290 Analyst, April, 1979, Vol. 104, p p . 290-298 Accuracy of Determination of Chloride in River Waters: Analytical Quality Control in the Harmonised Monitoring Scheme Analytical Quality Control (Harmonised Monitoring) Committee" Water Research Centre, Henley Road, Medmenham, Marlow, Buckinghamshire, SL7 2HD The Department of the Environment, in collaboration with the Regional Water Authorities, has initiated a Scheme for the Harmonised Monitoring of the Quality of Inland Fresh Waters in England and Wales. The Scottish Development Department has introduced a similar scheme in Scotland in collaboration with the Scottish River Purification Boards. To achieve the required comparability of results from all laboratories involved, each labora- tory takes part in an analytical quality control (AQC) scheme; this work is co-ordinated by the Water Research Centre.The general approach adopted to AQC has been described, and this paper presents the tests made and results obtained in the determination of chloride in river waters. Broadly, each of the ten participating laboratories achieved total errors not greater than *ZO% of the chloride concentration for concentrations greater than 5 mg 1-1 of chloride. Keywords : River-water analysis ; chloride determination ; accuvacy of results ; inter-laboratory comparability ; analytical quality control The scheme for the Harmonised Monitoring of the Quality of Inland Fresh Water has been described recently in detail.1 It is intended to provide objective data on river water quality so that accurate assessments can be made of long-term trends in the qualities of rivers and of the amount of material discharged by them to the sea.The Scheme complements moni- toring carried out for regional or local purposes and one of its essential aims is to achieve comparability of the results from all participating laboratories. To that end, special investi- gations have been made to establish suitable locations for sampling and to define the necessary sampling frequencies. Sampling procedures have been harmonised, and to ensure that subsequent analyses do not introduce unacceptably large errors, each participating laboratory carries out a specially designed programme of tests to ensure that their analytical results are of adequate accuracy for the Scheme. The Water Research Centre (WRC) is under contract to the Department of the Environment to advise on and co-ordinate this analytical quality- control (AQC) programme.The need for, and details of an approach to, a planned AQC system for this and similar schemes have been discussed in some detail elsewhere in this issue.2 In view of the growing interest in achieving comparable results from each of a number of laboratories, it was thought useful to describe the AQC work for the Harmonised Monitoring Scheme and to present the results obtained for different determinands. This paper considers the determination of chloride and subsequent papers will deal with other determinands of general importance in rivers. Chloride was chosen for the first study of the Committee because serious problems in achieving the required accuracy were not expected and it was considered, therefore, that this would facilitate the introduction of the AQC programme.The work presented in this paper was originally published as a WRC Technical R e p ~ r t . ~ Organisation of the Work A Committee was formed to plan the collaborative work, and has representatives? from the Department of the Environment, Scottish Development Department, each Regional * Communications concerning this paper should be addressed to D. J. Dewey, at the Water Research t The names of representatives a t the time the work reported here was carried out are given in the Centre. Appendix.ANALYTICAL QUALITY CONTROL (HARMONISED MONITORING) COMMITTEE 291 Water Authority (RWA), Scottish River Purification Boards and the WRC. This Com- mittee decided to adopt the approach to the AQC described elsewhere in this issue,2 each determinand being studied in two phases.Phase (i). One laboratory in each of the ten RWAs and one in Scotland participated, the WRC acting as the co-ordinating laboratory2 ; eleven laboratories were generally involved. Phase (ii). After satisfactory results have been obtained in Phase (i), those laboratories act as co-ordinators of tests within RWAs and Scotland. Certain RWAs are not involved in this phase because all analyses for Harmonised Monitoring are made by one laboratory. This paper deals only with Phase (i). Required Analytical Accuracy The following requirements were agreed by the Committee to represent the targets a t which to aim2; maximum tolerable bias, 10% of the chloride concentration or 0.5 mg 1-1 of chloride, whichever is the greater; and maximum tolerable total standard deviation, 5% of the chloride concentration or 0.25 mg 1-1 of chloride, whichever is the greater.Analytical Quality Control The approach followed was exactly as presented previously2; no attempt is made here, therefore, to explain the reasons underlying the various activities described below. The South West Water Authority was unable to participate in the tests reported here though it has now undertaken the work. The participating laboratories were : Anglian Water Authority, Welland and Nene River Division Laboratory, Oundle ; Northumbrian Water Authority, Headquarters Laboratory, Gosforth; North West Water Authority, Mersey and Weaver River Unit Laboratory, Warrington ; Severn-Trent Water Authority, Regional Laboratory, Finham ; Southern Water Authority, Resource Planning Laboratory, Winchester ; Thames Water Authority, Thames Conservancy Division Laboratory, Reading; Welsh Water Authority, Dee and Clwyd River Division Laboratory, Chester ; Wessex Water Authority, Bristol Avon Divisional Laboratory, Saltford; Yorkshire Water Authority, Headquarters Laboratory, Leeds ; and Forth River Purification Board, Headquarters Laboratory, Edin- burgh.The sequence of participating laboratories in the above list does not relate to the order of numbering of laboratories in the Tables. Choice of Analytical Methods Participating laboratories each chose a method they thought capable of achieving the required accuracy.The methods involved were (i) manual Mohr t i t r a t i ~ n , ~ (ii) manual mercury(I1) nitrate - diphenylcarbazone titration4 and (iii) semi-automatic continuous flow spectrophotometric methods based on the procedure in reference 5 [mercury(II) thiocyanate - iron(II1) salt]. One of the laboratories initially selected a commercially available instrument for the coulometric titration of chloride. Preliminary tests showed, however, that the discrimina- tion with which the instrument response could be read did not allow achievement of the required precision at low concentrations (below 10 mg 1-l). This procedure was, therefore, abandoned, and the laboratory adopted method (iii). Within-laboratory Precision and Recovery Tests Following any preliminary tests considered necessary by the laboratories, each then carried out the same programme of tests to assess the precision of, and certain sources of bias in, its results.Laboratories using method (iii) checked that their calibration graphs were linear. All laboratories, on each of 10 days, carried out in random order a batch of analyses consisting of two blank determinations and two portions of each of the following solutions, two standard solutions, a river water and the same river water after addition of a known amount of chloride. Each laboratory collected its own sample of river water from a local source and the same sample was used throughout the tests. The amount of chloride added to the spiked sample varied from laboratory to laboratory, but was in the range 0.2-0.5 Cu, where C, is the upper concentration limit of a laboratory’s method.Each laboratory292 ANALYTICAL QUALITY CONTROL (HARMONISED MONITORING COMMITTEE) : Analyst, Vol. 104 prepared its own standard solutions from its own stock standard solution just before each batch of analyses. The concentrations of the two standards used for the tests were usually 0.1 C, and 0.9 C,. On completion of the tests, each laboratory analysed its results statistically to obtain estimates of the within-batch (sw), between-batch (sb) and total (st) standard deviations,6 where st = d ( s i + sg). The values of st for the two standards, the river water and the spiked river water were compared with the appropriate target value using the F-test,2 and were accepted as satisfactory provided st was not significantly greater (9 = 0.05) than the target.2 The results from these tests are summarised in Table I.For the particular solutions used the chloride concentrations were such that the target for precision was a relative total standard deviation not greater than 5%. This value was exceeded in only three instances, Laboratories 1 and 7 river water and Laboratory 6 standard solution (20 mg 1-1 of chloride). However, in those instances the observed relative standard devi- ations were only marginally, and not significantly, greater than 5%. The results were, therefore, considered acceptable. It is further of interest to note that estimates of precision varied considerably between laboratories, even for the analysis of standard solutions of similar concentration.Of course some of these differences reflect random uncertainty in the estimates of precision. However, some differences between estimates are sufficiently large to indicate real differences in the precision of results obtained by individual laboratories. This reflects differences in the choice of method used and its application in individual laboratories. However, these differences were not important in the context of this exercise and, it is emphasised again, results were within the targets set other than the exceptions discussed above. Each laboratory also calculated the recovery, R, of chloride from the spiked river water, where R = 100 (cs - C,)/A, and c, and c, are the mean concentrations found for the spiked and unspiked river water, respectively, and A is the equivalent concentration of chloride added to the spiked sample, allowance being made for the slight dilution of the sample caused by the addition of a standard solution of chloride.It was agreed that recoveries would be considered acceptable if R was not significantly worse (t-test, p = 0.05) than the closer of the two values, 95 and 105%. These results are also summarised in Table I, which shows recoveries between 95 and 104% with an over-all mean of 99.6%. None of the individual recoveries were significantly worse than 95 or 105% and the results were, therefore, considered acceptable. The satisfactory completion of these tests in all laboratories indicated that within-laboratory precision was adequate and the next stage of the AQC work was started.Accuracy of Standard Solutions To ensure that differences in the chloride concentrations of laboratories’ standard solutions did not cause important between-laboratory bias, the following tests were made. The WRC prepared a standard solution of sodium chloride (1 000 mg 1-1 of chloride) and portions were sent to all laboratories. Each laboratory then accurately diluted portions of the WRC solution and its own standard solution to nominally the same concentration, which was usually at or near C, so as to achieve the smallest relative standard deviation. Sufficient replicate analyses of both the diluted standards were made in one batch of analyses so that if there were a difference of 2% in their true concentrations, a statistically significant difference (95% confidence level) between the two means would be found.The required number of analyses on each standard was calculated statistically using the estimates of within-batch standard deviation obtained in the preceding stage of the work.2 On completion of the tests, each laboratory compared the mean values obtained for each standard (t-test, p = 0.05) and the results are summarised in Table 11. None of the labora- tories’ standards differed from the WRC by as much as 2%, the maximum observed difference was 0.4% and the over-all mean difference was 0.1%. These results were consideredsatis- factory. Each laboratory then set up a preliminary statistical quality control chart2 based on the analysis of a standard solution in each subsequent batch of analyses. These charts are intended to aid the continuing, long-term assessnient of accuracy in each laboratory and are not further discussed here.TABLE I RESULTS FROM WITHIN-LABORATORY PRECISION TESTS Laboratory No.and analytical method* Sample Standard solution 1 . Standard solution 2 . River water . . . Spiked river water . Parameter Chloride concentration/mg 1-I Relative total standard deviation,? yo Total standard deviationtlmg 1-l Chloride concentration/mg 1-I Relative total standard deviation,? % Total standard deviationt/mg 1-l Chloride concentration/mg 1-I Relative total standard deviation,? yo Total standard deviationtlmg 1-l Chloride concentration/mg 1-l Relative total standard deviation,? yo Total standard deviationtlmg 1-' Mean recovery of chloride from spiked river water,$ % r 1, M 2, M 2.6 (0.0) 0.65 (0.0) 0.26 0.20 0.52 0.49 6.1 54 5.1 1.3 0.31 0.68 25 50 200 250 106 99 0.36 0.59 0.38 0.58 100.0 f 0.1 101.5 f 1.1 3, MN 20 1.5 0.29 180 0.23 0.42 47 1.2 0.55 139 0.50 0.70 99.2 f 0.5 4, MN 5, SA 8 20 1.5 3.6 0.12 0.71 0.18 0.89 0.14 1.6 0.57 2.2 0.09 0.94 0.11 1.9 0.07 2.4 80 180 16 42 65 124 99.8 f 0.1 94.7 f 1.9 6, SA 20 6.0 1.2 1.2 2.1 40 180 2.2 0.89 1.2 1.6 104.2 f 1.0 137 7, SA 8, SA 9, SA 30 5 5 3.7 3.8 3.4 1.1 0.19 0.17 270 45 45 1 .o 0.33 0.93 2.7 0.15 0.42 5.4 0.92 3.1 0.97 0.24 0.71 18 26 23 111 45 36 1.4 0.42 2.8 1.5 0.19 1 .o 98.6 f 1.0 96.4 f 1.2 102.6 f 3.1 * M = Mohr titration'; MN = mercury(I1) nitrate titratiod; SA = semi-automatic spectrophotometric method based on reference 5 (see Choice of Analytical Methods). t The total standard deviations have between 9 and 19 effective degrees of freedom.' $ 95% confidence limits of the mean recovery are also given.10, SA' 4.2 0.84 20 180 0.89 1.6 53 2.9 1.5 144 0.96 1.4 99.1 f 0.8294 ANALYTICAL QUALITY CONTROL (HARMONISED MONITORING COMMITTEE) : Analyst, VoZ. 104 TABLE I1 COMPARISON OF PARTICIPANTS' AND WRC STANDARD SOLUTIONS Mean difference between laboratory Laboratory and WRC standards,* % 1 -0.16 f 0.05 2 -0.05 f 0.29 3 0.33 f 0.96 4 0.35 f 0.17 5 0.14 & 0.99 6 0.09 f 0.83 7 0.30 f 0.57 8 0.07 f 1.00 9 0.07 f 0.58 10 0.00 f 0.55 Mean 0.11 * The 95% confidence limits of the mean difference are also given. Tests of Between-laboratory Bias To complete this initial phase of AQC, direct checks of any between-laboratory bias were made as follows.The WRC prepared and distributed portions of a standard solution (46.0 mg 1-1 of chloride) and samples of two different river waters to all laboratories, each participant being given only a broad indication of the concentration of these solutions. One of the river samples was a hard, lowland water, and the other was of the soft, moorland type. Each laboratory analysed each solution once on each of 5 days, and then calculated the mean of the five results and its 90% confidence limits (obtained from the five results only) for each solution. These results were then returned to the WRC for examination for any evidence of bias. For each solution, tests were made to decide if the mean result from any laboratory could be regarded as a statistical outlier.' No such outliers were indicated ( p = 0.05) for the two river waters.For the standard solution, the result from Laboratory 5 just achieved signifi- cance, but neglect of this result only changed the over-all mean from 46.0 to 45.7 mg 1-1 of chloride. In view of this small difference and because no reasons for regarding the result as suspect were known, it was thought better not to reject the result of Laboratory 5. The results in Table I11 show that the confidence intervals of the mean results of different laboratories for a given solution do not all overlap, i.e. , a certain amount of between-laboratory bias exists. To assess whether or not the bias of any laboratory exceeded the target value of loyo, the following procedure was used.2 Denoting the mean and its 90% confidence limits for a given solution and the ith laboratory by Xi hi, the upper limit (95% confidence) for the bias of the laboratory, U , was calculated as follows: if Zi > T , 100 (Zt + hi - T ) T U = or, if Zi < T, 100 (Xi - hi - T ) T U = where T is the true concentration of chloride in the standard solution or the over-all mean concentration of chloride in the river waters. For this purpose, the over-all means for the river waters were calculated from the means of individual laboratories, no weighting for the precisions of the means being used.The values for U are also given in Table 111, which shows that they were usually substantially less than the maximum tolerable bias of 10%. In only two instances did U exceed the target and then only marginally (Laboratories 6 and 8 for river water B).However, the mean results of these two laboratories were within 5% of the over-all mean for that sample and their values of U for the other two solutions were much smaller than 10%. It was decided, therefore, that the results from all laboratories could be regarded as satisfactory.RESULTS FRO1 TABLE I11 TESTS OF BETWEE ABORATORY BIAS Mean of all Sample Parameter 1, M 2, M 3, MN 4, MN 5, SA 6, SA 7, SA 8, SA 9, SA 10, SA tories Laboratory No. and analytical method* c r , labora- Standard chloride solution, 46.0 mg I-' Mean chloride contenttlmg 1-l 46.0 f 0.1 45.0 & 0.3 46.0 f 0.4 45.9 f 0.1 48.0 + 0.9 46.0 f 1.2 45.8 + 1.6 45.0 f 0.2 46.1 + 0.4 46.2 + 0.4 46.0 River water A .. Mean chloride contentt/mg 1-l 97.9 f 0.2 97.4 f 0.5 97.0 f 0.3 96.7 f 0.2 94.6 f 1.0 95.7 f 1.5 97.2 & 1.6 95.2 f 0.3 97.1 f 0.2 8 96.57 River water B . . Mean chloride contentt/mgl-l 8.52 f 0.17 8.28 & 0.13 8.80 0.23 8.39 & 0.11 8.90 f 0.40 8.24 f 0.60 8.60 -+ 0.52 9.02 f 0.43 8.48 f 0.38 8.60 f 0.52 8.587 Upper limit for bias,$ % f0.2 - 2.7 fO.9 -0.5 + 6.3 12.7 - 3.8 -2.6 +1.2 +1.4 - Upper limit for bias,$ yo + 1.6 +1.4 +0.8 + 0.4 - 3.0 - 2.4 + 2.4 -1.6 + 0.8 § - Upper limit for bias,$ % - 2.7 - 5.0 + 5.2 - 3.5 +8.4 -11.0 +6.2 +10.1 - 5.6 +6.3 - Mean upper limit for bias for the three above solutions, % - .0.4 - 2.1 + 2.3 - 1.2 + 3.9 - ,5.4 +1.6 + 2.0 - +3.9 - 1.2 * M = Mohr titration'; MN = mercury(I1) nitrate titration4; SA = semi-automatic spectrophotometric method based on reference 5 (see Choice of Analytical Methods).t The goo/, confidence limits are also given for each mean. $ See text for method of calculating the upper limit (95% confidence) for bias. 5 This laboratory originally conducted the work on another semi-automatic system and repeated all the tests on changing to the present system. Insufficient sample of river water A remained for tests fl The values obtained by the WRC for river waters A and B were 96.2 and 8.48 mg 1-l of chloride, respectively. with the new system. t9 W 01296 ANALYTICAL QUALITY CONTROL (HARMONISED MONITORING COMMITTEE) : Analyst, VoZ. 104 At this point the objectives of the initial AQC programme had been achieved, and a similar programme was started for the next determinand of interest, ammoniacal nitrogen.Routine AQC To attempt to ensure that the required accuracy of results for chloride is maintained, AQC is now an integral part of the routine analyses for the Harmonised Monitoring Scheme. As mentioned above, prime reliance for this purpose is placed on within-laboratory AQC using statistical quality control charts. However, to obtain direct checks of between- laboratory bias, portions of a river-water sample are distributed occasionally by the WRC to all laboratories. The first four such tests (over a period of approximately 2 years) are of value in indicating the efficiency of the AQC work, and are, therefore, summarised in Table IV. Each of these tests was carried out as described in the previous section except that, for the fourth test, the five replicate analyses were all made in one batch of analyses.Table IV shows that the upper limit (95% confidence level) for bias is usually less than 10% and that the mean of the individual upper limits is less than 10% for each laboratory. Overall, therefore, it appears that reasonably satisfactory accuracy has been maintained. However, the need for continuing care and emphasis on AQC is indicated by the fact that certain laboratories appear to be rather prone to bias close to or slightly exceeding the target. Therefore, of the 40 sample - laboratory combinations, eight provide upper limits for bias of greater than 10%; of these eight instances, Laboratory 10 gave three, and Laboratories 5 and 9 each gave two.Each laboratory calculated the standard deviation from the five results for each sample. Of the first three tests (for which the relative total standard deviation could be calculated), no laboratory obtained any values significantly greater (j5 = 0.05) than the target of 5% and only four of the values were greater than 5%. It appears, therefore, that precision was maintained reasonably well. TABLE IV RESULTS FROM ROUTINE TESTS OF BETWEEN-LABORATORY BIAS River water C River water D River water E River water F (September 1975) (April 1976) (March 1977) (December 1977) 7-p- Mean Mean Mean Mean chloride Upper limit chloride Upper limit chloride Upper limit chloride Upper limit Mean upper content/ for bias, content/ for bias, content for bias, content/ for bias, limit for Laboratory mg 1-1 % mg 1-1 % mg 1-l % mg 1-I % bias, yo 1 2 3 4 5 6 7 A 9 10 Meant 17.5 + 3.5 58.0 + 2.6 29.8 + 5.3 18.1 + 5.7 56.8 - 1.2 29.4 + 3.3 17.6 + 3.9 56.2 - 2.8 27.4 - 6.3 16.5 - 4.9 56.0 - 2.1 29.0 + 1.6 15.6 - 14.4 59.0 + 4.2 26.0 - 12.5 15.4 -15.5 54.4 - 8.5 28.8 + 5.8 11.4 + 4.1 56.6 - 3.3 28.8 + 3.4 16.8 - 6.4 57.0 + 0.2 29.6 + 3.6 18.7 +11.1 58.8 + 3.9 31.6 +12.2 65.0 +16.8 26.4 -16.9 - 57.0 - 28.7 18.6 17.2 - +11.0 37.4 +2.5 37.1 + 1.7 34.6 - 7.2 35.6 -3.5 34.0* -7.3 37.0* +0.9 40.0* + 9.1 36.0* -1.8 37.02 + 0.9 36.8 + 1.5 36.7 - + 3.5 +2.4 - 3.1 - 2.2 - 7.5 -4.3 + 3.3 - 1.2 + 7.0 + 3.1 - * The five results were reported as identical.t No laboratory’s results were rejected in calculating the over-all mean for samples C, E and F; the result of Laboratory 10 was rejected for sample D.Discussion The detailed results provide evidence on a number of points relevant to the design of AQC schemes for a group of laboratories, e.g., the dependence of standard deviation on the con- centration of the determinand and the relative importance of different sources of error. However, discussion of such aspects is best deferred until the results for other determinands and analytical techniques have been published. One point is worth a brief mention here, namely the suitability of the accuracy targets for chloride. The targets for systematic and random errors imply that the tolerable total error (95% confidence level) of individual analytical results is 20% of the chloride concentration or 1 mg 1-1 of chloride, whichever is the greater.The former target may well be thought rather lax, particularly as chloride is usually regarded as a determinand for which good accuracy is readily achieved. It is interesting, therefore, to consider the consequences ofApril, 1979 ACCURACY OF DETERMINATION OF CHLORIDE I N RIVER WATERS 297 reducing the maximum tolerable errors to half the values used in this work. If the experi- mental estimates of errors in this paper were compared with these reduced targets in the manner described above, the following targets would not be achieved : precision tests from Laboratories 1, 6 and 7 ; recovery tests from Laboratories 5 and 6; accuracy of standard solutions from Laboratories 3, 5 and 8; and between-laboratory bias from Laboratories 3, 5, 6, 7, 8, 9 and 10.In contrast, with the targets adopted in this work, only two instances occurred where the targets were exceeded, i.e., Laboratories 6 and 8 in the tests of between-laboratory bias. This analysis suggests that substantially greater effort would be required in most laboratories to achieve the smaller targets routinely. Hence, the idea that better accuracy than that aimed for in this work can readily be achieved would seem to be incorrect even for chloride, which is normally regarded as being determined with good accuracy and precision. The approach to AQC described in this paper aims progressively to identify and, if necessary, control particular sources of error so that a permanently sound basis is established for routine achievement of the required accuracy.2 The individual tests are, therefore, designed to provide many opportunities for unsuspectedly large errors to reveal themselves.Of all the tests described above, none gave positive evidence that a target was not achieved, although there were several instances where the results indicated that a target could have been exceeded, e.g., the bias of Laboratories 6 and 8 for river water B could have exceeded the target of 10% (see Table 111). Such uncertainties are bound to arise as a result of random errors, and in this situation, a partially subjective judgement on the basis of the complete set of tests is necessary. In this way, it is considered reasonable to conclude that all laboratories achieved the required accuracy in the preliminary AQC work.The results of subsequent tests (see Table IV) indicate that this accuracy has been maintained over a period of almost 3 years. So far as is known, this is the first time that such an achievement has been reported for the analysis of river waters. The success of the work is attributed to two main factors: (i) the suitability of the analytical methods adopted by the laboratories; and (ii) the sequential approach followed in the AQC programme. The latter involves a relatively large amount of work in each laboratory and a rather long period is necessary to complete all tests. These are disadvantages in the approach, but they are counterbalanced by its ability to ensure that the required accuracy is achieved. Most of the participating laboratories had no previous experience of this type of work, but all agreed that the approach should be retained for tests on other determinands and it is hoped to report the results in subsequent papers.Conclusions The targets chosen for accuracy and precision of results appear to be suitable for the Harmonised Monitoring Scheme and capable of achievement for the river waters tested. All participants achieved the required accuracy during the tests. To ensure that the position is maintained, continuing care is needed, and subsequent analytical quality control will, in addition to normal precautions, be based mainly on the use of quality-control charts and the analysis of samples distributed at regular intervals by the WRC. The work has confirmed that each of three types of method [i.e., (i), (ii) and (iii) in the section Choice of Analytical Methods] are capable of achieving the target accuracy.Valuable experience of this collaborative work has been gained. The participating laboratories have commented favourably on the approach adopted and on the co-ordination of the work by the WRC. This approach is now being applied to successive studies of other determinands, the results for which will be reported in subsequent papers. For those Authorities with more than one participating laboratory, the tests within each area (Phase ii) need to be completed before the over-all situation can be regarded as satis- factory. The results in this paper suggest that no undue problems are to be expected in this second stage of testing. The work has demonstrated a procedure for ensuring permanent comparability of results from many laboratories, Already some of the Authorities have taken advantage of this by extending the procedure to laboratories not directly concerned with the Harmonised Moni- toring Scheme.Such extensions are recommended.298 ANALYTICAL QUALITY CONTROL (HARMONISED MONITORING) COMMITTEE Appendix The following were members of the Analytical Quality Control (Harmonised Monitoring) Committee at the time of the initial work on chloride: Mr. A. L. Wilson, Chairman (Water Research Centre), Dr. E. A. Simpson, Secretary (Department of the Environment), Mr. M. J. Beard (Southern Water Authority), Mr. J. R. Borland (Welsh Water Authority), Mr. R. V. Cheeseman (Water Research Centre), Mr. N. Croft (Yorkshire Water Authority), Dr. B. T. Croll (Anglian Water Authority), Mr. D. V. Hopkin (Thames Water Authority), Mr. J. G. Jones (Wessex Water Authority), Mr. P. Kingslan (Department of the Environ- ment), Mr. J. C. Lambie (Scottish Development Department), Mr. B. Milford (South West Water Authority), Mr. P. Morries (North West Water Authority), Mr. B. D. Ravenscroft (Northumbrian Water Authority), Mr. D. Rodda (Water Data Unit), Mr. J. E. Saunders (Welsh Office), Dr. K. C. Wheatstone (Severn-Trent Water Authority), and Mr. T. William- son (Forth River Purification Board). 1 . 2. 3. 4. 5. 6 . 7 . References Simpson, E. A., J . Inst. Wat. Engrs Scient., 1978, 32, 45. Wilson, A. L., Analyst, 1979, 104, 273. Analytical Quality Control (Harmonised Monitoring) Committee, “Accuracy of Determination of Chloride in River Waters,” Technical Report TR 27, Water Research Centre, Medmenham, Bucks., 1976. Department of the Environment, “Analysis of Raw, Potable and Waste Waters,” HM Stationery Office, London, 1972, pp. 73-76. Zall, D. M., Fisher, D., and Garner, M. D., Analyt. Chew., 1956, 28, 1665. Wilson, A. L., Talanta, 1970, 17, 31. Davies, 0. L., and Goldsmith, P. L., Editors, “Statistical Methods in Research and Production,” Fourth Revised Edition, Oliver and Boyd, Edinburgh, 1972, pp. 49-50. Received September l l t h , 1978 Accepted November 2nd, 1978