PM SIP Meeting Workgroups Sep26 060926revC

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 regionally transported material (primarily sulfate, nitrate and carbon) precursors converted to PM over the urban area diffuse sources within urban area point sources within urban area

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 (*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial (**) Nonferrous Metals: Arnold includes steel processing NR = factor not resolved NR NR NR NR NR Different data collection and analysis methods (especially carbon); consistent source apportionment methodology

Slide21: 

Sulfate Factor (*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial (**) Nonferrous Metals: Arnold includes steel processing NR = factor not resolved NR NR NR NR NR Sulfate factor... Is this gradient from sulfate ion concentration, or from other species present in the sulfate factor?

Slide22: 

Nitrate Factor (*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial (**) Nonferrous Metals: Arnold includes steel processing NR = factor not resolved NR NR NR NR NR Nitrate factor... Is this gradient from nitrate ion concentration, or from other species present in the nitrate factor?

Slide23: 

Mobile Source Factor (*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial (**) Nonferrous Metals: Arnold includes steel processing NR = factor not resolved NR NR NR NR NR Mobile source factor… gradient seems backwards; highest in suburbs and lowest in urban core.

Slide24: 

Soil / Crustal Material Factor (*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial (**) Nonferrous Metals: Arnold includes steel processing NR = factor not resolved NR NR NR NR NR Soil/crustal factor… difficult to assess consistency due to admixing with other sources (see footnote) ?

Slide25: 

Steelmaking Factor (*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial (**) Nonferrous Metals: Arnold includes steel processing NR = factor not resolved NR NR NR NR NR Steel production… relatively large at East St. Louis but small at Blair; not resolved at Arnold

Slide26: 

Nonferrous Metals Processing Factor (*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial (**) Nonferrous Metals: Arnold includes steel processing NR = factor not resolved NR NR NR NR NR Nonferrous metals (zinc, lead, copper)… in aggregate similar contributions across al three sites

Slide27: 

Biomass Burning Factor (*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial (**) Nonferrous Metals: Arnold includes steel processing NR = factor not resolved NR NR NR NR NR 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

Slide28: 

“Carbon-Rich Sulfate” Factor (*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial (**) Nonferrous Metals: Arnold includes steel processing NR = factor not resolved NR NR NR NR NR Carbon-rich sulfate factor… 20% of mass at East St. Louis… what does it represent?

Slide29: 

Representations of Carbonaceous Material Total Carbon (TC)

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 … unresolved OC individual OC compounds EC

Slide33: 

Carbon in the Hopke Apportionments East St. Louis Arnold Blair IMPROVE carbon fractions NIOSH OC/EC

Slide34: 

Apportionments with NIOSH OC/EC at all Sites (*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial (**) Nonferrous Metals: Arnold includes steel processing NR = factor not resolved NR NR NR NR NR NR Carbon-rich sulfate factor primarily distributed to sulfate and nitrate… largely regionally transported carbon?

Slide35: 

Apportionments with NIOSH OC/EC at all Sites (*) Soil: Arnold includes separate Ca-rich factor; Blair = soil + non-soil industrial (**) Nonferrous Metals: Arnold includes steel processing NR = factor not resolved NR NR NR NR NR NR Intraurban gradients still exist! Regional plus local contributions and/or measurement artifacts?

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) N = 213 maximum difference is 4% (ESL vs. Blair)

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 As expected, fine particulate matter sulfate is highly coupled between the two sites. September 5-6

Slide41: 

Daily-Integrated PM-2.5 Sulfate 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 5 September 6 Two days – representing the highest sulfate concentration at each respective site – appear to be outliers. These samples are actually adjacent days…

Slide42: 

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

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) +10% +12% +22% 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 thermal carbon fractions OC/EC … speciated organics

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 PM(ESL) >> PM(Arnold) during calm conditions Are only local emissions pooling or do all PM components increase? ESL wind speed < 0.5 m/sec Large jump in fine PM mass at East St. Louis compared to Arnold

Slide64: 

Thermo SHARP PM2.5 Mass Monitors at East St. Louis and Arnold PM(ESL) >> PM(Arnold) during calm conditions Are only local emissions pooling or do all PM components increase? Small excess at Arnold under advective conditions at ESL; often observed with winds from the east/southeast

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 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... * includes “non-soil industrial” contributions soil factors inconsistently resolved across sites admixed with other sources at Blair Separate calcium-rich factor resolved at Arnold

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 Arnold site marked with “A”