Evidence for Positive and Negative Evidence for Positive and Negative Organic Sampling ArtifactsOrganic Sampling Artifacts

John G. Watson (john.watson@dri.edu)

Judith C. ChowL.-W. Antony Chen

Desert Research Institute, Reno, NV

Presented at:IMPROVE–CSN Carbon PM Monitoring Workshop

University of California, DavisJanuary 22, 2008

Definition of Organic Sampling ArtifactDefinition of Organic Sampling Artifact

• Fundamental: the difference between filter-based organic matter (OM) and “actual” OM in the atmosphere.

• Practical: the difference between filter-based OM and Teflon-membrane filter OM, which is used to determine PM mass concentration.

Atmospheric Organic Volatility Atmospheric Organic Volatility Categories Span a ContinuumCategories Span a Continuum

PL0 at 20 ºC

10-1 Torr

10-8 Torr

H2O: 17.54

Benzo(e)pyrene: 4.3 10-8

Fluorene: 1.9 10-3

Volatile

Semi-Volatile

(SVOC)

Non-Volatile HULIS, WSOC

Organic Sampling ArtifactsOrganic Sampling Artifacts

• Positive sampling artifact: SVOC is volatilized “before” captureby filters

• Negative sampling artifact: SVOC is volatilized “after” captured by filters

Quartz- or other filter material

Backup fiber CIG Absorbent

Particle (P)

• Particle and gas are in a dynamic equilibrium!

Gas Molecule

CIG: Charcoal-impregnated glass-fiber filter

Early Reports of Negative ArtifactEarly Reports of Negative Artifact• Commins, B.T. (1962). Interim report on the study of techniques for determination of

polycyclic aromatic hydrocarbons in air. Report No. Monograph 9. Prepared by National Cancer Institute.

• Lee, F.S.; Pierson, W.R.; and Ezike, J. (1980). The problem of PAH degradation during filter collection of airborne particulates - An evaluation of several commonly used filter media. In Polynuclear Aromatic Hydrocarbons: The Fourth International Symposium on Analysis, Chemistry and Biology, A. Bjorseth and A.J. Dennis, Eds. Battelle Press, Columbus, OH, pp. 543-563.

• Schwartz, G.P.; Daisey, J.M.; and Lioy, P.J. (1981). Effect of sampling duration on the concentration of particulate organics collected on glass fiber filters. J. Am. Ind. Hyg. Assoc., 42:258-263.

• Galasyn, J.F.; Hornig, J.F.; and Soderberg, R.H. (1984). The loss of PAH from quartz fiber high volume filters. J. Air Poll. Control Assoc., 34(1):57-59.

• van Vaeck, L.; van Cauwenberghe, K.; and Janssens, J. (1984). The gas-particle distribution of organic aerosol constituents: measurements of the volatilisation artifact in Hi-Vol cascade impactor sampling. Atmos. Environ., 18:417-430.

• Coutant, R.W.; Brown, L.L.; Chuang, J.C.; Riggin, R.M.; and Levis, R.G. (1988). Phase distribution and artifact formation in ambient air sampling for polynuclear aromatic hydrocarbons. Atmos. Environ., 22:403-409.

• Eatough, D.J.; Sedar, B.; Lewis, L.; Hansen, L.D.; Lewis, E.A.; and Farber, R.J. (1989). Determination of semivolatile organic compounds in particles in the Grand Canyon area. Aerosol Sci. Technol., 10:438-449.

Early Reports of Positive ArtifactEarly Reports of Positive Artifact• Cadle, S.H.; Groblicki, P.J.; and Mulawa, P.A. (1983). Problems in the

sampling and analysis of carbon particulate. Atmos. Environ., 17(3):593-600.

• McDow, S.R. (1986). The effects of sampling procedures on organic aerosol measurement. Ph.D. Dissertation, Oregon Graduate Center, Beaverton, OR.

• Fung, K.K. (1988). Artifacts in the sampling of ambient organic aerosols, S. Hochheiser and R.K.M. Jayanty, Eds. Air Pollution Control Association, Pittsburgh, PA, pp. 369-376.

• Watson, J.G.; Chow, J.C.; Richards, L.W.; Andersen, S.R.; Houck, J.E.; and Dietrich, D.L. (1988). The 1987-88 Metro Denver Brown Cloud Air Pollution Study, Volume II: Measurements. Report No. 8810.1F2. Prepared for Greater Denver Chamber of Commerce, Denver, CO, by Desert Research Institute, Reno, NV.

Operational Definitions of Particulate OC FParticulate OC = Total – APositive or Negative Sampling Artifact

Filter-Adsorbent (FA) AQF

A or FQFDDenuder-Filter-Adsorbent or -Filter (DFA or DFF)

A E

Electrostatic precipitator (EA)

Filter-Filter-Adsorbent (FFA)

QBQQF A

Several Methods to Compensate for Several Methods to Compensate for Positive ArtifactPositive Artifact

• Do nothing and assume it is zero

• Denude organic gases before sampling and assume it is zero

• Subtract the quartz lab blank

• Subtract the quartz field blank

• Subtract the back half of the filter

• Subtract the quartz backup behind quartz

• Subtract the quartz backup behind Teflon

• Calculate the intercept of OC vs. mass as mass approaches zero (Solomon’s method)

• Subtract weighted ions and elements from mass, assume remainder is carbon. Excess measured carbon is positive artifact (Frank’s SANDWICH)

• Subtract low temperature fractions

IMPROVE Acquires Backup Filters and IMPROVE Acquires Backup Filters and Field BlanksField Blanks

MORA

YOSE

HANC

CHIR

SHEN

OKEF

MORA

YOSE

HANC

CHIR

SHEN

OKEF

FRES

PUSO

PHOE

TONT

WASH

DOSO

BIBE

MORA

YOSE

HANC

CHIR

SHEN

OKEF

MORA

YOSE

HANC

CHIR

SHEN

OKEF

FRES

PUSO

PHOE

TONT

WASH

DOSO

BIBE

• The six circled sites are locations where backup filters are acquired ~6% of the time

• The eight square sites are collocated IMPROVE and STN/CSN sites.

IMPROVE has a Large Number of IMPROVE has a Large Number of Analyzed Blanks and Backup FiltersAnalyzed Blanks and Backup Filters

Between 1/1/2005 and 12/31/2006:• 44,016 samples from the IMPROVE network were

analyzed for OC and EC following the IMPROVE_A protocol

• 959 (2.2% of the total) field blanks were collected at 187 sites (including six collocated sites).

• 1,406 backup filters (i.e., QBQ) were acquired at six sites (i.e., MORA, YOSE, HANC, CHIR, SHEN, and OKEF).

Blank Levels Blank Levels Vary by Season Vary by Season

0

10

20

30

40

50

60

70

80

90

100

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34

Total Carbon Concentration (µg/filter)

Nu

mbe

r o

f Site

s

SummerWinter

IMPROVE Blank Filter (bQF) Loading 1/1/2005 - 12/31/2006

Blank Levels Don’t Depend on Average Blank Levels Don’t Depend on Average Carbon LevelsCarbon Levels(1/05 – 12/06)(1/05 – 12/06)

Averaged blank TC (bQF) compared with concurrent averaged front filter carbon loading in the IMPROVE network. (Only 77 sites with data from > 5 blanks are included.)

0

20

40

60

80

100

120

140

160

180

200

HAVO1

WHPA1

HALE1

DENA1

NOCA1

MELAX

MOZI1

CRLA1

ULBE1

HOOVX

LYBR1

REDW1

CANY1

HECA1

BOAP1

SENEX

CEBL1

BAND1

SAWE1

THBA1

INGA1

SHEN1

COGO1

FOPE1

EVERX

FLAT1

ELLI1

HEGL1

SAMA1

JOSH1

AGTI1

CABA1

CADI1

DOUG1

BRIG1

SIKE1

PITT1

FRES1

MONT1

Sampling Sites

Car

bo

n C

on

cen

trat

ion

( g

/fil

ter)

EC3

EC2

EC1

OC4

OC3

OC2

OC1

blk TC (BLKTC)

Act

ive

Sam

pli

ng

Pas

sive

D

epo

siti

on

Blank OC Levels Don’t Show a Spatial Blank OC Levels Don’t Show a Spatial PatternPattern

*Blanks Acquired between 01/05 and 05/06

Spring (March – May)

Summer (June – August)

Fall (September through November)

Winter (December through February)

IMPROVE Field Blanks IMPROVE Field Blanks Stay Longer than Those of Stay Longer than Those of

Other NetworksOther Networks (1/1/2005 – 12/31/2006)(1/1/2005 – 12/31/2006)

IMPROVE

STN/CSN

SEARCH

Blank Deposition Period

~every 7 days (once per week)

Varies (~1-15 minutes) with exceptions (~5-7 days)

Varies (~1-15 minutes)

0

10

20

30

40

50

60

70

80

90

100

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Average Organic Carbon Field Blank Concentration(µg/cm2)

Num

ber

of S

ites

IMPROVE(181 Sites)

0

20

40

60

80

100

120

140

160

180

200

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

Average Organic Carbon Field Blank Concentration

(µg/cm2)

Nu

mb

er

of

Sit

es

STN/CSN(239 Sites)

0

1

2

3

4

5

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5Average Organic Carbon Field Blank Concentration

(µg/cm2)

Nu

mb

er o

f S

ites

SEARCH(8 Sites)

IMPROVE Field Blank Carbon is IMPROVE Field Blank Carbon is Higher than that for STN/CSNHigher than that for STN/CSN

IMP_bQF: IMPROVE field blanks

STN_FB: STN/CSN field

blanks

STN_TB: STN/CSN trip blanks

PUSO1

0

1

2

3

4

10/16/2001 1/16/2002 4/16/2002 7/16/2002 10/16/2002 1/16/2003 4/16/2003 7/16/2003 10/16/2003

Bla

nk

Car

bo

n ( g

/cm

2) IMP_bQF

STN_FBSTN_TB

MORA1

0

1

2

3

4

10/16/2001 1/16/2002 4/16/2002 7/16/2002 10/16/2002 1/16/2003 4/16/2003 7/16/2003 10/16/2003

Bla

nk

Car

bo

n ( g

/cm

2 ) IMP_bQFSTN_FBSTN_TB

PHOE1

0

1

2

3

4

10/16/2001 1/16/2002 4/16/2002 7/16/2002 10/16/2002 1/16/2003 4/16/2003 7/16/2003 10/16/2003

Bla

nk

Car

bo

n ( g

/cm

2 ) IMP_bQFSTN_FBSTN_TB

TONT1

0

1

2

3

4

10/16/2001 1/16/2002 4/16/2002 7/16/2002 10/16/2002 1/16/2003 4/16/2003 7/16/2003 10/16/2003

Bla

nk

Car

bo

n ( g

/cm

2 ) IMP_bQFSTN_FBSTN_TB

IMPROVE Field Blank TC is Higher IMPROVE Field Blank TC is Higher than STN/CSN than STN/CSN (continued)(continued)

IMP_bQF: IMPROVE field blanks

STN_FB: STN/CSN field

blanks

STN_TB: STN/CSN trip blanks

WASH1

0

1

2

3

4

10/16/2001 1/16/2002 4/16/2002 7/16/2002 10/16/2002 1/16/2003 4/16/2003 7/16/2003 10/16/2003

Bla

nk

Car

bo

n ( g

/cm

2 ) IMP_bQFSTN_FBSTN_TB

DOSO1

0

1

2

3

4

10/16/2001 1/16/2002 4/16/2002 7/16/2002 10/16/2002 1/16/2003 4/16/2003 7/16/2003 10/16/2003

Bla

nk

Car

bo

n ( g

/cm

2 ) IMP_bQFSTN_FB

STN_TB

BIBE1

0

1

2

3

4

1/1/2005 4/1/2005 7/1/2005 10/1/2005 1/1/2006 4/1/2006 7/1/2006 10/1/2006

Bla

nk

Car

bo

n ( g

/cm

2 ) IMP_bQFSTN_FB

STN_TB

FRES1

0

1

2

3

4

1/1/2005 4/1/2005 7/1/2005 10/1/2005 1/1/2006 4/1/2006 7/1/2006 10/1/2006

Bla

nk

Car

bo

n ( g

/cm

2 ) IMP_bQFSTN_FB

STN_TB

But STN/CSN OC Artifact Correction is Higher But STN/CSN OC Artifact Correction is Higher than IMPROVE due to Lower Flow Rates and than IMPROVE due to Lower Flow Rates and

Larger Filter Area (Intercept method)Larger Filter Area (Intercept method)

0.0

0.5

1.0

1.5

2.0

2.5

Sea

ttle

, W

A

Mt.

Rai

nier

,W

A

Pho

enix

, A

Z

Ton

to N

atnl

Mon

, A

Z

Was

hing

ton,

DC

Dol

ly S

odds

Wld

rns,

WV

Fre

sno,

CA

Big

Ben

dN

P,

TX

ST

N-I

MP

TC

Int

erce

pt ( g

/m3 )

AllSpringSummer FallWinter

STN = a(IMPROVE) + b

0.0

0.5

1.0

1.5

2.0

2.5

3.0

TC OC EC OC1 OC2 OC3 OC4 EC1 EC2 EC3

Carbon Fractions

Co

nce

ntra

tion

(µg

/cm

2 )

IMPROVE Urban

IMPROVE Rural

STN/CSN Urban

SEARCH Non-Urban

SEARCH Urban

Fractions up to OC4 can be found on blank filters

More OC on Blanks is in Low Temperature More OC on Blanks is in Low Temperature OC Fractions, but there is also Blank OC at OC Fractions, but there is also Blank OC at

High TemperaturesHigh Temperatures

ImplicationsImplications

• Blank filter does not reach equilibrium with organic gases within a few minutes of atmospheric exposure (i.e., STN/CSN approach).

• At most ambient conditions, the blank filter is close to saturation with VOCs after the equilibrium is attained

• The equilibrium/saturation may depend on ambient temperature.

IMPROVE Blank OC and Backup OC Agree in IMPROVE Blank OC and Backup OC Agree in Winter, but Not in SummerWinter, but Not in Summer

0

2

4

6

8

10

12

14

16

18

20

1/1/02 1/1/03 1/1/04 1/1/05 1/1/06

carb

on o

n ba

ckup

filt

er o

r fie

ld b

lank

, ug

CS

FB

Series3

OC

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

1/1/02 1/1/03 1/1/04 1/1/05 1/1/06

carb

on o

n ba

ckup

filt

er o

r fie

ld b

lank

, u

g CS

FB

Series3

O1

0

1

2

3

4

5

6

7

8

9

10

1/1/02 1/1/03 1/1/04 1/1/05 1/1/06

carb

on o

n ba

ckup

filt

er o

r fie

ld b

lank

, ug

CS

FB

Series3

O3

QBQ and bQF OC agree well in winter, but more OC is found on QBQ in summer!

(Adapted from Warren White 2007)

QBQ

bQF

QBQ

bQF

QBQ

bQF

IMPROVE Negative Artifact is SmallIMPROVE Negative Artifact is Small

0

5

10

15

20

25

30

35

Spring Summer Fall Winter

OC

Co

nce

ntr

atio

n (m g

/fil

ter)

Average Intercept OCAverage QBQAverage bQF

Average across 163 IMPROVE sites; QBQ is only available at six sites.

OC = a(Mass) + b

*If volatilization (negative artifact) is negligible, we expect to see the average Intercept OC agree with QBQ or bQF OC (representing the positive sampling artifact).

A Conceptual ModelA Conceptual Model

(40)

(20)

0

20

40

60

80

100

120

140

160

Teflon Quartz Teflon Quartz

Summer Winter

OC

Con

cent

ratio

n (

m g/fi

lter)

pSVOC (volatilized)

pSVOC (retained)

pOC

VOC and gSVOC (adsorption)

• Teflon filter is not subject to positive sampling artifact

• More volatilization on Teflon filters, resulting in a higher negative sampling artifact

• Cannot rule out the volatilization from quartz-fiber filters

• Volatilization is stronger in summer than in winter

• The volatilized OC is not always recaptured by the backup filter (same for positive sampling artifact)

Key Question:Key Question:• Is the difference between QBQ and bQF OC due to positive

or negative sampling artifact?

Excess OC on the backup filter (with respect to the field blank) correlated well with ambient PM filter mass loading (from Jay Turner, 2006)

-10

-5

0

5

10

15

20

25

30

35

40

1 10 100 1000 10000

CP

CS

- F

BOC

QB

Q-b

QF

QF

Organic artifact may be estimated by Organic artifact may be estimated by slicing the bottom half of the quartz-slicing the bottom half of the quartz-

fiber filterfiber filter

•Filter slicer

Procedure:

1. Analyze a whole punch

2. Acquire another punch from the same filter and weight the whole punch

3. Slice the punch and weight each of the two halves

4. Analyze both halves for carbon concentration

5. Estimate sampling artifact by scaling carbon measured on the bottom-half filter to the whole filter

Similar OC between bottom half of QF and QBQSimilar OC between bottom half of QF and QBQ

SHEN1 2005/1/13 (Q89488)

0

10

20

30

40

50

60

70

0 1 2 3 4Slice Mass (mg)

Car

bon

Load

ing

(ug)

CHIR1 2005/4/7(Q94596)

0

5

10

15

20

25

0 1 2 3 4

Slice Mass (mg)

Car

bon

Load

ing

(ug)

Pattern of Sliced Filter Carbon Loading (I)

QF

QBQ

Original QF analysis

QFtop or QBQtop

QFbott or QBQbott

SHEN1 5/17/2005 (Q93898)

0

10

20

30

40

50

60

0 1 2 3 4Slice Mass (mg)

Car

bon

Load

ing

(ug) Front Filter

Backup Filter

YOSE1 2006/2/16 (R14098)

0

5

10

15

20

25

30

0 1 2 3 4Slice Mass (mg)

Car

bon

Load

ing

(ug)

Higher OC in bottom half of QF than QBQHigher OC in bottom half of QF than QBQ

Pattern of Sliced Filter Carbon Loading (II)

QF

QBQ

QFtop or QBQtop

Original QF analysis

QFbott or QBQbott

ConclusionsConclusions

• Blank levels are higher in summer, lower in winter, but have no consistent spatial pattern.

• Blank filter artifact contains high temperature OC (i.e., OC4 at 580 °C), suggesting changes in thermo/chemical properties of VOCs after adsorption.

• Short (a few minutes) blank filter exposure in CSN/STN and the SEARCH network underestimates actual positive OC artifact.

Conclusions Conclusions (continued)(continued)

• In rural areas and during winter, backup filters (QBQ) resemble blank filters (bQF) with respect to carbon loading, possibly due to less SVOC.

• Negative artifact may be more for Teflon than for quartz filters (especially in summer).

• OC artifact on the bottom-half of sliced filter (QFbott) are similar to or higher than backup filter (QBQ), and appear to differ by environment.