speciated particulate networks Mark Green Judith Chow John Watson Desert R esearch Institute Ann Dillner University of California at Davis Neil Frank and Joann Rice USEPA Office of Air Quality Planning and Standards ID: 781249
Download The PPT/PDF document "Carbon artifact adjustments for the IMPR..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Carbon artifact adjustments for the IMPROVE and CSN speciated particulate networks
Mark Green, Judith Chow, John Watson
Desert
R
esearch Institute
Ann
Dillner
University of California at Davis
Neil Frank and Joann Rice
USEPA Office of Air Quality Planning and Standards
National Air Quality Conference, San Diego, CA March 201
1
Slide2Introduction
Organic aerosol is a major contributor to PM
2.5
concentration, typically accounting for 25-50% of reconstructed fine mass
Organic aerosol % of fine mass (2000-2004)
Slide3Seasonal OC Concentrations
Winter
Spring
Summer
Autumn
Slide4CSN network map (continental US)
IMPROVE network map (Continental US)
Slide5Statement of the problemAnalysis of filter samples for carbon analysis done by thermal optical reflectance (TOR) for both IMPROVE and CSN networks
TOR heats filter in stages without oxygen and organic carbon volatilized and converted to CO2
Finally, 2% oxygen is added to combust elemental carbon
Because high temperatures are needed for TOR analysis, Teflon filter cannot be used
Quartz fiber filters are used to collect the aerosol for analysisQuartz fiber filters are known to react with organic gases causing sampling artifactsPositive artifact from adsorption of organic gases
Negative artifact from volatilization of particles off filter (e.g. as temperature increases during the day or after sample is collected).
Slide6Organic Sampling Artifacts
Positive
sampling artifact:
gas-phase adsorption onto quartz
Negative
sampling artifact: SVOC is volatilized “
after”
captured by filters
Quartz- or other filter material
Backup fiber
Particle (P)
Particle and gas are in a dynamic equilibrium!
Gas Molecule
CIG: Charcoal-impregnated glass-fiber filter
Slide7Treatment of artifact
Typically, positive artifact thought to be greater than negative artifact
SEARCH network uses denuders to remove organic gases upstream of filter- use a backup filter to capture gases volatilizing off front filter- negative artifact
However denuder approach add complexity and expense and also alters gas-particle equilibrium
IMPROVE network has used backup filters at a few sites to characterize positive artifactIMPROVE has subtracted monthly median backup filter OC concentrations at 6 sites to give a monthly “correction” to apply to all sites
CSN network is currently collecting back filters at all sites without denuders but has not determined how to use them for artifact correction
Desire a method consistent between networks and that can give continuity in time over >20 years of IMPROVE data
Slide8Average front filter (QF), backup filter (QBQ), and field blank OC concentrations (Top 1% QF excluded)
IMPROVE QBQ= 21% of QF,
bQF
=11.8% of QF
CSN QBQ= 15.8% of QF,
bQF
=6.7% of QF
Slide9Some possible methods for artifact adjustment1) Ignore potential positive and negative artifacts
2) Subtract representative OC concentration on field blank
3) Subtract representative OC concentration on backup filter
4) Use denuder to prevent positive artifact and ignore any negative artifact
5) Use denuder and backup filter to characterize negative artifact and field blanks for positive artifact6) Same as above, except without denuder
Add quartz filter behind
T
eflon and subtract (recommended by
McDow
&
Huntzicker (1990)1Only 1-3, 6 can be done with existing sampling set-ups/existing data1 Atmos. Environ., 24A,2563-2571
Slide10Data analysis methodsUse only sample site days with front filter (QF), blank filter (bQF
), and backup (QBQ) available
Look at relationships between QF and
bQF and QBQ and if they vary by geographic location or season to see if regional or seasonal artifact adjustments are called for
Consider IMPROVE and CSN data separately and then together (may expect differences because CSN mainly urban, IMPROVE mainly ruralUsed 1839 CSN samples 2008-2009Used 1387 IMPROVE samples Sep 2008- Feb 2010
Removed QBQ OC>1.2*QF OC
Slide11CSN Average
bQF
and QBQ OC by site show little geographic pattern (sites ordered from EPA Region 1 (left) to EPA Region 10 (right)
Slide12IMPROVE –
bQF
and especially QBQ show seasonal pattern (higher in summer)
CSN shows subtle seasonal patterns in
bQF
and QBQ
Slide13Backup filter OC (QBQ) proportional to front filter OC (QF). Logarithmic or power law fits work about equally well. QBQ scales approximately with square root of QF. IMPROVE better fit than CSN.
Estimating backup filter OC from front filter OC
Slide14Combined data set power law fit slightly better than logarithmic fit. Power law fit equations for CSN, IMPROVE, and combined similar- suggests little adverse impact from using combined equation for all data
Slide15Averaging the data clarifies the front/back filter OC relationship
Slide16How to proceed?
Use of curve fitting to estimate QBQ from QF gives less error than using median QBQ – so no good reason to continue using median QBQ for IMPROVE adjustment or to apply to CSN data
BUT- What does the QBQ OC
really
represent?
Alternate explanation – back filter collects positive and negative artifact
fit linear curve to QBQ
vs
QF
slope represents negative artifact proportional to QF
concnetration
intercept represents positive artifact
QF OC adjusted = QF + slope*QF
-
intercept
Blank filters represent a positive artifact independent of concentration
QF OC adjusted=
QF+slope
*QBQ-2*
bQF
(method gives small negative average adjustment for IMPROVE, small positive for CSN)
This is equivalent to SEARCH approach, except no denuder is present to remove organic vapors
Slide17Linear model for QBQ based on QF
(highest 1% of data excluded from figure and fit)
OCback
= .08*OCfront+0.13
Intercept about equal to field blank-
CSN linear model gives intercept of 0.23 compared to
bQF
of 0.16
Linear model less satisfactory than power or
ln
for CSN
Slide18SummaryMethod of artifact “correction” can affect OC concentrations up to 20% or so
Don’t currently have enough information to determine most appropriate correction
Backup filter OC concentration best represented by fit proportional to concentration BUT
We don’t really know what backup filter represents
Use of field blanks (may adjust seasonally, monthly, etc.) straightforward and consistent with artifact corrections for non-reactive compoundsWant consistent methodology among CSN and IMPROVE networks for comparability and ability to calculate urban/rural differences, etc.
Conference call later this month to try to discuss artifact correction approach for both networks