1. State of the Science on Sources of Carbonaceous Aerosols and Their Contribution to Regional Haze John G. Watson (johnw@dri.edu) Judith C. Chow Desert Research Institute, Reno, NV, USA Presented at: WRAP Workshop on Fire, Carbon, and Dust May 23-24, 2006 Sacramento, CA

2. Objectives Describe sources of carbonaceous aerosol Identify and evaluate methods to identify and quantify carbon emissions and source contributions Review progress on reconciling different carbon measurement methods and instruments

3. Carbon contributes much to poor visibility at many western sites (Yosemite National Park)

4. OC contributions vary spatially and differ by year, mostly due to fires (Annual Average OC for Western States)

5. Fire contributions vary with time during the year (Yosemite National Park)

6. Area and mobile sources contain more carbon than point sources

7. Emissions relevant to carbonaceous PM PM Fugitive dust from wind erosion, agricultural activities, construction, storage piles, and vehicle traffic on paved and unpaved roads. VOC Vegetation, surface coatings, fuel storage and distribution, solvents. PM, VOC Burning and cooking from stoves, charbroilers, trash, forest fires, and agricultural burning. PM, NO x, VOC Ducted exhaust from industrial facilities (e.g., coal- and oil-fired power stations, smelting, cement plants, chemical plants, petroleum extraction and refining, glass manufacturing, paper making, shipping). Vehicle exhaust from cars, trucks, motorcycles, and buses. Exhaust from non-road generators, small engines, non-road vehicles.

8. Potassium is a reaonsable indicator of fire contributions, but there’s still noise

9. Limited 14 C measuements also show much contemporary carbon (Bench, 2004)

10. Urban carbon levels decrease rapidly with distance (Summer and Winter OC averages from the CY2000 CRPAQS Nitrate at Fresno Summer Winter

11. There are many manmade burning sources CRPAQS 2000 Annual Average Summer Winter SummerWinter Annual OC Distribution

12. Issues for PM Carbon Emission Rates and Compositions Many carbon compounds are semi-volatile and condense or evaporate depending on vapor pressure, temperature, and surface availablility on other particles Different (certification) test methods for different source types result in different emission rates and compositions for the same equipment, fuels, and operating conditions Organic vapors adsorb onto quartz fiber filters used to measure carbon Size distributions, compositions, and gas/particle phases continue to change within and emissions inventory grid. Grid scaling affects equivalent emissions

13. Source Measurement Methods Hot stack sampling: Samples taken directly from exhaust duct at duct temperatures. Vehicle dynamometer testing: Simulate driving cycles on fixed roller. Continuous emissions monitoring: In-duct or on-board (motor vehicle) measure continuously Diluted duct sampling: Samples drawn into aging chamber and cooled with clean air. Vehicle on-road testing: Roadside or tunnel, integrated or individual vehicle samples, in-plume or remotely sensed. Source-dominated sampling: Samples taken at locations and times when a single source dominates ambient concentrations (e.g roadside, tunnel,.

14. Hot Stack Compliance Method 201/202 Filter/Impinger Methods EPA Methods PRE4 & 202 Filterable PM Condensable PM (<1 µm) PM10 and PM2.5 cyclones and filter (in-stack) V TT Filter Glass or Teflon ® probe liner (heated) Teflon ® tubing (heated) Sample gas is cooled to 60-70 °F in iced impingers Analysis: Organic extraction Titration of inorganic fraction Dry and weigh organic and inorganic residue SO 4 = and Cl - Analysis: Evaporation of rinses Gravimetric analysis Post Test Purge with N 2 or Air Range of chemical speciation techniques is limited due to high temperatures, moisture, interfering particles & gases

15. Dilution sampling better represents what gets is emitted to environment, allows more variables to be measured Stainless steel Cross-flow jet mixing Dilution Ratio >20:1 Residence time >10 sec Sample gas is cooled to ambient temperature by dilution with ambient air Flow meter PM10 Cyclone Hildemann,L.M., Cass,G.R. and Markowski,G.R. (1989) A dilution stack sampler for collection of organic aerosol emissions: Design, characterization and field tests. Aerosol Sci. Technol. 10(10-11):193-204.

16. Difference in PM 2.5 Mass between In-Stack and Dilution Sampling Gas-Fired Boiler - Field Data 0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 Run 1Run 2Run 3Run 1Run 2Run 3AP42 lb/MMBtu inorganic condensable (M202) organic condensable (M202) Filterable PM (M201A) PM2.5 (dilution) Dilution Method In-Stack Methods Chang,M.C. and England,G.C. (2004) Development of fine particulate emission factors and speciation profiles for oil and gas-fired combustion systems, Update: Critical review of source sampling and analysis methodologies for characterizing organic aerosol and fine particulate source emission profiles. Irvine, CA: GE Energy and Environmental Research Corp.

17. Mobile source certification requires dilution Dilution tunnel and sampling ports Put generator on wheels and move it and it is certified by dilution sampling Install the generator permanently and it is certified by hot stack sampling and has different emissions

18. Mobile and area source emissions may have more semi volatile PM components than point source emissions Lipsky,E.M. and Robinson,A.L. (2006) Effects of dilution on fine particle mass and partitioning of semivolatile organics in diesel exhaust and wood smoke. Environ. Sci. Technol. 40(1):155-162.

19. Difficulties with OC and EC Sampling and Analysis No common definition of EC for atmospheric applications –It’s not graphite, diamond, or fullerenes Light absorption efficiencies are not constant –They vary depending on particle shape and mixing with other substances OC and EC properties on a filter differ from those in the atmosphere OC gases are adsorbed onto the quartz filter at the same time that semi-volatile particles evaporate

20. Thermal Evolution Methods are Conceptually Simple Lavoisier's Oil Analysis "Traité Élémentaire de Chimie" (1789) vol. II, chap.VII, p. 493-501

21. Various Thermal/Optical Protocols }{

22. Temperature and Analysis Time in OC varies in Thermal/Optical Protocols *OGI OC performed at 600 ºC only *OGI Time is variable **HKUST-3 Time 150 s only

23. Temperature and Analysis Time in EC also varies for Thermal Analysis *HKUST-3 Time 150 s only

24. 11TOT10TOT 11bTOT 12TOT 11TOT10TOT 11bTOT 13TOR Schmid et al., 2001, Atmos. Environ. 35:2111-2121 Different Thermal Evolution Protocols Give Different EC Results, but TC Results Generally Agree!

25. Learned much in transition from old to new IMPROVE analyzers DRI Model 2001 Analyzer DRI/OGC Analyzer

26. Several variables might affect OC/EC split and carbon fractions Carrier gas composition Temperature ramping rates Temperature plateaus Residence time at each plateau Optical pyrolysis monitoring configuration/wavelength Standardization Oxidation and reduction catalysts Sample aliquot and size Evolved carbon detection method Carrier gas flow through or across the sample Location of the temperature monitor relative to sample Oven flushing conditions

27. Low-temperature protocol corrects for pyrolysis by reflectance (TOR), has low initial temperatures (120 and 250 ˚C), long residence time at each temperature (150-580 seconds), carbon peaks back to baseline, 550 ˚C max in He IMPROVE (Low-Temperature) Protocol

28. High-temperature protocol corrects for pyrolysis by transmittance (TOT), has high initial temperature (310 ˚C), fixed and short residence times (45-120 seconds), 900 ˚C max in He STN (High-Temperature) Protocol

29. EC differs within Protocol between Reflectance (TOR) and Transmittance (TOT) Pyrolysis Corrections Chow et al., 2001, Aerosol Sci. Technol. 34:23-34

30. IMPROVE-TOR and STN-TOR yield the Same EC. Why? Chow et al., 2001, Aerosol Sci. Technol. 34:23-34

31. Charring Correction is the Key (Chow et al., 2004) Charring could occur throughout the filter. Therefore, the transmittance is more influenced by the charred material within the filter than reflectance. EC is oxidized and removed earlier than POC in an oxidative environment.

32. Example of Negative OP Thermogram (Site: Sipsey Wilderness Area, AL 12/29/2004) Early Split Introduction of O 2

33. Zero OP in IMPROVE Samples between 2000-2004 % of Zero OP (12,730 out of 93,438 samples)

34. Carbon Fractions vary by Sources

35. Configuration of DRI Model 2001: Sample Holder 19.12 mm 2 mm ThermocoupleShield Bare Thermocouple Tip (unshielded) Sample Sample Holder 8.46 mm

36. Example of Multi-point Temperature Calibration Tempilaq G indicator used for temperature calibration: 121  2, 184  2, 253  3, 510  6, 704  8, and 816  9 C (Chow et al., 2005)

37. IMPROVE vs. IMPROVE_A* Thermal Protocol Original OGC/DRI Thermal Optical Analyzer (1986) IMPROVE_A (DRI Model 2001) IMPROVE (DRI/OGC) OC1140 ºC120 ºC OC2280 ºC250 ºC OC3480 ºC450 ºC OC4580 ºC550 ºC OP (POC)TOR/TOTTOR EC1580 ºC550 ºC EC2740 ºC700 ºC EC3840 ºC800 ºC *Implemented for samples acquired after January 1, 2005

38. Comparison of Carbon Fractions between Model 2001 IMPROVE_A and DRI/OGC IMPROVE Protocols

39. IMPROVE OC/EC Split not Affected by O 2 in He, but an early split might occur. Carbon Fractions not Affected by O 2 <40 ppm OC1 OC2 OC3 OC4 POC OC/EC Split EC1-POC EC2

40. Minerals increase EC oxidation rate at higher temperatures in 100% He atmosphere. More early splits for high temperature STN

41. NaCl and other catalysts affect carbon fractions and can cause an early split

42. More specific detectors can give more carbon fractions on existing samples (examples from a GC/MS detector at 275 degrees) Gasoline Diesel Roadside dust Coal power plant

43. How can we Maximize Utility of STN and IMPROVE for Different Purposes? Understand OC and EC –Report both TOT and TOR corrections –Report negative pyrolysis corrections –Report initial, minimum, and final reflectance and transmittance –Re-analyze fraction of samples on old analyzers Source attribution (also needed in source samples) –Define temperature plateaus that bracket dominant compounds –Use more specific detectors

44. Conclusions Very large carbon concentrations are often due to fire. These can be identified by spatial and temporal changes in carbon Urban carbon concentrations decrease rapidly with distance Need more specific markers for non-fire sources and for fire contributions at normal carbon levels OC/EC split appears to be independent of temperature of program. Oxygen in the carrier gas, catalysts such as NaCl, and mineral oxides affect carbon fractions and may cause a negative OC/EC pyrolysis correction More information can be obtained from the same samples with more specific carbon detectors