Uncertainties in atmospheric observations Wenche Aas EMEP/CCC

Uncertainties in atmospheric observations

Uncertainties in atmospheric observations

Wenche Aas EMEP/CCC

Wenche Aas, Measurement uncertainty. TFMM/TFEIP Dublin 22 Oct 2007

Sources of uncertaintiesSources of uncertainties

Sampling and analytical method Detection limit Interference Instrument drift, calibration Positive or negative artefact

Sampling procedure Contamination Temperature and period for storage Transport

Representativity. Local farming (NH3) Nearby roads (NOx; O3) Dust (PM, Ca..)

Lab- and field intercomparison

Field inter-comparison; model comparison

Repr. studies, i.e passive sampling. Model comparison


Monitoring programme:Level 1•Main ions in precipitation and in air •heavy metals in precipitations•ozone •PM10 and PM2.5 mass •meteorology

at ca 125 sites

Level 2, supersite (joint EMEP/GAW)• POPs• Heavy metals in air and aerosols• VOC• EC/OC, OC speciation• Mineral Dust• PM speciation incl. gas particle ratio

+ all level 1 activities

15-20 sitesBoth levels are mandatory by all Parties


Data quality objectivesData quality objectivesAcidifying and eutrophying compounds

o 15-25% uncertainty in annual average (10-15% for indiv meas.)

Heavy metals

o 30% uncertainty in annual average (15-25% for indiv meas.)

o 40 % uncertainty for As, Cd, Ni in the EU 4th DD (70% in dep)

o 50% uncertainty for Hg (total gas) in the EU 4th DD

POPs (not defined in EMEP)

o 50% uncertainty for PAH in the EU 4th DD (70% in dep)

PM (not defined in EMEP)

o 25% accuracy in continuous measurements EU 1st DD

Photooxidants (not defined in EMEP)

o: 15% accuracy in continuous O3, NOx measurements, EU 3rd DD


Lab intercomparisons annually, 2005Lab intercomparisons annually, 2005

Spread: 2RSD % Bias: RB %


WMO

ICP


0

1

2

3

4

5

Nov-02 Dec-02 Jan-03 Feb-03 Mar-03 Apr-03 May-03 Jun-03 Jul-03 Aug-03 Sep-03 Oct-03

g-S

/m3

Preila

NILU

y = 0.83x + 0.05

R2 = 0.960

2

4

6

0 2 4 6

Preila

NIL

U

all for slope

Preila (LT) using filterpack

SO2 field intercomparisonSO2 field intercomparisonZarra (ES) , abs (H202) and monitor

0

1

2

3

4

5

Jun-00 Aug-00 Oct-00 Dec-00 Mar-01 Apr-01

g-S

/m3

Reference filterpack

Absorption, H2O2

UV fluoresence monitor

0

2

4

6

8

10

0 30 60 90 120 150 180 210 240 270 300 330 360

julian day, 2000

gS

/m3

Filterpack

TCM

y = 1.18x - 0.05

R2 = 0.87

0

2

4

6

8

10

0 2 4 6 8 10TCM

filt

erp

ac

k

0

0.5

1

1.5

2

2.5

0 30 60 90 120 150 180 210 240 270 300 330 360

julian day, 2000

gS

/m3

Filterpack

TCM

y = 0.87x + 0.34

R2 = 0.26

0

0.5

1

1.5

2

2.5

0 0.5 1 1.5 2 2.5

TCM

filt

erp

ac

k

TCM ain Germany (historic data) at DE09 (left and DE03 (right)


SO2. UV fluoresence monitor, interference

0

2

4

6

8

samples from Jan 02 - Jan 03

g

-S/m

2

NILU - filterpackEMPA -not corrEMPA -corr for NOEMPA -corr for NO and H20

'

y = 0.40x + 0.24

R2 = 0.66

y = 0.69x + 0.13

R2 = 0.82

0

4

8

0 4 8

EMPA

NIL

U


NO2. Chemiluminisence (Mo converter)NO2. Chemiluminisence (Mo converter)

0

5

10

15

20

samples from Nov 01 - Nov 02

g-N

/m3

NILU -glass sinters

EMPA (Monitor)

y = 0.68x - 0.17

R2 = 0.77

0

10

20

0 10 20

EMPA

NIL

U

Not selective for NO2


QA flagQA flagLab flag

Field flag

Green: Bias < 10%Spread <20% (S,N)

Blue:Bias < 40%Spread <40% (S,N)

Green: Bias < 10%Spread <25%

Blue:Bias < 50%Spread <50%


QA Flag for main ions in 2005QA Flag for main ions in 2005In precip In air

Analysis in lab are in general better than 20% (both in air and precip)

Total uncertainty (field intercomp):

about half the measurements is better than 25%

rest better than 50%


Lab intercomparison of HM annuallyLab intercomparison of HM annually

Average per cent error, 2005


Hg intercomparison at DE02 in 2006Hg intercomparison at DE02 in 2006

-50

-40

-30

-20

-10

0

10

20

30

40

50

J G X D B H E I C A F

Dif

fere

nce

s in

%

Data quality objective within EMEP (accuracy in annual average)

Tekran

Tekran

Tekran

Tekran TekranManual

Manual

Manual

Other Other Other

Laboratory

2 9 7 3 1 4 8 5

20

18

16

14

12

10

8

VR

Vr (w-o)

(w-o)

(w-o)(w-o)

(bulk)(bulk) (bulk)

(bulk)

Data quality objective within EMEP (30% in annual average)

Ove

rall

wei

ghte

d m

ean

in n

g/L

SR = ± 5.8 ng/L

SR

Tot Hg(g) in air

Tot Hg in precip


Heavy metal deposition, CEN WG20Heavy metal deposition, CEN WG20

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0

WO Bulk WO Bulk WO Bulk WO Bulk

Pb As Cd Ni

g/m

2 d

ay

Funnel

Sampler

WO Bulk Avg SD rel SD nrCd 0.11 0.10 0.10 ± 0.02 20 % 17Pb 4.2 4.4 4.3 ± 0.7 15 % 17As 0.4 0.4 0.4 ± 0.1 13 % 17Ni 1.6 1.4 1.5 ± 0.6 43 % 16

Birkenes, wet only - bulk

Birkenes, comparing wet only (analysed at UBA) and bulk (NILU)


Parallell wet only, CEN WG 20Parallell wet only, CEN WG 20

Avg SD rel SD nr Avg SD rel SD nr Avg SD rel SD nr Avg SD rel SD nrCd 0.18 ± 0.03 18 % 11 Cd 0.05 ± 0.01 13 % 19 Cd 0.07 ± 0.01 9 % 17 Cd 0.44 ± 0.04 9 % 19Pb 7.4 ± 0.5 7 % 11 Pb 2.2 ± 0.3 15 % 19 Pb 3.2 ± 0.3 10 % 17 Pb 28 ± 3 11 % 19As 0.7 ± 0.1 19 % 11 As 0.18 ± 0.02 11 % 19 As 0.26 ± 0.02 6 % 17 As 0.72 ± 0.05 7 % 19Ni 2.8 ± 0.3 12 % 10 Ni 0.8 ± 0.1 13 % 19 Ni 1.3 ± 0.3 20 % 14 Ni 4.0 ± 0.4 11 % 19mm 34.7 ± 3.1 9 % 7 mm 15.7 ± 1.6 10 % 18 mm 15.7 ± 0.5 3 % 19 mm 16.4 ± 0.5 3 % 19

deposition, µg/m²/day deposition, µg/m²/day deposition, µg/m²/day

Birkenes Peyrusse Copenhagen

Between sampler uncertainty, Wet only

deposition, µg/m²/day

Duisburg

From To

Cd 9% 18%

Pb 7% 15%

As 6% 19%

Ni 11% 20%

Relative SD in deposition measurements:


POP lab intercomparison,2002POP lab intercomparison,2002

pp'-DDT

-100

-50

0

50

100

150

200

250

129A

121

105

118

113

103

131

131

104

123

108

110

116

102

109

101

115

Lab no.

Dev

. fro

m m

edia

n [

%]

10075

g-HCH

-80

-60

-40

-20

0

20

40

60

109

115

131

131

118

101

102

103

116

113

104

105

121

123

108

110

129A

Lab no.

Dev

. fro

m m

edia

n [

%]

PCB 153

-40

-20

0

20

40

60

80

100

120

140

113

121

105

131

131

109

129A

118

104

103

115

108

110

116

102

101

Lab. no.

Dev

. fro

m m

edia

n [

%]

113237

Phenanthrene

-50

-40

-30

-20

-10

0

10

20

30

40

50

101

129A

116

131

131

109

108

102

103

105

107

113

104

123

Lab no.

Dev

. fr

om

med

ian

[%

]

-58


Artefact in gas/particle for NArtefact in gas/particle for N

0

1

2

3

4

5

6

SO4 NO3 NH4

g/m

3

Artefact free fine

PM2.5

PM10

Artefact free total

0

1

2

3

4

5

6

SO4 NO3 NH4

g/m

3

Artefact free fine

PM2.5

PM10

Artefact free total

IT01, Jan 2007

IT01, June 2007

Artefact free measurements using denuders only done at Montelibretti

NO3 NH4 NO3 NH4Fine 53 % 27 % 82 % 64 %Total 31 % 50 %

June 2006 Jan 2007

Underestimation of N


Estimates of the positive artefact of OC in PM10 and PM2.5/PM1 -June 2006

Estimates of the positive artefact of OC in PM10 and PM2.5/PM1 -June 2006

0

10

20

30

40

50

60

70

NO56 IT01 NO01

Po

siti

ve a

rtef

act

OC

p/O

C (

%)

PM10

PM2.5 or PM1

QBQ-approach


Measurement and model intercomparisonMeasurement and model intercomparison

Cd in precip at FI17

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

1990,Jan 1992,Jul 1995,Jan 1997,Jul 2000,Jan 2002,Jul

Mod

Obs??

NO

ES


Uncertainties in trendsUncertainties in trends

First year Last Year Signific. % change Signific. % changeAT02 1980 2003 * -33 % *** -39 %AT04 1985 2003 ** -34 % 0 -13 %CH02 1980 2003 ** -28 % ** -30 %CZ01 1980 2003 * -33 % *** -37 %DE07 1980 2003 + -27 % * -33 %DE01 1980 2003 0 -20 % ** -27 %DE03 1980 2003 ** -44 % * -19 %DE05 1980 2003 ** -49 % *** -33 %DK05 1980 2003 * -32 % *** -36 %FI04 1980 2003 * -22 % *** -38 %FI09 1980 2003 0 -34 % *** -56 %FR03 1980 2003 0 -11 % 0 -24 %FR05 1980 2003 0 -36 % *** -55 %GB02 1980 2003 + -26 % 0 -8 %GB06 1985 2003 0 18 % 0 -24 %HU02 1980 2003 0 -26 % ** -32 %IE01 1985 2003 0 -16 % 0 -15 %IT01 1980 2003 0 -18 % + 49 %LT15 1980 2003 * -45 % * -28 %NO01 1980 2003 0 -17 % ** -24 %NO39 1980 2003 0 -4 % * -24 %PL02 1985 2003 * -38 % *** -42 %PT01 1980 2003 0 -1 % 0 41 %RU01 1985 2003 * -34 % 0 42 %SE05 1980 2003 * -23 % * -28 %CS05 1985 2003 + -25 % 0 41 %

measurmentsNO3 precip

ModelSO2

SO4 in air


Representativity, NO2

Representativity, NO2

Site Nr Obs Model Bias Spread R^2AT0002 2345 1.57 2.57 -0.39 1.76 0.57BE0032 3493 3.33 4.83 -0.31 2.42 0.69ES0007 3122 0.87 2.73 -0.68 7.71 0.15IT0001 3300 3.16 4.79 -0.34 2.55 0.22NL0091 2774 5.84 7.03 -0.17 2.45 0.68

NO2, 1999-2005

Comparing EMEP model and obs. in light of population density

ES07

IT01

NL91

AT02BE32


ConclusionsConclusions Many factors influence uncertainty in measurements

Methodology, sampling procedure, Representativity

Need to distinguish between uncertainty in one data point, in averages and trends; and distinguish between bias and spread

In general, the measurements are within DQO if reference methods are used and the site is representative, but there are exceptions

Need to have better control of artefacts, especially nitrogen gas/particle and EC/OC

More intercomparison of other species than main comp and HM are needed


Data quality objectivesData quality objectivesAcidifying and eutrophying compounds

o 15-25% uncertainty in annual average (10-15% for indiv meas.)

Heavy metals

o 30% uncertainty in annual average (15-25% for indiv meas.)

o 40 % uncertainty for As, Cd, Ni in the EU 4th DD (70% in dep)

o 50% uncertainty for Hg (total gas) in the EU 4th DD

POPs (not defined in EMEP)

o 50% uncertainty for PAH in the EU 4th DD (70% in dep)

PM (not defined in EMEP)

o 25% accuracy in continuous measurements EU 1st DD

Photooxidants (not defined in EMEP)

o: 15% accuracy in continuous O3, NOx measurements, EU 3rd DD

Documents

Uncertainties in atmospheric observations Wenche Aas EMEP/CCC