. The determination of low DO levels is especially important in corrosion studies. and in studies of natural and waste waters having an appreciable oxygen demand. These results suggest
that the Winkler method ~s probably more reliable than the membrane electrode techniques, when the DO is less than 1-2 g m-3, and that care should be taken when comparing results at this level obtained by different workers. This is particularly important when different methods have been used. The disparity between the Winkler and electrode method means for C H E M A Q U A 12 was larger than expected, as there is usually very good agreement between these methods at moderate to high DO levels (Reynolds. 19691. Response times for membrane-.electrodes are typically of the order of 10-20s at 5-6 g m - 3 DO, and there is usually no difficulty in carrying out measurements in this range and achieving accurate results ( ~0.05 gin-3). It is possible that in some cases aeration errors were introduced when the samples were removed from their bottles, especially if the samples were poured rather than siphoned, into suitable containers. All participants were advised of the proper sample handling procedure to minimize aeration errors, and the large samples distributed (2.25 I.I should have provided adequate volume for flushing of analysis vessels. It is difficult to understand how such errors would account for the difference between the two sets of results, and it is more probable that an additional consistent error, such as aeration caused by over-vigorous stirring. was responsible for the higher electrode results. In spite of the difference between the means for C H E M A Q U A 12. the overall standard of results for both methods was quite satisfactory as judged by the total error criterion for analytical acceptability (McFarren et al.. 1970) which rates both methods "'excellent" at this DO level. However. the results cast some doubt on whether, in practice, it is possible to distinguish reliably between DO levels of 5 g m - ~ and 6 g m -3, as is required in some regulatory water classifications. CONCLUSIONS
The tnterlaboratory comparison of DO measurement, at two concentrations, showed that a reasonable level of performance was achieved in most cases. and that the between-laboratory variability was about 7-10 times the precision attainable by each method. Those results obtained by membrane-electrodes were on average significantly higher than those measured by the Winkler method. It was not possible to discern whether this positive bias in the electrode method was caused by an inherent failing of the method, or if analysts were consistently making the same kind of error. It is thought that the latter is the case, and that sample aeration either during the stirring or sample transfer stages was responsible. The variability among analysts" results suggests care should be taken when comparing results obtained by different workers, and that too much importance should not be attached to small differences in DO without close examination of the methods used.
An interlaboratory study of dissolved oxygen in water
The results of the survey suggest that at D O levels of 1-2 g m -a or less, the Winkler method is more reliable than most conventional membrane-electrodes. REFERENCES ASTM (1971) Recommended practice for dealing with oullying observations. ASTM Designation E178-68 Carpenter J. H. (1965) The accuracy of the Winkler method for dissolved oxygen analysis. Limnol. Oceano#. 10, 135-140. Davies O. L. (1947) Statistical Methods tn Research and Production, pp. 55-70. Oliver & Boyd, London. Ekedahl G.. Junker P+ & Rondell B. 0975) lnterlaboratory study of methods for chemical analysis of water. J. War. Pollut. Control Fed. 47. 858-866. Kingsford M.. Stevenson C. D. & Edgerley W. H L. (1973) Collaborative tests of water analysis (the CHEMAQUA programme~. I. Sodium, potassium, calcium, magnesium, chloride, sulphate, bicarbonate, carbonate and conductivii). N.Z. J/Sci. 16, 895-902. McFarren E. F.. Lishka R. J. & Parker J. H. (1970) Criterion for judging acceptability of analytical methods. Analyt. Chem. 42. 358-365. Reynold J. F. (1969) Comparison studies of Winkler vs
325
oxygen sensor. J. Wat. Pollut. Control Fed. 41. 2002-2009. Simpson E. A. (1978)The harmonisation of the monitoring of the quality of inland freshwater. J. Inst. War. Engng Sciem. 32. 45-56. Standard Methods for the Exanunation of Water and Wastewater (1976) 14th Edn. Americap Public Health Association, New York, NY. Stevenson C. D.. Kingsford M. & Edgerley W. H, L. (1975) Collaborative tests of water analysis (the CHEMAQUA programme) 2, Nitrate nitrogen, reactive dissolved phosphorus, and total dissolved phosphorus. N.Z. JI ScL 16, 895-902. Timperley M. H. (1978) Collaborative tests of water analysis (the CHEMAQUA programme) 3. Trace metals, N.Z. JI Sci. 21. 557-564. U.S. EPA (1976) Quality Criteria for Water. Office of Water and Hazardous Materials, Washington, DC. Wales R. W. & Traversy W. J. (19721 lnterlaboratory quality control study number 2. Total phosphate organic nitrogen, nitrate nitrogen, and organic carbon. Inland Waters Branch. Dept. of Envim. Report Series 119. Ottawa. Canada. Wilhelm E., Battino R. & Wilcock R. J. 0977) Low pressure solubility of gases in liquid water. Chem. Rer. 77, 219-262.