Science of the Total Environment 409 (2011) 2384–2396

Contents lists available at ScienceDirect

Science of the Total Environment

j ourna l homepage: www.e lsev ie r.com/ locate /sc i totenv

Particulate emission factors for mobile fossil fuel and biomass combustion sources

John G. Watson a,⁎, Judith C. Chow a, L.-W. Antony Chen a, Douglas H. Lowenthal a,Eric M. Fujita a, Hampden D. Kuhns a, David A. Sodeman a,1, David E. Campbell a,Hans Moosmüller a, Dongzi Zhu a, Nehzat Motallebi b

a Desert Research Institute, 2215 Raggio Parkway, Reno, NV 89512, United Statesb California Air Resources Board, Research Division, 1001 I Street, Sacramento, CA 95812, United States

⁎ Corresponding author at: Desert Research Institute, 289512, United States. Tel.: +1 775 674 7046; fax: +1 7

E-mail address: john.watson@dri.edu (J.G. Watson).1 Present Address: San Diego Air Pollution Control Di

0048-9697/$ – see front matter © 2011 Elsevier B.V. Adoi:10.1016/j.scitotenv.2011.02.041

a b s t r a c t

a r t i c l e i n f o

Article history:Received 12 July 2010Received in revised form 23 February 2011Accepted 25 February 2011Available online 1 April 2011

Keywords:Emission factorsMotor vehicleBiomass burning

PM emission factors (EFs) for gasoline- and diesel-fueled vehicles and biomass combustion were measured inseveral recent studies. In the Gas/Diesel Split Study (GD-Split), PM2.5 EFs for heavy-duty diesel vehicles(HDDV) ranged from 0.2 to ~2 g/mile and increased with vehicle age. EFs for HDDV estimated with the U.S.EPA MOBILE 6.2 and California Air Resources Board (ARB) EMFAC2007 models correlated well with measuredvalues. PM2.5 EFs measured for gasoline vehicles were ~two orders of magnitude lower than those for HDDVand did not correlate with model estimates. In the Kansas City Study, PM2.5 EFs for gasoline-powered vehicles(e.g., passenger cars and light trucks) were generally b0.03 g/mile and were higher in winter than summer.EMFAC2007 reported higher PM2.5 EFs than MOBILE 6.2 during winter, but not during summer, and neithermodel captured the variability of the measured EFs. Total PM EFs for heavy-duty diesel military vehiclesranged from 0.18±0.03 and 1.20±0.12 g/kg fuel, corresponding to 0.3 and 2 g/mile, respectively. Thesevalues are comparable to those of on-road HDDV. EFs for biomass burning measured during the FireLaboratory at Missoula Experiment (FLAME) were compared with EFs from the ARB Emission EstimationSystem (EES) model. The highest PM2.5 EFs (76.8±37.5 g/kg) were measured for wet (N50% moisturecontent) Ponderosa Pine needles. EFs were generally b20 g/kg when moisture content was b20%. The EESmodel agreed with measured EFs for fuels with low moisture content but underestimated measured EFs forfuel withmoisture content N40%. Average EFs for dry chamise, rice straw, and dry grass werewithin a factor ofthree of values adopted by ARB in California's San Joaquin Valley (SJV). Discrepancies between measured andmodeled emission factors suggest that there may be important uncertainties in current PM2.5 emissioninventories.

215 Raggio Parkway, Reno, NV75 674 7009.

strict, San Diego, CA, USA.

ll rights reserved.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

Particulate matter (PM) emissions affect the Earth's climate(MacCracken, 2008a, 2008b), visibility (Chow et al., 2002; Watson,2002), surface soiling (Sabbioni and Brimblecombe, 2003; Sabbioniet al., 2003), crop productivity (Grantz et al., 2003), and human health(Chow et al., 2006; Mauderly and Chow, 2008; Pope, III and Dockery,2006).

Annual emission rates are compiled by states, provinces, andcountries (CARB, 2009a; Environment Canada, 2008; EPD, 2008; U.S.EPA, 2008a) in a bottom-up approach to estimate primary PM2.5 andPM10 (PM mass with aerodynamic diameters less than 2.5 and 10 μm,

respectively), carbon monoxide (CO), reactive organic gasses (ROG,sometimes termed total non-methane hydrocarbons [NMHC] orvolatile organic compounds [VOCs]), sulfur dioxide (SO2), oxides ofnitrogen (NOx), and sometimes ammonia (NH3). These inventoriesare usually expressed as tons/year or tonnes/year and are derived asthe products of emission factors (EFs) and activities for differentsource categories (Mobley et al., 2005). PM2.5 and PM10 massemissions can be sub-divided into chemical components by applyingsource profiles (Watson, 1984; Watson et al., 2008a), or the massfraction of each measured chemical component in primary emissionsfor each source category (CARB, 2009b; U.S. EPA, 2007).

The majority of PM mass from combustion sources such as engineexhaust and biomass burning is in the PM2.5 fraction (Lighty et al.,2000; Lloyd and Cackette, 2001), and emission rates and compositionshave changed as new fuels and combustion technologies have beenadopted (Chow, 2001). In 2006, mobile fossil fuel and biomasscombustion sources accounted for 16 and 47% of PM2.5 emissions,respectively, and 43 and 53% of black carbon (BC) emissions,

Fig. 1. Comparisons of measured diesel-fueled vehicle PM2.5 emission factors (EFs) forthe Hot City-Suburban route (HCS) driving cycle during the Gas/Diesel Split Study withMOBILE 6.2 and EMFAC 2007 model estimates for the Federal Test Procedure (FTP)cycle for each diesel group. See Table 1 for vehicle identification codes and compositeinformation. Composites in each diesel group (light/medium heavy-duty, heavy heavy-duty, and urban bus) are ordered by the average vehicle model year. Error barsassociated with the Gas/Diesel Split Study data indicate measurement uncertainties.

2385J.G. Watson et al. / Science of the Total Environment 409 (2011) 2384–2396

respectively, in California (Chow et al., 2010). PM2.5 EFs weremeasured in the Gas/Diesel Split Study (GD-Split; Fujita et al.,2007a, 2007b), the Kansas City Study (Kishan et al., 2006; Namet al., 2008; U.S. EPA, 2008b, 2008c), the Strategic EnvironmentalResearch and Development Program (SERDP; Watson et al., 2008b),and the Fire Laboratory at Missoula Experiment (FLAME; McMeekinget al., 2008). Measurements from these studies are summarized,evaluated, and compared with those from California's emissioninventory. Because PM2.5 from engine exhaust and biomass burningare primarily composed of organic and elemental carbon (OC and EC),accurate emission estimates are needed to evaluate future climate-related emission control strategies (Bond and Sun, 2005; Jacobson,2002) as well as to attain National Ambient Air Quality Standards(Bachmann, 2007; Chow et al., 2007a) for PM2.5.

2. Emission characterization studies

2.1. Diesel and gasoline engine emission factors

The GD-Split Study measured exhaust from 53 light-duty vehicles(52 gasoline- and 1 diesel-fueled) and 34 light-, medium-, and heavy-heavy-duty diesel-fueled vehicles (HDDV). Dynamometer emissiontests were conducted at the Ralphs Grocery distribution center inRiverside, California, during the summer of 2001 (June 2–23 for light-duty gasoline- and diesel-fueled vehicles and from July 20 toSeptember 19 for HDDV). Emissions were sampled into a constant-volume sampler with continuous monitoring for CO, CO2, NMHC, andNOx, and integrated filter sampling for PM2.5 mass, elements, ions, OC,EC, and organic compounds. PM2.5 emission rates were estimatedusing the MOBILE 6.2 (Cook et al., 2007; U.S. EPA, 2008d) andEMFAC2007 (CARB, 2007) emission models under conditionscorresponding to those in the GD-Split Study tests (Fujita et al.,2007a, 2007b). The EMFAC2007 model considers technology groupand odometer mileage in addition to vehicle model year. The MOBILE6.2 model accounts for vehicle type and age but omits the influence offuel type, mileage, driving mode, and vehicle maintenance (Rakhaet al., 2003; McCarthy et al., 2006). The MOtor Vehicle EmissionSimulator (MOVES) model (http://www.epa.gov/otaq/models/moves/index.htm) improves on MOBILE 6.2, but was not available atthe time of this analysis. MOBILE 6.2 calculates EFs in grams pervehicle mile traveled (g/VMT) for PM2.5 mass, lead (Pb), sulfate(SO4

=), OC, and EC from gasoline- and diesel-engine exhaust, as wellas for brake and tire wear. EMFAC2007 estimates non-speciated PM2.5

and PM10 EFs (in g/VMT). To facilitate comparisons of modelestimates with dynamometer measurements, which only accountfor tailpipe emissions, MOBILE 6.2 PM2.5 EFs were calculated from thesum of Pb, SO4

=, OC, and EC emissions for gasoline- and diesel-fueledvehicles.

Gasoline-fueled vehicles were operated according to a modifiedCalifornia Unified Driving Cycle Schedule (UDC; DieselNet, 2008). TheUDC is more aggressive in terms of acceleration andmaximum speedsthan the Federal Test Procedure (FTP), especially during the hot-stabilized portion of the cycle. The gasoline-fueled vehicles weretested under “Warm Start” (WS) and “Cold Start” (CS) cycles. The Hot-City-Suburban (HCS) Heavy Vehicle Route and Highway Cycle (HW)were used in HDDV tests. Additional HDDV test cycles included theCold-City-Suburban (CCS) Heavy Vehicle Route, hot idle (ID) and coldidle (CID) periods, a City-Suburban Heavy Vehicle Route with JacobsBrake (CSJ), and a Heavy-Duty Urban Dynamometer Driving Schedule(UDDS). Busses were tested on the HCS andManhattan Cycle (MC) forTransit Busses cycles.

Observed and model-estimated PM2.5 EFs for diesel-fueledvehicles are presented in Fig. 1. Because some EFs from the GD-SplitStudy represented composites of exhaust frommore than one vehicle,the corresponding MOBILE 6.2 and EMFAC2007 model estimates arepresented as ranges. Table 1 includes 23 gasoline-fueled vehicle

sample composites tested under WS and CS cycles, and 17 diesel-fueled vehicle sample composites tested under various cycles. Becausethe HCS cycle is common for all heavy-duty diesel vehicles in thisstudy (Table 1), HCS PM2.5 EFs are compared with MOBILE 6.2 andEMFAC2007 model estimates (for the FTP cycle) in Fig. 1 with theunderstanding that EFs for other cycles may differ (Fujita et al.,2007b).

Fig. 1 shows that the EMFAC2007 model slightly overestimateddiesel-fueled vehicle emissions for GD-Split Study tests, especially forlow emitters, but the overall agreement was good (r2=0.8)considering the variability among individual vehicles. MOBILE 6.2underestimated measured diesel vehicle EFs (within an order ofmagnitude), and correlation with measurements was moderate(r2=0.63). Differences between minimum and maximum EF esti-mates by MOBILE 6.2 were small.

Both MOBILE 6.2 and EMFAC2007 estimated an increase in dieselvehicle PM2.5 EFs with vehicle age (i.e., the difference betweencalendar year [2001] and vehicle model year), as shown in Fig. 2.Inter-cycle comparisons of measured EFs for typical EMFAC2007medium heavy-duty vehicles (14,001–33,000 lbs) and heavy-heavy-duty vehicles (N33,000 lbs) are presented in Fig. 3. Both CCS and HCScycles produced similar EFs which were about double those of HWcycle EFs. Sample composite CI-9e (Table 1) on the UDDS (i.e., FTP)cycle produced an EF ~20% lower than those measured on the HCS orCCS cycles.

The GD-Split Study gasoline-fueled vehicles were either passengercars (LDA) or light-duty trucks (LDT). Their emissions were oftenmixed in a composite sample (Table 1). MOBILE 6.2 and EMFAC2007reported distinct EFs for LDA and LDT vehicles, resulting in a widerrange of EFs. Information on vehicle maintenance was not availableand is not reflected in MOBILE 6.2 and EMFAC2007 EF estimates. Thecomparisons in Fig. 4 show that measured EFs for the WS and CScycles were more variable than the modeled EFs, especially forvehicles manufactured before 1989 (Fig. 4a). A few high-emittingvehicles (often referred to as smokers) produced clear outliers (seefootnote to Fig. 4), and all vehicles manufactured after 1995 displayedlower measured than modeled EFs.

Fig. 5 compares the GD-Split Study PM2.5 EFs for gasoline-fueledvehicles under WS and CS cycles. EFs for CS (Fig. 5b) were higherthan those for WS (Fig. 5a) with a few exceptions. Modeled andmeasured diesel-fueled vehicle EFs from Fig. 2 are superimposed in

Table 1Descriptions of vehicle sample composites during the Gas/Diesel Split Studya.

Sample composite code Vehicle test cycleb MOBILE model vehicle categoryc EMFAC model vehicle categoryd Number of vehicles tested Range of vehicle model year Odometer range (miles)

SI_1_1 CS, WS LDGV/LDGT1 LDA/LDT1 4 1995–1997 23 K–59 KSI_2_1 CS, WS LDGV/LDGT1 LDA/LDT1 4 1995 32 K–83 KSI_3_1 CS, WS LDGV/LDGT1 LDA/LDT1 4 1995–1999 95 K–125 KSI_4_1 CS, WS LDGV/LDGT1 LDA/LDT1 4 1991–1992 52 K–134 KSI_5_1 CS, WS LDGV/LDGT1 LDA/LDT1 2 1984–1995 84 K–154 KSI_5_2 CS, WS LDGV/LDGT1 LDA/LDT1 6 1992–1995 103 K–216 KSI_6_1 CS, WS LDGV/LDGT1 LDA/LDT1 6 1991–1995 120 K–172 KSI_6_2 CS, WS LDGV/LDGT1 LDA/LDT1 2 1990–1991 149 K–160 KSI_6_3 CS, WS LDGV/LDGT1 LDA/LDT1 1 1992 160 KSI_7_1 CS; WS LDGV/LDGT1 LDA/LDT1 4 1986–1989 92 K–418 KSI_7_2 CS, WS LDGV/LDGT1 LDA/LDT1 1 1987 162 KSI_7_3 CS, WS LDGV/LDGT1 LDA/LDT1 1 1989 174 KSI_8_1 CS, WS LDGV/LDGT1 LDA/LDT1 2 1983–1984 197 K–248 KSI_8_2 CS, WS LDGV/LDGT1 LDA/LDT1 1 1985 212 KSI_8_3e CS, WS LDGV/LDGT1 LDA/LDT1 1 1984 167 KSI_9_1 CS, WS LDGV/LDGT1 LDA/LDT1 2 1979–1980 159 K–182 KSI_9_2 CS, WS LDGV/LDGT1 LDA/LDT1 1 1977 158 KSI_9_3 CS, WS LDGV/LDGT1 LDA/LDT1 1 1979 121 KSI_9_4e CS, WS LDGV/LDGT1 LDA/LDT1 1 1980 98 KSI_10_1 CS, WS LDGV/LDGT1 LDA/LDT1 1 1989 421 KSI_10_2 CS, WS LDGV/LDGT1 LDA/LDT1 1 1990 259 KSI_10_3 CS, WS LDGV/LDGT1 LDA/LDT1 1 1978 128 KSI_10_4e CS, WS LDGV/LDGT1 LDA/LDT1 1 1988 149 KLCI-11_1 CS, WS LDDT12 LDT1 1 1982 162 KCI-10 CCS, HCS, HW HDDV8 HHDV 3 1992–1993 109 K–842 KCI-11 HCS, HW HDDV8 HHDV 5 1994–1997 109 K–602 KCI-11e HCS, HW, CID, ID HDDV8 HHDV 1 1995 241 KCI-11n CCS, HCS, HW, ID HDDV8 HHDV 1 1994 NACI-12 CCS, HCS, HW HDDV8 HHDV 4 1998–2001 145 K–327 KCI-13.1 HCS, MC HDDBT UB 1 1992 519 KCI-13.2 HCS, MC HDDBT UB 1 1982 103 KCI-4r HCS HDDV3 LHDT2 1 2000 45 KCI-5 HCS, HW HDDV7 MHDT 1 1988 170 KCI-8r HCS HDDV5 MHDT 1 1999 15 KCI-9e CCS, HCS, CSJ, HW, CID, UDDS HDDV8 HHDT 1 1985 36 KCI-9n CCS, HCS, HW, ID HDDV7 MHDT 1 1985 501 KCI-Ia CCS, HCS, HW HDDV3 LHDT2 2 1989–1990 NACI-Ib CCS, HCS, HW HDDV2/HDDV3 LHDT1/LHDT2 5 1997–2000 NACI-II HCS, HW,ID HDDV6 MHDT 5 1995–1999 15 K–162 KCI-IIb HCS, HW HDDV6 MHDT 1 1995 151 K

a Fujita et al., 2007a; 2007b.b Driving cycles: CCS (Cold City-Suburban Route); CID (Cold Idle); CS (Cold Start Unified Driving Cycle [UDC]); CSJ (Hot-City-Suburban with Jacobs Brake); HCS (Hot City-Suburban Route); HW (Highway Cycle); ID (Idle); MC

(Manhattan Cycle for Transit busses); UDDS (Urban Dynamometer Driving Schedule); WS (Warm Start UDC).c MOBILE model category: LDGV (Light-Duty Gasoline Vehicle); LDGT1 (Light-Duty Gasoline Truck; weight class 1 [0–3000 lbs]); LDDT12 (Light-Duty Diesel Truck; combined weight class 1 and 2 [0–6000 lbs]); HDDV1 (Heavy-Duty

Diesel Vehicle; weight class 1 [0–8500 lbs]); HDDV2 (Heavy-Duty Diesel Vehicle; weight class 2 [8501–10,000 lbs]); HDDV3 (Heavy-Duty Diesel Vehicle; weight class 3 [10,001–14,000 lbs]) HDDV5 (Heavy-Duty Diesel Vehicle; weightclass 4 [14,001–16,000 lbs]); HDDV5 (Heavy-Duty Diesel Vehicle; weight class 5 [16,001–19,500 lbs]); HDDV6 (Heavy-Duty Diesel Vehicle; weight class 6 [19,501–26,000 lbs]); HDDV7 (Heavy-Duty Diesel Vehicle; weight class 7 [26,001–33,000 lbs]); HDDV8 (Heavy-Duty Diesel Vehicle; weight class 8 [N33,000 lbs]); HDDBT (Heavy-Duty Diesel Bus Transit).

d EMFACModel Category: LDA (Light-Duty Passenger Vehicle); LDT1 (Light-Duty Truck; weight class 1 [0–5750 lbs]); LHDT1 (Light Heavy-Duty Truck; weight class 1 [8501–10,000 lbs]); LHDT2 (Light Heavy-Duty Truck; weight class 2[10,001–14,000 lbs]); MHDT (Medium Heavy-Duty Truck; 14,001–33,000 lbs); HHDT (Heavy Heavy-Duty Truck; 33,001–60,000 lbs); UB (Urban Bus).

e High emitting vehicles; smokers.

2386J.G

.Watson

etal./

Scienceof

theTotal

Environment

409(2011)

2384–2396

Fig. 2. Comparisons between modeled (MOBILE 6.2 and EMFAC2007) and measureddiesel-fueled vehicle PM2.5 emission factors (EFs) for the Hot-City-Suburban route(HCS) driving cycle during the Gas/Diesel Split Study. Bubble diameter representsaverage vehicle age from smallest (0=model year 2001) to largest (19=model year1982).

2387J.G. Watson et al. / Science of the Total Environment 409 (2011) 2384–2396

Fig. 5 for comparison. Measured PM2.5 gasoline-fueled vehicle EFswere one to two orders of magnitude lower than those measured fordiesel-fueled vehicles, but neither MOBILE 6.2 nor EMFAC2007models reflect the large variability in measured gasoline-fueledvehicle EFs. Gasoline-fueled vehicle EFs may be more sensitive tofactors such as engine design and vehicle maintenance comparedwith diesel vehicles.

Prior studies (Gertler, 2005) suggest that gasoline-fueled vehiclesare responsible for a substantial portion of PM2.5 emissions from on-road mobile sources, because they outnumber diesel-fueled vehiclesand include high-emitting vehicles and operating cycles not capturedby certification tests. The Kansas City Study (Kishan et al., 2006; Namet al., 2008; U.S. EPA, 2008b, 2008c) during summer of 2004 andwinter of 2005 intended to better characterize these high-emittingvehicles and cycles. PM2.5 emissions were measured from ~100vehicles, divided into four age groups by model year (i.e., pre–1981;1981–1990; 1991–1995; and post–1995) for LDA and LDT. Emissionsmeasurements were made for each of the three phases of the UDC(i.e., WS, CS, and Hot-Stabilized), but composite EFs are reported forcomparison with the emission models.

Fig. 3. Comparison of measured PM2.5 emission factors (EFs) during the Gas/Diesel SplitStudy between a medium heavy-duty diesel truck (CI-9n) and heavy heavy-duty dieseltruck (CI-9e) under different driving cycles. Vehicle Driving Cycle: CCS (Cold City-Suburban Route); CCSF (Cold City-Suburban Route operated with Federal fuel); HCS(Hot City-Suburban Route); HW (Highway Cycle); ID (Idle); CSJ (Hot-City-Suburbanwith Jacobs Brake); UDDS (Urban Dynamometer Driving Schedule); CID (Cold Idle). SeeTable 1 for vehicle classification. Units are in g/mile except for ID and CID (g/min).

Table 2 summarizes the vehicle sample composites, modelyears, and categories (e.g., passenger vehicle or truck) assigned inthe MOBILE 6.2 and EMFAC2007 models. Measured and modeledPM2.5 EFs during summer and winter are compared in Fig. 6. Therewas a general decrease in PM2.5 EFs for later model year vehicles,although it is not clear whether this is related to improvedtechnology or reflects “…varying levels of vehicle deterioration”(U.S. EPA, 2008b, 2008c). The highest EFs were associated with thethree oldest vehicles (i.e., 1980 passenger vehicle [S5-5]; and 1985and 1989 trucks [S2-4 and S2-1, respectively],) tested duringsummer and three pre–1989 vehicles (i.e., W6-1, W6-4, and W2-1)tested during winter. In general, gasoline-fueled LDT had slightlyhigher PM2.5 EFs than did LDA. Fig. 7 compares modeled andmeasured PM2.5 EFs for summer 2004 and winter 2005. MeasuredEFs were higher in winter than summer. EMFAC2007 estimatedhigher EFs than MOBILE 6.2 during winter (Fig. 7b), but there wasno clear distinction in summer. Similar to the results for the GD-Split Study (Fig. 5), neither model captured the variability of themeasured PM2.5 EFs.

Non-road diesel vehicles were tested at a Marine Corps TrainingFacility in southern California during April 2007 as part of theSERDP Study (Watson et al., 2008b, Zhu et al., 2011). These heavy-duty vehicles serve as surrogates for off-road equipment used inconstruction and agriculture (e.g., graders, front loaders, tractors,and harvesters). Two Medium Tactical Vehicle Replacementvehicles (MTVR; Vehicles 1 and 2) and one Logistics VehicleSystem vehicle (LVS; Vehicle 12) were tested to measure total PM(with no size cut) EFs under loop and extended driving conditions(defined in Table 3) using in-plume measurement systems(Nussbaum et al., 2009, Zhu et al., 2009). Both driving cyclesrepresented relatively smooth operation with little hard accelera-tion and deceleration.

CO2 was measured with an LI-840 CO2/H2O Gas Analyzer (LicorBiosciences, Lincoln, NE; Watson et al., 2008b). Fuel-based EFs ingrams per kilogram of fuel (g/kg fuel) were estimated from PMmass and CO2 emissions assuming a diesel fuel carbon content of85.6%, following the approach of Moosmüller et al. (2003).Discounting the first MTVR sample (0.63 g/kg) as an outlier inTable 4, the average MTVR and LVS PM EFs were 0.18±0.03 and1.20±0.12 g/kg, respectively. The LVS emitted ~7 times more PMthan the MTVR. Higher LVS emissions are expected from its two-stroke engine.

Military vehicle fuel consumption data are needed to convertMTVR and LVS PM EFs (g/kg) into units of g/mile for cross-comparison among studies. However, fuel consumption was notmeasured during the 2007 SERDP Study. Fuel efficiency for the MTVRis rated at 3.8 miles per gallon (mpg) (http://www.oshkoshdefense.com/products/6/mtvr). Fuel efficiency for the baseline LVS is rated at2 mpg (http://www.globalsecurity.org/military/systems/ground/lvsr.htm). Applying a fuel efficiency of 2 mpg and assuming a dieselfuel density of 3.3 kg/gallon, the average equivalent MTVR and LVSPM EFs are 0.3 and 2 g/mile, respectively, similar to the average PM2.5

EF for HDDV (~1.8 g/mile) obtained during the GD-Split Study(Fig. 3). While there is considerable uncertainty in this comparison,the military vehicle PM EFs seem relatively low compared to those ofon-road HDDV.

2.2. Biomass burning emission factors

The Emission Estimation System Model (EES) (http://www.arb.ca.gov/ei/see/see.htm) estimates gaseous and PM emissions for wild-fires, prescribed burns, and wildland fires. The core of EES is the FirstOrder Fire Effects Model (FOFEM 4.0) that determines the fuel loadingcharacteristics for fuel components by vegetation type. EFs in EES arefunctions of fuel moisture (i.e., dry, moderate, and wet) and fuelcomponents, including: 1) litter; 2) small wood; 3) large wood; 4)

a

0

0.05

0.1

0.15

0.2

0.25

0.3

SI_

9_2

SI_

10_3

SI_

9_3

SI_

9_1

SI_

9_4

SI_

8_1

SI_

8_3

SI_

8_2

SI_

7_2

SI_

7_1

SI_

10_4

Vehicle Sample Composite

PM

2.5

Em

issi

on

Fac

tor

(g/m

ile)

MOBILE 6.2 Max

MOBILE 6.2 Min

EMFAC2007 Average

CS EF

WS EF

1977

1978

1979

1979

-198

0

1980

1983

-198

4

1984

1985

1987 19

86-1

989

1988

b

0

0.05

0.1

0.15

0.2

0.25

0.3

SI_

7_3

SI_

10_1

SI_

5_1

SI_

10_2

SI_

6_2

SI_

4_1

SI_

6_3

SI_

6_1

SI_

5_2

SI_

2_1

SI_

1_1

SI_

3_1

Vehicle Sample Composite

PM

2.5

Em

issi

on

Fac

tor

(g/m

ile)

MOBILE 6.2 Max

MOBILE 6.2 Min

EMFAC2007 Average

CS EF

WS EF

1989 19

89

1984

-199

5

1990

1990

-199

1

1991

-199

2

1992

1991

-199

5

1992

-199

5

1995

1995

-199

7

1995

-199

9

Fig. 4. Comparisons ofmeasured gasoline-fueled vehicle PM2.5 emission factors (EFs) under cold start (CS) andwarm start (WS) cycles during the Gas/Diesel Split StudywithMOBILE6.2 and EMFAC2007 estimates for model years: a) 1977–1988; and b) 1989–1999. See Table 1 for vehicle composite information. Composites are ordered by average vehicle modelyear. Model year(s) are shown above the bar in chronological order. Error bars indicate measurement uncertainties. The high-emitting vehicles are: SI_9_1 (model years 1979-1980,159,000-182,000 miles); SI_9_4 (model year 1980, 98,000 miles); SI_8_3 (model year 1984, 167,000 miles); SI_7_1 (model years 1986-1989, 92,000-418,000 miles); and SI_10_4(model year 1988, 149,000 miles).

2388 J.G. Watson et al. / Science of the Total Environment 409 (2011) 2384–2396

herb and shrub; 5) duff; and 6) canopy fuels. However, EFs from dryand wet fuels do not differ significantly for all fuel components. Aseparate set of EFs are used to estimate emissions from agriculturaland other management burns (http://www.arb.ca.gov/ei/see/mngdburnemissionfactors.xls), where EFs are provided for CO,VOCs, SO2, NOx, PM2.5, and PM10, but do not depend on fuel moisture,and are based on U.S. EPA (2006) and Jenkins et al. (1996). Hereafter,these EFs are noted as “San Joaquin Valley (SJV) EFs” because theyhave been used to develop the biomass burning emission inventoryfor California's SJV which is affected by different forms of biomassburning throughout the year (Chen et al., 2007b; Chow et al., 1992,2007b; Rinehart et al., 2006).

FLAME (Chakrabarty et al., 2006; Chen et al., 2006, 2007a;McMeeking et al., 2008; http://chem.atmos.colostate.edu/FLAME/)measurements were taken at the U.S. Forest Service Fire ScienceLaboratory (FSL) in Missoula, MT during November, 2003 (Pilot

Study), May, 2006 (Phase I), and June, 2007 (Phase II). Fresh fuelswere tested within one week of collection, while dried fuels wereprepared by long-term indoor storage. Before the experiment, fuelmoisture was determined by the weight difference prior to and afterheating the fuel to 80 °C for 24–48 h, and reported as the percentageof water with respect to dry fuel mass. Table 5 documents the 17biofuel types tested. Fuel-based EFs were quantified from in-plumePM2.5 CO and CO2 concentrations (Moosmüller et al., 2003). Carbonmass fractions for most biomes were ~0.4–0.5 (see carbon content inTable 5).

The FLAME PM2.5 EFs are classified into six EES categories,separated by moisture content (dry or wet), and compared with EESEFs in Fig. 8. The FLAME Study contains more detailed breakdowns offuel properties such as shrubs (e.g., chamise [Che] (Adenostoma sp.)and manzanita [Maz] (Arctostaphylos sp.)), leaves, and branches thandoes EES. Burning leaves producedmore particles per unit drymass of

Fig. 5. Comparison between modeled (MOBILE 6.2 and EMFAC2007) and measuredgasoline-fueled vehicle PM2.5 emission factors (EFs) during the Gas/Diesel Split Studyunder Federal Test Procedure (FTP) for: a) Warm Start (WS); and b) Cold Start (CS)cycles. Bubble diameter represents average vehicle age. Modeled and measured EFs fordiesel-fueled vehicle from Fig. 2 are plotted in the dashed circle for comparison.

2389J.G. Watson et al. / Science of the Total Environment 409 (2011) 2384–2396

fuel burned from both dry and wet (e.g., newly harvested or fresh)plants. The highest EFs resulted from burning fresh Ponderosa Pine(PP) (Pinus ponderosa) needles (76.8±37.5 g/kg), followed by freshmanzanita leaves (62.3 g/kg; single test) and fresh lodgepole pine(LP) (Pinus contorta) needles (56.2 g/kg, single test). PM2.5 EFs fromplant branch burning were b15 g/kg, even for fresh wood with amoisture content of ~70%. Considering the variability in burningdifferent fuels and fuel components, Fig. 8 shows that EES providesreasonable estimates for dry litter source (EES 1), dry small wood(EES 2), wet large wood (EES 3), dry herb and shrub (EES 4a (dry)),and dry duff burn (EES 5) EFs. However, EES underestimates PM2.5

EFs from wet herb and shrub (EES 4b (wet)), as well as from wetneedles from Ponderosa and lodgepole pine trees. Fresh pineneedles should not burn well during prescribed burns, whichfocus on fuels from the forest floor, but are vulnerable to wildfires. The PM2.5 EFs in EES should be updated with these recentmeasurements.

Fig. 9 shows that PM2.5 EFs varied from 2.38±1.38 g/kg (dryDambo grass) to 12.8 g/kg dry fuel (fresh tundra; single test). The EFsrepresent a wide variety of fuels with relatively low moisture content(b35%). As shown in Table 5, two pairs of samples were takenspecifically for the flaming and smoldering phases of dry chamise andrice straw burns. Each phase was distinguished visually according towhether flames were present. Smoldering combustion producedhigher EFs than did flaming conditions, as shown in Fig. 10 (3.2 vs.2.8 g/kg for chamise and 15.4 g/kg compared with 2.3 g/kg for ricestraw, respectively), consistent with the findings of Chen et al.

(2007a). The corresponding data from SJV overestimated EFs fromburning chamise (a species found in chaparral) but underestimatedEFs from burning rice straw and grass (grassland-type species) by afactor of three (Fig. 10). The effects of fuel moisture and combustionphase are being incorporated into emission models such as the FireEmission Production Simulator (FEPS; Anderson et al., 2004; http://www.fs.fed.us/pnw/fera/feps/).

3. Discussion

3.1. Comparison of measured and modeled emission factors

PM2.5 EFs for on-road gasoline and diesel vehicles and biomassburning are compared with modeled and/or empirical EFs fromCalifornia's PM2.5 emission inventory in Table 6. Comparisons are firstevaluated using the average of the absolute differences (AAD), i.e., theaverage of the percent differences between the non-default anddefault EFs divided by the default factors for specific source categories.For the 16 sets of on-road diesel-fueled vehicle EFs obtained underthe HCS cycle, the AAD, with respect to the EMFAC2007 model, is37.6%.

The Kansas City Study was conducted three to four years after theGD-Split Study. Gasoline-fueled vehicle EFs in the GD-Split Studywere classified into WS and CS cycles, while the Kansas City EFs wereclassified according to seasons (summer versus winter) based on thethree phases of the UDC (i.e., WS, CS, and Hot-Stabilized). The AAD forgasoline-fueled vehicle EFs ranged from 65.9% (Kansas City [winter])to 163.9% (Kansas City [summer]), two to four times the diesel vehicleAAD (37.6%) from the GD-Split Study.

Removing two extreme values (i.e., SI_9_4 and SI_10_4; see Fig. 4and footnotes in Table 1) from the GD-Split Study reduced the AADfrom 136.1% to 66.3%, and from 150.3% to 77.7% for the GD-Split WSand CS categories, respectively. Similarly, removing extreme values(i.e., S2_1, S2_2, W6_1, and W6_4; see Fig. 6 and footnotes in Table 2)from the Kansas City Study reduced the AAD from 163.9% to 107.1%and from 103.5% to 65.9% during summer and winter, respectively.Even with extreme values excluded, the AAD are much higher for thegasoline- than for the diesel-fueled vehicles. The best agreement withthe model for gasoline-fueled vehicles is found for the Kansas CityStudy during winter, which represents the more recent experimentaldata (February–March, 2005).

Both the California biomass burning EFs calculated from the EESmodel and the previous SJV measurements were based on fewermeasurements than were the on-road mobile source EFs, socomparisons of these with the FLAME biomass burning EFs may beless statistically significant. The largest AAD is 524.3% for wet freshcanopy fuels, followed by 245% for rice straw (Table 6). Combustionand fuel conditions, such as moisture content, were more variable forbiomass burning than for motor vehicle operating conditions.Emission models do not appear to accurately simulate the fullcomplexity of biomass burning.

Considering the large variability associated with biomass burning,modeled average EFs were compared with the measured average foreach subcategory. This was achieved by calculating the regressionslope of measured (y) against modeled (x) EFs with the interceptconstrained to zero. Both ordinary least-squares (OLS) regressionsand robust regressions (RR) are presented in Table 6. RR reduces theinfluence of extreme values using an iterative feedback algorithmdeveloped by Huber (2004). The standard error of the slope isreported for both cases.

The OLS and RR slopes of 103±9% and 87±7%, respectively, forthe GD-Split Study diesel EFs in Table 6 suggest that EMFAC2007overestimates diesel emissions for vehicles that are not high-emitters.The error is small (within 10%) in terms of category averages.EMFAC2007 also overestimates gasoline-fueled vehicle emissions, asindicated by RR slopes of less than 100%.When high-emitting gasoline

Table 2Descriptions of vehicle sample composites during the Kansas City Studya.

Sample composite code Vehicle test cycleb MOBILE model vehicle categoryc EMFAC model vehicle categoryd Number of vehicles tested Range of vehicle model year Odometer range (miles)

S5-5d FTP (Summer) LDGV LDA 1 1980 –

S6-1 FTP (Summer) LDGV LDA 1 1989 116 KS6-2 FTP (Summer) LDGV LDA 1 1989 209 KS6-3 FTP (Summer) LDGV LDA 1 1985 236 KS6-4 FTP (Summer) LDGV LDA 1 1986 36 KS7-1 FTP (Summer) LDGV LDA 2 1991–1994 169 K–214 KS7-2 FTP (Summer) LDGV LDA 3 1991–1994 32 K–185 KS7-3 FTP (Summer) LDGV LDA 1 1994 101 KS7-4 FTP (Summer) LDGV LDA 1 1991 226 KS8-1 FTP (Summer) LDGV LDA 5 1996–1998 45 K–131 KS8-2 FTP (Summer) LDGV LDA 5 1996–2000 40 K–148 KS8-3 FTP (Summer) LDGV LDA 5 1996–2003 24 K–146 KS2-1 FTP (Summer) LDGT1 LDT1 1 1989 161 KS2-2 FTP (Summer) LDGT1 LDT1 1 1985 30 KS2-3 FTP (Summer) LDGT1 LDT1 1 1989 132 KS2-4 FTP (Summer) LDGT1 LDT1 1 1985 47 KS3-1 FTP (Summer) LDGT1 LDT1 3 1995 74 K–113 KS3-2 FTP (Summer) LDGT1 LDT1 3 1990–1995 73 K–171 KS4-1 FTP (Summer) LDGT1 LDT1 3 1998–2003 19 K–131 KS4-2 FTP (Summer) LDGT1 LDT1 5 1999–2004 11 K–75 KW6-1 FTP (Winter) LDGV LDA 1 1988 207 KW6-2 FTP (Winter) LDGV LDA 1 1988 287 KW6-3 FTP (Winter) LDGV LDA 2 1989–1990 168 K–176 KW6-4 FTP (Winter) LDGV LDA 1 1989 62 KW7-1 FTP (Winter) LDGV LDA 2 1995 146 K–163 KW7-2 FTP (Winter) LDGV LDA 3 1991–1995 80 K–145 KW7-3 FTP (Winter) LDGV LDA 2 1994–1995 78 K–112 KW7-4 FTP (Winter) LDGV LDA 2 1993–1995 140 K–168 KW8-1 FTP (Winter) LDGV LDA 4 1996–2002 26 K–68 KW8-2 FTP (Winter) LDGV LDA 2 1997–1998 29 K–63 KW8-3 FTP (Winter) LDGV LDA 3 1998–2001 56 K–65 KW2-1 FTP (Winter) LDGT1 LDT1 1 1989 145 KW2-2 FTP (Winter) LDGT1 LDT1 1 1987 232 KW2-3 FTP (Winter) LDGT1 LDT1 1 1988 162 KW3-1 FTP (Winter) LDGT1 LDT1 3 1992–1995 85 K–136 KW3-2 FTP (Winter) LDGT1 LDT1 3 1993–1995 47 K–113 KW3-3 FTP (Winter) LDGT1 LDT1 1 1992 154 KW4-1 FTP (Winter) LDGT1 LDT1 5 1996–2004 14 K–66 KW4-2 FTP (Winter) LDGT1 LDT1 3 1998–2002 0 K–56 KW4-3 FTP (Winter) LDGT1 LDT1 3 1996–1997 125 K–146 K

a U.S. EPA, 2008b, 2008c.b FTP: Federal Test Procedure, includes the average of three cycles (Cold Start, Warm Start, and Hot Start); tests conducted during summer of 2004 and winter of 2005.c MOBILE model category: LDGV (Light-Duty Gasoline Vehicle); LDGT1 (Light-Duty Gasoline Truck; weight class 1 [0–3000 lbs]).d EMFAC model category: LDA (Light-Duty Passenger Vehicle); LDT1 (Light-Duty Truck; weight class 1 [0–5750 lbs]).

2390J.G

.Watson

etal./

Scienceof

theTotal

Environment

409(2011)

2384–2396

a

0

0.05

0.1

0.15

S5-

5

S6-

3

S6-

4

S6-

1

S6-

2

S7-

4

S7-

1

S7-

2

S7-

3

S8-

1

S8-

2

S8-

3

S2-

2

S2-

4

S2-

1

S2-

3

S3-

1

S3-

2

S4-

1

S4-

2

Vehicle Sample Composite

PM

2.5

Em

issi

on

Fac

tor

(g/m

ile)

PM

2.5

Em

issi

on

Fac

tor

(g/m

ile)

MOBILE 6.2 Max MOBILE 6.2 Min

EMFAC2007 Average FTP Emission Factor

Passenger Vehicles Trucks

1980

1985

1986

1989 19

8919

91

1991

1991

-199

419

94

1996

-199

8

1996

-200

0

1996

-200

3

1985

1985

1989

1989

1995 19

90-1

995

1989

-200

3

1999

-200

4

b

0

0.05

0.1

0.15

W6-

1

W6-

2

W6-

4

W6-

3

W7-

1

W7-

2

W7-

3

W7-

4

W8-

2

W8-

3

W8-

1

W2-

2

W2-

3

W2-

1

W3-

3

W3-

1

W3-

2

W4-

3

W4-

2

W4-

1

Vehicle Sample Composite

MOBILE 6.2 Max MOBILE 6.2 Min

EMFAC2007 Average FTP Emission Factor

Passenger Vehicles Trucks

1988

1988

1989

1989

-199

019

95 1991

-199

5

1994

-199

5

1993

-199

5

1997

-199

8

1998

-200

1

1996

-200

2

1987

1988

1989

1992 19

92-1

995

1993

-199

5

1996

-199

7

1998

-200

2

1996

-200

4

Fig. 6. Comparisons of measured gasoline-fueled vehicle PM2.5 emission factors (EFs) from the Kansas City Study with MOBILE 6.2 and EMFAC2007 estimates for: a) summer 2004and b) winter 2005. See Table 2 for vehicle composite information. Composites are ordered by average vehicle model year. Model year(s) are shown above the bars in chronologicalorder. Error bars indicate measurement uncertainties.

2391J.G. Watson et al. / Science of the Total Environment 409 (2011) 2384–2396

vehicles are not weighted less in the regression analysis, themeasured EFs equal or exceed EMFAC2007 estimates (i.e., OLS slopesare ≥100%). The only exception is for GD-Split WS vehicles (with anOLS slope of 80±40%), in which EFs were smaller than EMFAC2007

Table 3Summary of military diesel-fueled vehicles tested during the April 2007 Strategic Environm

Vehicle typea/IDnumber

Vehiclenumberb

Engine specification Gross vehicleweight (lbs)

Veye

MTVR/593901 1,2 Caterpillar C-12 (four stroke) 62,200 20MTVR/592995 1,2 Caterpillar C-12 (four stroke) 62,200 20LVS/550978 12 Detroit Diesel 8V92TA (two stroke) 32,000 20LVS/550978 12 Detroit Diesel 8V92TA (two stroke) 32,000 20

a The MTVR (Medium Tactical Vehicle Replacement) is a six-wheel drive, all-terrain vehicLVS (Logistics Vehicle System) is an eight-wheel drive all-terrain vehicle powered by a Det

b Vehicles 1 and 2 were Medium Tactical Vehicle Replacement vehicles (MTVR); Vehicletested to measure total PM (no size cut) EFs under loop and extended driving conditions u

c Loop Driving Cycle: An 1.8 km round-trip loop with driving time of 100–130 s over paved(2008b). This test loop has an approximately uniform slope resulting in equal time on 1.1cruising and less frequent acceleration and deceleration.

estimates. EMFAC2007 estimates are based on the FTP cycle whichincludes CS conditions. This study shows that gasoline-fueled vehiclesunder WS conditions emit only 36±11% (RR slope) of theEMFAC2007 estimates for the category average.

ental Research and Development Program (SERDP) Study.

hicle modelar/month

Odometerrange (miles)

Vehicle testcyclec

Number ofvehicles tested

Total numberof tests

02/10 6637 Loop driving 2 802/4 5306 Extended driving 2 206/8 377 Loop driving 1 306/8 377 Extended driving 1 1

le utilizing the Caterpillar C-12 turbo-charged, four-stroke, 12 l, 6-cylinder engine. Theroit Diesel 8V92TA turbo-charged, two-stroke, 12 l, 8-cylinder engine.12 was a Logistics Vehicle System vehicle (LVS; Vehicle 12); these three vehicles weresing in-plume measurement systems.and concrete surfaces was followed in each of the tests as documented inWatson et al.° uphill and downhill slopes. Extended Driving Cycle (10–21 min): More high speed

Fig. 7. Comparison between modeled and measured gasoline-fueled vehicle PM2.5

emission factors (EFs) during the Kansas City Study under the Unified Driving Cycle(UDC) with MOBILE 6.2 and EMFAC2007 model estimates for: a) summer 2004; and b)winter 2005. Bubble diameter represents average vehicle age from lowest (0=modelyear 2004) to largest (24=model year 1980).

Table 4PMa Emission Factors (EF) for military vehicles measured via on-board sampling for theStrategic Environmental Research and Development Program (SERDP) Study duringApril 2007b.

Vehicle Typec Vehicle Number Vehicle Test Cycled g/kg Fuele

PM EF

MTVR 1 Loop driving 0.63 f

MTVR 1 Loop driving 0.20MTVR 1 Loop driving 0.23MTVR 1 Loop driving 0.17MTVR 1 Loop driving 0.20MTVR 1 Extended driving 0.19MTVR 2 Loop driving 0.19MTVR 2 Loop driving 0.15MTVR 2 Loop driving 0.16MTVR 2 Extended driving 0.16LVS 12 Loop driving 1.38LVS 12 Loop driving 1.11LVS 12 Loop driving 1.18LVS 12 Extended driving 1.13

a Total PM (no size cut).b Watson et al. (2008b).c MTVR = Medium Tactical Vehicle Replacement. LVS = Logistics Vehicles System.d See Table 3 for description of test cycle.e Assumed a carbon content of 85.6% for diesel fuel.f Outlier, excluded from average.

2392 J.G. Watson et al. / Science of the Total Environment 409 (2011) 2384–2396

For biomass burning, the RR slopes indicate that the modeledEES and SJV EFs were lower than measured EFs for dry litter (229±91%), wet herb and shrub (253±129%), wet canopy (624±97%),rice straw (339±152%), and grass (222±99%; Table 6) fuels. Theresults are not statistically significant because the standard errorsof the RR slopes are large compared to those associated with mobilesource emissions. However, measured and model-estimated SJV EFsfor dry herb, dry duff, and wet wood burning are similar, with RRslopes of 81–97%. For biomass burning, the OLS and RR slopes aresimilar, but are based on a limited number of data points (N=2 to5 data pairs).

4. Conclusions

PM emission factors (EFs) for gasoline- and diesel-fueled vehiclesand biomass combustion measured in several recent studies arereported and compared with modeled and previously used values.The 2001 Gas/Diesel Split Study (GD-Split) measured PM2.5 emis-sions on light-duty gasoline and heavy-duty diesel vehicles usingmobile dynamometers. PM2.5 EFs were grouped according to vehiclemodel year. Measured PM2.5 EFs for diesel-fueled vehicles increasedwith vehicle age from ~0.2 to over 2 g/mile. The correlationsbetween measured and modeled PM2.5 EFs were 0.79 and 0.89 forthe U.S. EPA MOBILE 6.2 and ARB EMFAC2007 models, respectively.PM2.5 EFs for diesel vehicles operated under suburban driving cycles

were nearly double those of vehicles operated under highwayconditions.

Measured PM2.5 EFs for gasoline vehicles were considerablylower than PM2.5 EFs for diesel vehicles. Gasoline vehicle PM2.5 EFswere b0.05 g/mile for vehicle model year later than 1988. PM2.5 EFswere somewhat higher (~0.1 g/mile) and more variable for vehiclemodel years between 1977 and 1988. Neither MOBILE 6.2 norEMFAC2007 captured the variability in measured PM2.5 EFs forgasoline vehicles. This may reflect the lack of data on vehiclemaintenance, which is available to be used in EMFAC2007, but notMOBILE 6.2, modeling.

In the Kansas City Study, PM2.5 EFs were measured on 100gasoline-fueled passenger vehicles and light trucks during summer2004 and winter 2005. Measured PM2.5 EFs generally increasedwith vehicle age but, aside from several high emitters, most EFsmeasured were less than 0.03 g/mile. Measured EFs were alsohigher in winter than in summer. EMFAC2007 estimated signifi-cantly higher EFs than MOBILE 6.2 during winter but not duringsummer. As in the GD-Split Study, neither MOBILE 6.2 norEMFAC2007 completely simulated the range of measured andmodeled EFs.

The PM EFs for non-road-related activities are not well-repre-sented in real-worldmeasurements or in emissionmodels and are noteasily translated to the g/mile units used for on-road activities. TheSERDP non-road diesel emissions correspond to values of 0.3 and 2 g/mile, assuming a fuel consumption efficiency of 2 mpg and a fueldensity of 3.3 kg/gal, and are comparable in magnitude to the averageheavy-duty diesel PM2.5 EF of ~1.8 g/mile measured during the GD-Split Study.

The highest biomass burning PM2.5 EFs (76.8±37.5 g/kg) weremeasured for wet (N50% moisture content) Ponderosa Pine needles.When moisture content was less than 20% EFs were b20 g/kg. EESmodel EFs agreed well with measured EFs for fuels with lowmoisturecontent but underestimated measured EFs when moisture contentwas higher than 40%. Rice straw burning in the smoldering phaseproduced a much higher PM2.5 EF (15.4 g/kg) than when burning inthe flaming phase (2.3 g/kg). Average EFs for dry chamise, rice straw,and dry grass were within a factor of three of the values adopted byARB in the San Joaquin Valley (SJV).

The discrepancies between measured and modeled PM2.5 EFs forgasoline vehicles and burning of biomass with high moisture content

Table 5Summary of biomass burned during the Fire Laboratory at Missoula Experiment (FLAME) Study.

Biofuel types Biofuel components Number of samples Fuel moisture (%) C(%)/N(%)a Phase

Ceanothus 1 24.3 48/1.3Chamise Branches and leaves, dried 5 11.8–19.4 49/1 Separate burns for flaming

and smolderingBranches, dried 2 30.7–35.0Branches, fresh 3 23.9–50.0Leaves, dried 3 8.7–19.6Leaves, fresh 2 52.2–60.4

Dambo grass Leaves, dried 2 6.3 49/0.5 Mixed-phaseExcelsior(Shredded AspenWood Product)

Dried 2 5.9 48/0.07 Mixed-phase

Lignin 1 17 – Mixed-phaseLodgepolePine

Branches, dried 3 9.0–9.3 42–50/0.3–1.2 Mixed-phaseNeedles, fresh 1 76.4–90.6Needles, litter 2 13.7–15.6Needles, duff 2 20.1–24.1

Manzanita Branches, fresh 3 62.9–70.5 48/0.8 Mixed-phaseLeaves, dried 2 52.5–59.8Leaves, fresh 1 75.4–107.0

Montana Grass Leaves, dried 1 5.0–13.0 44/0.17 Mixed-phaseLeaves, fresh 5 17.5–94.0

Palmetto Leaves, fresh 3 5.0–7.1 51/1.0 Mixed-phasePonderosa Pine Branches (Large), dried 3 9.1–9.3 46–49/0.04–1.3 Mixed-phase

Branches (Large), fresh 3 63.0–72.0Branches (Small), dried 2 9.0–9.6Branches (Small), fresh 3 43.4–50.5Needles, dried 2 7.3Needles, fresh 3 57.5–60.7Needles, litter 13 9.2–10.5Needles and branche litter 1 10.3Needles, duff 4 13.9–14.7

Puerto Rico fern 1 12.8 46/0.4 Mixed-phaseRice Straw 5 8.1–10.1 39–46/0.6–0.9 Separate burns for flaming

and smolderingSagebrush Branches and leaves, dried 2 8.3 47–51/1.5–2.1 Mixed-phase

Foliage and sticks 1 9.1Tundra Core, fresh 1 113 31/0.5 Mixed-phaseJuniper Foliage and sticks, fresh 1 8.7 49–0.9 Mixed-phaseWax Myrtle Branches and foliage 1 13.6 48–53/1.1–1.4 Mixed-phaseWhite Pine Needles, dried 2 8.2 49–0.5 Mixed-phase

a Fuel carbon (C) and nitrogen (N) content, with respect to dry fuel mass.

2393J.G. Watson et al. / Science of the Total Environment 409 (2011) 2384–2396

implies that there may be large uncertainties in current PM2.5

emission inventories when there are substantial contributions fromthese sources. The results presented here provide guidance for placingbounds on PM2.5 emission estimates. To achieve more realisticemission estimates, emission models such as the MOtor VehicleEmission Simulator (MOVES) and the Fire Emission ProductionSimulator (FEPS) should incorporate information from this andother recent source measurement studies.

Acknowledgements

This work was primarily supported by the California Air ResourcesBoard (ARB) under contract 04–307. The statements and conclusionsin this paper are those of the contractor and not necessarily those ofARB. Other sources of support include the Department of EnergyOffice of Heavy Vehicles Technologies and FreedomCAR VehiclesTechnologies through the National Renewable Energy Laboratory(NREL) for the Gas/Diesel Split Study, U.S. EPA's Kansas City Studyunder Contract #GS 10F-0036K, the U.S. DOD's Strategic Environ-mental Research and Development Program ProjectWP-1336, and theJoint Fire Science Project through NPS Task J8R07060005. Themention of commercial products, their source, or their use inconnection with material reported herein is not to be construed asactual or implied endorsement of such products. Additional supportwas provided by the EPA STAR Grant RD-83108601-0. The authors

wish to thank Ms. Jo Gerrard of DRI for assisting with manuscriptediting and preparation.

References

Anderson G, Sandberg D, Norheim R. Fire emission production simulator (FEPS) user'sguide. 2004. U.S. Forest Service.

Bachmann JD. Will the circle be unbroken: a history of the US national ambient airquality standards — 2007 Critical Review. J Air Waste Manage Assoc 2007;57:652–97.

Bond TC, Sun HL. Can reducing black carbon emissions counteract global warming?Environ Sci Technol 2005;39:5921–6.

CARB. EMFAC2007, version 2.3. Calculating emission inventories for vehicles inCalifornia. Sacramento, CA: California Air Resources Board; 2007.

CARB. ARB's emissions inventory. Sacramento, CA: California Air Resources Board;2009a.

CARB. Speciation profiles used in ARB modeling. Sacramento, CA: California AirResources Board; 2009b.

Chakrabarty RK, Moosmüller H, Garro MA, Arnott WP, Walker J, Susott RA, et al.Emissions from the laboratory combustion of wildland fuels: particle morphologyand size. J Geophys Res -Atmospheres 2006;111:D07204.

Chen L-WA, Moosmüller H, Arnott WP, Chow JC, Watson JG, Susott RA, et al. Particleemissions from laboratory combustion of wildland fuels: in situ optical and massmeasurements. Geophys Res Lett 2006;33:1–4.

Chen L-WA, Moosmüller H, Arnott WP, Chow JC, Watson JG, Susott RA, et al. Emissionsfrom laboratory combustion of wildland fuels: emission factors and source profiles.Environ Sci Technol 2007a;41:4317–25.

Chen L-WA, Watson JG, Chow JC, Magliano KL. Quantifying PM2.5 source contributionsfor the San Joaquin Valley with multivariate receptor models. Environ Sci Technol2007b;41(8):2818–26.

Chow JC. 2001 Critical review discussion — diesel engines: environmental impact andcontrol. J Air Waste Manage Assoc 2001;51:1258–70.

Fig. 8. Comparison of FLAME PM2.5 emission factors (EFs) with those in the California Air Resource Board's Emission Estimation System (EES). FLAME EFs are classified into the sixEES categories and separated by dry andwet moisture content indicated by the boxes in the figure. EES EFs are shown by the black bar. The EES Categories are: 1) EES 1: Dry Litter; 2)EES 2: Small Dry Wood; 3) EES 3: Large Wet Wood; 4) EES 4a: Dry Herb and Shrub; 5) EES 4b: Wet Herb and Shrub; 6) EES 5: Dry Duff; and 7) EES 6 Wet Canopy Fuels. The biofueltypes are: 1) Che: Chamise; 2) Maz: Manzanita; 3) MTg: Montana Grass; 4) PP: Ponderosa Pine; and 5) LP: Lodgepole Pine.

2394 J.G. Watson et al. / Science of the Total Environment 409 (2011) 2384–2396

Chow JC, Watson JG, Lowenthal DH, Solomon PA, Magliano KL, Ziman SD, et al.PM10 source apportionment in California's San Joaquin Valley. Atmos Environ1992;26A(18):3335–54.

Chow JC, Bachmann JD, Wierman SSG, Mathai CV, Malm WC, White WH, et al. 2002Critical review discussion — visibility: science and regulation. J Air Waste ManageAssoc 2002;52:973–99.

Chow JC, Watson JG, Chen L-WA, Ho SSH, Koracin D, Zielinska B, et al. Exposure toPM2.5 and PAHs from the Tong Liang, China, epidemiological study. J Env Sci HealthPart A 2006;A41:517–42.

Chow JC, Watson JG, Feldman HJ, Nolan J, Wallerstein BR, Bachmann JD. Critical reviewdiscussion — will the circle be unbroken: a history of the U.S. National Ambient AirQuality Standards. J Air Waste Manage Assoc 2007a;57:1151–63.

Chow JC, Watson JG, Lowenthal DH, Chen L-WA, Zielinska B, Mazzoleni LR, et al.Evaluation of organic markers for chemical mass balance source apportionment atthe Fresno supersite. Atmos. Chem. Phys. 2007b;7(7):1741–2754.

Chow JC, Watson JG, Lowenthal DH, Chen L-WA, Motallebi N. Black and organic carbonemission inventories: review and application to California. J Air Waste ManageAssoc 2010;60:497–507.

Cook R, Touma JS, Fernandez A, Brzezinski D, Bailey C, Scarbro C, et al. Impact ofunderestimating the effects of cold temperature onmotor vehicle start emissions ofair toxics in the United States. J Air Waste Manage Assoc 2007;57:1469–79.

Fig. 9. FLAME PM2.5 emission factors (EFs) and corresponding fuel moisture (fuelmoisture of fresh tundra cores was not determined).

DieselNet. Emission test cycles: summary of worldwide engine and vehicle test cycles.Mississauga, ON, Canada: Ecopoint, Inc; 2008.

Environment Canada. Air pollutant emissions; 2008.EPD. Hong Kong air pollutant emission inventory. 2008. Hong Kong, China SAR, Hong

Kong Environmental Protection Dept.Fujita EM, Campbell DE, ArnottWP, Chow JC, Zielinska B. Evaluations of the chemicalmass

balance method for determining contributions of gasoline and diesel exhaust toambient carbonaceous aerosols. J Air Waste Manage Assoc 2007a;57:721–40.

Fujita EM, Zielinska B, Campbell DE, Arnott WP, Sagebiel JC, Mazzoleni LR, et al.Variations in speciated emissions from spark–ignition and compression–ignitionmotor vehicles in California's south coast air basin. J Air Waste Manage Assoc2007b;57:705–20.

Gertler AW. Diesel vs. gasoline emissions: does PM from diesel or gasoline vehiclesdominate in the US? Atmos Environ 2005;39:2349–55.

Grantz DA, Garner JHB, Johnson DW. Ecological effects of particulate matter. Environ Int2003;29:213–39.

Huber PJ. Robust statistics. Hoboken, NJ: John Wiley & Sons; 2004.Jacobson MZ. Control of fossil-fuel particulate black carbon plus organic matter,

possibly the most effective method of slowing global warming. J Geophys Res2002;107:ACH 16-1–ACH 16–22.

Jenkins BM, Turn SQ, Williams RB, Goronea M, Abd-el-Fattah H. Atmospheric pollutantemission factors from open burning of agricultural and forest biomass by windtunnel simulations. Volume 1: Final Report. PB–97-133037/XAB. 1996. Universityof California, Davis, Davis, CA.

KishanS, BurnetteAD, Fincher.S., SabischMA, CrewsW, SnowR, ZmudM,SantosRC,BrickaS, Fujita EM, Campbell DE, Arnott WP. Kansas City PM characterization study finalreport. ERG No. 0133.18.007.001. 2006. Austin, TX, Eastern Research Group, Inc.

Lighty JS, Veranth JM, Sarofim AF. Critical review: combustion aerosols: factorsgoverning their size and composition and implications to human health. J AirWasteManage Assoc 2000;50:1565–618.

Lloyd AC, Cackette TA. Critical review — diesel engines: environmental impact andcontrol. J Air Waste Manage Assoc 2001;51:809–47.

MacCracken MC. A summary of the 2008 critical review prospects for future climatechange and the reasons for early action. EM; 2008a. p. 40–5.

MacCracken MC. Critical review: prospects for future climate change and the reasonsfor early action. J Air Waste Manage Assoc 2008b;58:735–86.

Mauderly JL, Chow JC. Health effects of organic aerosols. Inhal Toxicol 2008;20:257–88.McCarthyMC, Eisinger DS, Hafner HR, Chinkin LR, Roberts PT, Black KN, et al. Particulate

matter: a strategic vision for transportation-related research. Environ Sci Technol2006;40:5593–9.

McMeeking GR, Kreidenweis SM, Baker S, Carrico CM, Chow JC, Collett Jr JL, Hao WM,Holden AS, Kirchstetter TW, Malm WC, Moosmüller H, Sullivan AP, Wold CE.Emissions of Trace Gases and Aerosols during the Open Combustion of Biomass inthe Laboratory. J Geophys Res 2009;(114):D19210. doi:10.1029/2009JD011836.

Mobley JD, Deslauriers M, Rojas-Brachos L. Improving emission inventories for effectiveair-quality management across North America — A NARSTO assessment. NARSTO-05-001. Pasco, WA: NARSTO; 2005.

2.8 3.2 3.0

8.5

2.3

15.4

7.9

2.5

8.2

21.5

6.7

0

5

10

15

20

25

30

35

40

Dry

Cha

mis

e (F

lam

ing)

Dry

Cha

mis

e (S

mol

derin

g)

Dry

Cha

mis

e (O

vera

ll)

Cha

parr

al (

SJV

)

Ric

e S

traw

(F

lam

ing)

Ric

e S

traw

(S

mol

derin

g)

Ric

e S

traw

(O

vera

ll)

Ric

e S

traw

(S

JV)

Dry

Mot

ana

Gra

ss

Fre

sh M

otan

a G

rass

Gra

ssla

nd (

SJV

)

PM

2.5

Em

issi

on

Fac

tor

(g/k

g D

ry F

uel

)

Type of Biofuel and Burn Phase

FLAME Study

SJV Study

Fig. 10. FLAME PM2.5 emission factors (EFs) (shown above each bar in g/kg dry fuel) from different burning phases measured during the FLAME Study and comparisons withemission factors from California's San Joaquin Valley (SJV).

Table 6Comparability of measured (y) and modeled (x) PM2.5 emission factors.

Study and source category for measuredemission factors (y)

Number ofdata pairs

Average AbsoluteDifference (AAD)%a

100×Ordinary Least Squares (OLS)Regression Slope±σ (%)b

100×Robust RegressionSlope (RR)±σ (%)c

Gas/Diesel Split Diesel (Hot City-Suburban route [HCS] mode) 16 37.6 103±9 87.2±7.2Gas/Diesel Split Gasoline (Warm Start [WS]) 23 136.1 80.1±39.7 36.2±10.5Gas/Diesel Split Gasoline (Warm Start [WS]) 21d 66.3Gas/Diesel Split Gasoline (Cold Start [CS]) 23 150.3 128±49 80.1±20.3Gas/Diesel Split Gasoline (Cold Start [CS]) 21e 77.7Kansas City Gasoline (FTP Summer) 19 163.9 111±34 86.8±25.6Kansas City Gasoline (FTP Summer) 17f 107.1Kansas City Gasoline (FTP Winter) 20 103.5 92.9±24.7 72.7±16.5Kansas City Gasoline (FTP Winter) 18g 65.9FLAME Dry Herb, Shrub 5 149 185±114 97.1±67.1FLAME Dry Litter 4 151 229±91 229±91FLAME Dry Wood 3 33.3 66.7±17.0 63.7±24.6FLAME Dry Duff 2 18.9 81.1±13.9 81.1±13.9FLAME Wet Herb, Shrub 5 186 255±100 253±129FLAME Wet Canopy Fuels 2 524.3 624±97 624±97FLAME Wet Wood 2 16.8 83.2±10.0 83.2±10.0FLAME Chaparral (Chamise) 3h 64.6 35.4±1.5 35.4±1.5FLAME Rice Straw 3h 245 339±152 339±152FLAME Grass 2i 122 222±99 222±99.4

a AAD=(100 × jY−Xj=X), where Y and X are themeasured andmodeled (2006 California Emission Inventory) emission factors, respectively. Emission factors for mobile sourceswere estimated with EMFAC2007. Emission factors for biomass burning were based on the Emission Estimation System (EES) model except for chaparral, rice straw and grass, whichwere based on SJV emission factors.

b Based on ordinary least square regression with zero intercept; σ is the standard error of the slope.c Based on robust regression with zero intercept to reduce the influence of outliers; σ is the standard error of the slope.d Two outliers, SI_9_4 and SI_10_4 for warm start, were removed from the comparison (see Fig. 4).e Two outliers, SI_9_4 and SI_10_4 for cold start, were removed from the comparison (see Fig. 4).f Two outliers, S2_1 and S2_4, were removed from the comparison (see Fig. 6).g Two outliers, W6_1 and W6_4, were removed from the comparison (see Fig. 6).h Emission factors represent both flaming and smoldering phases.i Emission factors represent dry and wet fuels.

2395J.G. Watson et al. / Science of the Total Environment 409 (2011) 2384–2396

Moosmüller H, Mazzoleni C, Barber PW, Kuhns HD, Keislar RE, Watson JG. On-roadmeasurement of automotive particle emissions by ultraviolet lidar and transmis-someter: instrument. Environ Sci Technol 2003;37:4971–8.

Nam EK, Fulper CR, Warila J, Somers JH, Michaels H, Baldauf RW, et al. Analysis ofparticulate matter emissions from light-duty gasoline vehicles in Kansas City.EPA420-R-08-010. Ann Arbor, MI: U.S. Environmental Protection Agency;2008.

Nussbaum NJ, Zhu D, Kuhns HD, Mazzoleni C, Chang M-CO, Moosmüller H, et al. The in-plume emissions test-stand: a novel instrument platform for the real-timecharacterization of combustion emissions. J Air Waste Manage Assoc 2009;59(12):1437–45.

Pope III CA, Dockery DW. Critical review: health effects of fine particulate air pollution:lines that connect. J Air Waste Manage Assoc 2006;56:709–42.

Rakha H, Ahn K, Trani A. Development of VT-micro model for estimating hot stabilizedlight duty vehicle and truck emissions; 2003. Washington, D.C.

Rinehart LR, Fujita EM, Chow JC, Magliano KL, Zielinska B. Spatial distribution of PM2.5

associated organic compounds in central California. Atmos Environ 2006;40(2):290–303.

Sabbioni C, Brimblecombe P. The effects of air pollution on the built environment.London: Imperial College Press; 2003.

Sabbioni C, Ghedini N, Bonazza A. Organic anions in damage layers on monuments andbuildings. Atmos Environ 2003;37:1261–9.

U.S. EPA. AP-42, Volume I: compilation of air pollution emission factors. Fifth Edition.Washington, D.C: U.S. Environmental Protection Agency; 2006.

U.S. EPA. National Emissions Inventory (NEI) air pollutant emissions trends data. 2008a.Research Triangle Park, NC, U.S. Environmental Protection Agency.

2396 J.G. Watson et al. / Science of the Total Environment 409 (2011) 2384–2396

U.S. EPA. Kansas City PM Characterization Study: Final Report, EPA420-R-08-009. EPAContractNo. GS10F-0036K. 2008b. AnnArbor,MI, Assessment and StandardsDivisionOffice of Transportation and Air Quality U.S. Environmental Protection Agency.

U.S. EPA. Analysis of Particulate Matter Emissions from Light-Duty Gasoline Vehicles inKansas City. EPA420-R-08-010. 2008c. Research Triangle Park, NC, U.S. Environ-mental Protection Agency.

U.S. EPA. MOBILE6. 2008d. Research Triangle Park, NC, U.S. Environmental ProtectionAgency.

U.S. EPA. SPECIATE Version 4.0. 2007. Research Triangle Park, NC, U.S. EnvironmentalProtection Agency.

Watson JG. Overview of receptor model principles. J Air Poll Control Assoc 1984;34:619–23.

Watson JG. Visibility: science and regulation. J Air Waste Manage Assoc 2002;52:628–713.

Watson JG, Chen L-WA, Chow JC, Lowenthal DH, Doraiswamy P. Source apportionment:findings from the U.S. supersite program. J Air Waste Manage Assoc 2008a;58:265–88.

Watson JG, Miller JW, Moosmüller H, Kuhns HD, Chow JC, Chang OMC, Nussbaum NJ,Mazzoleni C, Zhu D, Barber PW, KemmeMR, Cocker DR. Characterization of off-roaddiesel emissions of criteria pollutants. WP-1336. 2008b. Reno, NV, Desert ResearchInstitute, Reno, NV and CE-CERT, University of California, Riverside, CA.

Zhu DZ, Nussbaum NJ, Kuhns HD, Chang M-CO, Sodeman D, Uppapalli S, et al. In-plumeemission test stand 2: emission factors for 10-to 100-kW US military generators.J Air Waste Manage Assoc 2009;59(12):1446–57.

Zhu DZ, Nusssbaum NJ, Kuhns HD, Chang M-CO, Sodeman DA, Moosmüller H, et al.Real-world PM, NOx, CO, and ultrafine particle emission factors for military non-road heavy duty diesel vehicles. Atmos Environ 2011;45:2603–9.