ANALYST JANUARY 1984 VOL. 109 3 Results of an Inter-laboratory Analytical Quality Control Programme for Non-saline Waters The Severn Estuary Chemists’ Sub-Committee* The Severn-Trent South West Welsh and Wessex Water Authorities have a statutory responsibility for the quality of water in the Severn Estuary. A Joint Committee was formed to coordinate the activities of the Authorities in respect of the management of the water quality with special reference to the monitoring of polluting discharges entering the estuary. In order to carry out this duty effectively comparability of analytical results between the laboratories of the four Authorities was considered to be essential and a Chemists‘ Sub-committee was established to undertake this task. The Chemists‘ Sub-committee was required to establish within a period of 3 years and then to maintain an inter-laboratory analytical quality control programme for a total of 17 determinands.It was found that even for relatively simple determinands such as ammonia total oxidised nitrogen and chloride inter-laboratory comparability was not easy to achieve and that once achieved constant attention was necessary in order to maintain it. Hence the work required was much greater than at first envisaged and even so a perfect set of results was not possible for all determinands. This paper summarises the approach adopted illustrates some of the problems experienced and presents results of the inter-laboratory comparability finally achieved. Keywords Analytical quality control; inter-laboratory comparability; collaborative studies; river water analysis; sewage and trade effluent analysis The four Water Authorities that form the boundary of the Severn Estuary (Severn-Trent South West Welsh and Wessex) have statutory responsibilities regarding the quality of water in the estuary.There has been particular concern for many years about the pollution load entering the estuary via direct industrial discharges sewage effluents and rivers (these are termed “inputs”). In order to quantify these inputs with confidence it was necessary to establish comparability of analytical results between the various laboratories of the four authorities. A Chemists’ Sub-committee (CSC) (see Appen-dix) was established with the specific objective of achieving comparability of analytical results between all laboratories involved in the analysis of non-saline inputs to the estuary (the “Inputs Programme”).The CSC agreed that simply circulating samples to the participating laboratories for analysis by their current analyti-cal methods was unlikely to lead to comparability between the laboratories. Such exercises which have been widely used in other collaborative programmes1J and are commonly referred to as “round-robin” exercises have the disadvantage that they will provide information on the comparability achieved in a particular exercise only as measured by the over-all standard deviation. The round-robin exercise does not set objectives for precision and bias required nor will it readily enable the continuous performance of an individual laboratory to be assessed objectively nor enable the many possible sources of error in analytical methods to be identified and corrected.Therefore it was decided that a more systematic approach to comparability than round-robin exercises was required, ideally an approach that would enable the performance of each laboratory to be monitored objectively by comparison of results obtained with pre-set targets of maximum desirable total error (or accuracy). About the time of inception of the CSC such an approach had been proposed by the Water Research Centre (WRC) for achieving comparability within the group of laboratories participating in the Department of the Environment’s Harmonised Monitoring (HM) Scheme .374 The CSC felt that although this approach met its basic requirements a major disadvantage when applied to the * Correspondence regarding this paper should be addressed to: Dr.K. C. Wheatstone Severn-Trent Water Authority Lower Severn Division 141 Church Street Malvern Worcestershire WR14 2AN, UK. Severn Estuary Inputs Programme was that an exercise was held for each determinand in turn and consisted of several separate stages i.e. precision testing stage comparison of standards stage and bias testing stage each stage in an exercise being carried out sequentially on the successful completion of the previous stage. Consequently progress in achieving comparability between a group of laboratories although thorough was slow. Such a rate of progress was unacceptable in the context of the Inputs Programme for which inter-laboratory comparability was required for 17 determinands within 3 years.A compromise approach was adopted that combined the speed and simplicity of a round-robin exercise with many of the advantages of the HM Scheme approach. This paper provides a summary of the approach to comparability testing employed the problems experienced by individual laboratories in meeting the objectives and the level of inter-laboratory comparability finally achieved. A total of almost 40 exercises was necessary to achieve the comparability required; this paper provides only the results obtained in the final stages for each determinand. Full details of all exercises and associated work may be obtained on request. The Approach to Comparability Suite of Determinands The suite of determinands for which comparability between the participating laboratories was required for the Inputs Programme was as follows ammoniacal nitrogen total oxidised nitrogen silicate orthophosphate total phosphorus, chloride cadmium chromium copper iron lead mangan-ese mercury nickel zinc biochemical oxygen demand and organochlorine pesticides.Tolerable Error of Analytical Results After due consideration the tolerable errors of analytical results for the Inputs Programme were set as follows and this also enabled the results of exercises to be assessed objec-tively. In deciding the targets it was assumed that the total error of an analytical result is divided equally between systematic and random errors. This is an arbitrary but convenient sub-division of the total error that has been found to be acceptable in practice.4-4 ANALYST.JANUARY 1984. VOL. 109 ( a ) For metals and total phosphorus a maximum systematic error (or bias) for an individual analytical result of +20% of the concentration or half of the target limit of detection, whichever is the greater and a precision (as measured by the standard deviation) of +lo% of the concentration or one quarter of the limit of detection whichever is the greater. Target limits of detection are as follows: Determinand 1 Cadmium . . . . . . Copper . . . . . . . . Chromium . . . . . . Lead . . . . . . . . Nickel . . . . . . . . Zinc . . . . . . . . Iron . . . . . . . . Manganese . . . . . . Mercury . . . . . . Totalphosphorus . . .. Sewage and trade effluents 1.0pg1-1 10.0 pg 1-1 10.0 pg 1-1 10.0 pg I-’ 10.0 pg I-’ 10.0 pg 1- I 10.0 pg 1-1 10.0ug1-’ NA 0.1 mg 1-1 River water 0.2 pg I-’ 2.0 pg I-’ 2.0pgl-’ 2.0pgl-’ 2.0 pg I - ’ 0.1 1.181-1 NA NA NA 0.1 mg I-’ (b) For biochemical oxygen demand a maximum systematic error of k1.0mgl-1 and a precision of k0.5mgl-1 both relating to the sample whether diluted or not actually measured in the bottle. ( c ) For all other determinands the targets were k 10% of the concentration for maximum systematic error and +5’/0 of the concentration for precision. These tolerable errors may seem unduly large but experience has shown that for routine analysis they are realistic. Ideas that the total error of an analytical result is small (for example 5-10%) have not been confirmed in practice either in this or in other studies.4-7 Indeed McFarren et al.,7 when using results of inter-laboratory studies in order to categorise analytical methods objectively suggested that methods be judged excellent when the total error was f25% or less acceptable when the total error was +50% or less and unacceptable only when that total error was greater than The CSC had been given the requirement that the absolute minimum of analytical results should be reported as “less than” values for the purposes of the Inputs Programme. Therefore it was necessary to develop analytical methods and procedures that met the required percentage targets for all concentrations of determinands likely to be found in the “inputs” to the estuary.Consequently the methods employed by the CSC have limits of detection sufficiently small to allow the accuracy of an individual result to meet the targets regardless of the concentration. A good example to illustrate the reasoning behind this approach is provided by the results for lead in the River Severn at Hawbridge by Severn-Trent Water Authority’s routine water quality monitoring programme.8 For 1977-78, the mean concentration of lead was reported at “less than 0.04 mg 1-1” based on 25 samples (0.04 mg-1 being the limit of detection of the laboratory’s normal analytical method). At the same sampling point the mean flow over the same period was 9370 M1 d-1. From this it can be calculated that the mean mass load of lead in the River Severn at that point was “less than 374.8 kg d-1.” This is a statement that is unacceptably vague as the actual amount could be any value between 0 and 374.8 kg d-1.The analytical methods adopted for lead for the Inputs Programme consequently had limits of detection much smaller than the laboratory’s normal method which enabled real values to be reported for this determinand (and has in fact shown that the mean mass load of lead in the river at this point is about 40 kg d-1). +5O%. Analytical Quality Control Charts The first step in the CSC’s approach to comparability was to set up where not already in routine use analytical quality control charts in all laboratories for the determinands re-quired. By the use of such charts all laboratories were able to check routinely the precision of their analytical results.Comparability Exercise Format The comparability exercise format adopted by the CSC enabled all the stages i.e. precision testing comparison of standard solutions bias testing and where appropriate, spiking recovery to be carried out simultaneously and for several determinands at once rather than sequentially and individually as in the HM scheme. By taking the number of replicate batches of analyses as five it was possible for an inter-laboratory comparability exercise to be completed for a group of determinands within one working week. This convenience was offset by a lower degree of confidence in the results than if a larger number of replicates had been carried out. This was a calculated decision bearing in mind the conditions imposed by the time-scale required for the pro-gramme.The principles of the calculations employed were those used in the HM scheme.4.5 The “standard format” for comparability exercises was agreed to be the analysis by each laboratory of the following samples and solutions in duplicate and in random order on each of five days: a blank solution; the laboratory’s analytical quality control standard solu-tion; a standard synthetic solution of similar concentration to a standard synthetic solution of low concentration; a river sample; and an input other than a river-this varied and included sewage final effluent crude sewage and industrial efflu-ent. (b) ; Solutions (c)-(f) were prepared and circulated by the organis-ing laboratory.This format was followed for the “nutrient” and C1 determinands required for the Inputs Programme although for metals the format was varied slightly (see later). Check Exercises Once comparability had been achieved for a particular determinand or group of determinands check samples were circulated from time to time to all laboratories to ensure that this situation continued for the duration of the Inputs Programme. A river and another “input” sample were supplied for each such exercise and as these exercises were only checks on continuing comparability the format was reduced to duplicate analyses only for each determinand. These check exercises are still ongoing at approximately 3-monthly intervals. A summary of the application of the above approach to comparability testing is given below for all the determinands of interest in the Inputs Programme.Results and Discussion Determination of Metals in Sewage and Trade Effluents The first inter-laboratory comparability exercise for the metals of interest (Cd Cr Cu Fe Pb Mn Ni and Zn) involved the analysis of standard solutions samples and “spiked” samples by direct aspiration of the acidified samples into the atomic-absorption spectrophotometer. The results clearly demonstrated that such a method was unsatisfactory for most types of sample encountered being applicable only to concentrations of individual metals greater than about 500 pg 1-1 ANALYST. JANUARY 1984 VOL. 109 A more appropriate method was sought by laboratories testing various methods individually and which eventually culminated in an inter-laboratory comparability exercise objectively to compare and assess the candidate methods.The methods included concentration by evaporation solvent extraction and ion exchange prior to aspiration into the AAS instrument. From this exercise it was concluded that concen-tration by evaporation was the simplest and least tedious of the three methods and this was selected for further study. Early problems were experienced by some laboratories with the concentration by evaporation method owing to the lack of availability of a background absorption correction facility. Some laboratories also experienced problems with contamina-tion of the acid used for digestion while others found that their AAS instruments were in obvious need of servicing or even replacement.Each of these problems was identified and overcome in turn sometimes by the laboratory working alone and sometimes in collaboration with other laboratories. At every stage small exercises were held to monitor the progress of individual laboratories. The final inter-laboratory comparability exercise involved the analysis of all eight metals of interest in a circulated standard solution and a sewage effluent sample once each day for 4 days with all six laboratories using the standardised evaporation method. This method consisted of a 20-fold evaporation of the sample in the presence of nitric acid taking care to evaporate to fuming and not to allow the sample to boil dry. Correction for non-atomic (background) absorption was found to be essential for cadmium lead nickel and zinc.5 The results obtained from this exercise are given in Tables 1-4 from which it can be seen that most of the results were within the required targets. Where the targets were occasion-ally exceeded it was either only marginally or else it was due to an unacceptably large bias which in some instances was so large that the result was statistically rejected. These large outlying results were thought to be due to contamination of the sample during analysis and served to emphasise the care that needs to be taken in these determinations. For example, the high results obtained by laboratory 4 for cadmium and zinc for the sewage effluent sample were considered to be due to contamination.The results obtained for lead require special comment. Although all but one of the laboratories achieved acceptable results for the circulated standard the results for the sewage effluent fell into two distinct groups of concentration a factor of about two apart. The reasons for this are thought to be (a) contamination for laboratory 4 and possibly laboratory 1 and (b) omitting to use background absorption correction for laboratory 2. The lead concentration in the sewage effluent sample was checked by two independent laboratories using the same method who reported mean concentrations of 12.0 and 12.9 pg 1-1. This tends to confirm that the high results reported by laboratories 1 2 and 4 were erroneous. The low results for laboratory 3 for both standard and sample were caused by an incorrect laboratory standard.The results reported for chromium and iron for both the standard solution and the sewage effluent gave cause for concern. The range of results was large in all instances and no Table 1. Results for cadmium and chromium; Cadmium Chromium Circulated standard Circulated standard (true concn. 6.0 pg 1-1) Sewage effluent (true concn. 9.2 pg 1 - I ) Sewage effluent Mean Std. Rel. Max. Mean Std. Rel. Max. Mean Std. Rel. Max. Mean Std. Rel. Max. concn./ dev./ std. possible concn./ dev./ std. possible concn./ dev./ std. possible concn./ dev./ std. possible pg p! dev. bias dev. bias, 1 6.2 0.08 1.3 +4.9 0.9 0.05 5.6 ++ 8.3 0.29 3.5 -13.5 1.9 0.48 + ++ 2 5.7 0.04 0.7 -5.8 0.9 0.08 8.9 ++ 6.3 1.44 + ++ 2.1 0.63 + ++ 4.8 0.26 5.4 ++ 3 6.0 0.64 10.7* k12.5 0.7 0.15 + ++ 8.1 0.94 + ++ 4 5.4 0.56 10.4* -21.0** 4.8R 0.86 17.9** 9.8 0.96 9.8 +18.8 0.8 0.00 0.0 ++ 5 5.3 - - - 0.9 0.14 + ++ 4.5 - - - 2.0 0.50 + ++ 9.2 3.03* + ++ 0.5 0.33 + ++ 6 5.6 0.31 5.5 -12.7 1.0 0.08 8.0 ++ p! :! yo O/O tory 1 - 1 I-' % '% l5 1- y" Yo 1-1 I - Yo Y" 1-Labora- pg pg dev.bias p! dev bias, Mean 5.7 0.9 7.7 2.0 t Symbols as follows * result not significantly different from target (at 95% confidence level); * * result outside target; - single result only reported; + standard deviation within concentration target; + + maximum possible bias within concentration target; R statistically rejected value; ND not detected. Table 2. Results for copper and iron? Copper Iron Circulated standard (true concn.35.0 pgl-1) Sewage effluent Circulated standard (true concn. 466 pg 1- 1) Mean Std. Rel. Max. concn.1 dev.1 std. possible Labora- pg pg dev. bias, tory 1-1 1-1 ?Lo O/O 1 32.9 0.91 2.8 -9.1 2 35.8 2.7 7.5 +11.4 3 33.8 2.7 8.0 -12.5 4 32.1 0.65 2.0 -10.5 5 35.0 - -6 34.1 1.9 5.6 -9.0 -Mean 34.0 t For symbols see footnote to Table 1. Mean concn./ 1-1 29.2 31.9 30.2 26.1 25.0 28.9 Std. Rel. dev.1 std. I-' Yo 0.48 1.6 8.3 26.0** 0.46 1.5 0.82 3.1 1.6 6.4 1.5 5.2 28.6 pg dev., Max. possible bias, O/O +4.1 +45.7** +7.5 - 12.1 -19.2 +7.2 Mean concn.1 1-1 469 514 364 514 428 449 Std. Rel. dev.1 std. pg dev., I-' O/O 3.4 0.7 36.2 7.0 31.4 8.6 67.9 13.2* 16.1 3.6 456 - -Max.possible bias, O/O +1.5 +19.4 +27.4** -7.7 -29.8** -Sewage effluent Mean Std. Rel. Max. concn.1 dev./ std. possible I.18 dev. bias, 1-1 lY yo Yo 247 10.3 4.2 -14.3 244 3.3 1.4 -12.4 206 10.3 5.0 -29.2** 250 5.4 2.2 -11.1 357 37.4 10.5" +46.4** 342 34.7 10.2* +39.7** 27 6 ANALYST JANUARY 1984 VOL. 109 Table 3. Results for lead and manganese? Lead Manganese Circulated standard (true concn. 40 pg 1-1) Mean Std. Rel. Max. concn.1 dev.1 std. possible Labora- pg pg dev. bias, tory 1-1 1-1 Yo YO 1 40.4 2.78 6.9 +9.2 2 37.6 2.29 6.1 -12.7 3 35.0 3.1 8.9 -21.6** 4 38.9 4.53 11.6* -16.1 6 36.4 1.67 4.6 -13.9 - 5 39.6 - -Mean 38.0 t For symbols see footnote to Table 1.Sewage effluent Circulated standard (true concn. 42.6 pg 1-1) Sewage effluent Mean Std. Rel. concn.1 dev.1 std. 1-1 1-1 Yo 23.9R 1.11 4.6 38.3R 2.40 6.3 11.1 0.59 5.3 25.OR 3.2 12.8* 13.9 3.0* + 13.7 0.50 3.6 P8 Pg dev.9 12.9 Max. possible bias, Y O R R - 19.3 ++ +10.8 R Mean concn.1 1- p5 42.4 40.6 37.1 39.9 44.8 42.6 Std. Rel. dev.1 std. 1-1 Yo 1.75 4.1 1.89 4.6 3.28 8.8 1.14 2.9 0.71 1.7 41.2 Pg dev., - -Max. possible bias, YO -5.3 -9.9 -22.0** -9.5 k2.0 -Mean Std. Rel. concn.1 dev.1 std. pg dev. r 1-1 yo 39.0 0.41 1.0 30.1 4.50 15.0* 31.8 0.49 1.5 36.5 1.11 3.0 28.9 1.83 6.3 31.7 0.34 1.1 33.0 Max. possible bias, YO +19.6 -24.8** -5.4 +14.6 - 19.0 -5.2 ~ ~~ Table 4.Results for nickel and zinct Nickel Zinc Circulated standard (true concn. 8.8 pg 1-1) Sewage effluent Circulated standard (true concn. 92.6 pg 1-1) Sewage effluent Mean Std. Rel. Max. concn.1 dev. std. possible Labora- pg p dev. bias, tory 1-1 I- Yo O/O 2 6.8 0.29 4.3 ++ 3 6.4 0.42 6.6 ++ 4 9.7 0.75 7.7 +20.2** 1 8.1 0.75 9.3 -18.0 - 5 8.2 - -6 8.2 0.51 6.2 -13.6 Mean 7.9 t For symbols see footnote to Table 1. Mean Std. Rel. concn.1 dev.1 std. 1-1 1- Yo 5.3 0.29 5.5 5.3 0.65 + 3.8 0.13 3.4 5.5 1.27 + 5.0 0.96 + 4.4 0.85 + pg p dev.9 4.9 Max. possible bias, Y O +15.1 ++ ++ ++ ++ ++ Mean Std. Rel. concn.1 dev.1 std. P! dev. : 1- o/o 94.9 1.32 1.4 97.9 1.70 1.7 87.2 7.36 8.4 102.3 7.24 7.1 94.7 2.22 2.3 87.2 - -94.0 Max.possible bias, YO +4.2 +7.9 -15.2 +19.7 +5.1 -Mean concn.1 vg I-' 67.3 51.5 54.3 144.3R 49.4 58.0 Std. Rel. dev.1 std. I-' Yo 0.58 0.9 4.71 9.1 1.00 1.9 11.5 8.0 2.40 4.9 1.32 2.3 56.1 pg dev.7 Max. possible bias, Y O +21.2** -18.1 -5.3 - 17.0 +6.2 R common factor or reason could be found to which these differences could be attributed. This situation was clearly far from satisfactory and consequently one laboratory undertook a detailed investigation of the cause of the differences. The investigation showed that both chromium and iron could be subject to anomalous calibration graphs under certain flame conditions in the air-acetylene flame and that this was the probable cause of the differences.Details of the investigations and recommendations for overcoming the problem have been published.9Jo Determination of Metals in River Water For the determination of metals in sewage and trade effluents the evaporation and flame AAS method described above was considered satisfactory. However for the accurate determina-tion of cadmium lead copper nickel and zinc in river samples the target limits of detection obtainable by the evaporation method needed to be improved by a factor of about five. It was felt that only two analytical techniques were capable of meeting the limits of detection required namely concentration of the metals by means of a chelating resin followed by AAS and AAS with electrothermal atomisation.The chelating resin Chelex 100 had been widely used for the determination of low level metals in sea water,11-13 but for river waters the technique could not be directly applied because of the greater proportion of complexed and/or organically bound metals that would not be adsorbed into the resin. Consequently low results would be expected from samples of river waters. A method was eventually developed that overcame this problem involving treating the sample with potassium persulphate and nitric acid in order to break down the organic matter in the samples and to release the organically bound metals for analysis. The pH of the solution was adjusted to 6.3-6.7 and passed through a small column of Chelex 100. After rinsing to remove residual salts the chelated metals were removed from the column with a small volume of dilute nitric acid prior to analysis by flame AAS.Determination of low concentrations of metals by AAS with electrothermal atomisation presented difficulties for routine use owing to dissolved salts in natural samples which caused large suppressive interference effects (up to 80% suppression of the signal) particularly for lead. A method was developed whereby this interference was overcome by means of the addition of small amounts of lanthanum salts to the samples. Details of this method have been published.14-16 During the joint development of these two analytical techniques it was mutually beneficial to hold a series of small comparability exercises in order to monitor the progress and to check their accuracy.When the development work was completed it was decided that owing to the special precautions needed in analysing river samples the techniques were by no means suitable for routine use in all laboratories. Conse-quently only selected laboratories were equipped to carry out this analysis. The CSC also made available to another group involved in similar work the Humber Estuary ad hoc Group for Analy-tical methods and Inter-laboratory Testing details of the techniques developed. In-house precision testing was carried out by each of the laboratories together with coordinated tests on circulated river samples and standard solutions. A joint comparability exercise was held that involved fiv ANALYST JANUARY 1984 VOL.109 7 replicate determinations on the same day of a standard solution river water and river water spiked with the metals of interest (Lee Cd Cu Pb Ni and Zn). The river water was filtered through a 0.45-pm membrane filter into a large polythene container and the metals were preserved by the addition of 1 ml of concentrated nitric acid (low-in-metals grade) per litre of filtered sample. A bulk standard synthetic solution was also prepared and similarly preserved. The preserved river sample was divided into two, and to one aliquot was added a spike of a concentrated solution containing known amounts of the metals of interest. The river and spiked river samples and the synthetic solution were sub-sampled into small polyethylene bottles for distribu-tion to the participating laboratories.Two laboratories (numbers 3 and 5 ) used the Chelex 100 method the others used the electrothermal atomisation method. The results obtained are given in Tables 5-9 from which it can be seen that except for one precision result (laboratory 5 for zinc) and one bias result (laboratory 2 for lead) all laboratories met the targets for precision and bias on the circulated standard sample and spiked sample for all five metals. As these two laboratories were only marginally outside the target for only one sample each and all their other results were within the targets no further action was considered necessary. For the type of sample and levels of concentration of metals involved the results for the inter-laboratory exercise were considered to be very satisfactory overall.Mercury An initial survey of methods in use for mercury determina-tions indicated that all laboratories used the cold vapour electrothermal AAS measurement technique after reduction of mercury to the elemental form and the stripping of it from solution. A variety of reductants and measurement cell systems were in use and the first exercise was therefore designed to test the sensitivity of the measurement stage and the comparability achievable with inorganic mercury stan-dards. The results of this preliminary exercise demonstrated that comparability could be achieved for inorganic standards and indicated that the target detection limit was achieved by all laboratories. Various published methods for sample preservation and for the conversion of organically bound mercury to the inorganic form were tried and tested and a suitable procedure was adopted.Briefly this consists of taking about 100 ml of sample into a glass-stoppered borosilicate-glass bottle and adding Table 5. Results for cadmium Standard solution (true concn. 5.5 pg 1-1) River sample Mean Std. Rel. concn.1 dev.1 std. 1 5.4 0.14 2.6 2 5.5 0.11 2.0 3 5.4 0.29 5.4 4 5.6 0.44 7.9 5 6.2 0.00 0.0 Laboratory pg 1-1 pg 1-1 dev. ,% Mean 5.6 Max. possible bias YO -4.2 k1.9 -6.8 +9.4 + 12.7 Mean concn.1 1-1.2 1.2 1.2 1.4 1.5 Std. Rel. dev./ std. pgl-1 dev. YO 0.09 7.5 0.11 9.2 0.17 14.2* 0.12 8.6 0.05 3.3 1.3 * Result not significantly different from target (at 95% confidence level).Max. possible bias YO -14.3 - 15.7 -20.0 + 16.5 + 19.0 River sample + 5.5 pg 1-Mean concn.1 8.1 6.7 6.6 7.6 7.4 vg 1-Std. Rel. dev.1 std. pg 1-1 dev. YO 0.32 4.0 0.15 2.2 0.20 3.0 0.57 7.4 0.16 2.2 7.3 Max. possible bias YO +15.1 - 10.2 - 12.2 +11.5 +3.5 Table 6. Results for copper Standard solution (true concn. 24.0 pg 1-l) River sample River sample + 24.0 pg 1-1 Laboratory 1 2 3 4 5 Mean Mean Std. Rel. concn.1 dev.1 std. pg 1-1 pg 1-1 dev. '/O 21.2 1.41 6.7 21.5 0.36 1.7 24.2 1.21 5 .O 26.1 0.55 2.1 26.5 0.72 2.7 23.9 Max. possible bias YO -17.5 -12.1 +5.8 +11.3 +13.3 Mean Std. Rel. concn.1 dev.1 std. pgl-1 pgl-1 dev. YO 9.9 0.56 5.7 11.8 0.51 4.3 9.3 0.70 7.5 11.2 1.11 9.9 10.0 0.40 4.0 10.4 Max.possible bias YO -9.9 +18.1 - 17.0 + 17.8 -7.5 Mean concn .I 31.5 37.2 33.1 37.2 35.9 Yg 1-Std. Rel. dev.1 std. pg 1-1 dev. YO 1.35 4.3 0.40 1.1 1.09 3.3 0.60 1.6 1.35 3.8 35.0 Max. possible bias YO -13.1 +7.4 -8.4 +7.9 +6.2 Table 7. Results for leadt Standard solution (true concn. 18.0 pg 1-1) River sample River sample +18.0 pg 1-l Mean Std. Rel. concn.1 dev.1 std. pg dev., Laboratory 1-1 1-1 Y O 1 18.3 0.65 3.6 2 18.4 0.38 2.1 3 17.7 0.49 2.8 4 18.6 1.12 6.0 5 17.5 1.06 6.1 Mean 18.1 t For symbols see footnote to Table 1. Max. possible bias, Y O +5.6 +4.4 -4.3 +9.4 -8.9 Mean Std. Rel. concn.1 dev.1 std.1-1 I-' YO 1.6 0.38 + 1.8 0.36 + 1.9 0.42 + 1.0 0.00 0.0 0.4 0.00 0.0 pg dev.7 1.3 Max. possible bias, Yo ++ ++ ++ ++ ++ Mean concn./ 1-1 17.4 22.3 19.0 19.1 16.7 Std. Rel. dev.1 std. 1-1 Y O 0.27 1.6 0.49 2.2 0.54 2.8 0.74 3.9 1.26 7.5 18.9 pg dev., Max. possible bias, Y O -9.3 +20.4** +3.2 +4.8 - 18. 8 ANALYST JANUARY 1984 VOL. 109 Table 8. Results for nickel Standard solution (true concn. 38.0 pg I - ' ) River sample River sample + 38.0 pg 1 - I Mean Std. Rel. concn.1 dev.1 std. ELg pg dev.7 Laboratory 1-1 1-1 YO 1 35.7 1.52 4.3 2 34.4 1.08 3.1 3 33.7 1.98 5.9 4 39.4 1.52 3.9 5 42.4 1.23 2.9 Mean 37.1 Max. possible bias, O/O - 10.0 - 12.6 - 16.6 +7.6 + 14.7 Mean concn .I pg I-' 20.6 20.3 21.1 20.1 23.6 Std.Rel. dev.1 std. pg dev., I - ' Y O 1.52 7.4 0.76 3.7 0.67 3.2 1.78 8.9 1.48 6.3 21.1 Max. possible bias, YO -9.2 -7.2 k3.2 - 12.8 + 18.5 Mean Std. Rel. concn.1 dev.1 std. pg dev., I-' I-' Y O 53.3 0.40 0.8 54.8 0.68 1.2 52.6 0.49 0.9 61.2 1.79 2.9 62.5 2.28 3.6 56.9 Max. possible bias, YO -7.0 -4.8 -8.4 + 10.5 +13.6 Table 9. Results for zinc Standard solution (true concn. 47.0 pg 1-1) Mean concn.1 Laboratory 1- p9 1 45.4 2 53.4 3 48.1 4 52.3 5 49.3 ** Result outside target. Mean Std. Rel. dev.1 std. dev., 2.00 4.4 0.89 1.7 0.83 1.7 1.77 3.4 0.61 1.2 49.7 I- p9 Y O Max. possible bias, Y O -7.4 + 15.5 +4.0 +15.1 +6.2 River sample Mean concn.1 pg 1-1 32.4 32.4 26.1 31.2 28.8 Std.Rel. dev.1 std. dev., 1- p9 "/o 0.89 2.7 0.55 1.7 0.22 0.8 0.43 1.4 4.51 15.6** 30.2 Max. possible bias, Y O + 10.1 +9.0 - 14.3 +4.7 - 18.8 River sample +47.0 pg 1-1 Mean Std. Rel. Max. concn.1 dev.1 std. possible pg pg dev. bias, I-' 1-1 YO YO 81.5 4.82 5.9 +6.5 82.8 0.45 0.5 +3.0 84.9 2.69 3.2 1-8.2 80.8 75.8 0.76 1.0 -7.1 78.9 4.28 5.4 -7.4 Table 10. Results for mercury? Labora-tory 1 2 3 4 5 Mean Organic standard Inorganic standard (true concn. 0.936 pg 1-1) (true concn. 1 .000 pg I-') Mean concn.1 CLg I-' 0.953 0.958 1.064 0.930 1.018 Std. Rel.dev.1 std. I-' O/O 0.020 2.1 0.055 5.7 0.043 4.0 0.022 2.4 0.046 4.5 0.985 pg dev.3 Max. possible bias, YO +3.9 +8.0 + 18.0 -2.9 + 13.4 t For symbols see footnote to Table 1. Mean Std. Rel. concn.1 dev.1 std. dev., I- p rg Yo 0.975 0.044 4.5 1.064 0.037 3.5 1.294R 0'.063 4.9 0.959 0.030 3.1 1.040 0.071 6.8 1.010 Max. possible bias, YO -6.7 +9.9 -7.0 + 10.8 R River sample Mean Std. Rel. concn.1 dev.1 std. pg dev.7 :9 1-1 Yo 0.115 0.012 10.4 0.068 0.005 7.4 0.065 0.017 + 0.056 0.006 10.7 0.078 0.016 + 0.076 Max. possible bias, YO ++ ++ ++ ++ ++ River sample +0.936 pg 1-1 of organic Hg Mean Std. Rel. concn.1 dev.1 std. I-' I-' Yo 1.124 0.023 2.0 0.938 0.015 1.6 1.128 0.046 4.1 1.098 0.034 3.1 1.018 0.041 4.0 pg pg dev-7 1.061 Max.possible bias, YO +8.0 - 12.9 + 10.4 +6.5 -7.7 between 0.2 and 0.5 g of potassium persulphate (the amount depending on the expected organic content of the sample), followed by 1 ml of concentrated nitric acid (low-in-metals grade). The flask is stoppered and despatched to the labora-tory for analysis great care being taken to avoid contamina-tion. The sample is processed ready for analysis in the sample bottle by either bringing to the boil or if a concentration step is required evaporating five-fold. This pre-treatment converts any organic forms of mercury into the inorganic form suitable for analysis. The inter-laboratory comparability exercise was carried out on circulated standards and samples consisting of an inor-ganic mercury standard solution an organic mercury standard solution a river sample and the river sample spiked with organic mercury.The results obtained from five randomised replicate analyses in one batch are given in Table 10. With the exception of one bias result for the inorganic mercury standard solution (laboratory 3) all laboratories achieved the targets for both precision and bias for all the circulated solutions sample and spiked sample. Nutrients and Chloride The results obtained for the first inter-laboratory exercise for the nutrients ( i . e . ammonia total oxidised nitrogen silicate, orthophosphate and total phosphorus) and chloride were poor. Most laboratories failed to meet the targets for either the standard solutions or the samples or both for all determinands.However the advantages of the approach to comparability adopted were immediately obvious as in many instances it was possible to identify why individual labora-tories had failed to meet the targets and remedial action could then be taken. For example several laboratories failed to meet the targets for comparison of standard solutions and so clearly one major reason for the targets not being met was that the laboratory standards were inaccurate. Another factor was that most laboratories analysed the solutions using one range, typically a high range which was suitable for the sewage sample and the high concentration standard solution but completely inappropriate for the low-concentration river samples and solutions.Finally problems were identified wit ANALYST JANUARY 1984. VOL. 109 9 the precision achievable with certain analytical methods eg., distillation/Nesslerisation for ammonia and ion-selective elec-trodes for total oxidised nitrogen and the laboratories concerned were asked to investigate further and if necessary, change to an alternative method. After remedial action had been taken by the laboratories concerned the exercise was repeated with a fresh set of samples with similar concentrations. The results obtained were a considerable improvement over those obtained pre-viously although several laboratories still failed to meet the targets for some samples particularly the low-concentration standard solution and the river water sample.Further exercises were held to improve the quality of results so that the targets were met by the participants. Details of the results obtained from the final inter-laboratory exercise are con-sidered below. Ammonia For ammonia it can be seen that with the exception of laboratory 3 all laboratories achieved results that were either within or only just outside the required targets (Table 11). Laboratory 3 produced results well within the targets for precision but all the results for the river and low-concentration standards were outside the targets for bias. Clearly this laboratory had a fault in its low-concentration calibration standards during the period of the exercise; this was checked and found to be so and inter-laboratory comparability was confirmed in the next check exercise.Total Oxidised Nitrogen The results obtained for total oxidised nitrogen (Table 12) for precision and bias for the final exercise were all within the targets with the exception of the bias on the river sample for laboratory 6. Investigation revealed the cause of this large negative bias to be due to the problems with the automated continuous flow system used which occurred with real samples but not with synthetic solutions. It was overcome by the incorporation of a modified reducing reagent using hydrazine - copper - zinc instead of hydrazine - copper and changing the hydraulics of the analytical cartridge. This demonstrates the value of inter-laboratory comparability exercises involving real samples as this fault would not have been identified otherwise.In subsequent check exercises this fault has not reappeared. Silicate The results for the final exercise (Table 13) show that with the exception of one precision result and three bias results the targets have been achieved by all laboratories for all solutions. The one poor precision result and two of the bias results were only just outside the targets and taken overall the labora-tories’ results were considered acceptable. The laboratory that returned the bias result considerably outside the target checked its calculation but confirmed that the original result was correct. As their results for silicate on real samples in check exercises have been satisfactory it is considered that a simple arithmetic error was the probable cause of this spurious result.Orthophosphate The results for the final exercise (Table 14) were considered to be satisfactory although one laboratory just exceeded the bias target for its own low-level standard solution. As this laboratory achieved bias results well within the required targets for the circulated standards and samples its results were considered to be acceptable overall. For the river sample no bias result has been quoted for laboratory 2 as there was clear evidence from exchange of samples with another laboratory that their sample had been contaminated. Total Phosphorus The exercises involved the circulation of samples together with a standard solution of an organic phosphorus compound (adenosine-5-monophosphoric acid disodiuin salt; AMP).An organic phosphorus compound was used as it was considered to provide the best test of the laboratories’ digestion pro-cedures. If the results were satisfactory for this compound the procedures could be expected to be able to cope with the organic phosphorus compounds occurring in rivers and sewage effluents.17 The results obtained for the first exercise, however were far from satisfactory; all laboratories except one produced results considerably outside the precision and bias targets for the organic phosphorus standard. Most were also outside the targets for both the river and sewage effluent sample. Clearly the methods employed by the laboratories were unacceptable for total phosphorus determination. A closer inspection of results for the organic phosphorus standard revealed that most of the results were about only 50% of the actual concentration.The exception to this was one laboratory that had used a method developed by the Water Pollution Research Laboratory (WPRL). 18 This labor-atory’s results for the river and sewage effluent samples were also above the average. In view of this it was decided that all laboratories would standardise on the WPRL method. After a suitable interval of time to allow the laboratories to gain some experience with the method the exercise was repeated. The results obtained in the final exercise are given in Table 15 from which it can be seen that with only one small exception the results from all laboratories met the targets. The exception was laboratory 3 which had a result outside the bias target for its own standard solution (see also orthophos-phate above).This laboratory however obtained satisfac-tory results for all the other standards and samples and consequently no further action was considered necessary. Chloride The results obtained for chloride are summarised in Table 16 and were as expected generally acceptable. However the results were perhaps not as good as anticipated as two laboratories produced results outside both the precision and bias targets for the river sample. One of these laboratories also had the result for the sewage effluent outside the bias target. The reason for the outlying results was the same in both laboratories-the range of the continuous flow system em-ployed was unsuitable for the low concentrations found in this exercise.The normal working range in these laboratories could have been reduced to accommodate the low concentra-tions but then many routine samples would need dilution before analysis. After consideration of the implications of this for the laboratories’ normal workload together with the use made of chloride results in the Inputs Programme it was decided that a change of range was not appropriate. These results can be taken as confirming the choice of targets set for precision and bias which at first might seem unduly generous. Even for a relatively simple determinand such as chloride which is normally considered to be capable of precise and unbiased measurement problems have been experienced in achieving the targets.Biochemical Oxygen Demand Although biochemical oxygen demand (BOD) was not of primary importance for the Inputs Programme as oxygen deficiency is not a problem in the Severn Estuary it was considered that some knowledge of the mass input of BOD to the Estuary would be useful. Consequently one exercise wa Table 11. Results for ammonia (as N)t Laboratory's own standard Circulated standard (true concn. 0.68 mgl-1) Circulated standard (true concn. 4.0 mg 1-1) River sample Sewage Labora-tory 1 2 3 4 5 6 Mean Mean Std. concn.1 dev./ p p: 0.39 0.020 0.25 0.004 0.18 0.01 4.01 0.100 0.50 0.006 0.90 0.011 Rel. std. dev., YO 5.1* 1.6 5.6* 2.5 1.2 1.2 Max. poss . bias, YO -5.4 f0.9 - 12.9** +1.7 f0.7 k0.7 Mean Std.Rel. concn.1 dev.1 std. dev. p p yo 0.67 0.018 2.7 0.66 0.024 3.6 0.78 0.014 1.8 0.65 0.018 2.8 0.75 0.022 2.9 0.74 0.023 3.1 0.71 Max. poss. bias, YO -3.0 -5.0 +15.9** -6.0 + 12.2** 10.8** Mean concn.1 mg I - ' 4.05 4.13 4.03 3.96 4.16 3.90 Std. Rel. dev.1 std. mg dev., I-' Yo 0.25 6.2* 0.13 3.1 0.09 2.2 0.12 3.0 0.15 3.6 0.31 7.9** 4.04 Max. poss. bias, YO +4.9 +5.1 +2.1 -2.7 +6.2 -7.0 Mean Std. Rel. concn./ dev.1 std. dev., 0.36 0.02 5.6* 0.34 0.016 4.7 0.42 0.008 1.9 0.37 0.018 4.9 0.35 0.019 5.4* 0.34 0.019 5.6* ; yo 0.36 Max. poss. bias, Y O k3.2 -8.1 +18.0** +5.7 -5.8 -8.6 Mean Std. concn./ dev. ; p: 7.7 0.55 8.6 0.16 8.6 0.32 7.9 0.25 9.0 1.01 8.8 0.65 t For symbols see footnote to Table 1.Table 12. Results for total oxidised nitrogen (as N ) t Circulated standard (true concn. 3.4 mg 1 - 1 ) Laboratory's own standard Mean Std. Rel. Max. Mean Std. Rel. Max. concn.1 dev.1 std. poss. concn.1 dev.1 std. poss. Labora- mg mg dev. bias mg mg dev. bias, tory 1 - 1 1 - 1 Yo Yo I - I 1 - ' "/o O/" 1 5.10 0.167 3.3 -3.8 3.30 0.047 1.4 -3.7 2 1.97 0.049 2.5 -2.9 3.50 0.138 3.9 +5.3 3 0.99 0.031 3.1 -2.8 3.40 0.071 2.1 k1.2 4 15.9 0.22 1.4 -1.4 3.25 0.169 5.2* -7.3 5 0.96 0.012 1.3 -4.7 3.46 0.061 1.8 +2.8 6 4.46 0.037 0.8 -0.4 3.29 0.100 3.0 -4.9 Mean - 3.37 t For symbols see footnote to Table 1. Circulated standard (true concn. 18.3 mg 1-1) River sample Sewage Mean Std.Rel. concn.1 dev.1 std. mg dev. f3 1 - 1 yo 18.87 0.527 2.8 17.98 0.420 2.3 18.01 0.550 3.1 18.26 0.190 1.0 18.40 0.561 3.1 18.54 0.173 0.9 18.34 Max. poss. bias, YO +4.8 -3.1 -3.3 -0.8 +2.3 +1.9 Mean Std. Rel. concn.1 dev.1 std. mg mg dev., 1-1 I-' Yo 4.45 0.072 1.6 4.81 0.121 2.5 4.46 0.060 2.4 4.50 0.143 3.2 4.81 0.102 2.1 3.29R 0.073 2.2 4.61 Max. poss. bias, YO -4.4 +5.9 -4.6 -4.2 +5.6 R Mean concn.1 mg I-' 18.49 16.89 17.03 17.75 17.79 17.23 Std. dev. p: 0.612 0.339 0.574 0.228 0.665 0.31 Table 13. Results for silicate (as Si)? Labora-tory 1 2 3 4 5 6 Mean Circulated standard Circulated standard Laboratory's own standard (true concn.0.60 mg 1 - 1 ) (true concn. 2.4 mg 1-1) Mean Std. concn.1 dev.1 I- " 7: 0.60 0.014 2.28 0.084 2.03 0.022 7.82 0.147 1.00 0.005 4.71 0.077 Rel. std. dev., YO 2.3 3.7 1.1 1.9 0.5 1.6 Max. poss. bias, YO -9.8 -3.0 +2.1 -3.3 k0.3 +1.2 t For symbols see footnote to Table 1. Mean concn.1 mg I - ' 0.56 0.59 0.57 0.58 0.57 0.62 Std. Rel. dev.1 std. mg dev., I- I O/O 0.041 7.3* 0.031 5.3* 0.022 3.9 0.045 7.8** 0.004 0.7 0.018 2.9 0.58 Max. poss. bias, YO - 10.6** -4.7 -7.1 -7.7 -5.4 +5.1 Mean Std. Rel. concn.1 dev.1 std. mg mg dev., I - ' I - ' Yo 2.41 0.126 5.2* 2.43 0.146 6.0* 2.51 0.048 1.9 2.61 0.089 3.4 2.19 0.041 1.9 2.38 0.075 3.2 2.42 Max.poss . bias, YO +3.5 +4.8 +5.7 -t 10.9** -9.7 -2.6 River sample Sewage Mean Std. Rel. concn.1 dev.1 std. dev., 2.02 0.104 5.2* 2.08 0.059 2.8 1.96 0.022 1.1 2.17 0.122 5.6* 0.97R 0.007 0.7 2.05 0.025 1.2 : :5 yo 2.06 Max. poss. bias, Y O -4.9 +2.6 -5.5 +8.8 -1.2 R Mean Std. concn.1 dev. 17.59 0.305 18.64 0.067 16.93 0.143 17.05 0.452 18.48 0.465 18.75 0.227 pf 7: Table 14. Results for orthophosphate (as P)? Laboratory's own standard Circulated standard (true concn. 0.78 mg 1-1) Circulated standard (true concn. 2.0 mg 1-1) River sample Mean Std. Rel. Max. concn.1 dev.1 std. poss. Labora- mg mg dev. bias, 1 1.00 0.017 1.7 k1.0 2 0.04 0.001 2.5 -4.0 3 0.09 0.004 4.4 -12.3** 4 1.55 0.021 1.4 -3.9 5 0.50 0.011 2.2 k1.3 6 2.25 0.027 1.2 k0.7 tory 1- I I- 1 o/o YO - Mean t For symbols see footnote to Table 1.Mean concn . I mg 1 - 1 0.80 0.81 0.78 0.75 0.77 0.73 Std. Rel. dev.1 std. mg dev., I-' Yo 0.030 3.8 0.039 4.8 0.024 3.1 0.025 3.3 0.030 3.9 0.017 2.3 0.77 Max. poss. bias, O/O k4.8 +6.7 k1.8 -5.7 -3.5 -7.7 Mean concn.1 mg 1-1 1.97 1.99 2.00 1.93 1.98 1.94 ~~~ ~ Std. Rel. dev.1 std. dev. 7 yo 0.03 1.5 0.04 2.0 0.06 3.0 0.14 7.3* 0.07 3.5 0.03 1.5 1.97 Max. poss. bias, YO -2.4 -1.7 k1.7 -7.6 -3.0 -3.9 Mean concn.1 ;: 0.35 0.55R 0.34 0.36 0.36 0.33 Std. Rel. dev.1 std. dev., 0.023 6.6* 0.022 4.0 0.017 5.0 0.010 2.8 0.008 2.2 0.015 4.6 0.35 p Y Table 15.Results for total phosphorus (as P)t Laboratory's own standard Circulated standard (true concn. 2.0 mg 1-1) Circulated standard (trueconcn. 9.5mg1-1) River sample Labora-tory 1 2 3 4 5 Mean Mean concn .I 1- m: 1.02 0.98 0.08 0.84 1.03 Std. dev.1 1- m: 0.051 0.013 0.008 0.025 0.018 Re1 . std. dev., Yo 5.0 1.3 10.0 3.0 1.7 For symbols see footnote to Table 1. Max. poss . bias , O/O - 10.0 -2.8 -24.6** - 17.4 +4.0 Mean concn.1 mg I-' 1.71 2.17 2.03 2.15 2.18 Std. Rel. dev.1 std. mg dev., 1-1 Yo 0.165 9.6 0.136 6.3 0.057 2.8 0.048 2.2 0.027 1.2 2.05 Max. poss. bias, Yo -19.3 + 12.4 +3.2 +8.9 +9.8 Mean concn.1 mg 1-1 8.93 9.78 9.66 9.96 9.87 Std.Rel. dev.1 std. dev. :? Yo 0.21 2.4 0.45 4.6 0.29 3.0 0.66 6.6 0.19 1.9 9.64 Max. poss . bias, O/O -7.3 +5.7 +3.5 +8.9 +5.1 Mean Std. Rel. concn.1 dev.1 std. dev. p7 p yo 1.24 0.10 8.1 1.13 0.14 12.4* 1.32 0.08 6.1 1.24 0.03 2.4 1.29 0.05 3.9 1.24 Table 16. Results for chloride? Laboratory's own standard Circulated standard (true concn. 83 mg 1 - 1 ) Circulated standard (true concn. 186 mg 1-l) River sample Labora-tory 1 2 3 4 5 6 Mean Mean Std. concn.1 dev.1 ; ;: 122.5 4.06 98.4 0.52 25.3 0.44 25.1 0.32 90.1 0.39 157.7 1.60 -Re1 . std. dev., YO 3.3 0.5 1.7 1.3 0.4 1.0 Max.poss . bias, YO -3.1 -1.9 +2.2 +1.1 +0.4 -2.0 Mean concn.1 mg I-' 82.8 81.2 81.8 83.3 81.4 83.8 Std. Rel. dev.1 std. dev., 3.85 4.6 0.88 1.1 1.00 1.2 0.66 0.8 0.52 0.6 1.19 1.4 82.4 ; yo Max. poss. bias, Y O -2.9 -2.8 -2.1 +0.8 -2.3 + 1.8 Mean concn.1 ;: 182.3 178.3 181.7 185.7 183.1 182.8 Std. Rel. dev.1 std. dev. y yo 4.26 2.3 1.30 0.7 0.88 0.5 0.86 0.5 0.55 0.3 2.39 1.3 182.3 Max. poss. bias, YO -3.3 -4.5 -2.6 -0.4 -1.7 -2.5 Mean concn.1 mg I-' 18.8 21.7 20.9 21.4 20.7 18.8 Std. Rel. dev.1 std. dev. 1.65 8.8** 0.93 4.3 0.24 1.1 0.42 2.0 0.35 1.7 1.98 10.5** 20.4 ;3 yo t For symbols see footnote to Table 1 ANALYST JANUARY 1984 VOL.109 13 carried out to assess inter-laboratory comparability. The exercise involved the analysis of a blank (unseeded) a blank (seeded) the laboratories’ own standard glucose - glutamic acid solution a river sample and a sewage effluent sample using the “standard format.” As BOD is an empirical determinand in which the analytical method used could seriously influence the result obtained it was agreed that all laboratories would use the same method.19 The unseeded blank was used to calculate the results for the sewage sample and the seeded blank for the synthetic standard solutions. To ensure that different results were not produced simply by inappropriate dilution the dilutions to be used by all laboratories for the samples and standards were agreed beforehand.Clearly for a test such as BOD it is not possible to prevent biological activity in the river and sewage samples by sterilisation and still obtain a BOD result. Consequently the samples deteriorated over the 5-day period of the exercise, which made it impossible to assess bias for these real samples. Estimates of inter-laboratory bias were made however using the synthetic standard solutions. Similarly because of sample instability it was not possible to assess correctly the total standard deviation for the river and sewage samples and this was estimated by using the actual within-batch standard deviation (S,) obtained together with a between-batch stan-dard deviation (S,) calculation as a relative proportion of the between-batch standard deviation of the standard solutions.Such a calculated total standard deviation is thought to be the nearest to a “real-life” situation as can reasonably be obtained. The results obtained for the exercise are given in Table 17, from which it can be seen that with the exception of standard deviation values for the two standard solutions from labora-tory 3 all precision and bias results for samples and standards are within the targets. The reasons for laboratory 3’s standard deviations being outside the targets were investigated and appeared to be associated with the blank as their seeded blank values were very high and variable ( i . e . ranging from 1.1 to 3.2 mg 1-1 compared with 0.2-0.5 mg 1-1 obtained by other laboratories).Causes of the high blank values were investigated by the laboratory concerned and corrected. Organochlorine Pesticides A comparability exercise for organochlorine pesticides was undertaken on a round-robin format with the objective of simply assessing the capability of the participating laboratories for identifying correctly a range of pesticides. All laboratories employed their normal method of analysis, which involved essentially solvent extraction and concentra-tion of the pesticides followed by gas-chromatographic separation with electron-capture detection. To aid the positive identification of eluted peaks each laboratory analysed the solvent extract on two different chromatographic colymns. Three samples were circulated for the exercise consisting of (a) a synthetic solution containing a mixture of y-HCH aldrin and endrin in light petroleum (boiling range 40-60 “C) (b) a sewage works final effluent (from a sewage works known to receive effluent from a pesticide manufacturer) and (c) river water downstream of (b).The results obtained are detailed in Table 18 from which it was concluded that generally the agreement between labora-tories both qualitative and quantitative was fairly good. The exercise had however identified a number of points in which improvements could be made the chief of these being that the stocks of pure pesticide standards at several laboratories should be increased and that fresh pesticide calibration standards should be regularly prepared in order to improve quantitative results.Conclusions The programme of work to establish and maintain compar-ability of results for 17 determinands between laboratories Table 17. Results for biochemical oxygen demand as measured in the bottle? Laboratory 1 2 3 4 5 6 Mean Laboratory’s own standard Circulated standard (true concn. 4.40 mg 1-1) (dilution 1 in 50) (trueconcn. 1.80 mgl-l) (dilution 1 in 50) Mean concn.1 mg 1-1 4.10 4.18 4.38 4.04 4.20 4.32 Std. dev . I mg 1-1 0.19 0.18 0.87** 0.24 0.16 0.23 4.20 Max. possible biaslmg 1-1 -0.41 -0.32 -0.52 -0.50 -0.29 -0.21 t For symbols see footnote to Table 1. Mean concn.1 mg I-’ 1.71 1.69 1.88 1.88 1.86 1.79 Std. dev.1 mg 1-1 0.21 0.23 0.96** 0.19 0.16 0.17 1.80 Max.possible biaslmg 1-1 -0.21 -0.24 +0.64 +0.19 +O. 15 -0.11 River sample (no dilution) Sewage effluent (dilution 1 in 5) Mean Std. concn.1 dev.1 mg 1-1 mg 1-1 3.74 0.35 4.07 0.30 2.80 0.74* 3.28 0.50 3.73 0.16 4.10 0.18 3.62 Mean Std. concn.1 dev.1 mg 1-1 mg 1-1 5.54 0.34 3.80 0.20 4.28 0.42 4.80 0.32 5.54 0.66* 3.56 0.20 4.59 Table 18. Results for organochlorine pesticides? Synthetic mixture Sewage effluent River sample y-HCHI Laboratory pg 1-1 1 30 2 25 3 31 4 15R 5 29 6 29 Aldrinl 28 25 30 25 29 27 EL&’ Eldrinl 50 40 41 23R 40 44 pgl-’ (u-HCHI ELg1-I NSO NS 0.21 0.04R 0.17 0.14 y-HCHI 0.20 0.18 0.23 0.02R 0.32 0.21 pg I-’ Dieldrin1 0.16 0.12 0.21 ND 0.20 0.16 ELg 1-Aldrinl ND ND ND 0.01 ND ND I.18 I-’ (u-HCHI I%-’ NS NS 0.02 ND 0.10 0.03 y-HCHI ccgl-’ 0.05 0.04 0.03 0.04 0.14R 0.04 Dieldrin1 0.05 0.03 0.06 ND ND 0.04 pg I-’ Aldrinl ND ND 0.007 0.02 ND ND I.Lgl-‘ t For symbols see footnote to Table 1.0 NS no standard available at laboratory 14 ANALYST JANUARY 1984 VOL. 109 participating in the Severn Estuary Inputs Programme was arduous and for almost all determinands was much more difficult to achieve than at first expected. In the course of the collaborative work many laboratories found their existing analytical methods to be unsatisfactory and discarded them in favour of alternative procedures.The performance of some laboratory instrumentation was found to be unsatisfatory and had to be replaced and some new standardised methods and analytical techniques had to be developed to meet the demands of the programme. In general the exercises have shown that comparability of results can be achieved only by careful and constant attention to detail even for relatively simple determinands such as ammonia and total oxidised nitrogen. Once achieved com-parability cannot be taken for granted and it is essential that check exercises be regularly held with real samples to ensure that the required standard is maintained. The approach to achieving comparability employed in this work has subsequently been adopted by several other organ-isations for their own inter-laboratory comparability pro-grammes.This paper is published by kind permission of the Technical Working Party of the Severn Estuary Joint Committee (Chairman W. F. Lester Severn-Trent Water Authority), whose encouragement throughout the duration of the pro-gramme is gratefully acknowledged. Our appreciation is also due to Morlais Owens (Welsh Water Authority) for his advice, guidance and encouragement in this work. Appendix The period covered by the work described in this paper coincided with a consolidation of laboratory facilities within each Water Authority and a concomitant increase in the level of instrumentation available within most laboratories. This has inevitably meant that during this period changes in membership of the Chemists’ Sub-committee and indeed several changes in the number of laboratories participating in the programme have taken place.The complete list of Members and laboratories is given below together with an indication of the period of time over which they participated. This list is not the same order of laboratories as the coded listings given in the tables of results in the paper. Dr. K. C. Wheatstone Chairman (Severn-Trent Water Authority Birmingham); Mrs. E. M. Hewson Secretary, 1976-79 (Severn-Trent Water Authority Birmingham); Mr. M. M. Day Secretary 1979 onwards (Severn-Trent Water Authority Birmingham); Mr. M. R. Barker (Wessex Water Authority Divisional Laboratory Poole) (1975-76); Mr. J. G. Jones (Wessex Water Authority Bath); Dr. R. F. Mantoura (Institute for Marine Environmental Research) (1975-76); Mr. B. Milford (South West Water Authority, Exeter Laboratory) (1975); Dr. A. W. Morris (Institute for Marine Environmental Research) (1975-76); Dr. C. Pattinson (Welsh Water Authority Marine Laboratory Ponthir) (1978 onwards); Mr. A. Poole (Wessex Water Authority Divisonal Laboratory Saltford); Dr. E. Salt (Wessex Water Authority, Divisional Laboratory Bridgwater); Dr. W. Simpson (Welsh Water Authority Marine Laboratory Ponthir) (1977-78), Dr. J. Stoner (Welsh Water Authority Brecon); Mr. F. Sweeting (Wessex Water Authority Divisional Laboratory, Saltford); Mr. R. Toft (South West Water Authority Exeter Laboratory) (1976 onwards); Mr. J. E. Tomkin (Welsh Water Authority River Division Laboratory Caerleon; also rep-resenting Bridgend Hereford and Llanelli River Divisional Laboratories); Mr. C. Triner (Wessex Water Authority, Divisional Laboratory Poole); and Mr. K. Wagstaff (Severn-Trent Water Authority Regional Laboratory Malvern). 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. References Water Research Centre “Standard Analytical Samples,” (a) “Results of First Distribution,” WRA TM91 1974; ( b ) “Results of Second Distribution,” WRC TM94 1974; (c) “Results of Third Distribution,” WRC TM97 1974; ( d ) “Results of Fourth Distribution,” WRC TM98 1974; (e) “Results of Distributions 5 6 7 8 and 9,” WRC TR 65 1977; Water Research Centre Medmenham. United States Public Health Service Analytical Reference Service “A Series of Reports on the Results of Sample Distributions Between 1956 and 1971,” Robert A. Taft Sanitary Engineering Centre Cincinnati OH. “Analytical Quality Control for a Group of Collaborating Laboratories Inter-laboratory test,” TM96 Water Research Centre Medmenham 1974. Wilson A. L. Analyst 1979 104,273. Analytical Quality Control (Harmonised Monitoring) Com-mittee Analyst 1979 104 290. “Accuracy Required of Analytical Results for Water Quality Data Banks,” TR34 Water Research Centre Medmenham, 1976. McFarren E. F. Lishka R. J. and Parker J. H . Anal. Chem. 1970,42 358. “Water Quality 1977/78,” Severn-Trent Water Authority, Birmingham 1978 p. 145. Thompson K. C. Analyst 1978 103 1258. Thompson K. C. and Wagstaff K. Analyst 1980 105 641. Riley J . P. and Taylor D. Anal. Chirn. Acta 1968 40 479. Florence T. M. and Batley G. E . Talanta 1976 23 179. Abdullah M. I. El-Rayis D. A. and Riley J. P. Anal. Chirn. Acta 1976 84 363. Thompson K. C . Wagstaff K. and Wheatstone K. C., Analyst 1977 102 310. ‘ Bertenshaw M. P. Gelsthorpe D. and Wheatstone K. C., Analyst 1981 106 23. Bertenshaw M. P. Gelsthorpe D. and Wheatstone K. C., Analyst 1982 107 163. Morris A. W. personal communication. “Laboratory Procedure No. 15,” Water Pollution Research Laboratory Stevenage 1972. Department of the Environment “Analysis of Raw Potable and Waste Waters,” HM Stationery Office London 1972 p. 85. Paper A2155 Received March 8th 1982 Accepted September 15th 198