Global isoprene sources and chemistry: constraints from atmospheric observations Daniel J. Jacob with Emily Fischer, Fabien Paulot, Lei Zhu, Eloïse Marais,

Global isoprene sources and chemistry: constraints from atmospheric observations

Daniel J. Jacob

with Emily Fischer, Fabien Paulot, Lei Zhu, Eloïse Marais, Chris Miller

and funding from NASA, HUCE

Volatile organic compounds (VOCs) in the atmosphere:carbon oxidation chain

VOC RO2

NO2

O3

organicperoxyradical

NO

h

carbonyl R’O2

h

OH + products

organic aerosol

ROOHorganicperoxide

OHHO

2

OH, h

OH

products

EARTH SURFACE

biospherecombustionindustry

deposition

Increasing functionality & cleavage• sources of organic aerosol• sources/sinks of oxidants (ozone, OH)

Volatile organic compounds (VOCs) in the atmosphere:effect on nitrogen cycle

NOx

CH3C(O)OO

OH

EARTH SURFACE

combustion deposition

Reservoirs for long-range transport of NOx

lightning

deposition

HNO3

peroxyacetylnitrate(PAN)

other organic nitrates

NOx

OH

deposition

HNO3

Long-range atmospheric transport

RO2N fixation

hours

Why is isoprene such an important VOC?

Met

hane

Isop

rene

Oth

er b

iosph

ere

Anthr

opog

enic

Biomas

s bu

rning

0

200

400

600Global emission, Tg C a-1

1. Large emission:

2. Oxidation generates suite of volatile reactive products:

Isoprene

OH

~1 hmultistep

• Formaldehyde• Other carbonyls• Dicarbonyls• Peroxides• Epoxides• Isoprene nitrates

Contribution of isoprene to PANfrom GEOS-Chem global 3-D chemical transport model

Emily Fischer, Harvard

Anthropogenic

Open fires

Isoprene

Other biogenic VOCs

%

January July

Sensitivity of nitrogen deposition to isoprene emission

Sensitivity for Cayuhoga National Park (Ohio) computed with the GEOS-Chem adjoint

Local isoprene emission suppresses N deposition, upwind emission increases it

Fabien Paulot, Harvard

of local NOx

emission)

Estimating isoprene emissions: bottom-up and top-down approaches

Bottom-up estimate from plant model:EISOP = f(plant type, phenology, LAI, T, PAR, water stress, …)

Isopreneoxidationproducts

Ecosystem observations

Atmospheric observations

Top-down estimate from Inversion of chemical transport model:EISOP = f(atmospheric concentrations,transport, chemistry)

Observing isoprene oxidation products from space:formaldehyde (HCHO) and glyoxal (CHOCHO)

Scattering by atmosphereand Earth surface

l1

l2 HCHO orCHOCHOabsorptionspectrum

l1

l2

GOME (1995-2001), SCIAMACHY (2002-2012),OMI (2004-), GOME-2 (2006-) instruments

• Spectral fitting yields “slant” columns of HCHO, CHOCHO along light path• Air mass factor from radiative transfer model converts slant to vertical columns

HCHOCHOCHO

Annual mean vertical columns from GOME-2, 2007-2008

HCHO CHOCHO

Relating HCHO columns to VOC emission

VOCi HCHOh (340 nm), OHoxidation

k ~ 0.5 h-1

Emission Ei

displacement

In absence of horizontal wind, mass balance for HCHO column WHCHO:

i ii

HCHO

y E

k

yield yi

but wind smears this relationship depending on VOC lifetime wrt HCHO production:

Local linear relationshipbetween HCHO column and E

VOCsource Distance downwind

WHCHOIsoprene

a-pinenemethanol

100 km

detection limit

HCHO is mainly sensitive to isoprene emission with smearing ~ 10-100 km

Past use of HCHO vs. EISOP relationship over US to constrain isoprene emission with OMI data

OMI HCHO (Jun-Aug 2006)

OMI-constrained isoprene emission

GEOS-Chem local relationship betweenHCHO column and isoprene emission

Model slope (2400 s) agrees with INTEX-A vertical profiles (2300),

PROPHET Michigan site (2100)

Palmer et al. [2003, 2006}, Millet et al. [2006, 2008]

Temperature dominates variability of EISOP seen by OMIcan’t pick up any other variable from multivariate correlations, case studies

Lei Zhu, Harvard1 5 10 151015 molecules cm-2

HCHO column,Jun-Aug 2005

2006

2007

2008

Correlation of monthly mean HCHO with air T

NE Texas, JJA 2005-2008Exponential fitMEGAN

Daily data in Southeast US binned by air temperature

290 295 300 305 310 K

285 290 295 300 K

turnoverat 307 K

After 2009 it’s curtains for OMI…but GOME-2 provides consistent continuity

GOME-2 HCHO, 2007 OMI

June

July

August

GOME-2 vs. OMI correlationmonthly data in SE US JJA 2007-2008

Lei Zhu, Harvard

OMI 13x24 km2 13:30GOME-2 40x80 km2 9:30

nadir pixel time

slope = 0.91r2 = 0.82

Using OMI HCHOto constrain isoprene emissions in Africa

MODIS leaf area index MODIS fire counts Earth lights AATSR gas flares

1015 molecules cm-2

OMI annual meanHCHO slant columns

2005-2009

• Observed HCHO distribution over Africa points to sources from (1) biosphere, (2) open fires, (3) oil and gas industry

• Africa accounts for 20% of global biogenic isoprene emissions in MEGAN inventory…but based on little in situ data

Aug-Sep

Marais et al., in press

1015 molecules cm-2

Isolating biogenic HCHO in the OMI data• Exclude open fire (and dust) influence using MODIS

fire counts, OMI absorbing aerosol optical depth

• Exclude oil/gas industry influence using AATSR gas flare product

MODIS fire count > 0

OMI smoke AAOD > regional threshold

BIOMASS BURNING INFLUENCE (exclude)

NO

NO

YES

YES MODIS fire counts

OMI smoke AAOD

NO

OMI dust AAOD > 0.1 DUST INFLUENCE

(exclude) YES OMI dust AAOD

AATSR fire count > 0 ANTHROPOGENIC

INFLUENCE (exclude)

YES AATSR fire counts

OMI dust AAOD > 0.1 DUST INFLUENCE

(exclude) YES OMI dust AAOD

NO

BIOGENIC

(retain)

Marais et al., in press

HCHO slant columnoriginal data

HCHO vertical columnbiogenic only

air mass factor

HCHO slant column

HCHO biogenic vertical column;8-day product with 1ox1o resolution

OH

NO

O

HH

HO2

O

HO OH

-IEPOX

formaldehydeh

Pathways for HCHO formation from isoprene oxidation

RO2

OH

OH

IsomerizationC1,5 -shift

ROOH

• high-NOx branch (RO2+NO) yields fast HCHO as 1st generation product

Peeters

Paulot

O

MVK

O

MACR

Epoxydiols [Paulot et al., 2009]

More recently proposed low-NOx pathways regenerate OH, produce HCHO:

Isomerization [Peeters and Muller, 2010]

standardGEOS-Chemmechanism

first-generationhigh-NOx

low-NOx

• low-NOx branch (RO2+HO2 ) yields slower HCHO, depletes OH

OH

Time-dependent HCHO yield from isoprene oxidation

DSMACC box model calculations

aging/smearing

Yield is sensitive to NOx , not so much to mechanism except at very low NOx

Marais et al., in press

Boundary layer NOx levels over Africa

Annual NO2 tropospheric columns, fire influences excludedSatellite observations Model % isoprene RO2 reacting with NO

(GEOS-Chem, July)

• Boundary layer NOx over Africa is typically 0.1-1 ppbv• Expect NOx dependence of HCHO yield, moderate smearing

Marais et al., in press

boundary layer

Testing HCHO-isoprene smearing with AMMA aircraft data

Flight tracks (Jul-Aug 2006)and MODIS leaf area index Latitudinal profiles below 1 km

WIND

• HCHO tracks isoprene with only ~50 km smearing

• But NOx measured in AMMA was relatively high (mean 0.3 ppb)

OMI HCHO

Marais et al., in press

WIND

Smearing produces“shadow” region 200-300 km downwind of rainforest

Marais et al., in press

OMI HCHO column

1015 molecules cm-2

WIND

July

Testing HCHO-isoprene smearingin longitudinal transect across Congo:high isoprene and low NOx

shadow

Relationship between HCHO column and isoprene emission

Model sensitivity S of HCHO column (ΔHCHO) to isoprene emission (ΔEISOP) as function of tropospheric NO2 column (NO2)

StandardPaulot

• Use S = ΔHCHO / ΔEISOP for local OMI NO2 to derive isoprene emission• Exclude “shadow” regions on basis of anomalously high S values

Marais et al., in press

Error analysis on inferring EISOP from satellite HCHO data

Slant HCHO column

20% (spectral fitting)

Vertical HCHO column

20% (clouds, vertical distribution, albedo)

Isoprene emission

Estimated errors (8-day data, 1o x1o resolution)

15% (chemical mechanism)25-60% (smearing)15% (NO2 column)

• Total error: 40% (high-NOx ), 40-90% (low-NOx ). Can be reduced by averaging• Smearing is dominant error component. Need to resolve transport!

Marais et al., in press

Isoprene emission (12-15 local time annual mean, 2006)

Comparison of OMI isoprene emissions to MEGAN

MEGAN is too low for equatorial forest, too high for savanna

Marais et al., in press

2005-2009 monthly variability of isoprene emissionfor evergreen broadleaf forest of central Africa

Eloïse Marais, Harvard

• Variability is small and weakly correlated to temperature and LAI• Need to address uncertainty in meteorological and LAI products!

EISOP , temperature

EISOP , LAIAVHRR

2005-2009 monthly variability of isoprene emissionin open deciduous broadleaf forest of s. Africa

• May-Sept dry season; LAI drops below 1 in Aug, driving EISOP down• Sept-Nov increase in LAI (greening) causes spike in EISOP • Wet season cloudiness causes T to decrease after Nov, driving EISOP down

even though LAI continues to increase• Suggests saturation of EISOP when LAI exceeds 1.5 Eloïse Marais, Harvard

EISOP , temperature

EISOP , LAI

Jan Jan Jan

AVHRR

Glyoxal from space as additional constraint on VOC sources

GOME-2

• Glyoxal sources in GEOS-Chem:55% isoprene, 24% acetylene, 7% aromatics, 8% fire emission, 2% monoterpenes

• Glyoxal lifetime ~1 h (photolysis)

Chris Miller, Harvard

• Operational data available from SCIAMACHY, GOME-2

• OMI retrieval in progress (Chris Miller, Harvard)

GEOS-Chem

Does glyoxal provide information complementary to HCHO?

GOME-2

GEOS-Chem

Glyoxal columns (Jun-Aug 2007) Glyoxal/HCHO column ratio

GOME-2 shows variability in glyoxal/HCHO ratio that GEOS-Chem doesn’t capture

Chris Miller, Harvard

Glyoxal production from isoprene

Observed fast production with 2-3% yield [Galloway 2011] – Dibble isomerization?

Chris Miller, Harvard

Dibble isomerization

first-generation

Tower data from CABINEX, northern Michigan (Jul-Aug 09)

Measured GEOS-Chem with EISOP /2

isoprene

Glyoxal

Pathways forglyoxal formation

Dibble

Observations by Frank Keutsch

Dibble isomerization is dominant model pathway for glyoxal formation

Chris Miller, Harvard

OH-aldehydes

Vision for the future: ecosystem monitoringAdjoint inversion of isoprene emission using geostationary satellite observations of HCHO and glyoxal

HCHO, glyoxalmeasurement

(x, t)1-km chemicaltransport model

inverse model

Emission

E( x’, t’)

• Geostationary observation diurnal information, higher precision daily data GEMS (Korea), 2017; Sentinel-4 (Europe), 2019; GEO-CAPE (US), 2020+• Adjoint inversion solve smearing problem, allow isoprene emission monitoring need to properly represent chemistry-transport coupling on scales of PBL mixing

Wind

boundary layermixing (~1 h)