Prakash V. Bhave, Ph.D.Physical Scientist

EMEP Workshop – PM Measurement & Modeling April 22, 2004

Measurement Needs for Evaluating Model Calculations

of Carbonaceous Aerosol

Policy-Relevant Questions*• What fraction of PM2.5 mass is carbonaceous?• What fraction of carbonaceous aerosol is primary

vs. secondary?• What are the source contributions to primary

carbon?• What fraction of secondary organic aerosol is

anthropogenic vs. biogenic?• What fraction of carbon in urban areas is

transported from upwind locations?

Existing models can answer these questions, but we need measurements to evaluate the answers

* NCEA Carbonaceous PM Workshop Series - 2004

• Reconstruction of carbonaceous mass Organic mass contains C, H, O, and N atoms Carbon is measured; rest is estimated OM/OC ratios of 1.2 – 2.5 have been proposed Ratio increases with age of aerosol

[Turpin & Lim, 2001] Speciated or FTIR measurements can help

• Evaluation approach In model formulation, OM/OC ratios are known Convert model predictions to OC, and evaluate

against carbon measurements

Carbonaceous Fraction of PM2.5

Carbonaceous Fraction of PM2.5

CMAQ Model Results – Average of 2001 Annual Simulation (TC/PM2.5,dry)

Annual Average PM 2.5 Specia tion P rofiles for U rban S ites(April 2002 through M arch 2003 D ata)

Sulfate

Am m onium

N itrate

TCM

Crusta l

6.41 18.85 31.29

Carbonaceous Fraction of PM2.5

STN data averaged from April ’02 – March ’03* Courtesy of Joann Rice, EPA/OAQPS

OC/EC inter-network inconsistencies (STN vs. IMPROVE) - STN OC data are not blank-corrected - Different thermal-optical protocols

Primary vs. SecondaryCMAQ–2001 Annual Average (Primary Carbon / Total Carbon)

• OCpri & OCsec cannot be measured directly

• Several indirect estimation methods exist• EC tracer method

Primary vs. Secondary

prisec

pripri

EC

OCECOCOC

EC

OCECOC

Estimate (OC/EC)pri from emissions/transport model[S.Yu, et al., 2004]

Make use of the plentiful, ambient OC and EC data

• OC/EC in source profiles must be consistent with the ambient monitors

• Semi-continuous OC/EC data are needed to check model predictions of diurnal OCsec

patterns Most models predict the OCsec peak at nighttime [Pun et al., 2003]

OC & EC data from SEARCH network (IMPROVE TOR method)

Primary vs. Secondary

OC/EC splits in the inventory inconsistent with ambient sampling protocols

Primary vs. Secondary

PM2.5 Weight %

1999 EmissionsTC (tons)

PrimaryOC/EC

SamplingProtocolNEI99 Source Profile OC EC

Non-road Diesel Exhaust 18.70% 74.11% 86,966 0.25 TOR

Agricultural Burning 53.24% 7.50% 73,052 7.10 various

Heavy-duty Diesel Exhaust 18.93% 75.00% 38,375 0.25 TOR

Non-road Gasoline Exhaust 65.50% 8.01% 17,076 8.18 TOT

Light-duty Gasoline Exhaust 47.35% 19.01% 10,535 2.49 TOR

Soil Dust 4.54% 0.37% 8,356 12.27 various

Paved Road Dust 14.73% 1.12% 6,641 13.15 TOR

Jet Fuel Combustion 24.34% 65.87% 6,350 0.37 unknown

Wood Waste Boilers 9.81% 20.19% 4,214 0.49 thermal

Natural Gas Combustion 50.00% 0.00% 3,865 - N/A

Solid Waste Combustion 0.57% 3.50% 2,954 0.16 TOR

Residual Oil Combustion 19.93% 19.33% 2,437 1.03 TOR

Wood Products - Drying 65.83% 4.39% 1,612 15.00 various

Fiberglass Manufacturing 28.00% 2.00% 1,474 14.00 thermal

Food & Agriculture Handling 30.00% 0.00% 1,353 - N/A

Other Sources N/A N/A 10,880 3.59 various

- CMAQ results using source apportionment capability (Aug. 1, 1999)

- Evaluations will provide direct feedback to emission inventory improvement

- Validated results can support control strategy development

Apportionment of Primary CarbonDiesel exhaust fraction Biomass combustion fraction

Needs for Model Evaluation• Data types

Source-specific organic tracers (e.g., levoglucosan, hopanes, cholesterol, etc.) [Schauer, et al.]

Primary biogenic carbon (e.g., carbohydrates, vegetative detritus) [M. Hernandez; W. Rogge, et al.]

Semi-continuous wood smoke source tracers?• Spatial resolution

Several urban sites (e.g., each Supersite) Some representative rural sites

• Temporal resolution 24h data at urban sites for ~1 month per season 2-6h composites at urban sites to check diurnal variation Monthly composites at rural sites to check seasonality

Apportionment of Primary Carbon

SOA: Biogenic vs. Anthropogenic

• What fraction of SOA is anthropogenic? Great uncertainty within model

parameterizations Nashville: July 16-18,1995 model

inter-comparison yields values of 10% - 40% [Pun et al., 2003]

• Uncertainties too large to justify controls directed specifically at anthropogenic SOA

Needs for Model Evaluation• Data types

14C can help provide a measure of biogenic SOA;need collocated wood smoke & vegetative detritus data

Source-specific SOA tracers [Edney et al.]

SOA: Biogenic vs. Anthropogenic

Reprinted from: Edney & KleindeinstOAQPS Model Eval Workshop, Chapel

Hill, Feb.10, 2004

CMAQ–2001 Annual Average (Anthropogenic fraction of OCsec)

SOA: Biogenic vs. Anthropogenic

Carbon fraction of PM2.5

• OC, EC data are plentiful; some network inconsistencies

• OM/OC uncertain, but not essential for model evaluation

Primary vs. secondary?

• Consistent definition of OC and EC across ambient networks and source data

• Semi-continuous OC & EC

Primary source apportionment

• Source-specific tracers

• Increase spatial & temporal resolution of organic tracer measurements

Anthropogenic vs. biogenic SOA

• Tracers for aromatic and monoterpene oxidation

• 14C collocated with wood smoke & detritus markers

Summary of Measurement Needs

Acknowledgements

• Atmospheric Modeling Division (NOAA/EPA)

• Emissions Monitoring & Analysis Division – Air Quality Modeling Group (OAQPS)

• Computer Sciences Corporation

Disclaimer Notice:This work has been funded wholly by the United States Environmental Protection Agency. It has been subjected to Agency review and approved for presentation.

Carbonaceous Fraction of PM2.5

17%

19%

6%6%

Carbon

Sulfate

Nitrate

Ammonium

Other

Urban Network (STN)PM2.5 = 10.5 g/m3

19%

17%

5%9%

Carbon

Sulfate

Nitrate

Ammonium

Other

CMAQ ModelPM2.5 = 11.5 g/m3

2001 Network Median Values (~7000 observations)Carbon Value is “blank-corrected” by 1 g/m3

Apportionment of Primary CarbonGasoline exhaust fraction Coal combustion fraction

- CMAQ results using source apportionment capability (Aug. 1, 1999)

Apportionment of Primary CarbonOil combustion fraction Natural gas combustion fraction

- CMAQ results using source apportionment capability (Aug. 1, 1999)

Apportionment of Primary CarbonFood cooking fraction Paved road dust fraction

- CMAQ results using source apportionment capability (Aug. 1, 1999)

Apportionment of Primary CarbonCrustal material fraction Miscellaneous source fraction

- CMAQ results using source apportionment capability (Aug. 1, 1999)

Fresno IndyS.L.

Tulsa

Missoula

SLC Bronx

Charlotte

Baltimore

Atlanta

Cleveland

Richmond

Birmingham

16 rural IMPROVE sites

13 urban STN sites

Local vs. Regional Contribution

Reprinted from: N. FrankOAQPS Model Eval Workshop, Chapel

Hill, Feb.10, 2004

Differences between urban (STN) and paired rural site (IMPROVE)

Fresno

Missoula

SLC

Tulsa

Birmingham

Indy

Cleveland

Charlotte

Richmond

Baltimore

Bronx

St Louis

Atlanta

Sulfa te :

0 .0 0.4 0.9

Am m onium :

0.0 0.9 1.9

N itra te :

0 .4 3.5 6.5

TC M (k=1.8):

2 .9 8.1 13.2

C rusta l:

0 .0 0.4 0.8

Differences between urban and paired rural site(s) - Carbonaceous mass dominates the “urban excess”

Local vs. Regional Contribution

Reprinted from: N. FrankOAQPS Model Eval Workshop, Chapel

Hill, Feb.10, 2004

CMAQ Results at 1km resolution (OC+EC)Along Pennsylvania - New Jersey border

Local vs. Regional Contribution

Urban contribution 5.4 g/m3

Urban contribution 10.1 g/m3