Analytische Qualitätssicherung Baden-Württemberg

Transcript

1 Analytische Analytische Qualitätssicherung Baden-Württemberg Proficiency Test 9/15 - WFD PAH with suspended solilds in surface water - anthracene, fluoranthene, naphtalene, benzo[a]pyrene, benzo[b]fluoranthene, benzo[k]fluoranthen, benzo[g,h,i]perylene, indeno[1,2,3-cd]pyrene provided by AQS Baden-Württemberg at Institute for Sanitary Engineering, Water Quality and Solid Waste Management, University of Stuttgart Bandtäle 2, Stuttgart-Büsnau, Germany Stuttgart, April 2016

2 Responsibilities: Scientific director AQS: Dr.-Ing. Dipl.-Chem. Michael Koch PT manager: Dr.-Ing. Frank Baumeister Sample preparation: Matthias Mischo AQS Baden-Württemberg at Institute of Sanitary Engineering, Water Quality and Solid Waste Management at University of Stuttgart Bandtäle Stuttgart-Büsnau Germany Tel.: +49 (0)711 / Fax: +49 (0)711 / info@aqsbw.de

3 PT 9/15 I LIST OF CONTENTS 2. GENERAL PT DESIGN SAMPLE PREPARATION SAMPLE DISTRIBUTION ANALYTICAL METHODS SUBMISSION OF RESULTS EVALUATION PROCEDURE ASSESSMENT EVALUATION EXPLANATION OF APPENDIX A EXPLANATION OF APPENDIX B EXPLANATION OF APPENDIX C EXPLANATION OF APPENDICES D AND E MEASUREMENT UNCERTAINTY TRACEABLE REFERENCE VALUES INTERNET Appendix A Anthracene... A-1 Fluoranthene... A-8 Naphtalene... A-14 Benzo[a]pyrene... A-20 Benzo[b]fluoranthene... A-26 Benzo[k]fluoranthene... A-32 Benzo[ghi]perylene... A-38 Indeno[1,2,3-cd]pyrene... A-44 Appendix B

4 PT 9/15 II Appendix C Anthracene... C-1 Fluoranthene... C-13 Naphtalene... C-25 Benzo[a]pyrene... C-37 Benzo[b]fluoranthene... C-49 Benzo[k]fluoranthene... C-61 Benzo[ghi]perylene... C-73 Indeno[1,2,3-cd]pyrene... C-85 Appendix D Anthracene... D-1 Fluoranthene... D-8 Naphtalene... D-15 Benzo[a]pyrene... D-22 Benzo[b]fluoranthene... D-29 Benzo[k]fluoranthene... D-36 Benzo[ghi]perylene... D-43 Indeno[1,2,3-cd]pyrene... D-50 Appendix E Anthracene... E-1 Fluoranthene... E-13 Naphtalene... E-25 Benzo[a]pyrene... E-37 Benzo[b]fluoranthene... E-49 Benzo[k]fluoranthene... E-61 Benzo[ghi]perylene... E-73 Indeno[1,2,3-cd]pyrene... E-85

5 PT 9/15 page 1 1. General This PT was provided by AQS Baden-Württemberg to check the chemical analysis of priority and other substances in the context of chemical monitoring for the European Water Framework Directive, especially for compliance of surface waters with Environmental Quality Standards (EQS). The EQS directive 2008/105/EG requires that these substances are determined as total concentrations in the whole water sample, including the content at and in the suspended solids.the PT samples contained suspended solids, because it is well- known that a high ration of the PAH can be found at and in these suspendes solids. In this round the following parameters in surface water with suspended solids were to be determined: Anthracene Fluoranthene Naphthalene Benzo[a]pyrene Benzo[b]fluoranthene Benzo[k]fluoranthene Benzo[g,h,i]perylene Indeno[1,2,3-cd]pyrene The PT met the requirements of the water framework directive and the related environmental quality standards for the determination of priority substances in surface water. The PT was executed and evaluated according to the requirements of DIN A45 or ISO/TS and ISO PT design Each participant received the following samples: 3 different surface water test samples at three concentration levels of the analytes in 1000 ml amber ground glass bottles. Two bottles of each sample were distributed for duplicate measurement. The samples were preserved by cooling. 3 different concentration levels/batches were produced. All participants received the same concentration levels. 3. Sample preparation The samples were based on a real surface water. The suspended solids were isolated with a high-speed centrifuge, freeze dried, grinded to a grain size of below 40 µm and analysed with GC-MS after Soxhlett extraction. For further use a stock solution of the suspended solids was produced. For the production of the batches/concentration levels, the surface water was filtered by using 5 µm and 1 µm filter cartridges to eliminate course particles. To reduce germs, the surface water was irradiated with ultraviolet light. For preparation of the samples, each bottle was filled with surface water, was spiked with the stock solution of the suspended solids and with a solution of the substances

6 PT 9/15 page 2 dissolved in a mixture of water and acetonitrile. The concentration of the suspended solids was ca. 20 mg/l per bottle. The samples contained ca. 0,2 µl/l acetonitrile. The concentrations of the analytes of the spiked samples were chosen according to the Environmental Quality Standard (Directive 2008/105/EG on environmental quality standards in the field of water policy). The samples were cooled directly after preparation and dispatched with freezer packs added to the packages. 4. Sample distribution Because of a damage of the centrifuge, the PT had to be postponed from November 2015 to January The samples were dispatched on 18 th January 2016 by the express service TNT. The participants were requested to start with the analysis one day after receipt of the samples at the latest.. 5. Analytical methods The participants were free to choose a suitable method, but following limits of quantification were required. Parameter Limit of quantification [µg/l] Anthracene 0,03 Fluoranthene 0,002 Naphthalene 0,03 Benzo(a)pyrene 0,0003 Benzo(b)fluoranthene 0,005 Benzo(k)fluoranthene 0,005 Benzo(g,h,i)perylene 0,0025 Indeno(1,2,3-cd)pyrene 0,0025 The limit originates from the requirements of article 4 of the commission directive 2009/90/EC of the European communities for the technical specifications for chemical analysis and monitoring of water status in the context of the water framework directive. The limit corresponds to 30% of the environmental standards of the directive 2013/39/EU on environmental quality standards in the field of water policy. For benzo(a)pyrene the limit was increased to reach a range which is currently achievable by chemical analysis. The limit for naphthalene was lowered since the AA-EQS value is unrealisticly high. The samples had to be analysed twice by the complete method (sample preparation and measurement). The participants were asked to report the results in the whole water sample as average values in µg/l with four significant digits. 6. Submission of results The deadline for the submission of results was on 12 th February 2016.

7 PT 9/15 page 3 7. Evaluation procedure The statistical evaluation was executed according to ISO with the combined estimator Hampel/Q-method, a method of robust statistics. The assigned value x pt, derived from the weighings of the spiked samples and, if necessary, the matrix content 1,2 was used for the assessment of the single values preferably. For the parameters anthracene, fluoranthene, naphtalene the matrix content was considered for the calculation of the assigned values, but not for the other parameter (see chapter Traceable reference values ). The standard deviation for proficiency assessment σ pt was calculated in accordance with the European QA/QC Directive: σ pt = 0,25 * x pt. A z-score was calculated for each measurement result derived from the assigned value x pt and the standard deviation for proficiency assessment: z score x i x pt pt The z-score was modified to a z U -score with a correction factor for proficiency assessment (as described in the standards mentioned above). The assessment of the results was as follows: z U 2,0 satisfactory 2,0 < z U < 3,0 questionable z U 3,0 unsatisfactory It is well known that PAH sorb very strongly to suspended solids. Hence, many laboratories reported too low values. Nevertheless we decided to use the described reference value as assigned value because of the following reasons: The Water Framework Directive and the respective daughter directives expect the measurement of the actual, true contents with a measurement uncertainty of maximum 50%. We are sure that the amounts of the PAH in the sample bottles are reflected correctly by the reference value. In consequence we had to use the reference values as assigned values. Just one laboratory used the method DIN EN 16691:2015 with solid-phase extraction (SPE) with SPE-disks combined with GC-MS, the method which was especially developed for this kind of analyses. The results of this laboratory in many cases were very close to the reference values. 1 Rienitz, O., Schiel, D., Güttler, B., Koch, M., Borchers, U.: A convenient and economic approach to achieve SI-traceable reference values to be used in drinking-water interlaboratory comparisons. Accred Qual Assur (2007) 12: Koch, M., Baumeister, F.: Traceable reference values for routine drinking water proficiency testing: first experiences. Accred Qual Assur (2008) 13:

8 PT 9/15 page 4 For your information only we also assessed your results by using consensus values derived from the Hampel-estimator (see appendices D and E) showing the comparability with the other results but not the trueness of the results. 8. Assessment There was no overall assessment of the proficiency test round, but the single values were assessed. The following values have been assessed as unsatisfactory : 1) Values, which were indicated with lower than limit of quantification, 2) Values, resulting from a subcontracting, 3) Values, which were submitted after the deadline of submission of results. 9. Evaluation Number of participants: 60 1 laboratory did not report any results. Number of reported values: 1365 Number of accepted values: 711 (52,1%) In the following figure the percentage of satisfactory, questionable, unsatisfactory and not submitted results are illustrated.

9 PT 9/15 page Explanation of Appendix A Appendix A contains for each parameter - parameter tables - a figure of participants means versus the spiked amount for the determination of the recovery rate and the matrix content - a figure of the relative standard deviations versus the concentrations - a figure of the tolerance limits in the PT versus the concentrations - the frequency of application of analytical methods - the method specific evaluation - a comparison of mean and reference values for each concentration level - a comparison of the relative standard deviations of the different methods - the statistical characteristics of the method specific evaluation - a tabular comparison of the means with the reference values and their uncertainties Parameter tables In these tables the following values for each concentration level are listed: assigned value expanded uncertainty of the assigned value in % (see chapter Traceable reference values ) standard deviation, calculated using robust statistical method (Q-method) standard deviation for proficiency assessment in µg/l for the calculation of z- scores (25 % of the assigned value) rel. standard deviation for proficiency assessment in % tolerance limits above and below in µg/l and % number of values in this level number of not satisfactory values ( questionable or unsatisfactory ) below and above the assigned value and the percentage of these values in total. Determination of recovery rate In the diagrams of the assigned values versus the spiked amount of analyte a linear regression line was calculated using a generalized least square regression which takes into account the uncertainties of the values in both directions. From these values the recovery rate for each parameter was determined (see diagrams). The slope of the line indicates the average recovery rate. The diagrams also contain the expanded uncertainty (k=2) of the mass values and the assigned values. Relative standard deviations and tolerance limits The diagrams for the rel. standard deviation vs. the assigned value show the values compared to the fixed standard deviation for proficiency assessment (horizontal line at 25%) and the concentration dependence. The diagrams with the tolerance limits show the limits in percent.

10 PT 9/15 page 6 Method specific evaluation For each parameter the methods used by the participants are shown in a diagram. In a second diagram for each method with a frequency of more than 5 %, values are sorted in 5 categories: too low results with z U -score < -2 low results with 2 z U -score < 1 correct results with 1 z U -score +1 high results with +1 < z U -score +2 too high results with z U -score > +2 Comparison of means and reference values for each concentration level Finally the mean value calculated from all results (used as assigned value) is compared with mean values calculated for all methods separately (in this case using the Hampel estimator described in ISO and ISO/TS 20612). Mean values were calculated only, if more than 7 results were within a z U -score-range of ± 2. The means are reported with their expanded uncertainty calculated according to ISO Explanation of Appendix B Participants were asked to report expanded uncertainties of their results on a voluntary basis. In this diagram for each parameter the reported uncertainties for all concentration levels with the reproducibility standard deviation (horizontal line) are displayed. Values which deviate from the reproducibility standard deviation with a factor more than 2 are usually not realistic. 12. Explanation of Appendix C In the last part of the report, for all concentration levels the results of all participants are illustrated. Confidentiality of participants is ensured by using lab codes. The lab codes were sent to participants with the certificates. In detail Appendix C contains: - a table with all data - figures with o all reported results o all z U -scores o all reported expanded uncertainties o all scores Table with all data The assigned value with the expanded uncertainty and the tolerance limits for the concentration level is illustrated in the table. For each participant the following data are given: lab code reported result measurement uncertainty of the value (if reported) -score for this value, calculated with the following formula x xpt 2 2 ulab ux pt, with

11 PT 9/15 page 7 x x pt = difference of the measured value from the assigned value u lab = standard uncertainty of the value, reported by the participant u = standard uncertainty of the assigned value x pt z U -score for proficiency assessment assessment of the value according to its z U -score Meaning of -scores: The assessment of -scores is similar to that of z-scores. If the data are normally distributed and the uncertainties are well estimated, -scores will lie between -2 and +2 with a probability of around 95 %. -scores are mainly influenced by the measurement uncertainties reported by the laboratory. Therefore -scores are usually not appropriate for the assessment of the reported results, unless the reported measurement uncertainty is checked for fitnessfor-purpose. Therefore we do not use the -scores for the assessment of the laboratories. Nevertheless -scores are appropriate to check the plausibility of the reported measurement uncertainty: If the z-score of a result is within the tolerance limit and the -score is outside, then the measurement uncertainty is underestimated. If the z-score is outside the tolerance limits and the absolute value of the -score is lower than two, then the requirements of the proficiency test were stronger compared with the reported measurement uncertainty. Diagrams of uncertainty data In the first figure for all lab codes the measurement uncertainty (together with the reproducibility standard deviation) is illustrated. The second figure shows the associated -scores. 13. Explanation of appendices D and E Appendix D and appendix E contain the same figures as appendix A and appendix C, but here the basis for the evaluation is the mean value (Hampel estimator). These data are only given for information and does not influence the assessment since they only reflect comparability between participants and not the trueness. 14. Measurement uncertainty 27 (45,8%) out of 59 laboratories with valid values reported measurement uncertainties. In total 614 (45%) out of 1365 valid values were given with the measurement uncertainty. The following table displays the number of values with measurement uncertainty against the accreditation status. Accreditation status of Number of Number of values with measurement the values values uncertainty accredited (50,7%) not accredited (36,1%) not specified (20,7%) We would like to put emphasis on the fact that reporting of measurement uncertainties in our PT scheme is absolutely voluntary. The only objective is to help all participants to reasonably handle measurement uncertainties and their estimation. The diagrams show that in most cases the reported uncertainties are realistic.

12 PT 9/15 page 8 If measurement uncertainties are underestimated values assessed as satisfactory in the PT ( z U 2), will have a large -score. > 2 means that the own requirements (defined in terms of estimated uncertainty) are not fulfilled. 140 (43,3%) of the 323 values reported with uncertainties and having a z U -score z U 2,0 had a -score > 2,0. This means that the requirements of the PT scheme have been fullfilled, but not the own requirements, the uncertainty is underestimated. 15. Traceable reference values Traceability of analytical results to national and international references is of increasing importance in all laboratories. This is not easy to realise for chemical analyses and often can only be done by analysing certified reference materials. But availability of these reference materials in the water sector is very limited. Therefore we try to provide reference values for the proficiency test samples, traceable to national and international references. Since our PT samples without exception are spiked, real water samples, reference values can be calculated from the sum of matrix content and spiked amount of analyte. For both summands traceable values and their uncertainty have to be determined. Thereby we assume that no unrecognised bias resulting from sample preparation and transport is present and that we recognise all uncertainty components. Determination of the spiked amount and its uncertainty All spiking of samples was controlled gravimetrically. Conversion to concentration was done by measuring the density of the resulting samples using a pycnometer. This procedure allows the preparation of a complete uncertainty budget. The first step is the specification of the measurand with a formula. This shows the links between the result and all influence quantities with: c lot m s_ss,b m s_ss,t m ss_dila,b m ss_dila,t m dila_dilb,b m dila_dilb,t m dilb_lot,b m dilb_lot,t m ss,b m ss,t concentration of the analyte in the lot resulting from the spike in g/l mass of the substance added for preparation of the stock solution in g (gross weight of the container) mass of the substance added for preparation of the stock solution in g (tare weight of the container) mass of the stock solution into the dilution A in g (gross weight of the container) mass of the stock solution into the dilution A in g (tare weight of the container) mass of dilution A into the dilution B in g (gross weight of the container) mass of the dilution A into the dilution B in g (tare weight of the container) mass of dilution B into the lot in g (gross weight of the container) mass of the dilution B into the lot in g (tare weight of the container) total mass of the stock solution in g (gross weight of the container) total mass of the stock solution in g (tare weight of the container)

13 PT 9/15 page 9 m dila,b m dila,t m dilb,b m dilb,t m lot,b m lot,t lot P K total mass of the dilution A in g (gross weight of the container) total mass of the dilution A in g (tare weight of the container) total mass of the dilution B in g (gross weight of the container) total mass of the dilution B in g (tare weight of the container) total mass of the lot in g (gross weight of the container) total mass of the lot in g (tare weight of the container) density of the lot in g/l purity of the substance buoyancy correction

14 PT 9/15 page 10 Based on this formula the uncertainty budget can be prepared and all components can be quantified. The following figure shows a typical distribution of the contributions, here for anthracene as an example. The main contribution results from the weighing of the stock solution and the dilution B. Further remarkable uncertainty components result from the weighing into the lot and the purity of the substance. The samples were prepared by spiking each bottle and therefore the term weighing into the lot is contributed to the mean value of the spikes in each bottle of a lot. The same can be applied to the total mass of the lot and the respective uncertainty. All used balances are annually checked by an accredited calibration laboratory and for each balance a calibration certificate is provided. The uncertainties of the weighings are taken from these certificates. The determination of the density was also made using weighings (of the pycnometer). Temperature measurement was made with a calibrated thermometer. The purity of the used anthracene was used from the certificate of the supplier (99%). The uncertainty was stated in the certificate with 0,5% as expanded uncertainty. With all these uncertainty components the combined uncertainty, as described in the EURACHEM/CITAC-Guide Quantifying Uncertainty in Analytical Measurement, was calculated using the sensitivity coefficients determined by partial derivation of the formula to the respective influence quantities. So traceability was assured by using calibrated balances and thermometers. Determination of the matrix content For the parameters anthracene, fluoranthene as well as naphtalene the matrix content was calculated as described below. The same quantity of suspended solids was added to the samples, therefore the amount in these suspended solids can be regarded as matrix content. In consequence the matrix content could be calculated from the mean values of the participants and the spiked amounts in a standardaddition-like way 3,4. The uncertainties of the spiked amounts were known from the uncertainty budgets. The expanded uncertainties of the mean values of participants result were calculated according to ISO (Statistical Methods for Use in Proficiency Testing by Interlaboratory Comparisons) as u mean 2 1, 25 s R n 3 Rienitz, O., Schiel, D., Güttler, B., Koch, M., Borchers, U.: A convenient and economic approach to achieve SI-traceable reference values to be used in drinking-water interlaboratory comparisons. Accred Qual Assur (2007) 12: Koch, M., Baumeister, F.: Traceable reference values for routine drinking water proficiency testing: first experiences. Accred Qual Assur (2008) 13:

15 PT 9/15 page 11 with: s R reproducibility standard deviation n number of data for this level 2 coverage factor for the expanded uncertainty 1,25 correction factor (according to ISO to be used for robust methods) The content of the matrix can be derived from a linear regression of means vs. spiked amounts. Since uncertainties of all data points were available for x- as well as y-direction a generalised least square regression was used as described in DIN EN The computer program B_LEAST (from BAM) was used for this purpose. With this method a value for matrix and its uncertainty are obtained. Because of statistical variation of the input values the calculated matrix content might result in a negative value. From a scientific point of view this of course is nonsense. In those cases the matrix content is set to zero. The lower end of the uncertainty range of the matrix content also might be negative. Therefore the expanded uncertainty of the matrix content was set to the matrix content itself in this case. The matrix content is not directly traceable to national or international references, but it does not considerably compromise the traceability of the final content due to its comparably low contribution. Determination of reference values with neglect of the water matrix content For the parameters benzo[a]pyrene, benzo[b]fluoranthene, benzo[k]fluoranthene, benzo[ghi]perylene as well as indeno[1,2,3-cd]pyrene the recovery rates were very low and the linearity between the added amounts and the mean values was not sufficient. Consequently implausible high matrix contents resulted from the calculation and were not considered for the determination of the reference values. In these cases, the reference values were calculated from the sum of the spiking of the samples with the respective analytes and the content resulting from the spiked suspended solids. For the preparation of the samples, a stock solution was prepared from suspended solids mixed with pure water (2 g/l). In each bottle 10 ml of this stock solution was added. The PAH concentration resulting from the suspended solids was known from analysis of the suspended solids. After spiking of each 10 bottles, 10 ml of the suspended solid stock solution was added into an aluminium cup and after drying the solid content was determined. The mean value and the standard deviation for each batch were calculated from 20 measurements and the standard deviation was used as uncertainty of the spiking with suspended solids. The uncertainties of the reference values were calculated by combination of the uncertainties of the added amounts of the stock solutions and the added suspended solids. 16. Internet The report is available on the following webpage:

16 page A-1 of A-49

17 page A-2 of A-49

18 page A-3 of A-49

19 page A-4 of A-49

20 page A-5 of A-49

21 page A-6 of A-49

22 page A-7 of A-49

23 page A-8 of A-49

24 page A-9 of A-49

25 page A-10 of A-49