Can a “state of the art” chemistry transport model simulate Amazonian tropospheric chemistry?

28 Pages·2011·36.99 MB·English

JOURNAL OF GEOPHYSICAL RESEARCH, VOL.116,D16302, doi:10.1029/2011JD015893,2011
“ ”
Can a state of the art chemistry transport model
simulate Amazonian tropospheric chemistry?
Michael P. Barkley,1 Paul I. Palmer,2 Laurens Ganzeveld,3 Almut Arneth,4,5
Daniel Hagberg,4 Thomas Karl,6 Alex Guenther,6 Fabien Paulot,7 Paul O. Wennberg,7,8
JingqiuMao,9Thomas P.Kurosu,10KellyChance,10 J.‐F.Müller,11 Isabelle DeSmedt,11
Michel Van Roozendael,11 Dan Chen,12,13 Yuxuan Wang,14 and Robert M. Yantosca15
Received2March2011;revised8April2011;accepted10May2011;published17August2011.
[1] We present an evaluation of a nested high‐resolution Goddard Earth Observing
System (GEOS)‐Chem chemistry transport model simulation of tropospheric chemistry
over tropical South America. The model has been constrained with two isoprene emission
inventories: (1) the canopy‐scale Model of Emissions of Gases and Aerosols from Nature
(MEGAN) and (2) a leaf‐scale algorithm coupled to the Lund‐Potsdam‐Jena General
Ecosystem Simulator (LPJ-GUESS) dynamic vegetation model, and the model has been
run using two different chemical mechanisms that contain alternative treatments of
isoprene photo‐oxidation. Large differences of up to 100 Tg C yr−1 exist between the
isopreneemissionspredictedbyeachinventory,withMEGANemissionsgenerallyhigher.
BasedonoursimulationsweestimatethattropicalSouthAmerica(30–85°W,14°N–25°S)
contributes about 15–35% of total global isoprene emissions. We have quantified the
modelsensitivitytochangesinisopreneemissions,chemistry,boundarylayermixing,and
soil NO emissions using ground‐based and airborne observations. We find GEOS‐Chem
x
has difficulty reproducing several observed chemical species; typically hydroxyl
concentrations are underestimated, whilst mixing ratios of isoprene and its oxidation
products are overestimated. The magnitude of model formaldehyde (HCHO) columns are
most sensitive to the choice of chemical mechanism and isoprene emission inventory.
WefindGEOS‐Chemexhibitsasignificantpositivebias(10–100%)whencomparedwith
HCHO columns from the Scanning Imaging Absorption Spectrometer for Atmospheric
Chartography (SCIAMACHY) and Ozone Monitoring Instrument (OMI) for the study
year 2006. Simulations that use the more detailed chemical mechanism and/or lowest
isoprene emissions provide the best agreement to the satellite data, since they result in
lower‐HCHO columns.
Citation: Barkley, M.P.,et al.(2011), Cana“stateof the art”chemistry transport model simulateAmazonian tropospheric
chemistry?, J. Geophys.Res.,116,D16302, doi:10.1029/2011JD015893.
1. Introduction themostproductiveanddiverseecosystemonEarth.Oneof
thereasonswhytheAmazonrainforestisimportanttoclimate
[2] The Amazon Basin, covering nearly 6 million square is because tropical vegetation emit a wide range of highly
kilometers and containing the world’s largest rainforest, is
reactive nonmethane biogenic volatile organic compounds
8DivisionofGeologicalandPlanetarySciences,CaliforniaInstituteof
1EOSGroup,DepartmentofPhysicsandAstronomy,Universityof Technology,Pasadena,California,USA.
Leicester,Leicester,UK. 9Geophysical Fluid Dynamics Laboratory, Princeton University,
2SchoolofGeoSciences,UniversityofEdinburgh,Edinburgh,UK. Princeton,NewJersey,USA.
3Earth System Sciences, Department of Environmental Sciences, 10Atomic and Molecular Physics Division, Harvard‐Smithsonian
WageningenUniversityandResearchCenter,Wageningen,Netherlands. CenterforAstrophysics,Cambridge,Massachusetts,USA.
4Department of Physical Geography and Ecosystems Analysis, 11BelgianInstituteforSpaceAeronomy,Brussels,Belgium.
GeobiosphereScienceCenter,LundUniversity,Lund,Sweden. 12DepartmentofEnvironmentalScienceandEngineering,Tsinghua
5AtmosphericEnvironmentalResearch,InstituteofMeteorologyand University,Beijing,China.
ClimateResearch,KarlsruheInstituteofTechnology,Karlsruhe,Germany. 13NowatDepartmentofAtmosphericandOceanicSciences,University
6NationalCenterofAtmosphericResearch,Boulder,Colorado,USA. ofCalifornia,LosAngeles,California,USA.
7DivisionofEngineeringandAppliedScience,CaliforniaInstituteof 14Ministry of Education Key Laboratory for Earth System Modeling,
Technology,Pasadena,California,USA. Center for EarthSystem Science,Institute for Global Change Studies,
TsinghuaUniversity,Beijing,China.
Copyright2011bytheAmericanGeophysicalUnion. 15DivisionofEngineeringandAppliedSciences,HarvardUniversity,
0148‐0227/11/2011JD015893 Cambridge,Massachusetts,USA.
D16302 1 of28
D16302 BARKLEY ET AL.:TROPOSPHERIC AMAZONIAN CHEMISTRY D16302
(BVOCs)intotheatmosphere.TheseBVOCsplayacritical [5] DuringthelastdecadesatelliteobservationsofHCHO
role in global atmospheric chemistry and climate as their columns, retrieved using ultraviolet (UV) absorption spec-
photochemical reactions influence the oxidation capacity of troscopy [Chanceetal.,2000;DeSmedtetal.,2008],have
theatmosphere[Poissonetal.,2000;MonsonandHolland, been used in several studies to map top‐down isoprene
2001], and the lifetimes and distributions of other key trace emissionsoncontinentalandglobalscales[see,e.g.,Palmer
gases, such as carbon monoxide (CO) and methane (CH ) etal., 2003, 2006; Shim et al., 2005; Fu et al., 2007; Millet
4
[Granier et al., 2000; Collins et al., 2002]. The most et al., 2007; Barkley et al., 2008; Stavrakou et al., 2009b].
important BVOC emitted by vegetation is isoprene since it RecentworkbyBarkleyetal.[2008],usingHCHOcolumn
accounts for about half of the total global BVOC budget datafromtheGlobalOzoneMonitoringExperiment(GOME)
[Guenther, 2002], and because of its influence on tropo- [European Space Agency, 1995; Burrows et al., 1999] and
spheric ozone [Jenkin and Clemitshaw, 2000; Sanderson ScanningImagingAbsorptionSpectrometerforAtmospheric
et al., 2003; Fiore et al., 2005] and its precursor role in Chartography (SCIAMACHY) [Bovensmann et al., 1999]
the formation of secondary organic aerosol [Claeys et al., instruments, identified potentially large gaps in our quanti-
2004; Kanakidou et al., 2005]. Isoprene emissions are also tative understanding of Amazonian isoprene emissions, and
relevant to carbon cycle studies as they represent a loss of unexplained observed seasonal variations [Barkley et al.,
fixedcarbonfromtheterrestrialbiosphere[Kesselmeieretal., 2009]. In contrast, SCIAMACHY top‐down estimates
2002] and a nonnegligible photochemical source of carbon derived by Stavrakou et al. [2009b] gave better agreement
dioxide[Folberthetal.,2005]. with the MEGAN inventory for the Amazon region; in that
[3] Despite the Amazon Basin being acknowledged as a study emissions were calculatedusing the Model for Hydro-
significant isoprene source [Guenther et al., 2006; Arneth carbonEmissionsbytheCanopy(MOHYCAN)[Mülleretal.,
et al., 2008], there have only been a few measurement 2008] forced with European Center for Medium‐Range
campaignstodatethathavetargetedthisregion(owinginpart WeatherForecast(ECMWF)meteorologicalanalyses.These
to inaccessibility), resulting in a limited number of in situ differences, in part, reflect the difficulties and uncertainties
measurements. Leaf and branch level measurements [Kuhn associated with inferring isoprene emissions from satellite
et al., 2002] offer insight to small‐scale processes, whilst measurements of HCHO columns. Critically, the accuracy
tower,balloonandairborneplatformsprovideinformationon ofthetop‐downemissionslargelydependsontheabilityof
localized emissions [e.g., Helmig et al., 1998; Karl et al., thechosenchemistrytransportmodel(CTM),theintermediary
2007; Kuhn et al., 2007]. However, for the Amazon Basin used to invert the retrieved HCHO columns, to accurately
asawhole,isopreneemissionsarepoorlyquantified.Bottom‐ simulatethechemistryanddynamicsoftherainforestatmo-
upemissions inventories,suchasthewidelyusedModelof sphere.Modelspatialresolutionisimportantinthisrespect,
Emissions of Gases and Aerosols from Nature (MEGAN) bothtocapturelocalizeddynamicalandchemicalprocesses,
[Guentheretal.,2006],arehighlyuncertainastheyrelyon and distinguish biogenic and pyrogenic contributions to the
upscaling sparse point measurements to landscape scales; observedHCHOcolumns[Barkleyetal.,2008].
whereecosystemdiversityishigh,asintropicalecosystems, [6] However, modeling Amazonian tropospheric chemis-
thisisespeciallydifficult.Moreover,theempiricalalgorithms tryisdifficult[Ganzeveldetal.,2002;vonKuhlmannetal.,
that drive variations in the bottom‐up emissions are mostly 2004; Butler et al., 2008]. Besides the large uncertainties
basedonstudiesofextratropicalplantspecies,whichmaynot associated with the bottom‐up BVOC emissions (typically
beapplicabletotropicalvegetation.Modelstudiesconducted >100%), the subsequent oxidation chemistry occurring in
atrelativelycoarsespatialscaleshavegenerallybeenunableto the prevalent low‐NO conditions is poorly understood and
x
accurately reproduce the sparse observations of Amazonian ofteninadequatelyparameterized.Inparticular,therehasbeen
isoprenefluxesandsurfaceconcentrationswithoutsomeform much emphasis placed on the underestimation of hydroxyl
of adjustment or scaling of the bottom‐up emissions [von (OH)concentrationsoverdensetropicalrainforestsbymost
Kuhlmann et al., 2004; Jöckel et al., 2006; Müller et al., CTMs and chemistry‐climate models [e.g., Butler et al.,
2008;Barkleyetal.,2008]. 2008]. Various plausible mechanisms to recycle or regen-
[4] Formaldehyde (HCHO), a short‐lived trace gas and erate OH through improved isoprene degradation schemes
high yield product of isoprene oxidation, can provide addi- have been proposed and compared with observational data
tional information on surface isoprene emissions on length [Butler et al., 2008; Lelieveld et al., 2008; Paulot et al.,
scalesoforder100km[Palmeretal.,2003].However,since 2009a;Stavrakou etal., 2010]. Given the large uncertainties
HCHOoriginatesfromavarietyofatmosphericandsurface of these chemical mechanisms much remains unresolved.
sources,carefuldisaggregationisneededtoaccuratelyderive Although significant effort has gone into improving our
satellite‐based (i.e., top‐down) isoprene emission estimates. understandingoftropicaloxidationchemistry, theinfluence
Globally, the largest source of HCHO is produced from the of physical and micrometeorological processes on reactive
oxidation of methane [Stavrakou et al., 2009a], though this gas exchange can be equally, if not more, important [Pike
only maintains ambient background concentrations, such as et al., 2010; Pugh et al., 2010]. Correct parameterizations
found in the remote marine atmosphere. Over land, BVOC ofbothin‐canopy(e.g.,soilNO emissions,deposition)and
x
oxidation over densely vegetated areas and anthropogenic above‐canopy (e.g., convection, turbulent mixing, segrega-
VOC oxidation over urban areas create strong regional tion effects) processes are essential to properly model the
HCHOenhancements,easilyobservedfromspace.Biomass tropical atmosphere [Ganzeveld and Lelieveld, 2004; Karl
burning and wild fires are also significant localized HCHO et al., 2004; Ganzeveld et al., 2008; Pugh et al., 2010].
sources,owingtodirectlyreleasedHCHOduringincomplete [7] In this manuscript we present the first comprehensive
combustion and from the oxidation of coemitted VOCs evaluation of a high‐resolution simulation of tropospheric
[AndreaeandMerlet,2001]. chemistry over tropical South America, performed by a
2 of 28
D16302 BARKLEY ET AL.:TROPOSPHERIC AMAZONIAN CHEMISTRY D16302
Figure 1. Schematic showing thedomain ofthe Amazon nested grid. The thick black line is theactual
boundaryofthenestedwindow;0.5°×0.667°gridcellsoutsidetheblacklinerepresentsthebufferzone
for theboundaryconditions. The GEOS‐Chem4° ×5° horizontal grid isshown by thedottedlight grey
lines.ThelocationoftheTROFFEEcampaign[Karletal.,2007]isshownastheblackcross(seesection4.1).
ThedomainoftheGABRIELcampaign[Stickleretal.,2007]isshowninsetbytheblackdashedline(see
section 4.2).
nested grid version of the widely used Goddard Earth theauxiliarymaterial.1Wediscussthepotentialimplications
Observing System (GEOS)‐Chem CTM [Bey et al., 2001]. resultingfromthemodelvalidationinsection5,andconclude
Given that in future work we want to use GEOS‐Chem to thepaperinsection6.
derive top‐down isoprene emissions for this region, and
knowing the large uncertainties associated with simulating 2. GEOS‐Chem
Amazonian BVOC emissions and tropospheric chemistry,
2.1. Overview
the objective of this paper is two fold. First, we assess the
ability of GEOS‐Chem to accurately simulate isoprene oxi- [9] GEOS‐Chemisaglobal3‐Dchemistrytransportmodel
dation chemistry and other observed key atmospheric con- [Beyetal.,2001]whichweuseheretosimulatetropospheric
stituentsovertheAmazonrainforest.Secondly,weevaluate chemistry over tropical South America. To reproduce the
the model’s usefulness to interpret satellite observations of trace gas distributions over Amazon rainforest at relatively
HCHO, or in other words, its suitability for inferring top‐ finespatialscaleswerunGEOS‐Chem(v8‐03‐01)inaone‐
down isoprene emission estimates. To achieve these objec- way high‐resolution nested grid mode. This nested grid
tiveswelimitourfocusonthemodelsensitivitytofourkey capability of GEOS‐Chem was first developed to study the
processes: surface BVOC emissions, chemistry, boundary east Asia region by Wang et al. [2004], and was more
layermixing,andsoilNO emissions.Todeterminetherel- recently updated by Chen et al. [2009]. Here we adapt the
x
ativeimportanceofeachprocessweusesurfaceandairborne model to be centered over the Amazon Basin, as shown in
observations fromprevious Amazon fieldcampaigns,along Figure1.Themodelhasahorizontalresolutionof0.667°×
with satellite observations of HCHO vertical columns, to 0.5°(longitude×latitude)whichisconsistentwiththeforcing
validatethemodeloutput.Thisworkisnovelinthatforthe meteorology taken from the Goddard Earth Observing
first time GEOS‐Chem will be forced with two contrasting System (GEOS‐5) of the NASA Global Modeling and
BVOC emission inventories to assess their influence on Assimilation Office (GMAO) [Rienecker et al., 2008]. The
Amazoniantroposphericchemistry. GEOS‐5meteorologicaldataareupdatedevery6hfor3‐D
[8] This manuscript is structured as follows. In section 2 variablesandevery3hforsurfacefieldsandmixingdepths.
we provide an overview of the GEOS‐Chem model and In the vertical coordinate, we run the model with 47 hybrid
outline the chemical mechanism. We discuss and compare eta levels extending from the surface to 0.01 hPa, with the
thetwoBVOCemissioninventoriesseparatelyinsection3. boundary layer up to 2 km resolved by 14 layers (with
In section 4 we evaluate the model using the surface/
airbornefieldmeasurementsandthesatelliteHCHOcolumn 1Auxiliary materials are available in the HTML. doi:10.1029/
data;detailsofthesatelliteHCHOretrievalsareprovidedin 2011JD015893.
3 of 28
D16302 BARKLEY ET AL.:TROPOSPHERIC AMAZONIAN CHEMISTRY D16302
Table 1. GEOS‐Chem Anthropogenic, Biofuels, and Selected lumped>C alkenesandisopreneareconsidered,withorganic
2
Biomass BurningEmissions Fromthe Amazon for 2006a peroxides recycled [see Palmer et al., 2003; Fiore et al.,
2005]. Details of non‐BVOC emissions are described sepa-
Biomass
Speciesb Units Anthropogenic Biofuels Burning rately in the auxiliary material; their annual emissions for
2006aresummarizedinTable1.Photolysisratesarecalcu-
CO Tg 15.70 15.53 50.85 latedusingtheFast‐JalgorithmofWildetal.[2000],which
NO TgN 1.01 0.16 0.49
x takes into account Rayleigh scattering as well as Mie scat-
Acetone TgC 0.04 0.03 0.22
Acetaldehyde TgC – 0.06 0.22 teringbyaerosolsandclouds.Drydepositionofaerosolsand
ALK4 TgC 1.52 0.05 0.09 gases are based on a standard resistance‐in‐series model
C2H6 TgC 0.18 0.14 0.41 [Wesely,1989]asdescribedbyWangetal.[1998].Weupdate
CH TgC 0.74 0.39 0.49
3 6 themodeltoincludethedepositionofsixextraspecies(iso-
CH TgC 0.28 0.06 0.34
H3CH8O Tg – 0.13 0.52 prene, methyl vinyl ketone (MVK), methacrolein (MACR),
MEK TgC 0.05 0.11 0.15 acetone, acetaldehyde and methyl hydroperoxide) based on
SO2 TgS – – 0.27 their Henry’s Law constant and a reactivity factor, using
NH3 TgNH3 – – 0.68 values consistent with the study of von Kuhlmann et al.
BC Tg – – 0.32
OC Tg – – 2.53 [2004]; Table 2 summarizes the dry deposition losses for
2006.Wetdeposition(rainoutandwashout)ofaerosolsand
aNOxemissionsfromsoils,fertilizers,andlightningare0.91,0.03,and gasesaredescribedbyLiuetal.[2001]andMarietal.[2000],
0.83TgN,respectively. respectively. In addition to the standard GEOS‐Chem
bSpeciesdefinitionasfollows:ALK4,C , alkanes;C H,ethane;CH ,
45 2 6 3 8 chemicalscheme,analternativemechanism(hereafterreferred
propane; MEK, >C ketones; C H , propene; BC, black carbon; OC,
3 3 6
organiccarbon. toastheCaltechmechanism)hasbeendevisedfollowingthe
detailed work of Paulot et al. [2009a, 2009b]. This scheme
includes a more explicit treatment of the production of
midpoints at approximately 70, 200, 330, 470, 600, 740, organicnitrates,acidsandepoxidesfromthephoto‐oxidation
880,1000,1160,1300,1440maltitudefora columnbased ofisoprene,andcontainsanextra13transportedspeciesand
at sea level). approximately 50 more photochemical reactions. The for-
[10] Tracers are transported using a semi‐Lagrangian mationofepoxidesisofparticularrelevancetotheAmazon
approach for advection [Lin and Rood, 1996] and a relaxed region,giventhatextraOHisregeneratedinlow‐NO con-
Arakawa‐Schubert (RAS) scheme for moist convection ditions[Paulotetal.,2009a]. x
[Moorthi and Suarez, 1992]. Two options are available for [13] We include a sensitivity test, discussed in section 4,
the vertical mixing of chemical tracers within the planetary to quantify the effect of artificial OH recycling as proposed
boundary layer (PBL): (1) a full‐mixing scheme which byLelieveldetal.[2008],bymodifyingwithineachscheme
ensuresemissionsandmixingratiosareconstantwithaltitude the reaction of first generation isoprene peroxy (ISOPO )
2
at each chemistry time step (60 mins) and (2) a nonlocal and hydroperoxyl (HO ) radicals, in the formation of iso-
scheme that can include “local” mixing between adjacent prene hydroxy peroxide2s (ISOPOOH) as follows:
layersand,dependingonthestabilityofthePBL,“nonlocal”
mixingduetoturbulenteddies.Thenonlocalschemeisbased ðR1Þ ISOPO2þ HO2! ISOPOOHþnOHþ...
on the formulation of Holtslag and Boville [1993] and was
implementedintoGEOS‐ChembyLinandMcElroy[2010]; In this study we set n = 2 given the uncertainty reported in
in their study it yielded a more realistic simulation of NO recent literature [Butler et al., 2008; Kubistin et al., 2010;
2
and O at midlatitudes. Pugh et al., 2010; Stavrakou et al., 2010]. Consistent with
3
[11] To provide appropriate boundary conditions to the thesepreviousstudies,wealsoreducetheratecoefficientof
nested Amazon grid, the GEOS‐5 meteorological data is the isoprene and OH reaction (here by 10%), owing to
degradedtoa4°×5°horizontalresolutionandaglobalsim- possiblesegregationeffectsduetoincompletemixing[Krol
ulationperformed,withthetracermixingratiossavedevery et al., 2000; Karl et al., 2007], acknowledging this rate
3 h (consistent with the temporal resolution of the surface reductionishighlyuncertainandlikelytovaryconsiderably
meteorology). The archived tracer mixing ratios in coarse within the real atmosphere.
grid cells adjacent to the nested domain, are then supplied [14] Owing to the recent updates made to the chemical
to a delineated buffer zone (of three 0.667° × 0.5° grid mechanism [e.g., Millet et al., 2010; Paulot et al., 2009a,
cells width) in an area‐weighting, grid‐filled procedure (as 2009b], we reassess the GEOS‐Chem HCHO yields from
described by Wang et al. [2004]) to give the dynamic isoprene oxidation using the Master Chemical Mechanism
boundary constraints. (MCM) [Jenkin et al., 2003; Saunders et al., 2003] as our
reference chemistry (see auxiliary material). The time‐
2.2. Chemical Mechanism dependentHCHOyieldsfromtheoxidationofapulserelease
[12] The standard GEOS‐Chem chemical mechanism pro- of 1 ppbv of isoprene are given in Table 3. They show that
videsarelativelydetailedtreatmentofcoupledO ‐NO ‐VOC under high‐NO conditions (≥1.0 ppbv) the new GEOS‐
3 x x
andaerosolchemistry[Horowitzetal.,1998;Beyetal.,2001; Chem yields are similar to those previously published by
Fiore et al., 2002; Park et al., 2004] integrated using the Palmer et al. [2003, 2006]. The short‐term HCHO yield
Kinetic Preprocessor (KPP) solver of Sandu and Sander (i.e., that achieved by the end of the first day) calculated
[2006]. Emissions from anthropogenic, biogenic and pyro- by GEOS‐Chem’s standard and Caltech mechanisms are
genic sources are provided, and the explicit photo‐oxidation within 20% of the MCM value and comprise 98% and
schemes of methane, ethane, propane, lumped >C alkanes, 91% of their ultimate yields, respectively.
3
4 of 28
D16302 BARKLEY ET AL.:TROPOSPHERIC AMAZONIAN CHEMISTRY D16302
Table 2. GEOS‐Chem DryDeposition Losses Over the Amazon factors. Note that for the Amazon region the emission fac-
Domainfor 2006Based ontheS Scenarioa torsaresolelybasedonabove‐canopymeasurements.These
FM
standardized emission capacities are coupled to static veg-
Amountc vd Lifetimee
Speciesb (Tg) (cmds−1) (days) etationmaps,withemissionvariabilitymodeledusingempir-
ical algorithms forced by external meteorology and seasonal
NO2 0.43 0.16 27 estimatesofleafarea.Alternatively,withinLPJ‐GUESSana-
O 80.06 0.25 97
3 loguestoleaflevelemissioncapacitiesareassignedperplant
PAN 0.61 0.15 71
Isoprene 0.06 0.01 1054 functionaltype(PFT),andshort‐termemissionvariabilityin
HNO 4.01 1.52 10 response to temperature and light modeled using semi-
3
H2O2 14.32 0.69 9 mechanistic algorithms that link emissions to their chlopro-
Acetone 1.08 0.17 103
plastic production. The scaling to canopy,seasonal changes
Acetaldehyde 0.30 0.51 24
in leaf area index, and PFT distributions are explicitly cal-
MVK 3.62 0.16 9
MACR 3.32 0.16 9 culated permitting ecosystem emissions to be estimated. In
PMN 0.23 0.15 11 sections 3.1 and 3.2, we outline the details of each model
PPN 0.02 0.15 151 andthenprovideacomparisonoftheirrespectiveemissions
RN 0.04 0.15 53
4 2 for the Amazon region in section 3.3.
HCHO 6.10 0.51 9
N O 0.02 1.52 39
2 5 3.1. MEGAN Emissions
MP 6.10 0.24 19
aSeeTable5. [18] The latest MEGANrelease (version 2.1) [Guenther
bSpecies definition as follows: MVK, methyl vinyl ketone; MACR, andWiedinmyer,2007;Sakulyanontvittayaetal.,2008]can
methacrolein; MP, methyl hydroperoxide; PMN, peroxy methacryloyl calculate emission rates for 138 chemical species using 20
nitrate;PPN,peroxypropionylnitrate;RN ,C , alkylnitrates. explicit and lumped chemical classifications. Here we focus
4 2 45
cNotethatlossesofisoprene,acetone,andacetaldehydeareinTgC. onlyonthe(explicit)species:isoprene,methylbutenol(MBO),
dAveragedepositionvelocityoverlandandocean.
and several monoterpene compounds (a pinene, b pinene,
eAveragetroposphericlifetimeagainstdrydeposition.
limonene, myrcene, sabinene, 3‐carene, and ocimene). The
emissions,E,ofthesecompoundsareparameterizedby
[15] Inlow‐NOxconditions(≤0.1ppbv),morerelevantto
our study domain, the short‐term HCHO yields computed E¼E0(cid:2)(cid:1)CE(cid:2)(cid:1)Age(cid:2)(cid:1)SM (cid:2)(cid:3); ð1Þ
by the standard and Caltech mechanisms are within about
10% of the MCM yield and by the end of the day have whereE arethebasalemissions(inmgofcompoundm−2h−1)
0
reached 84% and 78% of their ultimate yield, respectively. normalized to standard conditions (current air temperature =
This is a significant improvement on the studies of Palmer 303 K, photosynthetic active radiation (PAR) = 1500 mmol
et al. [2003, 2006] since previously the GEOS‐Chem m−2s−1,leafareaindex=5),whicharemultipliedbyemission
yield was only within about 40% of the MCM after 1 day. activity factors that simulate changes in the emission rate
[16] AlthoughtheGEOS‐ChemHCHOyieldsarebroadly owing to the changes in the canopy environment gCE, leaf
consistent with the MCM,the reader should be aware these ageg ,andsoilmoistureg .Inthisstudyweneglectthe
Age SM
modelvaluesonlyserveasanapproximateguide,sincethey effect of soil moisture (g = 1). Similarly, we assume
SM
reflectsignificantuncertaintiesinBVOCoxidationchemistry. standard conditions for any extra production or loss of the
Furthermore, within the real atmosphere the yield depends BVOC within the vegetation canopy by setting r = 1. For
on its local environment and is therefore likely to vary isoprene, this implies typical canopy losses of about 4%
considerably [Palmer et al., 2003]. [Guenther et al., 2006].
[19] In GEOS‐Chem two different versions of MEGAN
3. BVOC Emissions can be employed by switching on/off the relevant flags in
the model control file. There are subtle but important dif-
[17] SinceBVOCemissionsfromterrestrialvegetationare ferencesbetweenthesemodelversionswhichareessentially
the primary driver of tropical oxidation chemistry, it is
based on the calculation of the canopy term g . The two
essential to model their emissions as accurately as possible. CE
models are as follows:
MostCTMsemployonlyasingleBVOCemissioninventory,
[20] 1. The PCEEA model is a simplified parameterized
typically one of the Guenther et al. [1995, 1999, 2006]
canopy environment emission activity (PCEEA) algorithm
algorithms,owingtotheireaseofuse.Inthisworkwegoa
that is described in detail by Guenther et al. [2006] and
stepfurtherbyimplementingtwodifferentisopreneemission
GuentherandWiedinmyer[2007].Hereg iscalculatedby
inventoriesintoGEOS‐Chem,toquantifytheirimpactonthe CE
subsequentHCHOcolumndistributions.Thefirstinventory (cid:1) ¼(cid:1) (cid:2)(cid:1) (cid:2)(cid:1) ; ð2Þ
weuseisMEGAN[Guentheretal.,2006],whichiscanopy‐ CE T PAR LAI
scale model. The second inventory is a leaf‐scale emission where g , g and g are activity factors to account for
T PAR LAI
algorithmdevelopedbyArnethetal.[2007a,2010]whichis theeffectofcurrentandpastvariationsintemperature,light
coupled to the Lund‐Potsdam‐Jena General Ecosystem and leaf area on the emissions for the ‘whole’ canopy
Simulator (LPJ‐GUESS) dynamic vegetation model [Smith environment.
et al., 2001]. The main difference between these models is [21] 2.Thehybridalgorithmmodelusesacombinationof
thatMEGANisbasedonextrapolatingavailableleaf/branch the new parameterizations of Guenther et al. [2006] and
enclosure measurements and ecosystemobservations (using GuentherandWiedinmyer[2007],andsomeolderGuenther
a canopy model), to derive areal basal canopy emission et al. [1995, 1999] algorithms. This hybrid model uses an
5 of 28
D16302 BARKLEY ET AL.:TROPOSPHERIC AMAZONIAN CHEMISTRY D16302
explicit canopy model to calculate variations in light and temperaturefromtheClimateResearchUnitoftheUniversity
leafarea atfive sublayers (denotedbyl)withinthecanopy. of East Anglia (CRU, http://www.cru.uea.ac.uk/) were cor-
Here g is calculated by rected by equivalent variables from the GEOS‐5 forecast
CE
[Rieneckeretal.,2008]tomatchthechemistrysimulations.
X
(cid:1) ¼(cid:1) (cid:2) (cid:1) (cid:2)LAI; ð3Þ First, the grid points in the GEOS‐5 data set, at a longitude
CE T PAR′ l
spatial resolution of 0.667°, were matched with the corre-
l¼1;5
sponding closest grid point of the 0.5° CRU data set. Cor-
rections at each grid point were based on monthly average
where LAI is the cumulative leaf area index at layer l, and
l values for the two data sets for the period January 2005 to
gPAR′ is based on Guenther et al. [1999]. Note that (1) gT December2006.Basedontheobserveddifferencesbetween
calculated here is equivalent to the g in the PCEEA algo-
T the two data sets, a least squares fitting procedure was
rithm(equation(2))and(2)inbothapproacheswealsotake
adopted, applying a general sinus function. The GEOS‐5
into account the light dependency of monoterpenes based
(years2005–2009)andGEOS‐5adjustedCRUvalues(years
on the work of Sakulyanontvittaya et al. [2008].
1901–2004)werethenusedtofollowthestandardsimulation
[22] In standard GEOS‐Chem simulations, the basal protocolof500yearspin‐upperiodwithdetrendeddataand
MEGAN emission factors are regridded from a default
constant atmospheric CO concentration to compute equi-
0.5° × 0.5° grid to the GEOS‐Chem horizontal resolutions 2
libriumsoilandvegetationcarbonpools,followedbya20th
and driven by 3 hourly surface air temperatures (at 2 m
and early 21st century simulation of vegetation dynamics
height), and by diffuse and direct PAR from the GEOS‐5
and BVOC emissions [Sitch et al., 2003; Arneth et al.,
assimilation system. To simulate changes in vegetation we
2007a]. We call this scenario LPJ(GC). In addition, to pro-
usegriddedMODISobservationsofmonthlymeanleafarea
vide a reference point for the LPJ(GC) scenario, we also
index (LAI) made over 2000–2008 (version C5; default
conducted a simulation in which vegetation dynamics and
resolution:0.25°×0.25°)[Mynenietal.,2007].Theaverage
BVOCemissionsarebasedonthedefaultCRUmeteorology
leafareaindexforvegetatedareaswithineachgridcell,LAI ,
v alone,denotedLPJ(CRU).
is estimated by dividing the LAI by the fraction of the cell
covered by vegetation following the approach of Guenther 3.3. Comparison of MEGAN and LPJ‐GUESS
et al. [2006].
[26] Figures 2 and 3, which show the monthly mean
3.2. LPJ‐GUESS Emissions emissionsmapsfor2006andthetimeseriesofthemonthly
totals over 2005–2008, respectively, illustrate there are
[23] LPJ‐GUESSisadynamicglobalvegetationmodeling
substantial spatial and temporal differences in the BVOC
framework [Arneth et al., 2010] that combines process
emissions predicted by the MEGAN and LPJ‐GUESS
descriptions for plant and soil carbon and water cycling of
inventories.GiventhelargeuncertaintiesinmodelingBVOC
LPJ [Sitch et al., 2003] with explicit formulation of vege-
emissions [Arneth et al., 2008], we have to assume that all
tation resource competition for light and water and succes-
these estimates are plausible. The annual totals given in
sionaldynamics[Smithetal.,2001].Forglobalsimulations,
Table 4, show that for isoprene, the estimates lie almost
themodelrepresentsvegetationby10plantfunctionaltypes.
within a factor of 2 of one another, consistent with known
Among these, tropical forests and woodlands are simulated
uncertainties of tropical emissions [Guenther et al., 2006].
via a dynamically changing mix of the tropical broadleaf
Furthermore,themaximummonthlyemissions,whichrange
rain green and tropical broadleaf evergreen plant functional from2.4to9.7mgm2h−1during2005–2006,arealsocon-
types,aswellasunderstoreyvegetationthatcanbeeitherof
sistent with previously reported measurements [see, e.g.,
the C3 or C4 photosynthetic pathway.
Kuhn et al., 2007, and references therein]. The discrepancy
[24] Emissionsofisopreneandmonoterpenesarecalculated
betweenthemonoterpeneannualtotalsismoreconsiderable;
followingArnethetal.[2007a]andSchurgersetal.[2009a].
uncertainties in the basal emission rates and algorithm dif-
Leafproductionoftheseterpenoidsislinkedtophotosynthetic
ferencesarethelikelyorigin.Forexample,owingtothelack
electronflow,reflectingtheirmetabolicpathway[Niinemets
of storage, monoterpene emissions from broadleaf trees
et al., 1999]. For monoterpene emissions from broadleaf
onlyoccurduringthedayintheLPJ‐GUESSmodel,whereas
andherbaceousplantfunctionaltypes,LPJ‐GUESSassumes
inMEGANtheemissionsarecontinuous.
a production‐driven emission pattern, with little or no con-
[27] Focusing on isoprene, we find the MEGAN hybrid
tribution from leaf storage pools [Kuhn et al., 2004; Bäck
and PCEEA algorithms (not unexpectedly) produce very
etal.,2005;Schurgersetal.,2009b].Theshort‐termvaria-
similar emissionsthroughouttheyear,with veryhighemis-
tion of BVOC emissions is thus driven by variation in tem-
sionsinthedryseason(broadlyAugust–November)owingto
peratureandlight,relativelysimilartothatdescribedbymore
higherlightlevelsandslightlywarmertemperatures[Barkley
empiricalalgorithmssuchasMEGAN[Arnethetal.,2007b],
etal.,2008]. Typically,MEGANpredictsthehighestemis-
whereastheoverallemissiontotalsandseasonalpatternsare
sions along the Brazilian border with Peru and Bolivia,
also greatly influenced by variation in leaf area index and
owing to a large percentage of high emitting species (e.g.,
overall vegetation productivity, and PFT composition. As
bamboo forest) [Barkley et al., 2008]. Isoprene emissions
LPJ‐GUESSoperatesonadailytimestep,estimatesofsub-
from the MOHYCAN model [Müller et al., 2008], which
dailyvariationwerecreatedbyapplyingempiricaltempera-
also uses MEGAN, generally have a similar spatial distri-
ture and light algorithms [Guenther et al., 1995] operating
bution(aconsequenceofusingsamebasalemissionfactors)
at 3 hourly time step to the emission daily totals.
butmuchlowerannualtotals,mostlikelyowingtodifferent
[25] Since dynamic vegetation models need a spin‐up
choice of meteorology and canopy model, and through the
period,themeteorologicaldriverslight,precipitationandair
6 of 28
D16302 BARKLEY ET AL.:TROPOSPHERIC AMAZONIAN CHEMISTRY D16302
Figure2. Monthlymeanisopreneemissionsfor2006(inmgisoprenem−2h−1)calculatedbytheMEGAN
hybridandPCEEAalgorithmsandtheLPJ‐GUESSmodelforcedbythedefaultCRUmeteorology,LPJ
(CRU),andtheGEOS‐5meteorology,LPJ(GC).Forcomparison,themonthlymeanemissionsfromthe
MOHYCAN model [Müller et al., 2008]remapped totheGEOS‐Chemnested grid arealso shown.The
correlationoftheemissions,relativetotheMEGANhybridmodel,areshowninset.
7 of 28
D16302 BARKLEY ET AL.:TROPOSPHERIC AMAZONIAN CHEMISTRY D16302
Figure 3. Time series of the total monthly isoprene emissions (in Tg C) calculated by the MEGAN
hybrid (solid red line) and PCEEA (solid blue line) algorithms and the LPJ‐GUESS model forced by
the default CRU meteorology (solid green line) and the GEOS‐5 meteorology (solid purple line). The
purple and green dashed and dotted lines represent the LPJ emissions from tropical broadleaf evergreen
and rain green trees, respectively. For comparison, the monthly totals calculated by the MOHYCAN
model [Müller et al., 2008] are also shown (solid black line).
inclusionofthesoilmoistureactivityfactor(whichwesetto with higher emissions in the dry season and fairly constant
unity in equation (1)). Therefore, as Figure 2 clearly illus- emissionsotherwise.
trates,thewayMEGANisimplementedcanproducesizeable
differences between emission estimates, even on a regional 4. Model Evaluation
basis.
[28] The LPJ‐GUESS model produces lower emissions [29] Our validation strategy is based on using a combina-
tion of surface, aircraft and satellite observations to assess
everywhere compared with the hybrid/PCEEA estimates,
model performance. Each of these observing platforms are
though when forced with the CRU meteorology the model
representative of differing spatial and temporal scales and
yields monthly and annual totals that agree well with
MOHYCAN(Figure3).IntheLPJ‐GUESSsimulations,the through combined use we can effectively relate the surface
concentrations of isoprene and its oxidation products to the
largestsourcesofisoprenearefrombroadleafevergreenand
satelliteretrievaloftheHCHOverticalcolumn.Inthiswork
rain green trees; emissions from temperate tree species and
grassesaremuchlower(about2–5%ofemissiontotal).The we focus on the observations made from two campaigns:
(1) the Tropical Forest and Fire Emissions Experiment
highest emissions tend to occur toward southeastern Brazil,
(TROFFEE) [Yokelson et al., 2007] undertaken in August–
where emissions from broadleaf rain green trees dominate
during November–June. Outside this time period emissions September2004and(2)theGuyanasAtmosphere‐Biosphere
ExchangeandRadicalsIntensiveExperimentwiththeLearjet
frombroadleafevergreensaretypicallyhigherthanemission
fromraingreentrees,owingtoadecreaseinthelatter’sleaf (GABRIEL)[Stickleretal.,2007]performedinOctober2005.
area.Althoughonacontinentalscalethetotalemissionsfrom [30] Inadditiontotheseinsituobservationsweusesatellite
eitherLPJ‐GUESSsimulationshowlittleseasonalvariation, data fromSCIAMACHY andtheOzoneMonitoringInstru-
ment(OMI)[Leveltetal.,2006]toevaluateGEOS‐Chem’s
indicatingtheAmazonisaconstantemissionsource,aclear
abilitytomodeltheHCHOcolumndistributions,fromwhich
seasonal signal is evident for tropical broadleaf evergreens
Table3. Time‐Dependent HCHO Yields perCarbon Fromthe Oxidationof Isoprenea
HCHOYieldinHighNO Conditions HCHOYieldinLowNO Conditions
x x
ChemicalScheme Midday Afternoon Midnight 5Days Midday Afternoon Midnight 5Days
MCM 0.45 0.48 0.49 0.50 0.22 0.30 0.35 0.47
GEOS‐Chem(standard) 0.37 0.40 0.41 0.42 0.22 0.29 0.32 0.38
GEOS‐Chem(Caltech) 0.33 0.38 0.39 0.43 0.23 0.29 0.31 0.40
aHCHOyieldsarecalculatedusingtheMasterChemicalMechanism(MCM)[Jenkinetal.,2003;Saundersetal.,2003]andtheGEOS‐Chemstandard
andCaltechschemes(seesection2.2)andcorrespondtotheoxidationof1ppbvofisoprenereleasedat07:00withinatropicalenvironment.High‐and
low‐NO regimesaresimulatedbyholdingNO constantat1and0.1ppbv,respectively.Theafternoonyieldiscomputedat14:00;theyieldafter5daysis
x x
takentobetheultimateyield[see,e.g.,Palmeretal.,2006].
8 of 28
D16302 BARKLEY ET AL.:TROPOSPHERIC AMAZONIAN CHEMISTRY D16302
Table4. TotalAnnualIsopreneandMonoterpeneEmissionsFrom restrictourmodel‐satelliteanalysistotheyear2006tosave
the Amazon Nested Grid for 2005–2008 as Calculated by the computationaltime;2006waschoseninpreferenceto2005,
MEGANHybridandPCEEAAlgorithmsandtheLPJ‐GUESS owingtothestrongdroughtthatoccurredin2005[Zengetal.,
Model Using the GEOS‐5 Meteorology and the Original CRU 2008].
Meteorologya [31] To determine the relative importance of model pro-
cesses on the surface and column concentrations we per-
AnnualEmissions(TgC)
formed a comprehensive set of sensitivity simulations with
Isoprene Monoterpenes
various key parameters changed or adjusted, as outlined in
Algorithm 2005 2006 2007 2008 2005 2006 2007 2008 Table5.Boundaryconditionsforeachsimulationwerepro-
Hybrid 174 154 152 135 32 30 30 28 videdseparatelyforthetwoGEOS‐Chemchemicalmechan-
PCEEA 162 140 138 119 31 29 29 27 isms, owing to the different number of reactive species and
LPJ(GC) 73 75 69 73 5 5 5 5 transported tracers within each scheme. Scenarios using the
LPJ(CRU) 102 90 – – 7 6– –
same chemical mechanism, but with different parameters
MOHYCAN 102 100 – – – – – –
changed(e.g.,emissions)usedthesameboundaryconditions,
aLPJ(GC), LPJ‐GUESS model using the GEOS‐5 meteorology; LPJ toensureidenticaledgeconstraintsandtosavecomputational
(CRU), LPJ‐GUESS model using the original CRU meteorology. For time.Thespin‐upperiodforeachsimulation,irrespectiveof
comparisonthetotalisopreneemissionscalculatedbytheMOHYCAN
thechemicalscheme,originatedfromthesameinitialization
model [Müller et al., 2008] are also included. Methylbutenol (MBO)
emissionsarenegligible(<0.1TgC). timerelevanttoeachstudyperiod.Inthisanalysiswedefine
our arbitrary baseline simulation, S , to which other sce-
FM
narios are compared, as that using the standard chemical
atransferfunctionisderivedtoinferthetop‐downisoprene mechanism, full boundary layer mixing and with isoprene
emissions[Barkleyetal.,2008].Detailsoftheseinstruments emissions based on the MEGAN hybrid algorithm.
and their respective retrieval algorithms are provided in the
4.1. Comparison With TROFFEE Observations
auxiliary material. The three most important distinctions
between the sensors (relevant here) is the their coverage, [32] The ground‐based component of the TROFFEE
campaigntookplaceapproximately60kmNNWofManaus
overpass time, and ground pixel size. SCIAMACHY has a
groundpixelsizeof60×30km2anda10:00localoverpass in Central Amazonia. Measurements were performed on an
instrumented flux tower (Z14: 2.6°S, 60.2°W, 55 m height)
time, with global coverage achieved at the equator every
6 days. OMI has a pixel size which ranges 14 × 26 km2 between 14 and 29 September 2004. The tower was sur-
to28×160km2(dependingonviewinggeometry),a13:30 rounded by a forest of canopy height 30 m and with an
average leaf area index of about 5–6. A proton transfer
local overpass time and can achieve global coverage in a
reactionmassspectrometry(PTR‐MS)instrumentsituatedon
single day. Since OMI has a higher spatial resolution and
thetowerwasusedinconjunctionwiththeeddycovariance
bettertemporalcoveragethanSCIAMACHY,ithasalower
technique to infer the surface fluxes of isoprene and mono-
measurement uncertainty owing to improved sampling sta-
terpenes, and concentrations of isoprene and the sum of its
tistics and reduced cloud contamination. In this work we
Table5. Description of the GEOS‐ChemSensitivity Simulations
IsopreneEmissions
Chemical BoundaryLayer OH SoilNO
x
Scenario Mechanism Model Scaling MixingScheme Recyclinga Scalingb
S standard MEGAN(hybrid) 1.000 fullmixing no 1.00
FM
S standard MEGAN(hybrid) 1.000 nonlocal no 1.00
NL
S standard MEGAN(hybrid) 1.000 fullmixing no 1.25
FM(NOx)
S standard MEGAN(hybrid) 0.635 fullmixing no 1.00
FM(rIE)
S standard MEGAN(PCEEA) 1.000 fullmixing no 1.00
FM(PCEEA)
S standard MEGAN(hybrid) 1.000 fullmixing yes 1.00
FM(OH)
S standard MEGAN(hybrid) 1.000 nonlocal yes 1.00
NL(OH)
S standard LPJ(CRU) 1.000 fullmixing no 1.00
FM(LPJ(CRU))
S standard LPJ(GC) 1.000 fullmixing no 1.00
FM(LPJ(GC))
S standard MEGAN(hybrid) 0.635 fullmixing yes 1.00
FM(rIE,OH)
C Caltech MEGAN(hybrid) 1.000 fullmixing no 1.00
FM
C Caltech MEGAN(hybrid) 1.000 nonlocal no 1.00
NL
C Caltech MEGAN(hybrid) 1.000 fullmixing no 1.25
FM(NOx)
C Caltech MEGAN(hybrid) 0.635 fullmixing no 1.00
FM(rIE)
C Caltech MEGAN(PCEEA) 1.000 fullmixing no 1.00
FM(PCEEA)
C Caltech MEGAN(hybrid) 1.000 fullmixing yes 1.00
FM(OH)
C Caltech MEGAN(hybrid) 1.000 nonlocal yes 1.00
NL(OH)
C Caltech LPJ(CRU) 1.000 fullmixing no 1.00
FM(LPJ(CRU))
C Caltech LPJ(GC) 1.000 fullmixing no 1.00
FM(LPJ(GC))
C Caltech MEGAN(hybrid) 0.635 fullmixing yes 1.00
FM(rIE,OH)
aSeereaction(R1).
bScalingfactorisonlyappliedtosoilNOemissions;fertilizeremissionsareunscaled.
9 of 28
D16302 BARKLEY ET AL.:TROPOSPHERIC AMAZONIAN CHEMISTRY D16302
HCHOVCD(10:00LT)−162×10cm) b2.062.572.572.532.592.592.451.831.782.142.162.122.122.182.061.591.59 Dstandsfor %;onlyfour
( C 0
V 4
≤
nd er
MeanOH–(10:0014:00LT)−63(×10cm) –0.480.371.701.430.510.560.731.990.740.503.592.750.810.961.414.07 elationcoefficient,a cloudfractionalcov
r –0.140.32−0.060.100.130.120.09−0.120.030.13−0.07−0.150.020.000.02−0.09 Pearsoncorr olumnswith
aCampaign CRVMR PPA(%) –−14−10−46−36−14−23−49−68−35−29−64−52−34−43−62−77 (5)),risthe usingonlyc
ROFFEE MVK+MA B(%) –758362375533−373550−32−83513−25−56 (equation calculated
sFromtheT Mean±1SD(ppbv) 2.43±1.084.24±0.594.44±0.572.57±0.322.98±0.294.23±0.583.71±0.482.49±0.331.53±0.213.28±0.503.63±0.501.65±0.342.22±0.443.28±0.492.75±0.371.83±0.271.03±0.23 peakaccuracy eptember2004
n d S
Observatio r –0.140.13−0.010.110.150.240.130.010.130.140.060.300.140.220.090.07 isthepaire –over1724
A
‐dBased MR PPA(%) –−1968−5516−21−38−57−73−3256−70−8−35−51−67−82 (4)),PP averaged
MixingRatiosWithGroun IsopreneV Mean±1SDB(ppbv)(%) –4.60±2.737.86±1.767115.12±3.89228−3.93±1.29159.33±2.981037.59±1.73655.95±1.2029−3.99±1.0313−2.25±0.82515.97±1.653013.05±3.87184−2.31±1.00506.53±2.49425.65±1.6223−4.16±1.1410−2.66±0.9442−1.37±0.6570 ormalizedmeanbias(equation ACHY[DeSmedtetal.,2008]
d n M
nsan r –0.810.810.810.810.810.810.810.810.810.810.810.810.810.810.810.81 isthe SCIA
Table6.ComparisonofModeledSurfaceEmissio IsopreneEmissions BPPAMean±1SD−−21Scenarioh)(%)(%)(mgm ––Observations2.12±2.82−S2.80±2.913231FM−2.80±2.913231SNL−2.80±2.913231SFM(OH)−2.80±2.913231SNL(OH)−2.80±2.913231SFM(NOx)−2.22±2.46545SFM(PCEEA)−−1.78±1.851656SFM(rIE)−−1.78±1.851656SFM(rIE,OH)−2.80±2.913231CFM−2.80±2.913231CNL−2.80±2.913231CFM(OH)−2.80±2.913231CNL(OH)−2.80±2.913231CFM(NOx)−2.22±2.46545CFM(PCEEA)−−1.78±1.851656CFM(rIE)−−1.78±1.851656CFM(rIE,OH) aKarletal.[2007].VMRisthevolumemixingratio,Bverticalcolumndensities.bThemeanHCHOverticalcolumndensityretrievedbyobservationsmatchedthisselectioncriteria.
10of28