Deciphering the origins and transformations of atmospheric organics:

CalNex2010 Bakersfield

A. Goldstein, D. Gentner, G. Isaacman, D. Worton,

Y. Zhao, R. Weber, R. Sellon, A. Guha (UC Berkeley)

N. Kreisberg, S. Hering (Aerosol Dynamics Inc.)

B. Williams (Washington University, St. Louis)

T. Hohaus, A. Lambe, J. Jayne, L. Williams,

D. Worsnop (Aerodyne Research Inc)

J-L. Jimenez (University of Colorado)

L. Russell, S. Liu, D. Day (UC San Diego)

CalNex Bakersfield Science Team

CalNex Pasadena Science Team

SPECIAL THANKS TO:

John Karlik, Rick Ramirez

UC Cooperative Extension Kern County Staff

R. Cohen, S. Pusede UC Berkeley (PI and site organization)

Funding: Contracts #08-316 and #09-316

Bakersfield

Fig from R. Cohen

PM2.5 Nonattainment Areas in U.S.

Worst Cities (PM2.5)

1) Bakersfield, CA

2) Fresno, CA

3) Hanford, CA

4) Los Angeles, CA

5) Modesto, CA

6) Pittsburgh, PA

7) Salt Lake City, UT

8) Logan, UT

9) Fairbanks, AK

10) Merced, CA

U.S. EPA & American Lung Assoc. (2012)

Ozone Nonattainment Areas in U.S. Worst Cities (O3)

1) Los Angeles, CA

2) Visalia, CA

3) Bakersfield, CA

4) Fresno, CA

5) Hanford, CA

6) Sacramento, CA

7) San Diego, CA

8) Houston, TX

9) San Luis

Obispo, CA

10) Merced, CA

U.S. EPA & American Lung Assoc. (2012)

1980 1985 1990 1995 2000 2005 20100

50

100

150

200

250

Year

Days E

xceedin

g U

S 8

hr

O3 S

tandard

South Coast (LA)

San Joaquin Valley

SF Bay Area

Air Quality Control Has Been More Effective

in LA than San Joaquin Valley D

ays

Exc

eed

ing

US

8-h

r O

3 S

tan

dar

d

Year

Data from www.arb.ca.gov

CalNex2010 was designed to investigate why.

Differences in ozone and particle precursors?

Differences in chemistry?

NOx 2010 CARB Emission Inventory

Anthropogenic Emission Projections (%) (Bakersfield is in Kern County)

Kern County Los Angeles County

DIESEL TRUCKS, 50

OFF-ROAD EQUIPMENT,

12

MINING- CEMENT

MANUFACT., 8

PASSENGER CARS, 5

OIL & GAS PRODUCTION,

5

TRAINS, 5

OTHER, 16

DIESEL TRUCKS, 25

OFF-ROAD EQUIPMENT, 22

PASSENGER CARS, 17

TRAINS, 2

SHIPS AND COMMERCIAL

BOATS, 12

OTHER, 21

Data from www.arb.ca.gov

OIL AND GAS PRODUCTION,

34

PASSENGER CARS, 9 DIESEL

TRUCKS, 7

PESTICIDES, 7

CONSUMER PRODUCTS, 5

LIVESTOCK, 4

OFF-ROAD EQUIPMENT, 4

OTHER, 30

PASSENGER CARS, 25

DIESEL TRUCKS, 3

CONSUMER PRODUCTS, 19

OFF-ROAD EQUIPMENT,

12

OTHER, 42

Reactive Organic Gas (ROG)

2010 CARB Emission Inventory

Anthropogenic Emission Projections (%)

EXTREMELY DIFFERENT

Kern County Los Angeles County

Data from www.arb.ca.gov

Organic Aerosol

Global Scale

OVOC

Oxidation

by OH, O3, NO3

Direct

Emission

Terpenes

Nucleation or Condensation

Hydrocarbons

OC

Isoprene

Cloud

Processing

FF: ~5 TgC/yr

BB: ~11 TgC/yr

SOA: ~140 TgC/yr

Fossil Fuel Biomass

Burning

Heterogeneous Reactions

Oligomerization

Hallquist et al 2009

(Goldstein & Galbally 2007

Hallquist et al 2009

Heald et al 2010 )

MOSTLY SOA

DIESEL TRUCKS, 12

CEMENT MANUFACT, 8

FARM DUST, 7

RES. FUEL COMBUSTION, 7

WIND DUST, 6

ROAD DUST, 11

PASSENGER CARS, 2

OTHER, 46

Primary PM2.5

2010 CARB Emission Inventory

Anthropogenic Emission Projections (%)

Kern County Los Angeles County

Data from www.arb.ca.gov

Primary versus Secondary PM

SOAR 2005 Aerosol in Riverside:

Organic is dominant component of submicron aerosol.

Organic aerosol is ~70% secondary in summer. (Williams et al., 2010 ACP; Docherty et al, 2008 ES&T)

Questions for CalNex 2010 Bakersfield and LA:

Primary versus secondary PM?

Sources of Primary and Secondary Precursors?

Current Major Issues & Questions • Why are models under-predicting organic aerosol

(missing precursors or formation pathways)? [Hodzic et al. ACP (2010), de Gouw et al. JGR (2008), Dzepina et al. (2011)]

• Importance of diesel vs. gasoline vehicles in urban

areas (hotly debated issue)? [Bahreini et al. GRL (2012), Robinson et al. Science (2007), Weitkamp et al.

ES&T (2008)]

• Importance of non-traditional SOA precursors that

have not been measured (e.g. intermediate volatility

organic compounds)? [Robinson et al. Science (2007), Kroll & Seinfeld Atmos. Env. (2008), Jimenez et

al. Science (2009)]

How do we measure and model (represent) emission,

oxidation and fate of 1000’s of individual compounds?

Key Challenge: Atmospheric Organics

Williams et al 2006

TAG Aerosol Gas Chromatogram

“Unresolved Complex Mixture” (UCM)

50-80% of mass (e.g. Diesel) Schauer et al 1999

UCM

characterization

must be

improved

Tkacik et al (Robinson group) 2012

SOA production from diesel exhaust

Field Measurements of Organic Carbon Urban (Bakersfield & LA) and on-road tunnel measurements

Gasoline and Diesel fuel samples (52 across CA)

In-Situ GCMS C1-C17 VOCs

In-Situ TAG speciated PM2.5 organics & I/SVOCs

Offline Filter Analysis using novel techniques

(GCxGC-VUV-HRTOFMS)

Caldecott Tunnel, Oakland, CA Bakersfield, CA

Speciated Organics Observed Spans 15 orders of Magnitude in Volatility!

Gentner et al

Robinson et al 2007

TAG

Focus on quantifying “inferred” IVOCs/SVOCs

Robinson et al 2007

(flipped to match GC)

TAG

Gentner et al

Gas

Chromatograph

Mass

Spectrometer

Aerosol Collector

&

Thermal

Desorption Cell

Cyclone Precut

(PM2.5)

Filter

x

Valveless

injector

Den

ud

er

TAG

= Thermal desorption Aerosol Gas chromatograph

Hourly In Situ Measurements

Speciated Organics in Aerosols

Williams et al AS&T 2006, Goldstein et al J Chrom A 2009, Zhao et al AS&T 2013

CalNex Bakersfield:160 compounds

measured hourly by TAG

Semi-volatile organic

compounds

0.5

0.4

0.3

0.2

0.1

0.0

O:C

Rati

o

10-13 10

-11 10-9 10

-7 10-5 10

-3

Subcooled Vapor Pressure (atm, 25C)

Alkanes PAHs and Branched PAHs Hopanes Alkyl-cyclohexanes Lactones Esters Acids Carbonyls Others

Zhao et al ES&T In Press

Semi-volatile organic

compounds

0.5

0.4

0.3

0.2

0.1

0.0

O:C

Rat

io

10-13 10

-11 10-9 10

-7 10-5 10

-3

Subcooled Vapor Pressure (atm, 25ºC)

Alkanes PAHs and Branched PAHs Hopanes Alkyl-cyclohexanes Lactones Esters Acids Carbonyls Others Phthalic acid Pinonaldehyde 6, 10, 14-Trimethylpentadecanone

Examine Partitioning of Typical

Compounds

Zhao et al ES&T In Press

Are additional gas-to-particle partitioning

pathways important

beyond traditional partitioning theory?

Pankow, 1994; Jang et al., 2002; Tillmann et al., 2010

Fraction in particle phase > model prediction

Zhao et al ES&T In Press

Evidence for additional formation pathways:

Fraction in particle higher than model prediction

Pinonaldehyde

1) Chamber studies have shown that

dimers of pinonaldehyde can form in

the presence of sulfuric acid (Liggio

and Li, 2006)

2) Oligomerization increases with

acidity (Liggio and Li, 2006)

Phthalic acid

1) Organic acids can react with

ammonia (Na et al., 2007)

2) Phthalic acid ammonium salts

can be thermally desorbed

back to phthalic acid and

ammonia (Hajek et al.,1971)

Zhao et al ES&T In Press

What else does particle-phase

pinonaldehyde depend on?

Acidity doesn’t

increase as RH

increases

Fraction of

Pinonaldehyde in

particles increases

as RH increases

0.9

0.6

0.3

0.0

Fra

ctio

n i

n t

he

par

ticl

e phas

e

6050403020

RH (%)

1.30

1.25

1.20

1.15Cat

ion-t

o-a

nio

n r

atio

706050403020

RH (%)

Sampling Period I Sampling Period II

Zhao et al

ES&T In Press

Formation of particle-phase phthalic acid

in the atmosphere

Phthalic acid ammonium salts thermally desorb to phthalic acid and ammonia (Hajek et al.,1971)

Some of the particle phase phthalic acid was actually an ammonium salt.

Particle phase uptake of phthalic acid enhanced by availability of ammonia in Bakersfield.

1.0

0.8

0.6

0.4

0.2

Fra

ctio

ns

in t

he

par

ticl

e p

has

e

3530252015

Gas-phase Ammonia (ppbv)

R2=0.8

Zhao et al ES&T In Press

Partitioning

Take Home Messages

Additional pathways of gas-to-particle partitioning are

important beyond absorptive partitioning in SJV:

1) Reactions of phthalic acid with ammonia form

SOA

2) Reactive uptake of pinonaldehyde into particles

does not need the presence of inorganic acids.

Organic Aerosol (OA) PMF to Examine Sources

6 Contributing Factors Identified with TAG

0.18

0.09

0.00

He

xa

de

ca

ne

He

pta

de

ca

ne

Octa

de

ca

ne

No

na

de

ca

ne

Eic

osa

ne

He

ne

ico

sa

ne

Prista

ne

Ph

yta

ne

Ph

en

an

thre

ne

Flu

ora

nth

en

eP

yre

ne

No

nylb

en

ze

ne

Dim

eth

yln

ap

hth

ale

ne

Dim

eth

yln

ap

hth

ale

ne

2(3

H)-

Fu

ran

on

e, 5

-bu

tyd

ihyd

ro2

(3H

)-F

ura

no

ne

, 5

-pe

nty

ldih

yd

roD

ibe

nzo

fura

nB

en

zo

ph

en

on

e9

-Flu

ore

no

ne

Xa

nth

on

e4

-Hyd

roxy-9

-Flu

ore

no

ne

9.1

0-a

nth

race

ne

dio

ne

2-P

en

tad

eca

no

ne

, 6

, 1

0, 1

4-t

rim

eth

yl

2H

-1-B

en

zo

pyra

n-2

-on

eP

ino

na

lde

hyd

eD

ieth

ylp

hth

ala

teB

is(2

-me

thylp

rop

yl)p

hth

ala

teD

ibu

tylp

hth

ala

teP

hth

alic

Acid

Me

thylp

hth

alic

acid

0.100.050.000.12

0.06

0.00

0.600.300.00

Con

trib

utio

n t

o F

acto

r (e

ach

fa

cto

r p

rofile

su

ms t

o 1

)

0.200.100.000.30

0.15

0.00

Local POA

A mixture ofOA sources

SOA1

SOA2

SOA3

SOA4(Nighttime SOA)

Zhao et al

in review

OA PMF factor Diurnal Cycles

(Small)

(Morning local)

Primary

Secondary

(Daytime regional)

OA PMF Factor

Wind Roses

(small) (local)

(regional)

ADD MAP HERE

Daytime Wind

Nighttime Wind

(higher wspd,

regional)

Diurnal Cycle of OA Source Contributions

SOA fractions agree for TAG and AMS PMF’s

(a) Daytime SOA

TAG = SOA2 + SOA3

AMS = high O/C aromatic

+ high O/C alkane

+ petroleum SOA

(Liu et al., JGR 2012,

SOA from fossil fuel sources

contributes majority of OA)

(b) Nighttime SOA

TAG = SOA1 + SOA4

AMS = low O/C alkane +

nighttime OA.

TAG shifted right for clarity

Vertical bars = Std Deviation

Organic Aerosol PMF Analysis

Policy Relevant Messages

1) SOA accounts for approximately 75% of OA in

Bakersfield (TAG and AMS PMFs agree)

2) SOA has multiple different sources, with daytime

and nighttime SOA dominated by different primary

precursor sources, chemical transformation, and

partitioning processes.

3) Ammonia emission control could be important for

SOA control in San Joaquin Valley due to

formation of acid-ammonia salts.

Unresolved Complex Mixture

TAG

Need Better Baseline

Separation to Improve

Identification and

Quantification

Instrument Development

to Enhance Exploration

of Atmospheric Organics

Multi-dimensional Gas Chromatography with Vacuum

Ultraviolet Ionization and High Resolution Time of Flight

Mass Spectrometry (GCxGC-VUV-HRTOFMS)

Can Soft Ionization Improve Identifications?

GC-Vacuum Ultraviolet Ionization-MS

Electron Impact Ionization

(70 eV, harder ionization)

Lots of fragment ions

Hard to distinguish organic species

VUV Photoionization

(10.5 eV, softer ionization)

at Advanced Light Source

Very little fragmentation

Easy molecular identification

of organic species

eicosane EI

VUV

VUV

EI

Diesel Fuel Mass Spectrum versus GC retention time

eicosane

VUV

EI

alkanes

cycloalkanes

bicycloalkanes

benzenes

napthalenes

acenapthenes

higher order

PAHs

Diesel Fuel VUV Mass Spectrum versus GC retention time

“Classic Complex Hydrocarbon Mixture”

Former “UCM”

Isaacman et al Anal Chem 2012

Isomers Cleanly Separated

Isaacman et al Anal Chem 2012

alkanes

cycloalkanes bicycloalkanes

benzenes

methyl(B1)

dimethyl(B2)

trimethyl(B3)

tetramethyl(B4) n-alkane(B0) C19H40

Quantitatively Speciated Diesel Fuel

(No More Unresolved Complex Mixture)

Isaacman et al Anal Chem 2012; Gentner et al PNAS 2012

Chemical Speciation of Motor Vehicle Sources

• Diesel fuel fully characterized via

VUV soft ionization

mostly C10-C20, branched and cyclic

former “UCM” SOA yields largely

unstudied

• Gasoline 30% aromatics

mostly C4-C10

• Non-tailpipe gasoline almost all light

alkanes

mostly C4-C8

Gentner et al., PNAS 2012

Motor Vehicles

Precursor Emissions SOA Yields

~50% Aromatic

~99% Aromatic

100% Aromatic

Gentner et al., PNAS 2012

How Much SOA from Gasoline vs. Diesel?

(In the United States Diesel = Large Trucks)

Diesel emission factor ~2x gasoline

×Diesel SOA yield ~7x gasoline

Diesel forms 15x more SOA than

gasoline per liter of fuel burned

Diesel Forms 56-90% of Vehicular SOA

(90% in Bakersfield)

Gentner et al., PNAS 2012

Deciphering a Complex Mixture of Sources

Emissions

Mix of 1000’s of compounds

in the atmosphere Transport

Data!

(on 100’s of

compounds…)

• Profile of compounds emitted from each source is unique

• Source contributions calculated using source profiles and

atmospheric measurements

• Over-constrained source receptor model

Fossil Fuel Source Apportionment

Fossil Fuel Source Apportionment

Bakersfield Reactive Organic Gas (ROG)

Fossil Fuel Sources Receptor Model

• Petroleum associated gas major source of ROG emissions

Contributions to SOA are negligible since it’s comprised of

small hydrocarbons with very low SOA yields

Gentner et al., in prep

Locating Petroleum Gas Emissions

Use concentration data &

meteorological modeling to

locate emissions

Meteorological data is used

to calculate “footprint” (aka

back-trajectory) for each data

time point

Met footprints are weighted

by ground site concentration

data to construct map of

emissions for a region

Average footprint (or area of

influence) for 6 hrs before

transport to Bakersfield site

Average 6 hr Footprint

Location of

Oil/Gas Wells

Statistical Distribution of

Petroleum Gas Source

Gentner et al., in prep

Locating Petroleum Gas Emissions

Locating Dairy Emissions

Methane Non-Vehicular Ethanol

Policy Implications

Fossil Fuel VOC sources • Both gasoline & diesel vehicles are important urban sources

of fossil fuel SOA precursors, but diesel dominates (65-90%),

particularly in Bakersfield (90%)

• Further research on understudied SOA yields is necessary

• Fuel composition and potential SOA formation should be

considered in SOA control policies

• Gasoline dominates over diesel for emissions of organic

precursors to ozone [Gentner et al. ES&T, in review]

• Petroleum operations are a major source of ROG emissions

in Bakersfield (potentially important for ozone, but not SOA)

Gentner et al., PNAS (2012)

Thank you for listening!

No More

UCM

Diesel SOA > Gasoline SOA

Deciphering the

origins and

transformations of

atmospheric organics