310 likes | 475 Views
Resolving the sources of PM 2.5 in Georgia using emission and receptor based models. Amit Marmur, Di Tian, Byeong-Uk Kim, James Boylan 6 th Annual CMAS Conference, October 1-3, 2007. Overview. PM 2.5 non-attainment areas in Georgia CMAQ based PM 2.5 sensitivity analysis
E N D
Resolving the sources of PM2.5 in Georgia using emission and receptor based models Amit Marmur, Di Tian, Byeong-Uk Kim, James Boylan 6th Annual CMAS Conference, October 1-3, 2007
Overview • PM2.5 non-attainment areas in Georgia • CMAQ based PM2.5 sensitivity analysis • PM2.5 source-apportionment using a receptor based model, Positive Matrix Factorization (PMF) • Comparison between PMF and CMAQ source-apportionment results • Policy implications
Background: PM2.5 Attainment Status in Georgia • Atlanta, Macon, Floyd county, Chattanooga are in non-attainment of the annual PM2.5 NAAQS (15 g/m3): • Designation for the new daily standard (35 g/m3) not yet finalized, but no “new” non-attainment areas expected Annual PM2.5 non-attainment areas
CMAQ based sensitivity analysis • In preparation for the PM2.5 SIP, GA-EPD conducted a PM2.5 sensitivity analysis (see 2005 & 2006 CMAS conference presentations) • Based on this previous analysis, the following controls were considered: • SO2 controls at major power plants • Controls of primary carbon emissions (EC/OC) • (ammonia) • Controls of NOx and anthropogenic VOCs controls had a negligible effect on PM2.5 levels
Chemistry Source Impacts Air Quality Meteorology Receptor vs. Emissions-Based Models Emissions Inventory Source-compositions Receptor (monitor) Emissions-based Model (3D Air-quality Model) Receptor Model (e.g., PMF, CMB) (e.g., CMAQ, CAMx)
PM2.5 source-apportionment using receptor models Purpose: Identify the main contributors (sources) to measured concentrations of pollutants at a receptor site. Required input: Speciated ambient measurements, along with “knowledge” of “typical” emissions composition Ci - ambient concentration of specie i (g/m3) fi,j - fraction of specie i in emissions from source j Sj - contribution (source-strength) of source j (g/m3) n – total number of sources ei – error term to be minimized (to obtain best fit) Sj are the unknowns (Ci, fi,j, n – required input)
The PMF receptor model • Positive Matrix Factorization (PMF) is the most commonly applied factor analytical technique in recent years • Enhanced factor analysis, including constraints to prevent negative source contributions • Developed by Dr. Pentti Paatero at the University of Helsinki in Finland • In 2005, EPA released the EPA-PMF 1.1 model
Why apply PMF in SIP development? Why compare with CMAQ? • Identify the major sources affecting monitors at the non-attainment area via ambient data, rather than emissions inventories • Evaluate/expand findings from CMAQ sensitivity analysis • Often impractical to quantify all sources via a CMAQ-based sensitivity analysis (time/resources) • Improve emissions inventories • Analyze events leading to high daily PM2.5 concentrations (emissions vs. meteorology) • Quantify impacts of local sources (e.g., FS#8 monitor) • Evaluate model performance for SOA
PMF-based PM2.5 source-contributions Based on STN data, 2003-04
What we’ve learnt so far… • CMAQ sensitivity modeling suggests control of primary carbon (PC) emissions for PM2.5 SIP • Receptor modeling suggests mobile sources and biomass burning as major sources of PC PM2.5 • Other major components of PM2.5 are: • Sulfate: modeling capabilities and control strategies (CAIR) are fairly mature • SOA: understanding of formation mechanism and modeling capabilities are not as well developed • Most available evidence suggest majority of SOA is of biogenic origin (yields, C14 analyses), though some studies suggest anthropogenic origin (Sullivan & Weber, WSOC studies)
What needs further investigation… • How good is our understanding of PM2.5 impacts from mobile-sources, biomass burning? • How do CMAQ and PMF compare? • How does that affect policy development? • Can PMF assist in CMAQ-SOA evaluation? • Can PMF assist in understanding CMAQ’s (high) unspecified/crustal concentrations
Modeling methodology • CMAQ4.5 w/SOA mods*: annual brute-force runs based on the VISTAS 2002 G2 “actual” inventory (alga12km domain), to quantify the impacts of: • Mobile sources (on and off road) • Fires (Rx, wild, agricultural, land clearing, residential) • PMF analysis for 2002-2005 using speciated PM2.5 data from eight STN sites in Georgia • Analyses done for each site separately and using one combined dataset for all sites; fairly similar results • Comparison of source/factor contributions for • Mobile sources • Fires/ biomass burning • Crustals/Soil • SOA Tracked directly by CMAQ * - Morris et al., Atm. Env., 40, 4960-4972, 2006.
CMAQ-based quarterly PM2.5 contributions Jan-Mar Apr-Jun Jul-Sep Oct-Dec Mobile sources Scale of0.0- 4.0 Fires Crustals SOA g/m3
Comparison b/w CMAQ and PMF:Monthly averages, Atlanta STN site Fires Mobile sources Crustals/Soil SOA
Comparison b/w CMAQ and PMF:Daily contributions, Atlanta STN site Mobile, R=0.66 Fires, R=0.23 Jan-Mar, R=0.47/0.67Apr-Dec, R=0.19/0.37 SOA, R=0.56 Road, R=0.66Soil, R=-0.11
Correlations between sources and species:Atlanta STN site Highlighted values: R0.5
Temporal variability in fire emissionsVISTAS G2 actual emissions 3/23/02 EC (g/s)
Temporal variability in fire emissionsVISTAS G2 actual emissions 3/24/02 EC (g/s)
Temporal variability in fire emissionsVISTAS G2 actual emissions 3/25/02 EC (g/s)
Temporal variability in fire emissionsVISTAS G2 actual emissions 3/26/02 EC (g/s)
Contributions from various biomass-burning sources at the Atlanta SEARCH site * * PMF analysis by Kim&Hopke, Atm Env 38, 3349-3362, 2004
Contributions from various biomass-burning sources at the Atlanta SEARCH siteUsing EPA 2001 Rx burning emissions, instead of VISTAS 2002 (see Tian et al., CMAS 2006 presentation for details) * * PMF analysis by Kim&Hopke, Atm Env 38, 3349-3362, 2004
Contributions from various biomass-burning sources at the Atlanta SEARCH siteUsing EPA 2001 Rx burning emissions, instead of VISTAS 2002 (see Tian et al., CMAS 2006 presentation for details) * * Marmur et al., Atm Env 40, 2533-2551, 2006
Summary and future work • Moderate agreement b/w CMAQ and PMF estimates for mobile sources and SOA PM2.5 • Poor agreement for biomass-burning • CMAQ crustals overestimated, temporal variability suggest resuspended road dust • There are “issues” with any modeling approach: • PMF • Measurement uncertainties and limitations • Fixed source compositions • Temporal variability over-estimated • Point measruement / Impacts of local sources • Models-3 (CMAQ) • Uncertainties in emission rates • Temporal variability in emissions under-represented • Meteorology/mixing • Volume average
Summary and future work • Future work • PMF • Use of organic markers data • Markers for SOA (oxidation products) • Investigation of spatial representativeness • Models-3 (CMAQ) • Investigation into Rx burning emissions (and others) • Detailed temporal variability in fire emissions • Soil dust emissions as a function of wind speed, moisture; increased “near-source” removal of particles • Detailed mobile-sources activity • Policy implications • Regulatory needs precedes scientific understanding
Contact Information Amit Marmur, Ph.D.Georgia Dept. of Natural Resources4244 International Parkway, Suite 120Atlanta, GA 30354amit_marmur@dnr.state.ga.us 404-363-7072