FUEL-BASED PARTICULATE MATTER AND GASEOUS EMISSION FACTORS DETERMINED FROM VEHICLES IN , PA'S SQUIRREL HILL TUNNEL

Andrew Grieshop, Eric Lipsky and Allen Robinson Carnegie Mellon University, Pittsburgh, PA

Why a Tunnel Study? In short… • Vehicles are a major source of fine particles Fuel-based Emission Factors PM2.5 Emissions as a Function of Fleet Composition In general, emissions studies in traffic tunnels are and other pollutants: Similarly, the total PM mass and OC and EC emissions from the different vehicle classes can be used to provide a composite emissions profile of the 2.5 – Carbonaceous Aerosol (OC and EC) • Fuel-based EFs were determined (i.e. mg/kg fuel vs. mg/mile or mg/hp-hr) separated. In each case, multiple instruments/methods were used to determine the tunnel and full range of vehicles found in a geographical region. – Ultra-fine particles • Kg fuel determined through a mass balance on the carbon in fuel background concentrations during the sample periods. The results of each are shown here, along with This study was specifically undertaken to provide an – Gases (NOx, CO, VOCs, NH3, SO2, etc.) values from literature and a brief explanation of the methods used. OC/EC analysis was completed automotive traffic source profile for the Pittsburgh Air • Ideal, stoichiometric combustion of octane (gasoline): • Reliable, representative Emission Factors using NIOSH method 5040. Overall car and HDDV EFs are tabulated in the Conclusions. Quality Study (PAQS) EPA Supersite project. Along (EFs) are needed. with taking a small suite of gas-phase measurements, C8H18 + (12.5O2 + 47N2 ) ⎯⎯→ 8CO2 + 9H 2O + 47N2 600 – E.g. mgpollutant /kgfuel or mgpollutant /mile OC Emission Factor (mg OC/ kg fuel) Hildemann, 1991 Organic Carbon we concentrated on characterizing the composition, • Emission profiles can be determined for use • Actual combustion also yields: CO, carbonaceous PM, NOx, etc. •MOUDI samples size distribution and mass and number emission rates OC(TQQ) in source-receptor modeling. • CO < 2% of emitted carbon, PM << 1% 500 OC (BareQ) Kirchstetter, 1999 collected on Al foil emitted by automobiles and trucks (Heavy Duty Diesel OC (MOUDI) Rogge, 1993 • Emissions from different vehicle • Therefore, fuel mass can be determined with [CO ] and [CO] HDDV EFs from Literature substrates Vehicles – HDDV) in an urban traffic tunnel in classes/types can be determined. 2 Automobile EFs from Literature Pittsburgh, . Later work examined the • However, to calculate EFs from our data, we must: 400 TQQ: Y = 246*X + 35 (R2=0.32) •BareQ indicates – E.g. gasoline, heavy-duty diesel (HDDV), MOUDI: Y = 256*X + 2 (R2=0.82) effect of dilution on the measured PM mass. A light-duty (LDV), ‘smokers’ – Adjust concentrations for background Bare Q: Y = 299*X + 43 (R2=0.41) sample on quartz sampling of this work is presented here. – Account for the fraction of carbon in fuels 300 fiber filter (QFF) Sagebiel, 1997 •TQQ indicates QFF • Emission Factors calculated with: 200 corrected for positive artifact with mass

Rogge, 1993 from a second QFF The Squirrel Hill Tunnel ⎡ P − P ⎤⎡ MW ⎤ 100 EF = tun amb p w sampling behind a The Squirrel Hill Tunnel is a 4-lane highway tunnel on ⎢ ⎥⎢ ⎥ c Kirchstetter, 1999 Teflon membrane []()()CO − CO + []()CO − CO ()MW Gray, 1986 on the eastern edge of the City of ⎣ 2 tun 2 amb tun amb ⎦⎣ C ⎦ 0 filter Pittsburgh. It is 4,225 feet long, has a 2.5% up-grade in the 1600 westerly direction and carries both commercial and non- EC Emission Factor (mg EC/ kg fuel) Hildemann, 1991 (Dyno) – P is species concentration 1200 Kirchstetter, 1999 (Tunnel) Elemental Carbon commercial traffic. Our testing took place in the tunnel’s –wis carbon fraction in fuel: weighted average of gasoline (85%) and diesel fuel (87%) Rogge, 1993 (Dyno) •MOUDI data are the west-bound tube. The majority of the data collection took c –MW are molecular weights of species/carbon sum of stages with place in November of 2002, while later work on the effects p/c EC (Quartz) 400 EC (MOUDI) size cuts of 2.5 µm dilution of aerosol mass took place during the summer of 2 Quartz : Y = 445*X + 28.1 (R =0.63) and below . 2 2004. MOUDI : Y = 383*X + 3.2 (R =0.783) 300 HDDV EFs from Literature •Background Measurements taken in the tunnel were corrected for Automobile EFs from Literature correction based on background levels using data taken from remote sites: Study Average PM Tunnel Emission Factors scaled data from stations run by the Allegheny County Health Department Nitrate 200 Photo Source: Bridges and Tunnels of Allegheny County, PA; 2001 Bruce S. Cridlebaugh TEOM at a remote for gas concentrations and instruments on the CMU Low-speed High Truck 3% High-speed location The western portal of the Squirrel Hill Tunnel campus for PM measurements. (per kg fuel) Overall (Rush-hour) (Early Chloride (Mid-day) Ammonium 2% 100 Morning) 4% Sagebiel, 1997 (Remote Sensing) Other Inorganic Data collected during study: These measurements and samples were Sulfate Kirchstetter, 1999 (Tunnel) NOX (g/kg) 10.4 ± 1.9 11.0 ± 1.9 8.0 ± 1.4 17.9± 3.8 2% Hildemann, 1991 (Dyno) collected from the ventilation shaft above the 10% 0 Continuous Air Quality Measurements traffic tunnel (as shown in the schematic below). PM2.5 (mg/kg) 194 ± 28 174 ± 24 175 ± 23 253± 72 2500 Kirchstetter, 1999 (Tunnel) – Gases: CO2, CO, SO2, NOx, NH3 PM2.5 Emission Factor (mg PM2.5/ kg fuel) –PM: 30° C TEOM Mazzoleni, 2003 (Roadside) Total PM2.5 2.5 Exhaust OC (mg/kg) 58 ± 14 56 ± 10 51 ± 16 101 ± 20 2000 – Size/Number: SMPS and nano-SMPS Fresh •TEOM was run with To Ambient Squirrel Hill Tunnel Hildemann, 1991 (Dyno) – Traffic Count: PennDOT Sensors Air Supply EC (mg/kg) 83 ± 18 95 ± 18 66 ± 17 140±27 1500 Miguel, 1998 (Tunnel) the SES (Sample Schematic EC – Traffic Video: Highway Patrol video 42% Allen, 2001 (Tunnel) Equilibration System) NH (mg/kg) 260 ± 65 209 ± 85 265 ± 57 294 ± 30 PM (MOUDI) 3 OC 1000 2.5 at 30° C, and does Ventilation Tunnels Integrated Air Quality Samples 37% PM (TEOM) 2.5 not show good – Artifact-corrected samples for OC/EC ~ 17% ~ 19% ~ 11% ~ 36% 2 Sampling MOUDI : Y = 1079*X + 33.3 (R =0.753) agreement with – Filters for Organic Speciation (results 2 Location HDDV HDDV HDDV HDDV TEOM : Y = 267*X + 143.3 (R =0.220) MOUDI data (see pending) HDDV EFs from Literature Morris, 1998 (On road opacity) •Carbonaceous aerosol Automobile EFs from Literature below). – MOUDI for size-resolved mass and 500 OC/EC •Overall tunnel emission factors were calculated using a fuel- predominates, representing ~80% •Data is background- – Inorganic gas/PM samples weighted average of sampling period emission factors. of PM mass. corrected using data Traffic Tunnel •Average PM mass concentration – Filters for metals analysis (results Hildemann, 1991 (Dyno) collected from a pending) (West bound) •NOx, PM2.5, Elemental Carbon and Organic Carbon emission 3 in the tunnel is ~ 50 µg/m . Kirchstetter, 1999 (Tunnel) second TEOM at a factors are strongly influenced by the proportion of HDDV in Mazzoleni, 2003 (Roadside) – Canisters for VOC speciation 0 the fleet. remote site in 0% 50% 100% Pittsburgh. HDDV Fuel Fraction Diurnal Patterns in Traffic and Concentration Data Determining the Fraction of Fuel Used by Trucks These plots show time series of average daily traffic conditions and background-corrected pollutant • Fleet composition has large impact on EFs concentration - all exhibit consistent diurnal patterns. Integrated sample collection periods are highlighted Loss of Semi-volatile Mass Limits Usefulness of on the plots. • Can we separate emissions from different vehicle types? • Identify HD vehicles from video tape TEOM for Automotive PM Measurements Rush Mid-day 0.4 Early morning Mid-day 90 hour ⎛ 1 ⎞ Comparison of data taken by the ) Rush • All HD vehicles assumed diesel-powered f ⎜ ⎟ 3 Early morning truck⎜ ⎟ 80 MOUDI (which samples at ambient hour ⎝ mpg⎠ 0.3 • Fraction of fuel consumed by HDDV: %fuel = truck conditions) and the TEOM (which 60 HDDV ⎛ 1 ⎞ ⎛ 1 ⎞ f ⎜ ⎟ + f ⎜ ⎟ samples at 30° C) indicate significant ] (ug/m 0.2 truck⎜ ⎟ car⎜ ⎟ loss of PM mass in the TEOM. This mpg mpg 60 2.5 ⎝ ⎠truck ⎝ ⎠car ) Assumed: 6 mpg for trucks, 21 mpg for cars 3 plot shows that the TEOM under- 30 0.1 predicted PM mass concentrations

g/m 2.5 [PM µ in the tunnel by ~30%. The plot above shows that the variability is even more HDDV Fuel Fraction Fuel HDDV 0.0 40 0 pronounced during the early morning

2500 ( - TEOM hours, when a larger portion of the

80 2.5 fleet is HDDVs.

2000 Influence of Traffic Composition on NOx Emissions PM 20 PM2.5 - TEOM (µg/m3) This effect in the TEOM is presumably 60 2500 Background Corrected NOx (ppbv) 2 Comparing NOx emission factor and Y = 0.69*X (R = 0.71) 1500 2000 due to the loss to the vapor phase of (ppbv) fleet composition indicates vehicle 1:1 Line x 40 1500 the portion of engine exhaust PM 1000 1000 classes can be clearly separated. 0 made up of semi-volatile organics.

NO 500 0 20406080Later work has similarly confirmed the 20 0 Comparison of the diurnal average time 500 3 PM - MOUDI (µg/m ) effects of dilution on the measured PM series of NOx emission factor and fleet 2.5 40 Average NO Emission Factor (g NO /kg fuel) mass in the tunnel. 0 (mph) Speed Average 0 x x composition (as the portion of fuel used by 30 HDDV) clearly suggests that vehicle types 20 should be separable by combining these 800 4000 10 data sets. 0 Conclusions 600 3000 0.4 This is the case, as is shown in the lower Fuel-based EF HDDV Fuel Fraction The following were established during this figure of NO EF versus HDDV Fuel (per kg fuel) Cars HDDV 0.3 x study of vehicle emissions:

(ppmv) 400 2000 0.2 consumption fraction. A very strong 2 0.1 correlation is seen between the two •Clear diurnal patterns in pollutant concentrations, NO (g/kg) 4.6 ± 0.7 40.0± 3.5 CO 200 1000 0.0 variables. Additionally, interpolating the emission factors and traffic density and X

Vehicles Per Hour Per Vehicles 3:30 8:30 13:30 18:30 23:30 least-squares linear fit line to the 0% composition. Study Average +/- 1 sigma Time HDDV (only cars) and 100% HDDV (only 0 0 •Emission factors for the tunnel as a function of PM2.5 (mg/kg) 33 ± 48 1110± 280 Kirchstetter, 1999 (Tunnel) Heavy Duty Diesel Vehicles) levels shows 3:30 8:30 13:30 18:30 23:30 3:30 8:30 13:30 18:30 23:30 the time of day and associated fleet composition. that the results of our study are in good Cassier, 1995 (Tunnel) 40 agreement with a variety of emissions •A study average composition of tunnel particulate OC (mg C/kg) 27± 14 294 ± 84 Study Average Time Pierson, 1995 (Tunnel) Time Pierson, 1996 (Tunnel) studies (ranging from other tunnel studies which shows a predominance of carbonaceous

Pierson, 1997 (Tunnel)

/ kg fuel) / kg to dynamometer and remote sensing aerosol (~80%) and an average daily

+/- 1 sigma 2 studies). concentration of 50 µg/m3. EC (mg C/kg) 16± 16 430±95 Trends in concentrations : Trends in traffic: Note that the increased uncertainties •A clear separation of emissions from cars and

20 associated with the early morning period HDDV via the determination of fleet compositions The table above contains interpolated emission •CO2 level in the tunnel closely tracks traffic •Rush hour period has highest hourly traffic volume, implying that there is less dilution of volume and lowest average speed. (when a larger proportion of traffic is through inspection of traffic video. factors based on the known fleet composition during the various sample periods. They represent exhaust during high-traffic periods. HDDVs) are due to the CO2 and NOx Yanowitz, 2000 (Dyno) Tunnel EFs (g NO2 /kg fuel) •NOx, PM2.5, OC and EC emission factors for •Rush hour (7 AM to 9 AM) and the early Emission (g NO Factor Kirchstetter, 1999 (Tunnel) 2 levels in the tunnel being less elevated composites of data taken from the sources listed x Y = 35.2*X + 4.61 (R =0.850) these two vehicle classes. •NO levels show a less-clear trend. morning hours (12 AM to 6 AM) have the Kean, 2000 (Tunnel) HDDV EFs from Literature x NO above background levels. The mixing and above. Uncertainties were estimated based on the Automobile EFs from Literature lowest and highest proportion of fuel Durbin, 1999 (Dyno) •All concentrations show peaks during rush- dilution in the tunnel is far more variable •The TEOM is not an effective instrument for errors in the least-squares fit of pollutant emission consumed by HDDV’s, respectively. 0 factor versus traffic composition. hour period (7 AM to 9 AM) 0% 50% 100% under these conditions as well. measuring concentrations of fresh, automobile- HDDV Fuel Fraction sourced PM.