observational data analysis to support pm2 5 sip development l.
Download
Skip this Video
Loading SlideShow in 5 Seconds..
Observational Data Analysis to Support PM2.5 SIP Development PowerPoint Presentation
Download Presentation
Observational Data Analysis to Support PM2.5 SIP Development

Loading in 2 Seconds...

play fullscreen
1 / 75

Observational Data Analysis to Support PM2.5 SIP Development - PowerPoint PPT Presentation


  • 107 Views
  • Uploaded on

Observational Data Analysis to Support PM2.5 SIP Development. Jay Turner and Jen Garlock Department of Energy, Environment and Chemical Engineering Washington University. Modeling and Control Strategies Joint Workgroup Meeting Saint Louis, MO September 26, 2006.

loader
I am the owner, or an agent authorized to act on behalf of the owner, of the copyrighted work described.
capcha
Download Presentation

PowerPoint Slideshow about 'Observational Data Analysis to Support PM2.5 SIP Development' - paxton


An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.


- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript
observational data analysis to support pm2 5 sip development

Observational Data Analysis to Support PM2.5 SIP Development

Jay Turner and Jen Garlock

Department of Energy, Environment and Chemical Engineering

Washington University

Modeling and Control Strategies Joint Workgroup Meeting

Saint Louis, MO

September 26, 2006

St. Louis 8-Hour Ozone and PM2.5 State Implementation Plan (SIP) Workgroup

slide2

Motivation & Objectives

  • St. Louis – Midwest Supersite program has collected a wealth of data for fine particle physical and chemical properties
  • Together with the state/local routine monitoring data, there is substantial information to support in PM2.5 SIP planning for the St. Louis area
  • Data analysis is needed to complement chemical transport modeling (CTM)
    • CTM model validation and diagnostics
    • Weight of evidence approach to control strategy development
slide3

BRIEF Summary of the St. Louis Supersite

  • Four year campaign, core monitoring site in East St. Louis, IL
    • Two years of intensive measurements (5/2001 – 5/2003)
    • Two years of measurements with a subset of the initial monitoring platform (6/2003 – 3/2005)
  • Data collection and analysis to support:
    • Development and evaluation of monitoring methods
      • e.g. “Development of a Wet Chemical Method for the Speciation of Iron in Atmospheric Aerosols", B.J. Majestic et al. (2006) Environmental Science & Technology
    • Exposure and health effects studies
      • e.g. "Association of ventricular arrhythmias detected by implantable cardioverter defibrillator and ambient air pollutants in the St Louis, Missouri metropolitan area", D.Q. Rich et al. (2006) Occupational and Environmental Medicine
    • Source apportionment and SIP planning
      • e.g. "Source Identification of Airborne PM-2.5 at the St. Louis - Midwest Supersite", J.H. Lee et al. (2006) Journal of Geophysical Research - Atmospheres
slide4

SIP Planning Support Grant to WUSTL

  • Coordination
  • Organic Carbon Source Apportionment
  • Data Harmonization & Episodes Analysis
  • Urban / Rural Contrast & Intraurban Variability
  • Transport Regimes Analysis
  • Refinements to PM2.5 Mass Apportionment
  • Soil / Road Dust Characterization

Subcontractors:

  • University of Wisconsin (Schauer group)
  • Sonoma Technology, Inc.

Performance period August 1, 2006 through July 31, 2007

(however, effort must be substantially front-loaded)

slide5

SIP Planning Support Grant to WUSTL

  • Coordination
  • Organic Carbon Source Apportionment
  • Data Harmonization & Episodes Analysis
  • Urban / Rural Contrast & Intraurban Variability
  • Transport Regimes Analysis
  • Refinements to PM2.5 Mass Apportionment
  • Soil / Road Dust Characterization

Subcontractors:

  • University of Wisconsin (Schauer group)
  • Sonoma Technology, Inc.

Performance period August 1, 2006 through July 31, 2007

(however, effort must be substantially front-loaded)

slide6

Data Harmonization & Episodes Analysis

  • Develop a “harmonized” data set for East St. Louis for 2002 to be used in chemical transport model validation
    • Reconcile inconsistencies in the data streams
      • Integrated sampling versus continuous monitoring
      • Multiple methods to measure the same parameter
  • Develop a conceptual model for fine particulate matter in STL
    • Emphasis on factors affecting PM2.5 mass
    • Start by examining two-to-three episodes in detail
      • Sulfate episode: August 27 – September 10, 2002
      • Nitrate episode: December 3 – December 17, 2002
      • Carbon episode (?)
    • Significant collaboration between the data analysis and modeling teams
      • PSAT analysis by Morris group (Environ)
    • Modeling by Kleeman group (UC-Davis)
slide7

Example – December 2002 Nitrate Episode

  • Speciation network filter 24-hour integrated filter nitrate
slide8

Example – December 2002 Nitrate Episode

  • STL Supersite daily 24-hour filter nitrate
slide9

Example – December 2002 Nitrate Episode

  • STL Supersite hourly nitrate (Particle-into-Liquid Sampler)
slide10

Example – December 2002 Nitrate Episode

  • STL Supersite hourly nitrate (Particle-into-Liquid Sampler)
    • frequent midday decreases, not captured by filter data
slide11

Example – August/September 2002 Sulfate Episode

  • 24-hour integrated sulfate (red) and hourly sulfate, East St. Louis
slide12

Example – August/September 2002 Sulfate Episode

  • Continuous sulfate measurements also conducted in Reserve, KS during this time period!
slide13

Semicontinuous Sulfate - Data Quality

  • 24-hour average semicontinuous sulfate versus 24-hour integrated filter sulfate

East St. Louis, IL

August 2002 through January 2003

slide14

Conceptual Model for Urban Area Fine PM Mass

point sources within urban area

diffuse sources within urban area

precursors converted to PM over the urban area

regionally transported material (primarily sulfate, nitrate and carbon)

slide15

Intraurban Variability in Fine PM

  • Factors Contributing to Spatial Variability in PM2.5 Concentrations within Urban Areas*:
      • local sources of primary PM (or fast-reacting precursors)
      • topographic barriers separating sites
      • transient emissions events
      • meteorological phenomena
      • differences in the behavior of semi-volatile components
      • measurement error
  • Data from multiple monitors within the urban area can be used to infer intraurban spatial variability in urban PM burdens
    • Must interpret data in light of differences in monitor makes and models (e.g. TEOM vs BAM vs SHARP)

*Pinto, J.P., Lefohn, AS., Shadwick, D.S. Journal of Air and Waste Management Association, 54,440-449, 2004

slide16

STL Source Apportionment Studies

  • Receptor modeling – explain observational data collected at a monitoring site (receptor)
    • linear combinations of source contributions that can “best” explain the observations
slide17

STL Source Apportionment Studies

  • Receptor modeling – explain observational data collected at a monitoring site (receptor)
    • linear combinations of source contributions that can “best” explain the observations
  • Chemical Mass Balance (CMB)
    • all significant sources identified and their emission profiles (fingerprints) are known (does not require emission rates)
slide18

STL Source Apportionment Studies

  • Receptor modeling – explain observational data collected at a monitoring site (receptor)
    • linear combinations of source contributions that can “best” explain the observations
  • Chemical Mass Balance (CMB)
    • all significant sources identified and their emission profiles (fingerprints) are known (does not require emission rates)
  • Receptor Modeling (e.g. PMF)
    • Examine covariance in large observational data sets
    • Reduction in variables to yield a set of “factors” which can explain most of the variance in the data
    • Hopefully these factors represent discernible source categories (factor loadings similar to fingerprints or have other distinguishing features)
      • Confirm using meteorology data, in some case pinpoint specific sources
    • Factors can be admixtures of contributions from multiple sources… no constraints by the actual emission fingerprints
slide19

STL Fine PM Mass Apportionment Studies

* Version of PMF to be determined

** Sensitivity studies and refinements to the apportionment of Lee, Hopke and Turner (2006)

Acknowledgement: Mike Davis (EPA Region VII) for draft synthesis of the contemporary STL PM2.5 mass apportionment studies

slide20

Reconciling the Hopke Group (Clarkson) Apportionments

Different data collection and analysis methods (especially carbon); consistent source apportionment methodology

NR NR NR NR

NR

(*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial

(**) Nonferrous Metals: Arnold includes steel processing

NR = factor not resolved

slide21

Sulfate Factor

Sulfate factor... Is this gradient from sulfate ion concentration, or from other species present in the sulfate factor?

NR NR NR NR

NR

(*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial

(**) Nonferrous Metals: Arnold includes steel processing

NR = factor not resolved

slide22

Nitrate Factor

Nitrate factor... Is this gradient from nitrate ion concentration, or from other species present in the nitrate factor?

NR NR NR NR

NR

(*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial

(**) Nonferrous Metals: Arnold includes steel processing

NR = factor not resolved

slide23

Mobile Source Factor

Mobile source factor… gradient seems backwards; highest in suburbs and lowest in urban core.

NR NR NR NR

NR

(*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial

(**) Nonferrous Metals: Arnold includes steel processing

NR = factor not resolved

slide24

Soil / Crustal Material Factor

Soil/crustal factor… difficult to assess consistency due to admixing with other sources (see footnote)

?

NR NR NR NR

NR

(*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial

(**) Nonferrous Metals: Arnold includes steel processing

NR = factor not resolved

slide25

Steelmaking Factor

Steel production… relatively large at East St. Louis but small at Blair; not resolved at Arnold

NR NR NR NR

NR

(*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial

(**) Nonferrous Metals: Arnold includes steel processing

NR = factor not resolved

slide26

Nonferrous Metals Processing Factor

Nonferrous metals (zinc, lead, copper)… in aggregate similar contributions across al three sites

NR NR NR NR

NR

(*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial

(**) Nonferrous Metals: Arnold includes steel processing

NR = factor not resolved

slide27

Biomass Burning Factor

Biomass burning… not resolved at Blair, not resolved in published East St. Louis apportionment but subsequent work by Hopke group suggests it can be resolved

NR NR NR NR

NR

(*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial

(**) Nonferrous Metals: Arnold includes steel processing

NR = factor not resolved

slide28

“Carbon-Rich Sulfate” Factor

Carbon-rich sulfate factor… 20% of mass at East St. Louis… what does it represent?

NR NR NR NR

NR

(*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial

(**) Nonferrous Metals: Arnold includes steel processing

NR = factor not resolved

slide30

Representations of Carbonaceous Material

  • Total Carbon (TC)
  • Two fractions
    • Organic carbon (OC)
    • Elemental carbon (EC)

OC

EC

slide31

Representations of Carbonaceous Material

  • Total Carbon (TC)
  • Two fractions
    • Organic carbon (OC)
    • Elemental carbon (EC)
  • Eight fractions
    • Five OC fractions
    • Three EC fractions

OC

EC

slide32

Representations of Carbonaceous Material

  • Total Carbon (TC)
  • Two fractions
    • Organic carbon (OC)
    • Elemental carbon (EC)
  • Eight fractions
    • Five OC fractions
    • Three EC fractions
  • Speciated Organics
    • Number of compounds depends on method

individual OC compounds

unresolved OC

EC

slide33

Carbon in the Hopke Apportionments

Arnold

Blair

East St. Louis

NIOSH OC/EC

IMPROVE carbon fractions

slide34

Apportionments with NIOSH OC/EC at all Sites

Carbon-rich sulfate factor primarily distributed to sulfate and nitrate… largely regionally transported carbon?

NR NR NR NR NR

NR

(*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial

(**) Nonferrous Metals: Arnold includes steel processing

NR = factor not resolved

slide35

Apportionments with NIOSH OC/EC at all Sites

Intraurban gradients still exist! Regional plus local contributions and/or measurement artifacts?

NR NR NR NR NR

NR

(*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial

(**) Nonferrous Metals: Arnold includes steel processing

NR = factor not resolved

slide36

Intraurban Variability - Sulfate

  • Sulfate at Arnold, East St. Louis, and Blair (City of St. Louis) during the St. Louis Supersite daily measurements period (include only days with a valid measurement at all three sites)

maximum difference is 4% (ESL vs. Blair)

N = 213

slide37

East St. Louis Sulfate

  • Examine day of week patterns for evidence of local contributions to observed sulfate.
  • Day of week analysis for possibly not robust using 1-in-3 day (e.g. STN) data because there is strong seasonality and, within a season, episodic behavior

STL Supersite, East St. Louis

Daily 24-hour integrated fine PM sulfate by the Harvard-EPA Annular Denuder System (HEADS)

April 2001 – May 2003

slide38

East St. Louis Sulfate – Day of Week

  • Represent a given day’s sulfate by the ratio of its concentration to the weekly average, centered on that day (following Millstein, Harley and Hering, IAC Meeting, September 2006, nitrate analysis)
  • median = black line
  • mean = red line
  • circles = 5th / 95th percentiles

No discernible day of week trends (as expected)

 sulfate concentration essentially all regional, sulfate factor varies due to carbon loadings on this factor

slide39

Monitoring Locations: 8/17/01 – 11/20/01

St. Louis Supersite core site,

East St. Louis, IL

URBAN

St. Louis Supersite satellite site,

Park Hills, MO

RURAL

East St. Louis (IL) is approximately 3 km east of the City of St. Louis (MO) central business district. Park Hills (MO) is a predominantly rural site ~100 km south/southwest of the St. Louis urban core.

slide40

Daily-Integrated PM2.5 Sulfate

September 5-6

As expected, fine particulate matter sulfate is highly coupled between the two sites.

slide41

Daily-Integrated PM-2.5 Sulfate

September 5

Park Hills, MO (rural) - PH versus

East St. Louis, IL (urban) - ESL

8/17/2001 – 11/20/2001

N = 90

Avg ESL = 3.27 mg/m3

Avg PH = 3.13 mg/m3

ESL / PH (urban/rural)…

- Ratio of Site Avg = 1.04

- Avg of Daily Ratio = 1.13

- Geo Mean of Daily Ratio = 1.05

September 6

Two days – representing the highest sulfate concentration at each respective site – appear to be outliers. These samples are actually adjacent days…

slide42

Using Semicontinuous Data at the Urban Site to Reconcile the Apparent Sulfate Outliers

A short-duration (~24 hr) sulfate event passed through the St. Louis on the evening of 9/5 and morning of 9/6. Surface wind data and air mass back trajectories suggest this air mass passed through the rural site about ½ day earlier and thus was largely contained within the 9/5 sample

24-hour sulfate event – very different compared to the aforementioned characteristic multiday pattern!

… an example where the use of daily 24-hour integrated samples might be misleading.

slide43

Intraurban Variability - Nitrate

  • Nitrate at Arnold, East St. Louis, and Blair (City of St. Louis) during the St. Louis Supersite daily measurements period (include only days with a valid measurement at all three sites)

+12%

+22%

+10%

N = 206

slide44

East St. Louis Nitrate

  • Examine day of week patterns for evidence of local contributions to observed nitrate.
  • Again, methodology must account for seasonal and episodic variations which can confound day-of-week analyses

STL Supersite, East St. Louis

Daily 24-hour integrated fine PM nitrate by the Harvard-EPA Annular Denuder System (HEADS)

April 2001 – May 2003

slide45

East St. Louis Nitrate – Day of Week

  • Represent a given day’s nitrate by the ratio of its concentration to the weekly average, centered on that day (following Millstein, Harley and Hering, IAC Meeting, September 2006)
  • median = black line
  • mean = red line
  • circles = 5th / 95th percentiles

Nitrate lowest on Mondays, followed by Sundays and Tuesdays

Modulation of nitrate by weekend/weekday differences in local emissions

slide46

PM2.5 Mass Apportionment Status

  • Sensitivity and related studies are underway for East St. Louis apportionment
  • Repeat modeling for all sites
    • Consistent modeling methodology (EPA PMF)
    • NIOSH OC/EC data for East St. Louis to be consistent with STN sites
    • Use all available data for the modeling but report out annual-average (not study-average) contributions
    • Model each site individually, and combine all sites into a single model
slide47

Refinements to PM2.5 Mass Apportionment (6)

  • Refined interpretation of existing apportionments and also refined apportionments; certain aspects already discussed
  • Example of refined interpretation of an existing apportionment
    • Carbon-rich sulfate (CRS) factor in the East St. Louis apportionment by Lee, Hopke & Turner (2006)
    • ~20% of the PM2.5 mass observed at East St. Louis
    • Relatively high EC/OC ratio in the factor profile suggests unaged, and therefore likely local, carbon
    • Compare total carbon apportionments for
      • IMPROVE carbon fractions
      • IMPROVE OC/EC
      • NIOSH OC/EC
slide48

Total Carbon Distribution Across Factors- three methods for representing carbon -

Carbon fractions: TC(CRS, nitrate, sulfate) = 1.8 mg/m3

IMPROVE OC/EC: TC(nitrate, sulfate) = 1.5 mg/m3

NIOSH OC/EC: TC(nitrate, sulfate) = 1.5 mg/m3

CRS LIKELY REGIONAL!

slide49

Interpretation of Carbon-Rich Sulfate Factor

  • Factor profile predominantly carbon, some sulfate
    • Relatively high EC/OC ratio suggests unaged carbon and thus likely local sources
    • However, modeled apportionments using different representations for carbon suggests the factor represents regional sources
  • Reconcile East St. Louis TC apportionment with urban/rural contrast, August-November 2001 measurements (Park Hills)
  • Assume TC at Park Hills is the regional contribution
    • Add in the modeled TC apportioned to STL local sources (all factors except sulfate, nitrate)
    • Can we reconstruct the observed TC at East St. Louis?
      • Two scenarios, carbon-rich sulfate as a local contribution or as a regional contribution
slide50

Carbon-Rich Sulfate Factor as Regional Source

  • Despite the relatively high EC/OC ratio in the CRS factor, treat as regional rather than local source… good reconstruction!
slide51

Carbon-Rich Sulfate Factor as Regional Source

  • Carbon-Rich Sulfate factor for East St. Louis often underestimates measured Park Hills TC; however, the modeled ESL TC often underpredicts the observed ESL TC
slide52

Urban/Rural Contrast

  • Data from paired urban/rural monitors can be used to infer regional versus local contributions to urban PM burdens
  • East St. Louis and Park Hills: Mid-Aug. to mid-Nov. 2001
    • Lead analysis by WUSTL
  • Blair/Arnold and Bonne Terre: 2003-present
    • Lead analysis by Sonoma Technology, Inc.
  • Rao et al. (2003)
  • March 2001 - February 2002
  • TCM = total carbon material (1.8 times the organic carbon mass)
  • rural concentration (regional contribution) on bottom, urban excess on top
slide53

Urban/Rural Contrast – STN Organic Carbon

  • Comparing Blair (City of St. Louis – urban) to Bonne Terre (rural), there is an OC urban excess at Blair on virtually every sampling day

May - September only, 2003 & 2004

Assuming urban plumes do not impact the rural site, then nearly 100% urban excess for the summer months!

slide54

Urban/Rural Contrast – STN Organic Carbon

  • We are currently cleaning up the STN data for the entire period available to perform annual estimates.
    • e.g. 23 carbon samples in the Arnold January 2003 – March 2005 data set deemed suspect based on data validation checks

February 2003 – March 2005

Assuming urban plumes do not impact the rural site, then nearly 100% urban excess on an annual basis.

PRELIMINARY – CONTINUE SANITIZING THE DATA (ALSO NEEDED FOR PMF APPORTIONMENT)

slide55

Urban/Rural Contrast – STN Organic Carbon

  • Urban/rural contrast results for OC depends upon appropriate blank corrections to the data.
    • Current approach subtracts 0.9 mg/m3 for all samples for all sites based on our interpretation of field- and trip-blank data for each site Is this approach robust? What are the sensitivities of the results?
  • Methodology assumes urban plume does not impact the rural site and the urban and rural sites are bathed in the same regional air mass... need examine these assumptions.
slide56

Organic Carbon Apportionment - Schauer group (U. Wisconsin) -

  • Carbonaceous matter is a significant fraction of the ambient fine PM burden in STL
  • PM mass apportionments using thermal carbon fractions or OC/EC data does not provide a robust apportionment of carbon
  • Organic carbon mass apportionment using organic molecular marker data is needed
slide57

Towards more specificity in representing carbon

OC/EC

speciated organics

thermal carbon fractions

slide58

Organic Carbon Apportionment - Schauer group (U. Wisconsin) -

  • Carbonaceous matter is a significant fraction of the ambient fine PM burden in STL
  • PM mass apportionments using thermal carbon fractions or OC/EC data does not provide a robust apportionment of carbon
  • Organic carbon mass apportionment using organic molecular marker data is needed
  • Schauer group (University of Wisconsin – Madison) is performing carbon apportionments for East St. Louis
    • OC apportionment by CMB completed
    • OC apportionment by PMF currently being optimized
    • EC apportionment by PMF recently initiated
slide59

Primary OC Apportionment by CMB

  • East St. Louis, 1-in-6 day data with organic speciation by extraction-GCMS, June 2001 – May 2003
  • CMB apportionment assumes we know all of the primary OC sources and have representative source profiles!
slide60

Preliminary OC Apportionment by PMF

  • PMF modeling to compare and contrast with CMB results
    • not an optimized apportionment of OC
  • Eight factors resolved, in order of decreasing contribution
    • Resuspended soil factor… relatively poor agreement with CMB daily contributions; need regional dust profiles
    • Mobile source factor… relatively good correlation with sum of CMB mobile source factors daily contributions
    • Wood combustion factor… very good correlation with CMB daily contributions
slide61

Preliminary OC Apportionment by PMF

  • PMF modeling to compare and contrast with CMB results
    • not an optimized apportionment of OC
  • Eight factors resolved, in order of decreasing contribution
    • Resuspended soil factor… relatively poor agreement with CMB daily contributions; need regional dust profiles
    • Mobile source factor… relatively good correlation with sum of CMB mobile source factors daily contributions
    • Wood combustion factor… very good correlation with CMB daily contributions
    • Secondary organic aerosol factor
    • Two point source factors
    • Two winter combustion factors
  • Good correlation does not imply similar mass concentrations apportioned to the source category!
    • Premature to use the quantitative results
  • OC apportionment by PMF currently being optimized to obtain refined and possibly likely quantitative apportionment

THESE SOURCES NOT INCLUDED IN CMB ANALYSIS!

slide62

A few additional items…

  • Returning to intraurban variability, can gain insights from the hourly fine mass monitoring data?
slide63

Thermo SHARP PM2.5 Mass Monitors at East St. Louis and Arnold

Large jump in fine PM mass at East St. Louis compared to Arnold

ESL wind speed < 0.5 m/sec

  • PM(ESL) >> PM(Arnold) during calm conditions
    • Are only local emissions pooling or do all PM components increase?
slide64

Thermo SHARP PM2.5 Mass Monitors at East St. Louis and Arnold

Small excess at Arnold under advective conditions at ESL; often observed with winds from the east/southeast

  • PM(ESL) >> PM(Arnold) during calm conditions
    • Are only local emissions pooling or do all PM components increase?
slide65

Microscale Meteorological Effects at Arnold?

  • Examine data and source apportionment results for possible artifacts from such effects
slide66

Microscale Meteorological Effects at Arnold?

  • Examine data and source apportionment results for possible artifacts from such effects
slide67

Intraurban Variability and Microscale Conditions at the Arnold Site

  • KEY POINT: Care must be exercised in comparing and contrasting PM2.5 mass apportionments for different monitoring sites in the STL area
  • Lee and Hopke (2006) performed PM2.5 mass apportionments by PMF for Blair Street site and Arnold site
    • Mobile Source contributions
      • Blair = 2.8 mg/m3
      • Arnold = 4.0 mg/m3
      • Why 43% greater at Arnold compared to Blair?
    • Conditional probability plot for Gasoline Engine factor at Arnold (80% of gasoline + diesel contributions)…
slide68

Gasoline Factor at Arnold

Lee and Hopke, 2006

  • Conditional Probability Plot (presumably Lambert Airport meteorology data)…
  • Lobe to the southeast points towards a nearby industrial park, including a large aluminum can manufacturing facility.
    • Are emissions mobile source or industrial) from this zone being admixed into the gasoline factor?
    • Or, is this a result of the poor ventilation with winds from the E/SE?
slide69

Transport Regimes Analysis

  • Emission Impact Potential (EIP) Analysis
    • To be performed by Sonoma Technology, Inc.
    • Merge air mass histories (HYSPLIT) and emissions field to interpret observed PM burdens
      • Conventionally SO2 and NOx; expand to include NH3
  • Examine factors driving year-to-year differences in regional contributions to STL PM burdens (especially sulfate)
    • Hypothesis: different patterns in synoptic weather
    • Perform clustering on air mass back trajectories
    • Examine frequency distributions of the various air mass patterns
    • Examine relationships to PM and species concentrations after adjusting for local influences such as stagnations
slide70

3D PATH Analysis - Preliminary Clustering

- St. Louis 72-hr HYSPLIT,

1200 CST arrival at 50% of modeled mixing layer depth

- Radius of Proximity = 9

- Presented in order from highest to lowest frequency

slide71

Soil / Road Dust Characterization

  • Case already made for generating local soil profiles towards interpreting the OC apportionment
  • For the mass apportionment...
  • soil factors inconsistently resolved across sites
  • admixed with other sources at Blair
  • Separate calcium-rich factor resolved at Arnold

* includes “non-soil industrial” contributions

slide72

Local Soil / Road Dust Profiles Needed

  • East St. Louis soil factor has no calcium!
    • most of the Ca is in the diesel factor
  • Will collect local soil samples, suspend in a chamber and sample PM2.5, then analyze for chemical composition

no Ca in ESL soil factor

slide73

Summary

  • Ultimate goal is a defensible control strategy
    • Analyzing the observational data to provide technical support towards that effort
  • Have already conducted numerous data analyses and are currently placing that work in a context relevant to culpability assessments and control strategy development
    • Work over the next few months will explicitly focus on SIP support
    • Thereafter, consider modeling and measurements to address uncertainties and gaps in our understanding of the factors driving fine PM burdens in STL
slide75

Lee & Hopke (2006) Gasoline Factor at Arnold

Lee and Hopke, 2006

  • Arnold site marked with “A”