You are on page 1of 57

8

Flemish Environment Agency


Section Air

Comparative PM10 and PM2.5 Measurements in Flanders (Belgium) Period 2008-2009

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

DOCUMENT DESCRIPTION
Title Comparative PM10 and PM2.5 Measurements in Flanders, Period 2008-2009 Authors Section Air: Jordy Vercauteren, Christine Matheeussen and Edward Roekens Department Air, Environment and Communication, Flemish Environment Agency Summary This report describes the results of PM10 and PM2.5 comparison tests between automated monitors and the gravimetric reference method that were carried out at different sites in Flanders during the period 20082009. Way of referring VMM (2010) Comparative PM10 and PM2.5 Measurements in Flanders, Period 2008-2009. Publisher Philippe DHondt, Department Head Air, Communication, Flemish Environment Agency Report available from VMM-info A. Van De Maelestraat 96 9320 Erembodegem Belgium Tel: +32 (0)53 72 64 45; fax: +32 (0)53 71 10 78 e-mail: info@vmm.be Depotnumber D/2010/6871/010 Environment and

Flemish Environment Agency

www.vmm.be

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

SUMMARY
Although the current European reference methods for measuring PM in ambient air (EN12341 for PM10 and EN14907 for PM2.5) are based on a 24-hour collection and gravimetric analysis, the monitoring networks are required to use continuous automated monitors to provide real-time information. Since these monitors use different operation principles than the reference method the networks have to demonstrate that the automated instruments can produce results that are comparable with the reference method. This process is called Demonstration of Equivalence and mainly consists of executing comparative measurements between the reference method and the automated monitors. The advised (but strictly speaking not mandatory) procedure for this can be found in the Demonstration of equivalence of ambient air monitoring methods 1 . In addition to this document there is an Excel file that automatically calculates certain calibration and uncertainty parameters. During this 2008-2009 project more than 2300 comparisons were made between daily values of the reference method and automated instruments of the types ESM and FDMS, both for the PM2.5 and the PM10 fraction. Comparisons were carried out at 10 different sites: Aarschot (rural) Moerkerke (rural) Oostrozebeke (industrial) Roeselare (industrial) Zwevegem (suburban) Evergem (industrial) Zwijndrecht (suburban) Retie (rural) Borgerhout (urban background) Gent (urban background)

A full overview of all comparisons is given in the next table:


Table 1: Campaign overview

1 2 3 4 5 6 7

Location Moerkerke Aarschot Roeselare Evergem Zwijndrecht Zwevegem Borgerhout

fraction PM10 PM10 PM10 PM10 PM10 PM10 PM10 PM10 PM10 PM2.5 PM2.5 PM2.5 PM2.5 PM2.5

monitor ESM ESM ESM ESM ESM ESM ESM FDMS FDMS ESM ESM FDMS FDMS FDMS

reference filter Quartz Quartz Quartz Quartz Quartz Quartz Quartz Quartz Quartz Teflon Teflon Teflon Teflon Teflon

period 25.10.2008 - 31.12.2009 25.10.2008 - 31.12.2009 25.10.2008 - 31.12.2009 25.10.2008 - 31.12.2009 25.10.2008 - 31.12.2009 25.10.2008 - 31.12.2009 11.01.2009 - 31.12.2009 25.10.2008 - 31.12.2009 26.06.2009 - 31.12.2009 01.01.2009 - 31.12.2009 05.03.2009 - 31.12.2009 16.06.2009 - 31.12.2009 05.03.2009 - 31.12.2009 25.04.2009 - 31.12.2009

# dp 50 62 67 68 69 69 357 67 181 356 277 176 283 248

8 Oostrozebeke 9 Retie 10 11 12 13 14 Borgerhout Gent Retie Gent Borgerhout

For the reference measurements a sequential Leckel SEQ 47/50 sampler was used. As reference filter for PM2.5 a switch was made from the quartz fibre Whatman QM-A to the teflon membrane Pall Teflo because this filter type is known to suffer far less from artefacts (e.g. absorption of water) and

http://ec.europa.eu/environment/air/quality/legislation/pdf/equivalence.pdf www.vmm.be 5

Flemish Environment Agency

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

EN14907 allows the use of teflon filters 2 . For the PM10 fraction (where the EN12341 prescribes quartz as mandatory filter composition) a switch was made from the Whatman QM-A (95% quartz) to the Pall Tissuquartz 2500 QAT-UP (100% quartz) because this filter has a better reputation in the scientific community and because recent European research (in which VMM played an important role) indicated that the Pall filter suffers less from artefacts than the Whatman. This also appeared from the new comparison between the mass concentration on Pall quartz and teflon filters that were being operated in parallel at two sites. On average, the result for quartz was only 1.9% higher than for teflon, compared to 8% in the previous study (with Whatman QM-A and Pall Teflo). In addition, the field blanks for the Pall quartz filters (on average 79 g) were also lower than for the Whatman quartz filters that were used in the previous study (on average 110 g). The field blank for teflon (48 g) did not change compared to the previous study. These observations indicate that it is likely that the change in reference filters will have led to higher quality and slightly lower reference values than before. The comparisons between automated monitors and reference measurements were used to calculate calibration functions and factors. Because the intercept (determined using orthogonal regression) was always quite small, it was decided to only use calibration factors. For this the preference was given to the factor that was calculated from the ratio between the average of the two data sets because this, per definition, gives the best result for the calculation of the overall average. For ESM in the PM10 fraction an overall factor of 1.25 was calculated. This new factor is lower (9%) than the previous factor in use (1.37). After recalculation with the new factor the overall dataset and 3 six of the seven subsets pass the uncertainty criterion at the limit value of 50 g/m . The new factor is in very good agreement with the situation in The Netherlands where a factor of 1.25 is also currently used by RIVM for the ESM in their nationwide network. For the PM2.5 ESM monitors there was little difference between the instruments in Borgerhout (1.28) and Gent (1.26) and an overall factor of 1.27 was calculated. This new factor is 15% lower than the factor of the previous study (1.46) which was based on the Whatman QM-A as reference. After recalculation of the results with the new factor the overall PM2.5 dataset and the two subsets pass the criterion for the uncertainty at the tentative 3 PM2.5 limit value of 30 g/m3. For the five investigated FDMS monitors there turned out to be problems with three instruments. The involved monitors appeared to overestimate PM concentrations compared to the reference values. In contrast to the two monitors that did give good results, the three monitors were operated in a small trailer, indicating that there is a problem with the FDMS trailer setup. Probably the thermal conditioning in the trailers was not stable enough for the sensitive FDMS, an issue that has been reported by other networks too 4 . The three trailer datasets were not considered for further calculation and the problem is currently being investigated. Since the trailers are still in test phase and have not been used for routine monitoring there is no indications of problems regarding the quality of FDMS data that was reported to the public. This is also demonstrated by the fact that for the other two monitors calibration factors very close to 1 were calculated. Both the PM10 FDMS dataset (in Oostrozebeke) and the PM2.5 FDMS dataset (in Borgerhout) pass the criterion for the uncertainty at the limit value without correction. In the future, comparisons will remain necessary to confirm the new factors and to get a better understanding of differences from year to year and from site to site.

Since 2009 VMM uses the Pall Teflo filter for all reference measurements of PM2.5 (e.g. also for calculation of the average exposure indicator) 3 Since there is no daily limit value for PM2.5, CEN TC264 WG15 recently proposed a tentative limit value just for the purpose of equivalence demonstration 4 Unofficial communications at CEN TC264 WG15

Flemish Environment Agency

www.vmm.be

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

SAMENVATTING
Hoewel de huidige Europese referentiemethoden voor het meten van PM in buitenlucht (EN12341 voor PM10 en EN14907 voor PM2,5) gebaseerd zijn op een 24-uurs monsterneming en gravimetrische analyse, dienen de netwerken toch continue automatische monitoren te gebruiken om real-time gegevens te kunnen aanbieden aan het publiek. Aangezien deze monitoren gebaseerd zijn op andere meetprincipes dan de referentiemethode moeten de netwerken aantonen dat deze toestellen resultaten kunnen geven die vergelijkbaar zijn met die van de referentiemethode. Dit proces staat bekend als het aantonen van equivalentie (Demonstration of Equivalence) en behelst vooral het uitvoeren van vergelijkende oefeningen tussen de referentiemethode en de automatische monitoren. De aanbevolen, maar strikt genomen niet verplichte, procedure staat beschreven in het 5 document Demonstration of equivalence of ambient air monitoring methods . Bij dit document hoort tevens een Excel bestand voor het automatisch berekenen van kalibratiefuncties en onzekerheden. Tijdens deze studie in 2008 en 2009 werden in totaal meer dan 2300 vergelijkingen gemaakt tussen dagwaarden van de referentiemethode en automatische monitoren van het type ESM en FDMS, voor zowel de PM2.5 als de PM10 fractie. De vergelijkingen werden uitgevoerd op 10 verschillende locaties: Aarschot (landelijk) Moerkerke (landelijk) Oostrozebeke (industrieel) Roeselare (industrieel) Zwevegem (voorstedelijk) Evergem (industrieel) Zwijndrecht (voorstedelijk) Retie (landelijk) Borgerhout (stedelijke achtergrond) Gent (stedelijke achtergrond)

Een volledig overzicht van alle vergelijkingen wordt weergegeven in volgende tabel:
Tabel 2: Campagne overzicht

1 2 3 4 5 6 7

Plaats Moerkerke Aarschot Roeselare Evergem Zwijndrecht Zwevegem Borgerhout

fractie PM10 PM10 PM10 PM10 PM10 PM10 PM10 PM10 PM10 PM2,5 PM2,5 PM2,5 PM2,5 PM2,5

monitor ESM ESM ESM ESM ESM ESM ESM FDMS FDMS ESM ESM FDMS FDMS FDMS

referentie filter kwarts kwarts kwarts kwarts kwarts kwarts kwarts kwarts kwarts teflon teflon teflon teflon teflon

periode 25.10.2008 - 31.12.2009 25.10.2008 - 31.12.2009 25.10.2008 - 31.12.2009 25.10.2008 - 31.12.2009 25.10.2008 - 31.12.2009 25.10.2008 - 31.12.2009 11.01.2009 - 31.12.2009 25.10.2008 - 31.12.2009 26.06.2009 - 31.12.2009 01.01.2009 - 31.12.2009 05.03.2009 - 31.12.2009 16.06.2009 - 31.12.2009 05.03.2009 - 31.12.2009 25.04.2009 - 31.12.2009

# dp 50 62 67 68 69 69 357 67 181 356 277 176 283 248

8 Oostrozebeke 9 Retie 10 11 12 13 14 Borgerhout Gent Retie Gent Borgerhout

Voor de referentiemetingen werd gebruikt gemaakt van een sequentile Leckel SEQ 47/50 sampler. Als referentiefilter voor PM2,5 werd er overgeschakeld van de kwartsfilter Whatman QM-A naar de

http://ec.europa.eu/environment/air/quality/legislation/pdf/equivalence.pdf www.vmm.be 7

Flemish Environment Agency

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

Pall Teflo teflon membraanfilter omdat deze filter beduidend minder last heeft van artefacten (bv. opname van water) 6 . Voor de PM10 fractie (waar EN12341 kwarts voorschrijft als filtermateriaal) werd er overgeschakeld van Whatman QM-A (95% kwarts) naar Pall Tissuquartz 2500 QAT-UP (100% kwarts) omdat deze filter beter staat aangeschreven in wetenschappelijke kringen en omdat recent Europees onderzoek (waarin VMM een belangrijke rol speelde) aangaf dat de Pall kwartsfilter minder problemen heeft met artefacten. Dit bleek ondermeer ook uit de nieuwe parallelmetingen van de massaconcentratie op Pall kwarts en op Pall teflon op twee plaatsen. Gemiddeld was de concentratie op de Pall kwartsfilters slecht 1,9% hoger dan op de Pall teflon filter, terwijl dit in een vorige studie (met de Whatman QM-A kwartsfilter en Pall Teflo teflonfilter) 8% bedroeg. Ook de veldblanco voor de Pall kwarts filter (gemiddeld 79 g) was lager dan voor de Whatman kwarts filters die gebruikt werden in de vorige studie (gemiddeld 110 g). De veldblanco voor teflon was ongewijzigd (48 g) ten opzichte van de vorige studie. Er kan dus gesteld worden dat de aanpassing van de referentiefilters waarschijnlijk tot kwalitatief betere en iets lagere referentiewaarden vergeleken met het verleden zal gezorgd hebben. Uit de vergelijking tussen de monitoren en de referentiemetingen werden kalibratiefunctie en factoren berekend. Omdat de intercept bij de kalibratiefuncties (bepaald door middel van orthogonale regressie) steeds vrij klein was, werd er besloten om (net als in het verleden) enkel met kalibratiefactoren te werken. Hierbij werd de voorkeur gegeven aan de factor die berekend werd op basis van de verhouding van de gemiddelden van de twee datasets omdat dit per definitie het beste resultaat geeft voor de berekening van het algemeen gemiddelde. Voor de vergelijkingen van de PM10 ESM monitoren werd een algemene factor van 1,25 berekend. Deze nieuwe factor is lager (9%) dan de vorige factor die gebruikt werd (1,37). Na omrekening met de nieuwe factor voldoet de totale dataset en zes van de zeven subsets aan de voorwaarde voor 3 onzekerheid bij de PM10 grenswaarde van 50 g/m . De nieuwe factor vertoont een zeer goede overeenkomst met de situatie in Nederland waar RIVM momenteel ook een factor 1,25 gebruikt in zijn landelijk meetnet. Voor de PM2,5 ESM monitoren was er weinig verschil tussen de toestellen in Borgerhout (1,28) en Gent (1,26) en werd er een algemene factor van 1,27 berekend. Deze nieuwe factor is 15% lager dan die uit de vorige studie (1,46). Na omrekening met de nieuwe factor voldoen de totale dataset en de twee subsets aan de voorwaarde voor onzekerheid bij de test-grenswaarde 7 van 30 g/m3. Bij de vijf onderzochte FDMS monitoren werden er problemen vastgesteld bij drie monitoren. De betrokken monitoren bleken de PM concentraties aanzienlijk te overschatten ten opzichte van de referentiemeting. In tegenstelling tot de twee monitoren die wel goede resultaten gaven, bevonden deze drie monitoren zich in een trailer, wat doet vermoeden dat er een probleem is met de FDMS trailer setup. Mogelijk was de thermische conditionering in de trailers onvoldoende stabiel voor de gevoelige FDMS instrumenten (een probleem dat ook al door andere netwerken werd vastgesteld 8 ). De drie datasets werden niet verder meegenomen bij de berekeningen en het probleem wordt verder onderzocht. Aangezien de trailers nog steeds in testfase zijn en nog niet gebruikt werden voor routine metingen is er ook geen aanwijzingen dat er kwaliteitsproblemen zijn met de FDMS meetwaarden die aan het publiek werden meegedeeld. Iets wat ook blijkt uit het feit dat voor de twee andere monitoren kalibratiefactoren zeer dicht bij 1 gevonden worden. Zowel de PM10 FDMS dataset (in Oostrozebeke) als de PM2.5 FDMS dataset (in Borgerhout) voldeden zonder omrekening aan de voorwaarde voor onzekerheid bij de grenswaarde. In de toekomst zullen vergelijkende metingen nodig blijven om de nieuwe kalibratiefactoren te bevestigen en om een beter inzicht te krijgen in verschillen van jaar tot jaar en van plaats tot plaats.

Sinds 2009 gebruikt VMM de Pall Teflo als referentiefilter voor alle PM2.5 metingen (bv. ook de berekening van gemiddelde blootstellingsindicator) 7 Omdat er voor PM2,5 geen daggrenswaarde bestaat werd er door CEN TC264 WG15 een testwaarde van 30 3 g/m voorgesteld die uitsluitend bedoeld is voor dit soort vergelijkende metingen. 8 Officieuze mededeling in CEN TC264 WG15

Flemish Environment Agency

www.vmm.be

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

TABLE OF CONTENT Summary ................................................................................................................... 5 Samenvatting ............................................................................................................ 7 Table of content........................................................................................................ 9 List of figures...........................................................................................................10 List of tables ............................................................................................................11 Introduction..............................................................................................................13 1 Experimental setup .............................................................................................15
1.1 Sampling Sites ............................................................................................................................ 15 1.1.1 Aarschot (N035)................................................................................................................ 15 1.1.2 Moerkerke (N012) ............................................................................................................. 16 1.1.3 Oostrozebeke (OB01) ....................................................................................................... 17 1.1.4 Roeselare (M705) ............................................................................................................. 18 1.1.5 Zwevegem (N052) ............................................................................................................ 19 1.1.6 Evergem (R731)................................................................................................................ 20 1.1.7 Zwijndrecht (R815)............................................................................................................ 21 1.1.8 Retie (RT01)...................................................................................................................... 22 1.1.9 Borgerhout (R801) ............................................................................................................ 23 1.1.10 Gent (R701) ...................................................................................................................... 24 1.2 Instruments.................................................................................................................................. 25 1.2.1 Leckel SEQ 47/50 ............................................................................................................. 25 1.2.2 ESM FH 62 I-R.................................................................................................................. 25 1.2.3 Series 8500 FDMS System............................................................................................... 25 1.3 Weighing and handling of gravimetric samples .......................................................................... 27 1.4 Data treatment............................................................................................................................. 28 1.4.1 Validation and Quality Control .......................................................................................... 28 1.4.2 Outlier removal.................................................................................................................. 28 1.4.3 Calibration factor/equation calculation.............................................................................. 28 1.5 Campaign Overview .................................................................................................................... 30 1.5.1 Time & instruments ........................................................................................................... 30 1.5.2 Meteorological data........................................................................................................... 30

2 The gravimetric reference method.....................................................................31


2.1 History of filter choice at VMM .................................................................................................... 31 2.2 Filter choice for this project ......................................................................................................... 31 2.3 Field comparisons & tests ........................................................................................................... 32 2.3.1 Filter comparison: quartz vs teflon filter ............................................................................ 32 2.3.2 Field blanks ....................................................................................................................... 34

3 Comparisons with automated monitors ............................................................35


3.1 3.2 3.3 3.4 Time & scatter plots PM10 - ESM ............................................................................................... 36 Time & scatter plots PM10 - FDMS............................................................................................. 44 Time & scatter plots PM2.5 - ESM .............................................................................................. 46 Time & scatter plots PM2.5 - FDMS............................................................................................ 48

4 Demonstration of equivalence ...........................................................................51


4.1 Overview ..................................................................................................................................... 51 4.2 Suitability of datasets .................................................................................................................. 51 4.3 Between sampler uncertainty...................................................................................................... 51 4.4 Reference vs. automated ............................................................................................................ 51 4.4.1 Uncertainty at the limit value before calibration .............................................................. 52 4.4.2 Calibration factors & equations ......................................................................................... 52 4.4.3 Discussion of calibration ................................................................................................... 54 4.4.4 Uncertainty at the limit value after calibration ................................................................. 55

5 Conclusions.........................................................................................................57

Flemish Environment Agency

www.vmm.be

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

FIGURES
Fig 1: The rural monitoring station at Aarschot ...................................................................................................... 15 Fig 2: The rural monitoring site at Moerkerke ........................................................................................................ 16 Fig 3: The industrial site at Oostrozebeke ............................................................................................................. 17 Fig 4: The industrial site at Roeselare ................................................................................................................... 18 Fig 5: The suburban / industrial site at Zwevegem ................................................................................................ 19 Fig 6: The industrial site at Evergem ..................................................................................................................... 20 Fig 7: The suburban/industrial site at Zwijndrecht ................................................................................................ 21 Fig 8: The rural background site at Retie............................................................................................................... 22 Fig 9: The urban background monitoring site at Borgerhout.................................................................................. 23 Fig 10: The urban background site at Gent ........................................................................................................... 24 Fig 11: The different instruments used in by VMM (b.Grimm and e.TEOM were not investigated in this project) . 26 Fig 12: Weighing cupboard.................................................................................................................................... 27 Fig 13: Meteorological data, Gent-Tolhuiskaai (M701) .......................................................................................... 30 Fig 14: Ratio of quartz vs teflon concentration for Retie and Borgerhout .............................................................. 32 Fig 15: Scatter plot of teflon vs quartz filters for Retie ........................................................................................... 33 Fig 16: Scatter plot of teflon vs quartz filters for Borgerhout.................................................................................. 33 Fig 17: Concentrations and daily ratios for PM10 ESM in Moerkerke.................................................................... 36 Fig 18: Scatter plot for PM10 ESM in Moerkerke................................................................................................... 36 Fig 19: Concentrations and daily ratios for PM10 ESM in Aarschot ...................................................................... 37 Fig 20: Scatter plot for PM10 ESM in Aarschot ..................................................................................................... 37 Fig 21: Concentrations and daily ratios for PM10 ESM in Roeselare .................................................................... 38 Fig 22: Scatter plot for PM10 ESM in Roeselare ................................................................................................... 38 Fig 23: Concentrations and daily ratios for PM10 ESM in Zwevegem ................................................................... 39 Fig 24: Scatter plot for PM10 ESM in Zwevegem .................................................................................................. 39 Fig 25: Concentrations and daily ratios for PM10 ESM in Evergem ...................................................................... 40 Fig 26: Scatter plot for PM10 ESM in Evergem ..................................................................................................... 40 Fig 27: Concentrations and daily ratios for PM10 ESM in Zwijndrecht .................................................................. 41 Fig 28: Scatter plot for PM10 ESM in Zwijndrecht ................................................................................................. 41 Fig 29: Concentrations and daily ratios for PM10 ESM Borgerhout (limited dataset) ............................................ 42 Fig 30: Scatter plot for PM10 ESM in Borgerhout (limited dataset) ....................................................................... 42 Fig 31: Concentrations and daily ratios for PM10 ESM in Borgerhout (full dataset) .............................................. 43 Fig 32: Scatter plot for PM10 ESM in Borgerhout (full dataset) ............................................................................. 43 Fig 33: Concentrations and daily ratios for PM10 FDMS in Retie.......................................................................... 44 Fig 34: Scatter plot for PM10 FDMS in Retie......................................................................................................... 44 Fig 35: Concentrations and daily ratios for PM10 FDMS in Oostrozebeke............................................................ 45 Fig 36: Scatter plot for PM10 FDMS in Oostrozebeke........................................................................................... 45 Fig 37: Concentrations and daily ratios for PM2.5 ESM in Borgerhout.................................................................. 46 Fig 38: Scatter plot for PM2.5 ESM in Borgerhout................................................................................................. 46 Fig 39: Concentrations and daily ratios for PM2.5 ESM in Gent............................................................................ 47 Fig 40: Scatter plot for PM2.5 ESM in Gent........................................................................................................... 47 Fig 41: Concentrations and daily ratios for PM2.5 FDMS in Retie......................................................................... 48 Fig 42: Scatter plot for PM2.5 FDMS in Retie........................................................................................................ 48 Fig 43: Concentrations and daily ratios for PM2.5 FDMS in Borgerhout ............................................................... 49 Fig 44: Scatter plot for PM2.5 FDMS in Borgerhout .............................................................................................. 49 Fig 45: Concentrations and daily ratios for PM2.5 FDMS in Gent ......................................................................... 50 Fig 46: Scatter plot for PM2.5 FDMS in Gent ........................................................................................................ 50

Flemish Environment Agency

www.vmm.be

10

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

TABLES
Table 1: Campaign overview ................................................................................................................................... 5 Tabel 2: Campagne overzicht.................................................................................................................................. 7 Table 3: Campaign overview ................................................................................................................................. 30 Table 4: Overview of filters used for gravimetric sampling .................................................................................... 32 Table 5: Comparison between teflon and quartz filters.......................................................................................... 32 Table 6: Statistical parameters regarding quartz and teflon field blanks................................................................ 34 Table 7: Uncertainties at the limit values before calibration................................................................................... 52 Table 8: PM10 - ESM calibration factors ............................................................................................................... 53 Table 9: PM10 - FDMS calibration factors ............................................................................................................. 53 Table 10: PM2.5 - ESM calibration factors ............................................................................................................ 53 Table 11: PM2.5 - FDMS calibration factors .......................................................................................................... 54 Table 12: Uncertainties at the limit values after calibration.................................................................................... 55

Flemish Environment Agency

www.vmm.be

11

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

Introduction
Measuring particulate matter Considering the health issues of particulate matter (PM), the importance of accurate and reliable PM data can hardly be overestimated. The fact that PM is not defined by a chemical formula and that therefore PM is specified solely by a reference method (in Europe: EN12341 for PM10 and EN14907 for PM2.5) creates a real challenge not often encountered with other pollutants. The reference methods require PM to be collected on a filter during 24 hours. Based on the mass difference of the filter and the sampled volume the PM concentration can be calculated. The major drawback of this approach is that it only provides 24-hour values and this with a certain delay because filters have to be collected from the field and weighed under controlled conditions in the lab. As the EU also requires the Member States to inform the public with real-time information on PM levels the use of continuous automated PM monitors is unavoidable. The automated monitors use different operation principles than the reference methods. Demonstration of equivalence For the results of the automated methods (called candidate methods) to be accepted, the Member States have to demonstrate that these results are equivalent to those of the reference method. The advised (but strictly speaking not mandatory) procedure for this can be found in the Demonstration of equivalence of ambient air monitoring methods 9 . The most important point in this document is the aspect of (a minimum number of) field comparisons between the candidate method and the reference method. In addition to this document there is a European Excel file that automatically calculates certain calibration and uncertainty parameters. At a European equivalence workshop in Italy 10 (May 2007) several issues were discussed and a number of suggestions where done to update the current document, some of which were taken into account in the latest edition. Issues with the reference method Apart from the problems with the equivalence procedure there is a second, and possibly even more important problem, namely the uncertainty within the current European reference method. Recent studies (e. g. VMM 2007 11 ) have demonstrated that the uncertainty of several aspects of the reference method EN12341 has been underestimated. The two most important being the influence of the filter type and brand and the effect of filter conditioning before weighing. Previous VMM studies A first VMM comparison study 12 was carried out in 2001 and 2002. Two general calibration factors were calculated: 1.37 (for ESM FH 62 I-N and ESM FH 62 I-R at 40 C) and 1.47 (for TEOM at 40 C). Cellulose nitrate (CN) was used as reference filter (in those days a previous short comparison campaign had shown no significant difference between CN and a quartz type filter). With these two values VMM was applying some of the highest calibration factors in Europe. In 2003 and 2004 a second study 13 was done. For the ESM FH 62-I-R (now with dynamic heating) a factor of 1.39 was calculated while the TEOM (now at 50 C) gave a factor of 1.49. Both were calculated versus a reference method which still used cellulose nitrate filters. Taking into account the uncertainty and variation on the factors and the small difference with those of the first study VMM decided to keep using the factors 1.37 and 1.47. A third study in 2005-2006 used quartz fibre filters for the reference method and different types and brands of filters were compared. For certain campaigns the differences between different brands of quartz filters were as high as 29%. Continuous automated monitors were compared with two variations of the reference method (one with Machery-Nagel filters and one with Whatman filters). The range of calibration factors for ESM and TEOM more or less covered the factors in use (1.37 and 1.47

http://ec.europa.eu/environment/air/quality/legislation/pdf/equivalence.pdf Demonstration of Equivalence of Ambient Air Monitoring Methods, 2-4 May 2007, JRC-IES, Ispra Italy 11 VMM (2007), Comparative PM10 Measurements in Flanders, Period 2005-2006 12 VMM (2003). Vergelijkende PM10-metingen, periode 2001-2002, Vlaamse Milieumaatschappij, February 2003 13 VMM (2005). Vergelijkende PM10-metingen, periode 2004, Vlaamse Milieumaatschappij, April 2005
10

Flemish Environment Agency

www.vmm.be

13

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

respectively). Because of this and the fact that the PM10 reference method was (and still is) being updated by CEN VMM again decided not to update its factors. When taking into account the expanded uncertainty at the limit value criterion (< 25%) only TEOM-FDMS managed to pass without applying a calibration factor. ESM passed after applying a calibration factor and TEOM only passed after applying a calibration factor for the MN dataset. The fourth intercomparison was linked to the PM10 chemical characterisation project 14 that was carried out from September 2006 until September 2007 at six locations in Flanders. Every sixth day simultaneous sampling was carried out at all the sites. At each location two PM10 Leckels were placed: one with quartz filters (Whatman QM-A) and one with teflon filters (Pall Teflo). It turned out that overall the mass on the Whatman QM-A filters (which were pre-fired) was 8% higher than on the teflon filters. Again the range of the calibration factors for ESM and TEOM covered the factors in use, so no change was made to the existing procedure. Early 2007 VMM started measuring PM2.5 concentration with Whatman QM-A filters on four locations, which also allowed to do some comparison calculations. For ESM a factor of 1.46 was found, which is currently being used. This project (2008-2009) In this comparison study the mass concentration data of a new PM10 chemical characterisation project was used just like in 2006-2007. The Chemkar hotspots project was carried out at seven sites, five of which were (industrial or suburban) hotspots and two were background sites. Additionally, since the beginning of 2009 the gravimetric method was used to determine PM2.5 and PM10 mass concentrations at a handful of sites. Most of this data was also used to compare with automated PM monitors.

14

VMM (2008), Chemkar PM10: Chemische karakterisatie van fijn stof in Vlaanderen, 2006-2007

Flemish Environment Agency

www.vmm.be

14

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

Experimental setup

1.1 Sampling Sites


The follow sites were used for comparison exercises in 2008-2009: Aarschot (rural) Moerkerke (rural) Oostrozebeke (industrial) Roeselare (industrial) Zwevegem (suburban) Evergem (industrial) Zwijndrecht (suburban) Retie (rural) Borgerhout (urban background) Gent (urban background)

1.1.1 Aarschot (N035) The monitoring site in Aarschot (population approx. 30 000) is considered a rural background site (though rural is a relative concept in the densely populated region of Flanders). A highway (E313) is about 1.5 km S of the station and the city centre is less then 1 km away. The station lies between two small roads (both at about 75 m from the station).

Fig 1: The rural monitoring station at Aarschot

Flemish Environment Agency

www.vmm.be

15

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

1.1.2 Moerkerke (N012) The monitoring site in Moerkerke (population approx. 3 000) is also considered a rural background site. It is located in an agricultural area and lies 10 km NE of the city of Bruges (population approx. 120 000) and 2 km NE of the small village of Moerkerke. The site is about 1.5 km SW of the highway E34.

Fig 2: The rural monitoring site at Moerkerke

Flemish Environment Agency

www.vmm.be

16

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

1.1.3 Oostrozebeke (OB01) The site in Oostrozebeke (population approx. 7 500) is located in a suburban area but is considered an industrial site since it is less than 100 m S of a floorcovering manufacturer (Orotex) and about 500 m NE from a large wood processing plant (Spano). Some of the highest PM10 concentrations in Flanders are measured at this site.

Fig 3: The industrial site at Oostrozebeke

Flemish Environment Agency

www.vmm.be

17

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

1.1.4 Roeselare (M705) The monitoring site in Roeselare (population approx. 60 000) is located on the outskirts of the city, near the canal of Roeselare in an industrial area. Next to the site is a busy road with frequent heavy traffic. About 200 m to the E is a transhipment site. Less than 200 m NW of the site is a suburban area and about 300 m to the N is a galvanization plant (Belcroom). Some of the highest PM10 concentrations in Flanders are measured at this site.

Fig 4: The industrial site at Roeselare

Flemish Environment Agency

www.vmm.be

18

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

1.1.5 Zwevegem (N052) The site in Zwevegem (population approx. 25 000) is located in a suburban area but also has an industrial influence. Its located 1 km SW of a large company specialized in advanced metal transformation and advanced materials and coatings (Bekaert). The site is also about 1 km to the SE of the busy highway E17 and about 4 km E of the city centre of Kortrijk (population approx. 75 000). PM10 concentrations at this site are generally above average.

Fig 5: The suburban site at Zwevegem

Flemish Environment Agency

www.vmm.be

19

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

1.1.6 Evergem (R731) The site in Evergem (population approx. 30 000) is located in a suburban area but is considered an industrial site since it is close to the harbour of Gent (about 1 km to the S) with a high number of industrial plants, e.g. less than 1 km to the SE is a company specialised in production of Molybdenumalloys (Sadaci NV) and a company producing TiO2 (Kronos Europe NV), 1 km to the S lies a (gas) power plant and 2 km to the E lies the Gent coal terminal. The city of Gent (population approx. 250 000) lies about 8 km SW of the site. Some of the highest PM10 concentrations in Flanders are measured at this site.

Fig 6: The industrial site at Evergem

Flemish Environment Agency

www.vmm.be

20

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

1.1.7 Zwijndrecht (R815) The site in Zwijndrecht (population approx. 20 000) is considered a suburban site in the centre of the small town with possible influence from the industrial activity in the Antwerp harbour. The site is located about 40 m N from the railroad connecting Gent to Antwerp, 700 m NE of the highway E17, 3 km S of the harbour of Antwerp and about 4 km W of the city of Antwerp (population approx. 500 000).

Fig 7: The suburban site at Zwijndrecht

Flemish Environment Agency

www.vmm.be

21

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

1.1.8 Retie (RT01) The monitoring site in Retie (population approx. 10 000) is considered a rural background site. The site is located in a meadow next to a forested park area (Prinsenpark). The only known potential sources are a regional road 500 m to the SE and a small industrial area about 1 km to the E.

Fig 8: The rural background site at Retie

Flemish Environment Agency

www.vmm.be

22

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

1.1.9 Borgerhout (R801) The monitoring site at Borgerhout can be considered an urban background site with a significant influence of traffic. The station is about 30 m from a busy road (Plantin en Moretuslei) and less than 1 km from the busy city ring around Antwerp (R1). Its location is within the city ring of Antwerp (population approx. 500 000) at the eastern side.

Fig 9: The urban background monitoring site at Borgerhout

Flemish Environment Agency

www.vmm.be

23

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

1.1.10 Gent (R701) The monitoring site at Gent (population approx. 240 000) is an urban background site located in a park on the NE side of the inner city. The site is also about 1 km S of a transhipment site at the southern end of the harbour.

Fig 10: The urban background site at Gent

Flemish Environment Agency

www.vmm.be

24

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

1.2

Instruments

1.2.1 Leckel SEQ 47/50 The Leckel SEQ 47/50 was used in the Flemish implementation of the PM reference method (EN12341 and EN14907). Based on the outcome of various intercomparison campaigns all over Europe, this low-volume gravimetric instrument - though strictly not the reference method - can be considered a variation on a theme (as stated at the 2007 Demonstration of Equivalence of Ambient Air Monitoring Methods workshop held in Ispra, Italy by the Joint Research Centre (JRC) European Commission) 15 . The Leckel is also the reference instrument used by the JRC in their European comparison programs. For the intercomparison campaigns the instruments were generally equipped with a set of 14 filters (on non-measuring days dummy filters were used). Filters were collected every 2 weeks. Gravimetric samplers were always placed outside on the roof of the monitoring cabin with the exception of the monitoring sites Oostrozebeke (on a small platform next to the station) and Retie (at ground level). Technical info: Air flow: 2.3 m3/h at ambient conditions PM inlet: EU-type, no heating Filter material during these comparative measurements: quartz fibre (Pall Tissuquartz 2500 QAT-UP) or teflon membrane (Pall Teflo 2 m) Filter change: at 0h00 UT Sampling time: 24 h 1.2.2 ESM FH 62 I-R For general PM10 monitoring (which is done by the Telemetric network) the Flemish Environment Agency uses the ESM FH 62 I-R monitor (also known as FAG monitor). This instrument uses the technique of -attenuation (the attenuation of -rays by a filter is directly related to the amount of mass on the filter). To avoid condensation of water on the filter the inlet is heated. This process not only leads to the loss of water, but also of certain semi-volatile compounds such as ammonium nitrate. Technical info: Air flow: 1 m3/h at ambient conditions Inlet: US-type, dynamic heating, slightly above ambient temperature, for PM2.5 a sharp cut cyclone is placed between the PM10-head and the monitor Filter material: glass fibre Beta source: 85Kr 1.2.3 Series 8500 FDMS System The FDMS (Filter Dynamics Measurement System) consists of a TEOM 1400ab system with an addon unit that allows correction for the loss of mass due to volatilization. The instrument works in two cycles of each six minutes. During one cycle the base PM10 is recorded, more or less comparable to the Series 1400ab TEOM. During the next cycle the air is sent over a zero-filter (at 4 C) prior to passing through the sampling filter and the mass decrease on the sampling filter (= volatized PM) is recorded. The combination of the results of both cycles allows for the calculation of total PM10. Technical info: Air flow through sampling head: 1 m3/h at ambient conditions Air flow through filter: 0.18 m3/h at ambient conditions Inlet: US type, heating: 30 C + Nafion dryer, for PM2.5 a sharp cut cyclone is placed between the PM10-head and the monitor

CEN TC264 WG15 recently agreed to allow sequential samplers like the Leckel as a reference sampler in the upcoming revision of the European PM reference method.

15

Flemish Environment Agency

www.vmm.be

25

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

Filter material: teflon coated glass fibre (Pallflex TX40)

Fig 11: The different instruments used in by VMM (b.Grimm and e.TEOM were not investigated in this project) a: Leckel SEQ47/50 b:Grimm180 c: ESM FH 62 I-R d: Series 8500 FDMS System e: TEOM Series 1400ab

Flemish Environment Agency

www.vmm.be

26

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

1.3 Weighing and handling of gravimetric samples


For both the PM10 and PM2.5 filters, the more stringent weighing criteria of the PM2.5 reference method (EN14907) were used. This implies that filters are weighed twice before sampling and twice afterwards. Blank filters are placed for at least 48 h in a (custom made) climate controlled weighing cupboard (inside a temperature controlled room) prior to the first weighing, 48 h later a second weighing is carried out. The same procedure is used for sampled filters. Filters are collected every two weeks. Transport from the field to the lab is done in the Leckel sampling canisters which are placed inside special cases. Transport time (from Leckel to controlled conditions) generally ranges from 30 minutes to approximately 5 hours. Upon arrival in the lab the filters are taken out of the filter rings and placed immediately in the climate controlled weighing cupboard.

Fig 12: Weighing cupboard

Technical info: Balance: Sartorius M5P (resolution 1 g) Weighing cupboard temperature: 20 1 C Weighing cupboard humidity: 50 2%

Flemish Environment Agency

www.vmm.be

27

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

1.4 Data treatment


1.4.1 Validation and Quality Control Gravimetric reference method During weighing all filters were visually inspected for any irregularities. Filters were also considered invalid when the difference between 1st and 2nd weighing was too high: - For PM2.5: higher than 40 g (blank filters) or 60 g (sampled filters), according to EN14907. - For PM10: higher than 80 g (blank filters) or 120 g (sampled filters), according to own VMM criteria The flow of the Leckel instruments was checked every two weeks with a rotameter. When a deviation of the flow was more than 3% the flow was readjusted to 2.3 m3/h. It later turned out that the rotameters do not give a correct reading at low temperatures due to differences in expansion/shrinking of the float and the tube, therefore the small flow corrections that were made in the field were considered not justified and were recalculated. The total sampled volume is calculated by the instrument based on its running time. Whenever the sampling time/volume was less than 95% of the normal value (24h, 55m3) the data were considered invalid. Automated monitors The standard routine validation procedure for the automated monitors was used during the campaigns. Daily average concentrations were calculated using half hourly values by means of XR software. Average values are rounded to the nearest whole number by the software. Although this can create a significant bias for small concentrations, the average effect will be negligible. The 24-h average values for automated monitors with less than 80% of half hourly values (# < 39) were considered invalid. 1.4.2 Outlier removal In addition to the removal of data because of technical reasons, the Grubbs outlier test was carried out at the 99% level on the data series of 1) the reference-candidate absolute difference and 2) reference/candidate ratio. 1.4.3 Calibration factor/equation calculation Various methods can be used to describe the relationship between two PM monitors. In this work three different techniques that treat the data in a symmetrical way were applied. 1. Orthogonal regression (OR) This is the technique suggested by the Demonstration of equivalence of ambient air monitoring methods document 16 . In ordinary linear regression, the goal is to minimize the sum of the squared vertical distances between the data points and the corresponding points on the fitted line. In orthogonal regression (also referred to as total least squares) the goal is to minimize the orthogonal (perpendicular) distances from the data points to the fitted line (y= bx + a). Because of that, orthogonal regression is a technique that leads to a symmetrical 17 treatment of both variables. This is preferable because the experimental errors in the PM reference method are in the same order of magnitude as the errors in the candidate method.

16 17

http://ec.europa.eu/environment/air/quality/legislation/pdf/equivalence.pdf A symmetrical approach leads to the same results when the datasets are exchanged

Flemish Environment Agency

www.vmm.be

28

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

Syy Sxx

Syy Sxx 2 4(Sxy) 2


2 Sxy

Sxx ( xi x) 2

Sxy ( xi x).( yi y )

Syy ( yi y ) 2

a y bx
For the application of this method the EU Excel spreadsheet was used 18 . 2. Orthogonal regression forced through origin When the normal orthogonal regression leads to an intercept that is not significantly different from 0 it is justifiable to force the regression through the origin, thereby describing the relationship between the two methods by a single value or calibration factor. Even if the intercept is small but significantly different from 0 one could opt to use this technique, although it is then up to the user to decide whether the benefits of using one single factor outweigh the small mathematical error that is made. For the application of this method an in-house Excel spreadsheet was used, based on the following formula 19 (where b is the slope of the regression line):

tan 2b

2 xi yi
2 i

yi2

factor b 1
3. Ratio of averages A very simple, but effective way of calculating a calibration factor is taking the ratio of the averages of both methods. This is a symmetrical approach for which data points in the higher concentration range have somewhat more weight than those in the lower range. Calculating the calibration factor like this also gives per definition the best results for calculating the annual average value.

factor

x y

18 19

http://ec.europa.eu/environment/air/pdf/test_equivalencev31004.xls http://mathforum.org/library/drmath/view/68362.html

Flemish Environment Agency

www.vmm.be

29

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

1.5

Campaign Overview

1.5.1 Time & instruments In this project more than 2 300 comparisons were made between 24h reference values and 24h values of automated instruments. Ten sites were involved and two types of monitors were evaluated, both in the PM10 and PM2.5 fraction.
Table 3: Campaign overview

1 2 3 4 5 6 7

Location Moerkerke Aarschot Roeselare Evergem Zwijndrecht Zwevegem Borgerhout

fraction PM10 PM10 PM10 PM10 PM10 PM10 PM10 PM10 PM10 PM2.5 PM2.5

monitor ESM ESM ESM ESM ESM ESM ESM FDMS FDMS ESM ESM

reference filter Quartz Quartz Quartz Quartz Quartz Quartz Quartz Quartz Quartz Teflon Teflon

period 25.10.2008 - 31.12.2009 25.10.2008 - 31.12.2009 25.10.2008 - 31.12.2009 25.10.2008 - 31.12.2009 25.10.2008 - 31.12.2009 25.10.2008 - 31.12.2009 11.01.2009 - 31.12.2009 25.10.2008 - 31.12.2009 26.06.2009 - 31.12.2009 01.01.2009 - 31.12.2009 05.03.2009 - 31.12.2009

# dp 50 62 67 68 69 69 357 67 181 356 277 176 283 248

8 Oostrozebeke 9 Retie 10 11 Borgerhout Gent

12 Retie PM2.5 FDMS Teflon 16.06.2009 - 31.12.2009 13 Gent PM2.5 FDMS Teflon 05.03.2009 - 31.12.2009 14 Borgerhout PM2.5 FDMS Teflon 25.04.2009 - 31.12.2009 Note: at sites 1, 2, 3, 4, 5, 6 and 8 sampling was done every sixth day 1.5.2 Meteorological data

Fig 13 shows the meteorological data for the VMM meteo station in Gent-Tolhuiskaai (M701) during the project. These data should be seen as indicative since the data were not collected at the sites themselves and daily averages values were used.
35 100 90 80 precip. T wspeed RH 70 60 50 10 40 5 30 0 01/10/08 01/11/08 01/12/08 01/01/09 01/02/09 01/03/09 01/04/09 01/05/09 01/06/09 01/07/09 01/08/09 01/09/09 01/10/09 01/11/09 01/12/09 20 10 0 % RH

30

25

20 mm - C - m/s

15

-5

-10

Fig 13: Meteorological data, Gent-Tolhuiskaai (M701)

Flemish Environment Agency

www.vmm.be

30

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

The gravimetric reference method

2.1 History of filter choice at VMM


Up to 2005 cellulose nitrate filters were used for gravimetric sampling (and heavy metal determination). This was because: - At the time the tested cellulose filters had shown to have better blank values for heavy metals than the tested quartz fibre filters. - No significant difference in PM mass capture between cellulose filters and quartz fibre filters had been observed at a short comparison campaign. Therefore they were thought to be equivalent to quartz filters. - Cellulose filters were (and still are) less fragile. Additional information regarding artefacts on cellulose filters and the availability of quartz fibre filters with low metal content led to new tests on quartz fibre filters. Based on metal blank levels and price Macherey-Nagel QF10 filters were chosen as reference filters. During the PM10 comparison campaign 2005-2006 different type of filters (quartz fibre, teflon, cellulose nitrate) were tested and also different brands of quartz fibre filter (Machery-Nagel, Whatman, Pall, Schleiger & Schuell). This study clearly showed that different filters can lead to significant differences in sampled PM mass. For one campaign the difference between different brands of quartz filters was as high as 29%. For the previous (2006-2007) PM10 and PM2.5 study Whatman QM-A was chosen for gravimetric sampling because EN12341 (PM10) only allows quartz as reference filter and because at the time other European networks (like RIVM-The Netherlands) and JRC-Ispra also used this type of filter. For PM10 gravimetric sampling was also done on teflon filters (Pall Teflo), these filters were used for elemental analysis within the chemical characterisation project.

2.2 Filter choice for this project


PM10 Recent tests by VMM and other members of CEN TC264 WG15 have indicated that Whatman QM-A filters are not the best possible filters for PM10 sampling (e.g. due to irreversible uptake of water vapour). Since the networks in the USA and a high amount of institutes in the scientific community currently prefer Pall Tissuquartz filters, VMM decided to switch to this type and brand. These filters are also 100% quartz (unlike Whatman QM-A which contains 5% borosilicate as binder), so there is no discussion whether this filter is allowed for reference sampling according to EN12341. PM2.5 Although the previous (and first) PM2.5 comparison was done with Whatman QM-A as reference filter VMM decided to use Pall Teflo teflon filters for all PM2.5 reference measurements starting 2009 (e.g. also for the gravimetric determination of the PM2.5 average exposure indicator). This choice was made because the PM2.5 reference method EN14907 allows the use of other types of filter materials such as teflon and it is known that this material is hardly affected by water vapour so more reliable mass data can be obtained 20 . The only drawback of this type of filter is that the back pressure can become too high when the filter load is too high. This occurred on a couple of days in January 2009 when one of the highest PM episodes in recent years occurred.

20

Since artefacts on quartz filter can be higher than 1 g/m3, quartz filters might cause problems for the determination of the reduction of the PM2.5 average exposure indicator, for which small changes in concentration should be measurable. www.vmm.be 31

Flemish Environment Agency

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

Table 4: Overview of filters used for gravimetric reference sampling

Company Pall Pall

Type Quartz fibre Teflon membrane

Name/ Code Tissuquartz 2500 QAT-UP Teflo 2 m

Fraction PM10 PM2.5

2.3 Field comparisons & tests


2.3.1 Filter comparison: quartz vs teflon filter Simultaneous PM2.5 sampling with both quartz and teflon filters was carried out at two sites: the rural background site of Retie and the urban background site of Borgerhout. Fig 14 shows that the ratio between the results of the two filters does not show any clear dependence on the time of year, while Fig 15 and Fig 16 show that there is a good correlation over the full concentration range.
Table 5: Comparison between teflon and quartz filters

site Retie Borgerhout

Orthogonal regression T = 1.02 Q - 0.38 g/m3 T = 1.01 Q - 0.90 g/m3

Ratio of averages 1.007 1.049

Average difference 0.11 g/m3 0.78 g/m3

The relation between the two types of filters is presented in Table 5. Compared to the last study (2006-2007), when quartz filters gave on average 8% more PM10 mass than teflon filters, this time the differences were much smaller. In Retie the PM2.5 concentrations on quartz filters were on average 0.7% higher than on teflon filters, while in Borgerhout this was 4.9%. Overall the difference was 1.9%, although this number is somewhat biased since there were only 141 data points from Borgerhout versus 340 from Retie. When only the period was used in which both sites had data an overall average difference of 3.8% was found (2.5% in Retie and 4.9% in Borgerhout). The fact that the difference between quartz and teflon was smaller in this project can have multiple reasons: another type of quartz filter no additional pre-heating of the quartz filters a different PM fraction
2.00 1.80 1.60 1.40 1.20 1.00 0.80 0.60 0.40 0.20 0.00 01 okt 09 01 nov 09 01 aug 09 01 mei 09 01 dec 09 01 mrt 09 01 apr 09 01 jun 09 01 sep 09 01 jan 09 01 feb 09 01 jul 09

Q/T (Retie) Q/T(Borgerhout)

Fig 14: Ratio of quartz vs teflon concentration for Retie and Borgerhout

Flemish Environment Agency

www.vmm.be

32

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

70

60

50 Ret_T (g/m3)

40

30

20

10

0 0 10 20 30 40 50 60 70 Ret_Q (g/m3)

Fig 15: Scatter plot of teflon vs quartz filters for Retie

70

60

50 Bor_T (g/m3)

40

30

20

10

0 0 10 20 30 40 50 60 70 Bor_Q (g/m3)

Fig 16: Scatter plot of teflon vs quartz filters for Borgerhout

Flemish Environment Agency

www.vmm.be

33

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

2.3.2 Field blanks The PM2.5 reference method EN14907 states that if the blank masses are more than 60 g for LVS (equivalent to a difference in measured concentration of 1 g/m at nominal flow and 24h sampling) the reason shall be investigated. However is does not say that these values should be considered invalid, nor that they can be subtracted. In fact this point is one of the main issues in the current revision of the PM reference method. In the VMM network field blanks remain in the instrument for the full 14 days that a batch spends in the field. From the start they are placed in the right container that stores the sampled filters. Blank values are checked during validation but are not subtracted. A range of statistical parameters regarding both quartz and teflon field blanks are presented in Table 6.
Table 6: Statistical parameters regarding quartz and teflon field blanks

Min (g) Quartz Teflon 10 10

P5% (g) 40 10

Median (g) 80 30

Mean (g) 79 48

P95% (g) 120 107

Max (g) 150 150

% > 60 g 68% 32%

# 85 59

As in the previous project, the quartz filters give a higher field blank than the teflon filters, probably due to a higher uptake of water (and other gasses?). Compared to the previous project, the teflon field blank is very constant (average 2006-2007: 47 g) but the quartz field blank is significantly lower this time (average 2006-2007: 110 g). This is probably due to the different type of quartz filter and the fact that no additional pre-heating was carried out in this project. The current results for the quartz field blank are also in good comparison with the results of the European comparison results of the Joint Research Centre in Ispra 21 . The mean (median) equivalent concentration of the field blanks (assuming a sampled volume of 55 m3) are 1.44 (1.45) g/m3 for quartz and 0.87 (0.55) g/m3 for teflon. The mean (median) difference between the two types of filters is 31 g (50 g), which would results in a difference in equivalent concentration of 0.56 (0.91) g/m3. This is close to the average difference found in 2.3.1. Although is it not sure whether these results should be compared since a field blank (through which no air actively passes) may behave differently to a sampled filter (through which 55 m3 of air passes).

21

Oral communication at CEN TC264 WG15

Flemish Environment Agency

www.vmm.be

34

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

3
Note:

Comparisons with automated monitors


The diagonal lines in the scatter plots present the 1:1 ratio. Concentrations are all in g/m3.

Flemish Environment Agency

www.vmm.be

35

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

3.1 Time & scatter plots PM10 - ESM


ESM Moerkerke
200 REF 180 160 140 PM conc. (g/m3) 120 100 80 60 40 20 0 01 okt 08 01 nov 08 01 dec 08 01 jan 09 01 feb 09 01 mrt 09 01 apr 09 01 mei 09 01 jun 09 01 jul 09 01 aug 09 01 sep 09 01 okt 09 01 nov 09 01 dec 09 01 jan 10 ESM REF/ESM 1.60 1.40 1.20 1.00 0.80 0.60 0.40 0.20 0.00 1.80 2.00

Fig 17: Concentrations and daily ratios for PM10 ESM in Moerkerke

100 90 80 70 60 AMS 50 40 30 20 10 0 0 10 20 30 40 50 Ref


Fig 18: Scatter plot for PM10 ESM in Moerkerke

60

70

80

90

100

Flemish Environment Agency

www.vmm.be

REF/ESM

36

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

ESM Aarschot
200 REF 180 160 140 PM conc. (g/m3) 120 100 80 60 40 20 0 01 okt 08 01 nov 08 01 dec 08 01 jan 09 01 feb 09 01 mrt 09 01 apr 09 01 mei 09 01 jun 09 01 jul 09 01 aug 09 01 sep 09 01 okt 09 01 nov 09 01 dec 09 01 jan 10 ESM REF/ESM 1.60 1.40 1.20 1.00 0.80 0.60 0.40 0.20 0.00 1.80 2.00

Fig 19: Concentrations and daily ratios for PM10 ESM in Aarschot
100 90 80 70 60 AMS 50 40 30 20 10 0 0 10 20 30 40 50 Ref 60 70 80 90 100

Fig 20: Scatter plot for PM10 ESM in Aarschot

Flemish Environment Agency

www.vmm.be

REF/ESM

37

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

ESM Roeselare
200 REF 180 160 140 PM conc. (g/m3) 120 100 80 60 40 20 0 01 okt 08 01 nov 08 01 dec 08 01 jan 09 01 feb 09 01 mrt 09 01 apr 09 01 mei 09 01 jun 09 01 jul 09 01 aug 09 01 sep 09 01 okt 09 01 nov 09 01 dec 09 01 jan 10 ESM REF/ESM 1.60 1.40 1.20 1.00 0.80 0.60 0.40 0.20 0.00 1.80 2.00

Fig 21: Concentrations and daily ratios for PM10 ESM in Roeselare
100 90 80 70 60 AMS 50 40 30 20 10 0 0 10 20 30 40 50 Ref 60 70 80 90 100

Fig 22: Scatter plot for PM10 ESM in Roeselare

Flemish Environment Agency

www.vmm.be

REF/ESM

38

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

ESM Zwevegem
200 REF 180 160 140 PM conc. (g/m3) 120 100 80 60 40 20 0 01 okt 08 01 nov 08 01 dec 08 01 jan 09 01 feb 09 01 mrt 09 01 apr 09 01 mei 09 01 jun 09 01 jul 09 01 aug 09 01 sep 09 01 okt 09 01 nov 09 01 dec 09 01 jan 10 ESM REF/ESM 1.60 1.40 1.20 1.00 0.80 0.60 0.40 0.20 0.00 1.80 2.00

Fig 23: Concentrations and daily ratios for PM10 ESM in Zwevegem
100 90 80 70 60 AMS 50 40 30 20 10 0 0 10 20 30 40 50 Ref 60 70 80 90 100

Fig 24: Scatter plot for PM10 ESM in Zwevegem

Flemish Environment Agency

www.vmm.be

REF/ESM

39

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

ESM Evergem
200 REF 180 160 140 PM conc. (g/m3) 120 100 80 60 40 20 0 01 okt 08 01 nov 08 01 dec 08 01 jan 09 01 feb 09 01 mrt 09 01 apr 09 01 mei 09 01 jun 09 01 jul 09 01 aug 09 01 sep 09 01 okt 09 01 nov 09 01 dec 09 01 jan 10 ESM REF/ESM 1.60 1.40 1.20 1.00 0.80 0.60 0.40 0.20 0.00 1.80 2.00

Fig 25: Concentrations and daily ratios for PM10 ESM in Evergem
100 90 80 70 60 AMS 50 40 30 20 10 0 0 10 20 30 40 50 Ref 60 70 80 90 100

Fig 26: Scatter plot for PM10 ESM in Evergem

Flemish Environment Agency

www.vmm.be

REF/ESM

40

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

ESM Zwijndrecht
200 REF 180 160 140 PM conc. (g/m3) 120 100 80 60 40 20 0 01 okt 08 01 nov 08 01 dec 08 01 jan 09 01 feb 09 01 mrt 09 01 apr 09 01 mei 09 01 jun 09 01 jul 09 01 aug 09 01 sep 09 01 okt 09 01 nov 09 01 dec 09 01 jan 10 ESM REF/ESM 1.60 1.40 1.20 1.00 0.80 0.60 0.40 0.20 0.00 1.80 2.00

Fig 27: Concentrations and daily ratios for PM10 ESM in Zwijndrecht
100 90 80 70 60 AMS 50 40 30 20 10 0 0 10 20 30 40 50 Ref 60 70 80 90 100

Fig 28: Scatter plot for PM10 ESM in Zwijndrecht

Flemish Environment Agency

www.vmm.be

REF/ESM

41

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

ESM Borgerhout (dataset limited to days of hotspot project)


200 REF 180 160 140 PM conc. (g/m3) 120 100 80 60 40 20 0 01 okt 08 01 nov 08 01 dec 08 01 jan 09 01 feb 09 01 mrt 09 01 apr 09 01 mei 09 01 jun 09 01 jul 09 01 aug 09 01 sep 09 01 okt 09 01 nov 09 01 dec 09 01 jan 10 ESM REF/ESM 1.60 1.40 1.20 1.00 0.80 0.60 0.40 0.20 0.00 1.80 2.00

Fig 29: Concentrations and daily ratios for PM10 ESM Borgerhout (limited dataset)
100 90 80 70 60 AMS 50 40 30 20 10 0 0 10 20 30 40 50 Ref 60 70 80 90 100

Fig 30: Scatter plot for PM10 ESM in Borgerhout (limited dataset)

Flemish Environment Agency

www.vmm.be

REF/ESM

42

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

ESM Borgerhout (full dataset)


200 REF 180 160 140 PM conc. (g/m3) 120 100 80 60 40 20 0 01 okt 08 01 nov 08 01 dec 08 01 jan 09 01 feb 09 01 mrt 09 01 apr 09 01 mei 09 01 jun 09 01 jul 09 01 aug 09 01 sep 09 01 okt 09 01 nov 09 01 dec 09 01 jan 10 ESM REF/ESM 1.60 1.40 1.20 1.00 0.80 0.60 0.40 0.20 0.00 1.80 2.00

Fig 31: Concentrations and daily ratios for PM10 ESM in Borgerhout (full dataset)
160 140

120

100 AMS 80

60

40 20

0 0 20 40 60 80 Ref 100 120 140 160

Fig 32: Scatter plot for PM10 ESM in Borgerhout (full dataset)

Flemish Environment Agency

www.vmm.be

REF/ESM

43

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

3.2 Time & scatter plots PM10 - FDMS


FDMS Retie
200 REF 180 160 140 PM conc. (g/m3) 120 100 80 60 40 20 0 01 okt 08 01 nov 08 01 dec 08 01 jan 09 01 feb 09 01 mrt 09 01 apr 09 01 mei 09 01 jun 09 01 jul 09 01 aug 09 01 sep 09 01 okt 09 01 nov 09 01 dec 09 01 jan 10 FDMS REF/FDMS 1.60 1.40 REF/FDMS 1.20 1.00 0.80 0.60 0.40 0.20 0.00 1.80 2.00

Fig 33: Concentrations and daily ratios for PM10 FDMS in Retie

100 90 80 70 60 AMS 50 40 30 20 10 0 0 10 20 30 40 50 Ref


Fig 34: Scatter plot for PM10 FDMS in Retie

60

70

80

90

100

Flemish Environment Agency

www.vmm.be

44

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

FDMS Oostrozebeke
200 REF 180 160 140 PM conc. (g/m3) 120 100 80 60 40 20 0 01 okt 08 01 nov 08 01 dec 08 01 jan 09 01 feb 09 01 mrt 09 01 apr 09 01 mei 09 01 jun 09 01 jul 09 01 aug 09 01 sep 09 01 okt 09 01 nov 09 01 dec 09 01 jan 10 FDMS REF/FDMS 1.60 1.40 REF/FDMS 1.20 1.00 0.80 0.60 0.40 0.20 0.00 1.80 2.00

Fig 35: Concentrations and daily ratios for PM10 FDMS in Oostrozebeke

100 90 80 70 60 AMS 50 40 30 20 10 0 0 10 20 30 40 50 Ref


Fig 36: Scatter plot for PM10 FDMS in Oostrozebeke

60

70

80

90

100

Flemish Environment Agency

www.vmm.be

45

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

3.3 Time & scatter plots PM2.5 - ESM


ESM Borgerhout
200 180 160 140 PM conc. (g/m3) 120 100 80 60 40 20 0 REF ESM REF/ESM 0.60 0.40 0.20 0.00 2.00 1.80 1.60 1.40 1.20 1.00 0.80

01 jan 09

01 feb 09

01 mrt 09

01 apr 09

01 mei 09

01 jun 09

01 jul 09

01 aug 09

01 sep 09

01 okt 09

01 nov 09

01 dec 09

01 jan 10

Fig 37: Concentrations and daily ratios for PM2.5 ESM in Borgerhout

100 90 80 70 60 AMS 50 40 30 20 10 0 0 10 20 30 40 50 Ref


Fig 38: Scatter plot for PM2.5 ESM in Borgerhout

60

70

80

90

100

Flemish Environment Agency

www.vmm.be

REF/ESM

46

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

ESM Gent
200 REF 180 160 140 PM conc. (g/m3) 120 100 80 60 40 20 0 ESM REF/ESM 1.60 1.40 1.20 1.00 0.80 0.60 0.40 0.20 0.00 1.80 2.00

01 jan 09

01 feb 09

01 mrt 09

01 apr 09

01 mei 09

01 jun 09

01 jul 09

01 aug 09

01 sep 09

01 okt 09

01 nov 09

01 dec 09

01 jan 10

Fig 39: Concentrations and daily ratios for PM2.5 ESM in Gent
100 90 80 70 60 AMS 50 40 30 20 10 0 0 10 20 30 40 50 Ref 60 70 80 90 100

Fig 40: Scatter plot for PM2.5 ESM in Gent

Flemish Environment Agency

www.vmm.be

REF/ESM

47

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

3.4 Time & scatter plots PM2.5 - FDMS


FDMS Retie
200 180 160 140 PM conc. (g/m3) 120 100 80 60 40 20 0 REF FDMS REF/FDMS 1.60 1.40 REF/FDMS 1.20 1.00 0.80 0.60 0.40 0.20 0.00 2.00 1.80

01 jan 09

01 feb 09

01 mrt 09

01 apr 09

01 mei 09

01 jun 09

01 jul 09

01 aug 09

01 sep 09

01 okt 09

01 nov 09

01 dec 09

01 jan 10

Fig 41: Concentrations and daily ratios for PM2.5 FDMS in Retie

100 90 80 70 60 AMS 50 40 30 20 10 0 0 10 20 30 40 50 Ref


Fig 42: Scatter plot for PM2.5 FDMS in Retie

60

70

80

90

100

Flemish Environment Agency

www.vmm.be

48

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

FDMS Borgerhout
200 180 160 140 PM conc. (g/m3) 120 100 80 60 40 20 0 REF FDMS REF/FDMS 2.00 1.80 1.60 1.40 REF/FDMS 1.20 1.00 0.80 0.60 0.40 0.20 0.00

01 jan 09

01 feb 09

01 mrt 09

01 apr 09

01 mei 09

01 jun 09

01 jul 09

01 aug 09

01 sep 09

01 okt 09

01 nov 09

01 dec 09

01 jan 10

Fig 43: Concentrations and daily ratios for PM2.5 FDMS in Borgerhout

100 90 80 70 60 AMS 50 40 30 20 10 0 0 10 20 30 40 50 Ref


Fig 44: Scatter plot for PM2.5 FDMS in Borgerhout

60

70

80

90

100

Flemish Environment Agency

www.vmm.be

49

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

FDMS Gent
200 REF 180 160 140 PM conc. (g/m3) 120 100 80 60 40 20 0 FDMS REF/FDMS 1.60 1.40 REF/FDMS 1.20 1.00 0.80 0.60 0.40 0.20 0.00 1.80 2.00

01 jan 09

01 feb 09

01 mrt 09

01 apr 09

01 mei 09

01 jun 09

01 jul 09

01 aug 09

01 sep 09

01 okt 09

01 nov 09

01 dec 09

01 jan 10

Fig 45: Concentrations and daily ratios for PM2.5 FDMS in Gent

100 90 80 70 60 AMS 50 40 30 20 10 0 0 10 20 30 40 50 Ref


Fig 46: Scatter plot for PM2.5 FDMS in Gent

60

70

80

90

100

Flemish Environment Agency

www.vmm.be

50

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

Demonstration of equivalence

4.1 Overview
The most important criteria for PM2.5 and PM10 in the current equivalence document are: The between sampler uncertainty of the Reference Method shall be smaller than 2 g/m3 The between sampler uncertainty of the Candidate Method shall be smaller than 3 g/m3 A minimum of 4 comparisons of at least 40 measurements at a minimum of 2 sites The expanded uncertainty of the candidate method at the limit value shall be smaller than 25% Even when the candidate method passes the equivalence test a correction may be applied to improve the accuracy of the candidate method. When the candidate method is not accepted as equivalent method it is permitted to apply a calibration factor or term resulting from the regression equation obtained for the full dataset. Note: For PM10 the daily limit value of 50 g/m3 shall be used. For PM2.5 a daily limit value does not exist, but a tentative value of 30 g/m3 was recently proposed by CEN TC264 WG15 solely for the purpose of equivalence demonstration.

4.2 Suitability of datasets


The document demonstration of equivalence states that of the full dataset at least 20% of the results shall be greater than 50% of the limit value (50 g/m for PM10, 30 g/m3 as indicative daily limit value for PM2.5). This criteria is met for all comparisons.

4.3 Between sampler uncertainty


The Reference method In 2005 an intercomparison test with two Leckel instruments gave a between sampler uncertainty of 0.46 g/m . Although a different type of filter (quartz; Machery-Nagel QF10) was used at that time, we assume that the uncertainty value is still valid and that the criterion for the reference method (< 2 g/m3) is met. Automated samplers In 2004 an intercomparison test with two ESM instruments gave a between sampler uncertainty of 1.44 g/m3, so the criterion for the candidate method is met. For TEOM-FDMS VMM has not yet carried out between sampler uncertainty experiments. But since these instruments have been featured in foreign equivalence studies we refer to those for this aspect in the equivalence demonstration.

4.4 Reference vs. automated


The second part of the equivalence demonstration, focuses on the relation between the reference and the automated monitor (AM) and the expanded uncertainty of the AM at the limit value. This expanded uncertainty should be lower than 25% at the appropriate limit value, which implies that the uncertainty should be lower than 12.5% (a coverage factor k=2 can be used to calculate the expanded uncertainty in view of the large number of experimental values). There are three possible scenarios for the comparison between an automated method and the reference method: 1. The results of the AM meet the criterion 2. The results of the AM meet the criterion after applying a correction factor or equation 3. The results of the AM do not meet the criterion

Flemish Environment Agency

www.vmm.be

51

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

4.4.1 Uncertainty at the limit value before calibration Table 7 show the relative uncertainties at the appropriate limit value (50 g/m3 for PM10 and the tentative value of 30 g/m3 for PM2.5). Except for one station with an ESM monitor and two stations with an FDMS monitor none of the instruments meet the uncertainty at the limit value criterion without calibration.
Table 7: Uncertainties at the limit values before calibration

site 1 2 3 4 5 6 7 Moerkerke Aarschot Roeselare Evergem Zwijndrecht Zwevegem Borgerhout

fraction PM10 PM10 PM10 PM10 PM10 PM10 PM10 PM10 PM10 PM2.5 PM2.5 PM2.5 PM2.5 PM2.5

monitor ESM ESM ESM ESM ESM ESM ESM FDMS FDMS ESM ESM FDMS FDMS FDMS

reference filter Quartz Quartz Quartz Quartz Quartz Quartz Quartz Quartz Quartz Teflon Teflon Teflon Teflon Teflon

Uncertainty at limit value 11.1 % 18.7 % 14.8 % 17.5 % 18.2 % 19.1 % 23.1 % 8.1 % 19.2 % 21.4 % 20.6 % 39.2 % 16.5 % 7.3 %

8 Oostrozebeke 9 10 11 12 13 14 Retie Borgerhout Gent Retie Gent Borgerhout

4.4.2 Calibration factors & equations In the tables below the factors or equations that were calculated for the different (sets of) comparisons are listed. The factors should be interpreted as follows: Reference value = factor * automated value (+ constant) Note: When comparing the results in the tables below, one should keep in mind that the data for the different locations are not entirely comparable because of the differences in sampling days between the campaigns.

Flemish Environment Agency

www.vmm.be

52

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

PM10 - ESM
Table 8: PM10 - ESM calibration factors

site Moerkerke Aarschot Roeselare Zwevegem Evergem Zwijndrecht Borgerhout lim. Borgerhout all All sites
PM10 - FDMS

# data pairs 50 62 67 69 68 69 59 357 742

Orthogonal regression factor 1.04 1.18 1.11 1.23 1.16 1.14 1.33 1.27 1.20 constant +2.48 +1.92 +1.83 -0.67 +0.65 +2.63 -0.64 +1.01 +1.04

Orthogonal regression through origin factor 1.13 1.25 1.17 1.21 1.19 1.22 1.32 1.29 1.25

Ratio of averages 1.16 1.27 1.19 1.21 1.19 1.24 1.31 1.30 1.25

Table 9: PM10 - FDMS calibration factors

Site Retie Oostrozebeke


PM2.5 - ESM

# data pairs 181 67

Orthogonal regression factor 0.85 0.96 constant -0.35 +2.45

Orthogonal regression through origin factor 0.84 1.03

Ratio of averages 0.84 1.04

Table 10: PM2.5 - ESM calibration factors

Site Borgerhout Gent All data

# data pairs 356 277 633

Orthogonal regression factor 1.22 1.20 1.20 constant +0.73 +0.96 +0.82

Orthogonal regression through origin factor 1.25 1.25 1.25

Ratio of averages 1.26 1.28 1.27

Flemish Environment Agency

www.vmm.be

53

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

PM2.5 - FDMS
Table 11: PM2.5 - FDMS calibration factors

site Retie Borgerhout Gent

# data pairs 176 248 283

Orthogonal regression factor 0.81 0.96 0.95 constant -3.35 +0.36 -0.79

Orthogonal regression through origin factor 0.65 0.98 0.92

Ratio of averages 0.63 0.98 0.91

4.4.3 Discussion of calibration Since the constant factors for the comparisons are all quite low (< 5 g/m3) 22 we consider it justified to apply a calibration factor instead of a correction function based on slope and intercept. Although both the orthogonal regression and the ratio of averages provide a symmetrical approach we prefer the use of the ratio of averages since this gives per definition the best result for the calculation of the overall average. As the previous tables show, the differences between the two methods are very small. In further discussion the factor based on ratios of averages will be used unless stated otherwise. PM10 - ESM Except for the full year dataset in Borgerhout the comparisons were carried out in parallel th (every 6 day) at all sites. This has the advantage that differences in factors can largely be attributed to differences in local conditions and are not effected by differences in meteorological conditions between the comparison campaigns. The instrument in Moerkerke behaved quite different than the other stations. From July to September the ESM gave higher values than the reference method. For the moment it is still unclear what the reason for this is. The suspicious timeframe was not taken into account. Even without these data pairs the factor for Moerkerke was lower than expected. For the full dataset a factor of 1.25 was found. This factor is the same factor that is currently used by RIVM in the Netherlands for the ESM in their nationwide network. The new factor 1.25 is somewhat lower (9%) than the factor that was used before: 1.37. This is probably due to the filter choice. Although the previously used filters were perfectly valid according to the European reference method (EN12341) recent European research (in which VMM played an important role) indicate that the current filter (from Pall) suffers less from artefacts and that this filter can therefore provide a higher quality reference value. PM2.5 - ESM Both Borgerhout (1.26) and Gent (1.28) have a similar factor. For the combined dataset a factor of 1.27 is found. This factor will be used for all sites in Flanders. Compared to the factor found in the previous project (1.46) the new factor is 15% lower. This is most probably mainly due to the filter choice but could partially have been influenced by year to year variability and the types of sites in this project, which were both urban background. In the previous project a high factor at the rural site in Houtem contributed to a high overall factor.

22

The value of 5 g/m3 dates back to an earlier version of a European document dealing with correction factors: Correction factors and PM10 measurements, ETC/ACC Technical paper 2004/4

Flemish Environment Agency

www.vmm.be

54

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

PM10 & PM2.5 FDMS To our knowledge all countries in Europe use a calibration factor of 1.00 (= no correction) for FDMS instruments. For two sites a calibration factor close to 1.00 was found: Oostrozebeke (PM10: 1.04) and Borgerhout (PM2.5: 0.98) For the three other comparisons quite low factors were found (0.84 for PM10 and 0.63 for PM2.5 in Retie and 0.91 for PM2.5 in Gent), indicating that the FDMS instruments overestimate the PM concentrations at these sites. In all three suspicious cases the FDMS instrument was operating inside a small trailer, which suggest that there is some technical issue with the FDMS trailer setup. This setup is still in testing phase and the results are not yet used to inform the public. Possibly there is a problem with the stability of the environmental conditions (the time plots in section 3.2 and 3.4 show a seasonal effect on the daily correction factor for the three 23 datasets). Similar issues have been reported in other networks . The FDMS trailer datasets are therefore not taken into account. Due to this the criteria for the minimum number of comparisons is no longer met for this study. VMM will continue to use a factor 1.00 for the FDMS and will further investigate the issues with instruments operating in trailers. 4.4.4 Uncertainty at the limit value after calibration Using the factors that were decided upon in the previous section the original datasets (that were taken into account) were calibrated and the uncertainty was recalculated. Except for Moerkerke all datasets now meet the uncertainty criterion (< 12.5%).
Table 12: Uncertainties at the limit values after calibration

site Moerkerke Aarschot Roeselare Evergem Zwijndrecht Zwevegem Borgerhout All sites Oostrozebeke Borgerhout Gent Borg. + Gent Borgerhout

fraction PM10 PM10 PM10 PM10 PM10 PM10 PM10 PM10 PM10 PM2.5 PM2.5 PM2.5 PM2.5

monitor ESM ESM ESM ESM ESM ESM ESM ESM FDMS ESM ESM ESM FDMS

factor 1.25 1.25 1.25 1.25 1.25 1.25 1.25 1.25 1.00 1.27 1.27 1.27 1.00

Uncertainty at limit value 16.4 % 7.6 % 10.6 % 12.3 % 11.8 % 8.7 % 7.9 % 8.8 % 8.1 % 10.7 % 8.6 % 9.8 % 7.3 %

23

Unofficial communications at CEN TC264 WG15

Flemish Environment Agency

www.vmm.be

55

Comparative PM10 and PM2.5 measurements in Flanders, 2008-2009

Conclusions
A total of over 2300 comparisons were carried out between October 2008 and December 2009. Both the ESM and FDMS monitors were tested in the PM2.5 and PM10 fraction. To reduce the effects of artefacts in the reference method the reference filters were changed compared to the previous study. The reference was changed from Whatman QM-A to Pall Tissuquartz 2500 QAT-UP for PM10 and to Pall Teflo 2 m (a teflon membrane filter) for PM2.5. The change in reference filter appears to have led to higher quality and somewhat lower reference values. Compared to the previous VMM studies the difference between the teflon and quartz reference filters was smaller. Overall the quartz filter gave only 2% more than the teflon filters. This seems to confirm the fact that the Pall quartz filters that were used in this study suffer less from artefacts than the Whatman quartz filters that were used in the previous study. The field blanks for the Pall quartz filters were also lower than before. The average field blank was 79 g for quartz filters and 48 g for teflon filters, which is equivalent to 1.4 3 3 g/m and 0.9 g/m respectively. Although most of the quartz field blanks were higher than 60 g (and indicative value for investigation in the reference standard) these values are in good agreement with findings by others (e.g. the European Joint Research Centre in Ispra). Because calibration functions based on orthogonal regression all had small intercepts, it was decided to use calibration factors. These factors were calculated based on the ratio of averages of the two datasets. This approach was preferred because it gives, per definition, the best result for the calculation of the overall average. Also, the difference between this approach and orthogonal regression through the origin was always very small. For ESM in the PM10 fraction an overall factor of 1.25 was calculated. This new factor is lower (9%) than the previous factor in use (1.37). After recalculation with the new factor the overall dataset and six of the seven subsets pass the uncertainty criterion at the limit 3 value of 50 g/m . The new factor is in very good agreement with the situation in The Netherlands where a factor of 1.25 is also currently used by RIVM for the ESM in their nationwide network. For ESM in the PM2.5 fraction a factor of 1.27 was calculated. This factor was 15% lower than the previous factor. After recalculation with the new factor the overall dataset and the two subsets pass the uncertainty criterion at the tentative limit value of 30 g/m3. For FDMS the three instruments that were operated in small trailers (which are still in testing phase and have not been used for routine monitoring) showed an overestimation of the PM concentrations, indicating that there probably are technical issues with the FDMS trailer setup (possibly unstable environmental conditions). These datasets were considered invalid and the issue is further investigated. For the remaining FDMS instruments no calibration was needed (= a factor of 1.00) to pass the uncertainty criteria at the PM10 and PM2.5 limit values. Therefore VMM will continue to use the FDMS instruments without calibration.. In the future a continuation of comparisons will be necessary to evaluate the new factors and to get a better understanding of differences from year to year and from site tot site.

Flemish Environment Agency

www.vmm.be

57

You might also like