PM2.5 measurement results with the reference method and Modelling for 2008 in The Netherlands Ronald Hoogerbrugge, Dave de Jonge* and

Jan Matthijsen# * GGD Amsterdam #PBL

Status end 2007 • automatic PM2.5 measurements using TEOM • TEOM: equivalence with reference not shown (presumably difficult) • several partners (RIVM, GGD Amsterdam, DCMR, provences) preferred to buy the same instrument • Careful selection and testing necessary • Impossible to perform before 1/1/2008 • however new directive requires starting at 1/1/2008 • Results shall be comparable until 2020 in order to demonstrate 15-20% reductions

Status end 2007 Conclusion: Quality requirements can not be met within this period of time Decision: Start measurements with the reference method How to find the necessary equipment/ (weighing) capacity locations ? -> National corporation

Agglomerations + municipalities >100.000 Gesorteerd op inwonersaantal per 1-1-08 uitbreiding 2008 Inwonersaantalaantal stations aantal stations Indeling Amsterdam/Haarlem A'dam 1538435 2 2 Rotterdam/Dordrecht R'dam 1279364 2 2 Den Haag/Leiden 1061200 Den Haag 1 1 Utrecht 453619 Utrecht 1 Eindhoven 427000 Eindhoven 1 Kerkrade/Heerlen 240326 Heerlen 1 1 Tilburg 200819 Zuid Groningen 181171 Noord 1 1 Almere 179695 Noord Breda 170029 Zuid 1 1 Nijmegen 160215 Midden 1 Apeldoorn 155807 Midden Enschede 154427 Noord 1 Arnhem 142382 Midden Amersfoort 138026 Midden 's-Hertogenbosch 135183 Zuid 119607 Maastricht Zuid Zoetermeer 117502 Zuid 113857 Zwolle Noord Emmen 108711 Noord Ede 107274 Midden Totaal 7184649 8 12 Agglomoraties

8 is minimum number urban background measurement locations

Strategy/ corporation Area Aggl. Heerlen Zone Zuid Aggl. Den Haag Aggl. Rotterdam Zone Noord Aggl. Rotterdam Zaandam Amsterdam

Owner/ operator location RIVM Heerlen-Deken Nicolayestraat RIVM Breda-Bastenakenstraat RIVM Den Haag-Rebequestraat RIVM Rotterdam_Schiedamsevest RIVM Groningen-Neijensteinheerd DCMR Schiedam GGD-Amsterdam Zaandam GGD-Amsterdam Overtoom

type Urban background Urban background Urban background Urban background Urban background Urban background Urban background Urban background

Zone Zuid Zone Zuid Zone Midden Zone Midden Zone Noord

RIVM Prov. N. Brabant RIVM RIVM RIVM

Vredepeel Fijnaart Wieringerwerf Wekerom Kollumerwaard

regional regional regional regional regional

Aggl. Rotterdam Zone Zuid Aggl. Heerlen Zone Midden Aggl. Amsterdam

RIVM RIVM RIVM RIVM GGD-Amsterdam

Rotterdam-Bentinckplein Breda-Tilburgseweg Heerlen-Looijerstraat Breukelen A'dam A10 zuid

trafic trafic trafic trafic trafic

Measurement strategy 1/1/ 2008 8 urban background 5 traffic 5 regional In the course of 2008 extended with 4 until 12 urban background

Results Location 014 Amsterdam GGD 018 Amsterdam GGD 701 Zaandam GGD 131Vredepeel 136 Heerlen 137 Heerlen 230 Biest Houtakker 240 Breda 241 Breda 246 Fijnaart 404 Den Haag 418 Rotterdam 433 Vlaardingen 444 De Zilk 448 Rotterdam 10 Schiedam DCMR 538 Wieringerwerf 620 Cabauw 641 Breukelen 738 Wekeron 742 Nijmegen 807 Hellendoorn 934 Kollumerwaard 938 Groningen

Average

Type urban traffic urban regional traffic urban regional traffic urban regional urban urban traffic regional traffic urban regional regional traffic regional urban regional regional urban

Average availab 2008 le 14.8 82% 16.5 92% 14.3 89% 17.5 40% 20.1 25% 16.2 93% 17.2 11% 19.8 8% 17.6 93% 16.4 77% 17.0 87% 17.5 91% 17.3 65% 14.0 45% 18.3 77% 15.9 59% 13.1 22% 16.3 50% 16.1 13% 17.5 17% 22.1 10% 14.5 29% 13.4 27% 18.8 96%

16.8

14.8 16.5 14.3 17.5 20.1 16.2 17.2 19.8 17.6 16.4 17.0 17.5 17.3 14.0 18.3 15.9 13.1 16.3 16.1 17.5 22.1 14.5 13.4 18.8 regional 15.5

urban 17.1

traffic 18.0

Alternative approach for missing data Previous approach:

result missing data are skipped

Alternative approach :

impute missing data by value which is consistent with the level for that day and that location

xij = xi .

x. j x

with

∑x

∑

ij

xi . =

j =1:k

k

, x. j =

∑

xij

i =1:366

366

en

x=

xij

j =1:k ,i =1:366

366k

Uncertainty in yearly average is estimated with the assumption that correlation between the uncertainty of daily values may exist

Location 014 Amsterdam GGD 018 Amsterdam GGD 701 Zaandam GGD 131Vredepeel 136 Heerlen 137 Heerlen 230 Biest Houtakker 240 Breda 241 Breda 246 Fijnaart 404 Den Haag 418 Rotterdam 433 Vlaardingen 444 De Zilk 448 Rotterdam 10 Schiedam DCMR 538 Wieringerwerf 620 Cabauw 641 Breukelen 738 Wekeron 742 Nijmegen 807 Hellendoorn 934 Kollumerwaard 938 Groningen

Average

Type urban traffic urban regional traffic urban regional traffic urban regional urban urban traffic regional traffic urban regional regional traffic regional urban regional regional urban

Average uncertainty on Average (imputing limit (25 µg/m3) (skipping) ) 14.8 14.5 10% 16.5 16.2 5% 14.3 15.1 5% 17.5 18.4 20% 20.1 18.5 40% 16.2 16.3 6% 17.2 18.9 22% 19.8 18.2 35% 17.6 18.0 5% 16.4 16.7 8% 17.0 17.0 8% 17.5 18.2 5% 17.3 18.3 11% 14.0 13.2 18% 18.3 18.5 7% 15.9 16.5 9% 13.1 13.9 23% 16.3 15.9 19% 16.1 17.1 19% 17.5 17.4 23% 22.1 20.3 43% 14.5 15.1 26% 13.4 13.2 31% 18.8 18.7 5%

16.8

16.8

14.5 16.2 15.1 18.4 18.5 16.3 18.9 18.2 18.0 16.7 17.0 18.2 18.3 13.2 18.5 16.5 13.9 15.9 17.1 17.4 20.3 15.1 13.2 18.7 regional 15.9

urban 17.2

traffic 17.8

13 19

14

Average 2008 (μg/m3) rural

16 ± 4

urban 17 ± 3 traffic 18 ± 2

15 14 16 14 16 17 16 18 18 17 18

15 17

20 17

18 18

19

18

16

19

Multiple imputation (preliminary results) • • • • • • • •

Impute missing values with starting data Model data for first location using the other locations Impute missing data by prediction + random uncertainty Than 2nd location etc. After all locations (1 iteration) start again at first location 4 cycles assuming normal distributed model deviations 4 cycles assuming log normal distributed model dev. Continue until cycles mix

• Using extreme starting conditions (f.e. 1 cycle missing data are set at 100 and another at 1).

Location

Type

014 Amsterdam GGD 018 Amsterdam GGD 701 Zaandam GGD 131Vredepeel 136 Heerlen 137 Heerlen 230 Biest Houtakker 240 Breda 241 Breda 246 Fijnaart 404 Den Haag 418 Rotterdam 433 Vlaardingen 444 De Zilk 448 Rotterdam 10 Schiedam DCMR 538 Wieringerwerf 620 Cabauw 641 Breukelen 738 Wekerom 742 Nijmegen 807 Hellendoorn 934 Kollumerwaard 938 Groningen

urban traffic urban regional traffic urban regional traffic urban regional urban urban traffic regional traffic urban regional regional traffic regional urban regional regional urban

Average

average 2 stdev on multiple limit (25 imputation µg/m3) 14.6 16.3 14.9 18.3 19.3 16.4 18.0 19.2 17.8 16.8 16.6 18.1 17.9 13.2 18.6 16.8 14.2 16.6 17.4 17.5 20.5 14.6 14.4 18.8

dif with dif with single skipping imp.

1% 0% 1% 2% 3% 1% 3% 4% 1% 1% 1% 1% 2% 2% 1% 2% 5% 2% 3% 4% 6% 3% 8% 1%

16.9 stdev

-0.2 -0.1 0.7 0.8 -0.8 0.2 0.8 -0.6 0.2 0.3 -0.4 0.5 0.6 -0.8 0.3 0.9 1.1 0.3 1.3 0.0 -1.6 0.1 1.0 -0.1

0.1 0.1 -0.1 -0.1 0.8 0.1 -1.0 1.0 -0.2 0.0 -0.3 -0.1 -0.3 0.0 0.1 0.2 0.3 0.7 0.3 0.1 0.1 -0.5 1.2 0.1

0.2 0.7

0.1 0.5

Model calculation

Emission data Dispersion (OPS model) Calibration on measurements

Ontwikkeling van de jaargemiddelde concentratie fijn stof µg/m3

PM10: 2008 relatively low

50 45 40 35 30 25 20 15 10 5 0 1992

1994

1996

1998

2000

2002

Regionale achtergrondstations Stedelijke achtergrondstations Verkeersbelaste stations Grenswaarde

2004

2006

2008 Jaar

Comparison with targets 2008 measurement results are all below 25 μg/m3 and on other spots exceedences of 25 are unlikely Some exceedences of 20 are found both in the measurements and in the model calculation Measured average on 9 urban background locations is 17 μg/m3 (and might be close to 18 over 2009-2011 )

(Be aware that PM10 in 2008 was relatively low in The Netherlands)

Strategy ERT Start = 2009-2010-2011 End = 2018-2019-2020

To do currently selection of automatic monitor measurement strategy for automatic monitor Compared to PM10 focus measurement strategy even more on quality rather than quantity (small local gradients)

Conclusions PM2.5 measurements started at 1/1/2008 using reference method on >20 locations rural 16 ± 4 urban 17 ± 3 and traffic 18 ± 2 (based on PM10 -> 2008 may be relatively low)

Average results (µg/m3)

90 % availability -> loss of most data Simple imputation works well up to 30-40 % availability First results with Multiple imputation are promising even for very small data sets Thanks to colleagues at RIVM, GGD, DCMR and PBL etc.