Jun 27, 2001 - tropospheric NO2 for cloud-free conditions over the eastern United States and western. Europe represent ... GOME [Burrows et at., 1991, 1993, 1999] is a small scale version of the ... addition, two three-dimensional global chemistry transport ... standing of atmospheric radiative transfer, in particular, in the.
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 106, NO. D12, PAGES 12,643-12,660, JUNE 27, 2001
Global tropospheric NO2 column distributions' Comparing three-dimensional
model calculations
with
GOME
measurements
GuusJ.M. Velders, 1'2ClaireGranier, 2'3'4 Robert W. Portmann, 2KlausPfeilsticker, 2'5
MarkWenig, 5Thomas Wagner, 5UlrichPlatt, 5Andreas Richter, 6and JohnP. Burrows6 Abstract. TroposphericNO2 columnsderivedfrom the dataproductsof the Global Ozone MonitoringExperiment(GOME), deployedon theESA ERS-2 satellite,havebeencompared with modelcalculationsfrom two globalthree-dimensional chemistrytransportmodels, IMAGES and MOZART. The main objectivesof the studyare an analysisof the tropospheric NO2 dataderivedfrom satellitemeasurements, an interpretationof it andevaluationof its qualityusingglobalmodels,andanestimation theroleof NO2in radiativeforcing.The measuredand modeledNO2 columnsshowsimilarspatialand seasonalpatterns,with large tropospheric columnamountsoverindustrialized areasandsmallcolumnamountsover remoteareas.The comparisonof the absolutevaluesof the measuredandmodeled tropospheric columnamountsareparticularlydependent uponuncertainties in the derivation of thetropospheric NO2 columnsfrom GOME andthe difficultyof modelingtheboundary layerin globalmodels,bothof whicharediscussed below.The measured tropospheric columnamountsderivedfrom GOME data are of the sameorderas thosecalculatedby the MOZART model over the industrializedareasof the United Statesand Europe, but a factor of 2-3 largerfor Asia.The modeledtropospheric NO2columnsfromMOZART aswell asthe columnamountsmeasuredby GOME arein goodagreementwith NO2 columnsderivedfrom observed NO2 mixingratiosin theboundarylayerin easternNorthAmerica.The comparison of the modelsto the GOME dataillustratesthe degreeto whichpresentmodelsreproducethe hotspotsseenin theGOME data.The radiativeforcingof NO2hasbeenestimated fromthe calculatedtropospheric NO2 columns.The localmaximain theradiativeforcingof tropospheric NO2 for cloud-freeconditionsovertheeasternUnitedStatesandwestern
Europe represent 0.1-0.15 W m-2,whilevalues of0.04-0.1 W m-2areestimated ona
continentalscalein theseregions,of the sameorderof magnitudeasthe forcingof N20 and somewhatsmallerthanthe regionalforcingof tropospheric ozone.The globallyaveraged
radiative forcing oftropospheric NO2isnegligible, --0.005 W m-2. 1.
Introduction
troposphere [e.g., Chameideset al., 1992]. TherebyNO2 participatesin the control of the strongoxidant, 03, and the Nitrogen dioxide (NO2) plays a key role in tropospheric strongestatmosphericoxidizing agent, OH, and plays an chemistry.For example,its photolysisleadsto the formation importantpart in determiningthe oxidizingcapacityof the of ozone(03) and its reactionwith the hydroxylradical (OH) atmosphere [Loganet al., 1981]. Recently,it hasbeenrecogproducesnitric acid (HNO3). Theseprocesses are of impor- nized that NO2 contributes both directlyandindirectlyto the tance in both the planetary boundary layer and free radiative forcing of climate [Solomon et al., 1999]. In addition,it is known that in high concentration, NO2 causes respiratory problems for humans [e.g., Environmental Pro1AirResearch Laboratory, National Institute ofPublic Health and tectionAgency (EPA), 1998]. the Environment(RIVM), Bilthoven,Netherlands. As both the atmosphericsourcesand sinks of NO2 are 2Aeronomy Laboratory, NOAA,Boulder, Colorado. 3Serviced'Adronomie, CentreNationalde la Recherchestronglynon homogeneouslydistributed,the concentrationof Scientifique,Paris, France.
NO2 in the troposphere is highlyvariable,both spatiallyand temporally [e.g., Emmons et al., 1997]. In the boundarylayer Universityof Colorado,Boulder, Colorado. aboveindustrializedareas,valuesup to severaltensof ppbv 5Institut ftirUmweltphysik, University ofHeidelberg, Heidelberg,
4Cooperative Institute for Research in Environmental Science,
Germany.
are observed having a strong diurnal variation, whereas in
6Institute of Environmental Physics, University of Bremen,remote areasthe concentrationsare typically in the range of
Bremen,Germany.
Copyright2001 by the AmericanGeophysicalUnion. Papernumber2000JD900762. 0148-0227/01/2000JD900762509.00
10-200 pptv. Prior to the advent of space-basedremote sensinginstrumentation suchas the Global OzoneMonitoring Experiment (GOME), and because of its high intrinsic variability and consequentsparseset of in situ or remote sensingmeasurements,global knowledge about the amount 12,643
12,644
VELDERS ET AL.: GLOBAL TROPOSPHERIC NO2 COLUMN DISTRIBUTIONS
anddistributionof NO2 hasbeendifficult if not impossibleto obtain. In several recent studies [Horowitz and Jacob, 1999;
Levy et at., 1999; Penner et al., 1998; Wang et at., 1998; Kraus et at., 1996; Lamarque et at., 1996], chemistrytransport models have been used to calculatethe global threedimensionaldistributionof troposphericNO,• and study its sources,sinks,and atmospherictransport. GOME [Burrows et at., 1991, 1993, 1999] is a small scale
versionof the ScanningImagingAbsorptionSpectrometer for Atmospheric Chartography(SCIAMACHY) [Bovesmannet at., 1999] observingbetween232 and 793 nm and only in nadir viewing geometry.It measuresthe radianceupwelling from the atmosphereand the extraterrestrialsolar irradiance. Appropriateinversion of these data productsyields global informationaboutthe atmosphericabsorptionand scattering processesenabling the abundanceof trace constituentssuch as O3 and NO2 to be retrieved.
andobserved(GOME) NO2 columnamountsare compared. Thecontributions of thedifferentatmospheric modellayersto the total troposphericNO2 columnamountsand the different factorsaffectingthe simulatedNO2 columnsare discussedin section5, followedby a comparisonof the modeledNO2 data with observations in section6. The effectof theNO2 columns on the radiativeforcingof climateis probedin section7, and section8 summarizesthe major conclusions to be drawnfrom our study.
2. Tropospheric NO: Columns Retrieved From GOME
In this section, the GOME instrument and its measure-
mentsare described,the retrievalmethodsappliedto the GOME observations arediscussed, andtheresultant monthly averaged tropospheric NO2 verticalcolumndensities(VCDs) for 1997 are presented. As a largepart of the tropospheric
The GOME measurements havebeenusedin this studyto determine the NO2 troposphericcolumn distributionon a global scale.The retrieval of theseunique data is just now
NO2 column over polluted areasis believedto be locatedin the boundarylayer, the sensitivityof GOME data to uncer-
possiblewith the measurementsof the GOME instrument.In
taintiesin the retrievalof NO2 in the boundarylayer is
addition, two three-dimensional global chemistrytransport consideredand an estimateof the total tropospheric NO2 models IMAGES (Intermediate Model for the Global Evolu- VCD made,which accounts for cloudscoveringits source tion of Species) [Miitler and Brasseur, 1995; Pham et at., regionsduringtheGOME overpasses. 1995; Granier et al., 1996, 1999, 2000a, 2000b] and Tropospheric NO2 plumestypicallyoccuron regionaland MOZART (Model for Ozone and Related ChemicalTracers) continental scales,asshownin Plates1c and2 wherelocally [Brasseuret at., 1998; Haugtustaineet al., 1998] have been large amountsare observedover Europe, easternNorth used to investigatethe retrieved troposphericNO2 column America,partsof Asia, and the Johannesburg metropolitan amountsand help to interpretit. Both modelsuse the EDGAR areain SouthAfrica.More than90% of thetropospheric NO2 emissiondatabase[Otivier et at., 1996] for anthropogenic is believed to be emitted into continentalair, while the emissionsand the databasedevelopedby Miitter [1992] for remaining10%is generated overtheoceanby lightning,and natural emissions. The main differences between these models shipor aircraftemissions [e.g., Vatksand Vetders,1999].In are the sourceof the atmospheric wind fields(monthlymean contrast,stratospheric NO,•resultslargelyfrom the reactionof climatologicalfields for IMAGES and 6 hourlyfieldsfrom a nitrousoxide (N20) with excitedoxygenatoms.In addition, globalcirculationmodel for MOZART) and the representa- the relativelyshortlifetimeof NO,•in the troposphere, comtion of the boundary layer. MOZART has a less diffusive paredto that in the stratosphere, leadsto its tropospheric boundarylayer than IMAGES, resultingin a more realistic distributionappearingas plumes emanatingfrom source representation of the physicaland chemicalprocesses in the regions. lowestlayersof the model. Over populatedregionsof the continents the majorityof In this decade a number of satellite instruments will be
the NO2 column is released from and believed to be located
launchedinto low Earth orbit whoseprimaryobjectiveis to primarilyin the boundarylayer,with smalleramounts being investigatethe trace gas distributionsin the lower tropo- transported to or generatedin the free troposphere andstratosphere,e.g., SCIAMACHY on ENVISAT [Bovesmannet at., sphere.In the remote maritime atmospherethe only 1999, and references therein], GOME-2 on METOP, Ozone
Monitoring Instrument (OMI), and TroposphericEmission Spectrometer(TES) on EOS-CHEM. An adequateunderstandingof atmospheric radiativetransfer,in particular,in the boundarylayer, is requiredto interpretaccuratelythe slant columnamountsfrom thesenadirsoundinginstruments. In this studythe global distributionof tropospheric NO2
significant sourceof tropospheric NO2 is in-cloudlightning andtransport fromthecontinents. Analysisof bothmodelsof thestratospheric NO2behavior andtheUARSHALOENO2
dataproducts revealsthat the NO2 columnin the upper atmosphere hassignificant longitudinal homogeneity up to andbeyond latitudes of 65ø for thesamesolarzenithangle. Thisis explained by thefactthatthemajority of thisNO2is columnamountsretrievedfrom GOME level 1 dataproducts locatedin the middleand upperstratosphere, whereit is have been comparedwith NO2 columnamountsinferredfrom relatively unperturbed bylocalchanges in thetroposphere, but three-dimensional chemistrytransportmodels.Investigatingstronglyinfluencedby the solarzenithangle.In addition, the similaritiesand differencesin the magnitude,global advective mixingof NOrat suchaltitudes is relatively rapid distribution, and seasonal variations of the retrieved and compared to therateof itschemical production or loss. simulatedNO2 providesnovelinsightinto theglobalbehavior The assumptions that the upperatmospheric columnof of NO,•. Importantly,the GOME tropospheric NO2 columns NO2 is longitudinally homogeneous and thattropospheric
reveal "hotspots" of significance for the assessmentof regional,transnational, andintercontinental air pollution.
The algorithmsused to retrieve NO2 columnamountsfrom GOME
observations are discussed in section 2. Section 3
amounts of NO2 above the clean remote ocean are small
providethe physicalbasisenablingthe tropospheric and stratospheric NO2columnsto be separated fromtheinversion of nadir observationsof GOME or SCIAMACHY. Two
describes the chemistry transport models IMAGES and relatedbut independentapproaches have been usedin this MOZART. In section4 modeled(IMAGES and MOZART) study. TheybothutilizetheGOMElevel1 dataproducts: the
VELDERS ET AL.' GLOBAL TROPOSPHERIC NO2 COLUMN DISTRIBUTIONS
Vertical ColumnDensity,12.1.1997-18.1.1997
12,645
Estimateof the Stratosphere,12.1.1997 - 18.1.1997
o
•--• -?•
.... _.• ,•
-
, Estimate of the
•,_•,. • Stratosphere ,J j .
2 I• 3 4 Vertical Column NO2[10•$molec/cm2]
1.
5
0
1
2
I•
3
4
VerticalColumnNO2[10•smolec/cm2]
Calculation ofthe Difference
II
5
b
' --"lJ
0
4
C
8
12
16
20
Vertical Column NO 2[10 •$molec/cm2]
Plate 1. Flowchartof the stepsin the algorithmto separatethe stratospheric and tropospheric partsof the GOME NO2 column. (a) the land massesand cloudy pixels are maskedout and (b) the stratosphereis estimatedusinga normalizedconvolution.(c) The tropospheric residualis thenestimatedby calculatingthe differencebetweenPlates1a and 1b. Plate 1c showsthe GOME troposphericNO2 columnwith the correctair massfactorsdepending on solarzenithangle,groundalbedoandclouds.Notethedifferentscalesof themaps.
top of the atmosphereradiance (TOARAD), observedfrom the GOME viewing geometry,and the extra-terrestrialsolar irradiance(ETSIR).
Different sectionsof this spectrumare then directed toward the four spectral channels of GOME: each channel comprisinga grating,transmissiveopticsand a 1024 element diode array. In this manner the entire spectrumbetween232 2.1. GOME Instrument and Observations and 793 nm may be observed:the spectralresolutionbeing The GOME instrument, on board the second European -0.2 and 0.33 nm, below and above400 nm respectively. Research Satellite (ERS-2) comprises a scanning mirror, As GOME is optimizedfor the collectionof the upwelling spectrometer,thermal and electronicsubsystems[Burrowset radiationfrom the atmosphere directlyby the scanmirror,it is at., 1991, 1993, 1999]. Light from the atmosphereis collected necessaryto direct the extra terrestrial solar output over a by the scan mirror then focusedon the entranceslit of the diffuserplateto reduceits intensityprior to it beingreflected spectrometerby an off-axis parabolictelescope.After being by the scanmirrorinto the instrument.In thismannerboththe collimated, the light passesthough a predispersingprism upwelling radiance and the extraterrestrialirradianceare forming an intermediatespectrumwithin the instrument.The measuredby GOME. The signals,recordedby GOME, are predispersingprism is suchthat internalreflectionresultsin a dumpedto the ESA groundstationat Kiruna.Thesedataare polarized beam, which is directed towards the polarization processedfor ESA by the GOME Data Processorat the monitoring device (PMD). The latter comprise three German Atmosphericand Space ResearchProcessingand broadbanddetectors,which observeupwelling radiationfrom ArchivingCenter(DLR-PAC): the resultantlevel 1 products, the atmospherepolarized relative to plane within the instru- TOARAD and ETSIR, and level 2 total ozone column data ment in the wavelengthranges300-400 nm, 400-600 nm, and product are subsequentlybeing distributedto the scientific 600-800 nm, respectively.The main beam from the predis- community. ERS-2 was launchedon the April 20, 1995, into a sun persing prism forms a spectrum within the instrument.
12,646
a) GOME: NO2 column:January1997
b) GOME: NO2 column'April 1997
ß
: .....
:
Max USA: 33.3
0
2
4
6
Max Europe:41,5
8
10
12
14
Max Asia 43 3
16
18
20
>22
c) GOME: NO2 column:July 1997
Max USA' 16.7 0
2
4
6
10
12
14
16
'
Max USA.72'1.3 Max Europe: 16,7
0
2
4
6
8
10
12
14
Max As4a: 14.7
16
18
20
>22
d) GOME: NO2 column:October 1997
Max Europe:13.0
8
'
MaxAs•a:124
18
20
>22
Max USA. 18 4
0
2
4
NO2 (1.0e15 molecules/cm2)
6
8
Max Europe:18 0
10
12
14
Max Asia-16 4
16
18
20
>22
NO2 (1.0e15 molecules/cm2)
Plate 2. Monthly averagedtroposphericNO2 columnsas derivedfrom GOME measurements for (a) January, (b) April, (c) July, and (d) October 1997. The measurements are made around 1030 LT. The grid shown is 0.45ø x 0.45ø and derived from both cloudy and cloud free pixels.Blank pixels indicatethat there is no data. The maximum valuesthat occurover the United States,Europe,and Asia in a singlegrid cell are given below the plots. The maximumGOME NO2 valueson the IMAGES grid are (21.9, 30.6, 17.8) for (United States, Europe,Asia) for January:(13.1, 11.0, 6.5) for April, (12.2, 8.7, 7.7) for July, and (12.6, 9.8, 7.9) for October
(allin 10Ismolecules cm-2). a) IMAGES:NO2 column:January
b) IMAGES:NO2 column:April ..
.... ?.
ß i0.0
0.5
1,0
1.5
2.0
Max Europe:7.6 2.5
:
Max USA: 4,6 1.0
,
-:........
.
.................
0.5
, ......
..
3,0
3.5
4.0
MaxAsia:3.6 4.5
5.0 >5.5
c) IMAGES:NO2 column:July
0.0
......
:'a,........ 'i:i.•,,.: ' ' '
Max USA:7.7
[_
.•
1.5
2.0
....•
.
3.0
.
3,5
4.0
Max Europe.5.7
Max Asia:2.2
0.0
05
1.0
1,5
2.0
2.5
30
3,5
4.0
4.5
5.0 >5,5
d) IMAGES: NO2 column: October
MaxEurop•: 51 2.5
Max USA:5.6
1
.:.
Max As,a: 2.0 4.5
5.0 >5.5
NO2 (1.0e+15 molecules/cm2)
Max USA: 6.6
0.0
0.5
1.0
1.5
2.0
Max Europe:7.2
2.5
3.0
3.5
4.0
Max Asia:2.6
4.5
5.0
>5.5
NO2 (1.0e+l 5 molecules/cm2)
Plate3. Monthly averaged tropospheric NO2columns ascalculated byIMAGESfor(a)January, (b)April,(c) July,and(d)October. Datacorrespond to--1030LT. Themaximum values thatoccur overtheUnitedStates, Europe,andAsiain a singlegridcell aregivenbelowtheplots.Notethecolorscaleis 4 timessmallerthan
thescale used fortheGOMEplotsandthelowest level(0-0.5)hasbeenleftblanktoobtain clearer plots.
VELDERS ET AL.: GLOBAL TROPOSPHERICNO2COLUMN DISTRIBUTIONS synchronous orbit; having an equatorcrossingtime of 10:30 am in a descendingnode.The scanstrategyof GOME yields completeglobal coverageat the equatorin 3 days:the swath
Image-processing techniques(lFr)
12,647
applied to 3-day
compositeimages of the GOME NO2 observationhas been
usedextensively in thisstudy.Plate1 illustrates schematically
widthofGOMEbeing320x 40km2forthearraydetector and the individual IPT stepsused to discriminatebetween the
20 x 40 km2for thePMDs,whicharereadout16 timesfaster stratospheric and troposphericcontributionto the total NO2 than the arrays. More details about the design and the first resultsfrom GOME are providedelsewhere[Burrowset al.,
VCDs. For the marineregionsthe following procedurewas used.First, the continentsplus a band -200 km off shoreis 1999, and referencestherein]. maskedout to avoid regionsof NO2 emissionson and near coastlines.The stratosphericcolumn of NO2 VCDs is 2.2. Retrieval Algorithm for Tropospheric NO2 assumedto dominate the resultant region. A normalized convolution is then applied to the data (see dtihne et al. Using the differential optical absorption spectroscopy [1999] for details), enablingthe global stratospheric NO2 (DOAS) technique[e.g., Platt, 1994], it has previouslybeen shownthat the total column amountsof ozone (03) [Burrows distributionto be inferred.For theIPT algorithm,marineNO2 et al., 1998c] brominemonoxide(BrO) litegels et al., 1998; measurementsfor clear sky conditions were selected for Richter et al., 1998; lYagner and Platt, 1998], chlorine dioxide (OC10) [Burrowset al., 1998c;Eisinger et al., 1996; lYagner et al., 1999; Burrows et al., 1999], formaldehyde (HCHO) [Burrowset al., 1999] sulfurdioxide (SO2) [Eisinger and Burrows, 1998, and referencestherein], nitrogendioxide (NO2) [Burrows et al., 1998c; Leue et al., 1999], and water vapor(H20) [Noel et al., 1999] maybe retrievedfrom GOME irradiance
and radiance
measurements.
In addition
ozone and
aerosolprofiles may also be retrieved[1toogenet al., 1999, and referencestherein].
whichradiativetransportcalculations canbe performedvery precisely.For this selectionthe cloud fractionis derived from
theonboard-operated PMD instrument asdescribed by Wenig et al. [1999].
Plate lb showsthe "IPT stratospheric background"thus derived. The resulting troposphericNO2 VCDs were then obtained from the differencesin total atmosphericand inferredstratospheric NO2VCDs. A moredetaileddescription of the IPT retrievalof troposphericNO2 VCD from the total atmosphericNO2 can be found in the works of Leue [1999], Leueet al. [ 1999, 2001], andRichterand Burrows[2000]. In the IPT algorithm, the difference between the total (SCDtotal)and the stratosphericNO2 slant column densities
In this study the DOAS spectralwindow chosenfor NO2 retrieval from GOME is 425-450 nm [lYagner, 1999]. Preflight measurementsof the absorptioncross section of is attributedto tropospheric NO2.The tropospheric NO2, O3 [Burrows et al., 1998a, 1999], H20, and O4 and (SCDstrat) NO2 VCDs are then computed: calculationsof an effective Ring spectrum[e.g., Bussemer, 1993; Fountaset al., 1998] serve as input for the retrieval. VCDtrop = (SCDtotalSCDstrat) / AMFtrop ß The NO2 slant column densitiesderived by DOAS are conThe tropospheric air massfactors(AMFtrop) werederived verted into total atmosphericNO2 vertical column densities (VCDs) by applying a calculatedair mass factor (AMF) for using the radiative transfer model GOMETRAN [Rozanovet each observation.
The AMF describesthe weightedpath of the light through the atmospherecomparedto the vertical path. In general,the AMF is a function of the solar zenith angle, the selected wavelengthinterval, the vertical profile of NO2, amountsof aerosol, the surface spectralreflectanceor albedo, the cloud cover,and absorptionby the atmosphere. AMFs for this study have been calculated with the radiative transport models AMFFRAN [Marquard et al., 2000] for the total atmosphere and GOMETRAN [see Rozanovet al., 1997] for the troposphere.The albedo used for AMF calculationsin this study are thosederived usingby cloud free scenes,and an atmosphericmodel [Richter and Burrows,2000]. In order to retrieve the troposphericcolumn amountsof NO2 from GOME data, a priori knowledgeaboutthe behavior of the stratosphericNO2 is required.It is assumedthat stratospheric NO2 is longitudinally homogeneous and the troposphericNO2 in the clean remote troposphereabove the ocean is negligible. In one approach,known as the tropospheric excess method (TEM) [Burrows et al., 1999], troposphericNO2 in the clean remote troposphereover the oceans around 180ø longitude is assumedto be negligible.
al., 1997], assuming a well-mixedboundarylayerof average height1.5 km, containingthe entiretropospheric NO2 layer, an averagesurfacealbedobeing 5%, and a maritime aerosol
distribution assumed to be locatedin theboundarylayer. It shouldbe notedthat the resultsof the IPT algorithmare in goodagreement with thoseof an otherapproach, knownas the tropospheric excess method (TEM) [Burrows et al., 1998c, 1999; Richter and Burrows, 2000]. The main difference between TEM and IPT is that in the TEM algorithm cloud free pixels are selected,while in the IPT algorithmall pixels are used and cloud effects are taken into account a posteriori. Because a cloud threshold value is used in the TEM analysis, the residual cloud contaminationwill lead to an underestimation of troposphericNO2.
2.3. GOME's Sensitivity to BoundaryLayerNO2 Observations
Therelativepenetration of photons in theboundary layer, asobserved atthetopof theatmosphere, depends strongly on factors including thecloudcover,surface spectral reflectance (albedo), andtropospheric aerosol loading. Thustheweight-
ing of themeasurement sensitivity to thelowestlevelsin the Averagethreedays,weekly,monthly,andyearlyglobalmaps troposphere is a strongfunctionof the wavelength and canthenbe derived.In the TEM approach the percentage of viewing conditions. Thisresults in a variable sensitivity of the cloudcover within a sceneis estimatedusingthe higher retrievedDOAS NO2 absorption to the NO2 in the lowest spatiallyresolvedbroadband measurements by a threshold layerof theatmosphere, theboundary layer.Forexample, calalgorithm[Burrowset al., 1998b].In thismanner,composite culations using GOMETRAN show that for cloud free maps,typicallybetween60øN and60øS,showingthe tropo- conditions andoverhead sunroughly 20%of theincoming sphericcolumnof NO2for differentpercentages of cloudscan solarphotons at 437 nm arebackscattered by atmospheric
be obtained.
molecules beforereaching thesurface. Thebackscattering by
12,648
VELDERS ET AL.' GLOBAL TROPOSPHERIC NO2 COLUMN DISTRIBUTIONS is reflected in the air mass factor (Figure 1), which is small andnearlyindependentof solarzenith anglesbelow 80ø. Our AMF calculationsalso show that strongly absorbing aerosolslocated in the boundary layer further obscure the visibility of boundarylayerNO2 for space-borne observations.
1.4
1.2
1.0
•
o.8
15 0.6 0.4
0.2 0.0
0
15
30
45
60
75
90
105
Solar zenith angle (deg.)
Figure 1. Air massfactorof NO2 for a boundarylayer of 1.5 km at a wavelengthof 437.5 nm, assumingan albedoof 5% and a maritimeaerosollayer in the boundarylayer.
Multiple Mie scattering by aerosol or cloud particles, however,can also increasethe troposphericabsorptiondue to increasingopticalpath length.The AMF dependsstronglyon the surfacespectralreflectance,for example increasingthe surfacespectralreflectancefrom 4 to 8% increasesthe troposphericAMF by -30%. Optically thick clouds effectively obscure tropospheric NO2 located below the cloud top becausethey act as highly reflectingsurfaces,havingtypicallyan albedoin the range8095% [Kurosu et al., 1997]. As a result of their brightness, even a small cloud fraction within a GOME pixel may contribute significantlyto the signal receivedby the instrument, decreasingthe sensitivitytoward the detectionof boundary
layerNO2,but increasingthe relativeweightingof NO2 above aerosolvaries in the range of 1-10% and is dependenton type and distribution.As a result of this sensitivityto the surface spectralreflectance(albedo) in the blue spectralrange, -212% of the remaining 80% of the total incoming photonsare reflected from nonsnow covered land surfaces, arrd 4-8%
from the oceans [Richter and Burrows, 2000]. This behavior
2.4. Estimation of the Total NO2 Column
In this study, in order to estimatethe total tropospheric NO2 VCD all GOME observations(including also the cloudy regions)were takeninto account.This procedureenhancesthe
b) MOZART: NO2 column:April
a) MOZART: NO2 column:January
:•-
the cloud to that of the free troposphereor stratosphere.
•
•,-•.
.....
'i' ......... i Max USA 18 0.0
1.0
2.0
3.0
4.0
Max USA: 14.4
5.0
6.0
7.0
8.0
90
10.0>•1.0
0.0
1.0
2.0
3.0
4.0
Max Europe:13.4
5.0
6.0
7.0
8.0
0-0
1.0
2.0
3.0
4.0
Max Europe:10.7
5.0
6.0
7.0
8.0
Max Asia. 7.4
9.0
10.0>11.0
d) MOZART: NO2 column:October
c) MOZART: NO2 column:July
Max USA: 13.5
"•;....•........ ::.... :._•._:_' ...::......•........ i....../-
Max USA: 16.0
Max Asia:4.6
9.0
10.0 >11.0
NO2 (10e+ 15 molecules/cm2)
0.0
1.0
2.0
3.0
4.0
Max Europe:19.3
5.0
6.0
7.0
8.0
Max Asia:6.6
9.0
10.0 >11.0
NO2 (1.0e+ 15 molecules/cm2)
Plate 4. Monthly averagedtroposphericNO2 columnsas calculatedby MOZART for (a) January,(b) April, (c) July, and (d) October. Data shown correspondto -1030 LT. The maximum values that occur over the United States,Europe,and Asia in a singlegrid cell are given below the plots.Note the color scaleis 2 times smallerthan the scaleusedfor the GOME plots and the lowestlevel (0-1) hasbeen left blank to obtain clearer plots.
VELDERS ET AL.' GLOBAL TROPOSPHERIC NO2 COLUMN DISTRIBUTIONS
12,649
spectra)do not decreaseif severalobservations are averaged. However, by subtractingthe stratospheric columnfrom the tropospheric NO2. While, in principle,this can be done for total columnmostof the possiblesystematicerrorscancelout; individual GOME measurements,we here apply an average the remainingsystematicerror is estimatedto be _+10%.(3) correctionfor monthly averagedGOME measurementsof The mostimportantuncertaintiescomefrom the modelingof AMFs. The lack of preciseinformationabout tropospheric NO2. This approachwas chosenin order to the tropospheric reduce the complexity of the correction procedure. In the ground albedo and the troposphericNO2 height profile addition, there are still considerable uncertainties in the causeerrorsin the calculatedtroposphericNO2 VCDs which determinationof the individual cloud fractions using current can be up to -50% for a single observation.For monthly methods. averagestheseerrorspartly cancelout leaving an uncertainty The averagecorrectionappliedheremakesuseof assump- of-15%. (4) The uncertaintiescausedby cloudsconstitutethe tions of cloud fraction and albedo which are described below. largestcontributionto the error. For a single observationit They are consistentwith conditionsover NO2 sourceregions, might be up to -100%. For the monthly averagestheseerrors including in particular industrializedregions,but are consid- partly cancelout; we estimatethe remaininguncertaintyto be ered to inappropriateto other regions,e.g., desertsor snow -50%. This last uncertaintydominatesthe total error in the coveredregions.Thus the comparisonsin this studyfocus on retrievedtroposphericNO2 VCDs. total number of considered measurementsbut requires a correction
for
the
effects
of
clouds
on
the
retrieved
GOME troposphericcolumns of NO2 and the model results for regionswhere the aboveassumptionsare valid. From
GOME
measurements
it turned
out that
the total
atmosphericreflectance (including scatteringon the surface, and on aerosolsand molecules) is -25% at 437 nm, whereas the albedoabovecloudsis equalto or greaterthan 80%. Thus the geometric cloud fraction has to be weighted by these relative intensitiesto determinethe sensitivityof GOME for the troposphere(assumingthat the troposphericNO2 column is entirely below the clouds). For a geometriccloud fraction of 0.5 the relative contribution of the clear sky part of the GOME pixel to the total measuredsignalis -24%. Thus the respective tropospheric NO2 column density has to be correctedby about a factor of 4.2. For a geometric cloud fraction of 0.4 (0.6) the relative contributionof the clear sky part to the total signal is -31% (17%), resultingin correction factor of-3.2 (5.8). Assuminga Gaussianfrequencydistribution of the geometriccloud fractionsof GOME groundpixels (with a maximum at a cloud fraction of 0.5 and a standard deviationof 0.25), we then derive an averagecorrectionfactor of about4 which was appliedto troposphericNO2 derivedby the IPT algorithm,including all the cloudy pixels. No correction factorwas appliedto the Saharan,Arabian Peninsula,and Australian deserts,since these areasare almost always cloud free.
It shouldbe noted that the assumedfrequencydistribution of the cloud fractionsfor GOME groundpixels is somewhat arbitrarybut shouldprovide suitablevaluesfor many partsof the world (including,e.g., industrializedregions).Future work will yield more precise information on statisticsof GOME cloud fractions from improved GOME cloud determination algorithms. Comparingthe TEM NO2 VCD for all sceneswith those having30% cloud cover,indicatesthat the TEM NO2 VCD is typically a factor of 2 larger for the 30% case than over industrial source regions, where the majority of NO2 is emitted.These findings are in fair agreementwith the average cloudcorrectionappliedto the NO2 data derivedfrom the IPT algorithmas describedabove. The uncertainty in the troposphericNO2 columns from GOME dependsmainly on four factors.(1) The precisionof the NO2 DOAS fit, which is limited by photonand electronic noiseand can becomerelativelylarge for a singleobservation. Nevertheless,for monthlyaveragedvaluesthis error decreases to very small values (+_1%)and can thus be neglected.(2) Systematicerrorsin the NO2 DOAS fit (caused,for example, by uncertaintiesin the wavelengthcalibrationof the reference
2.5. Monthly Averaged NO2 Column Amounts for 1997 Plate 2 shows the inferred tropospheric NO2 columns (=VCDs) for the monthsof January,April, July, and October in 1997. As expected,the highest NO2 columnsoccur over industrialized areas in the Northern America, Europe, and China but hotspotshaving localizedpollutionin SouthAfrica, the ArabianPeninsula,and Japanare also clearlydiscernable. EnhancedtroposphericNO2 columns also occur in known regions of biomass burning and savannahfires in South America, Africa, and Indonesia (July and October). Much lower NO2 columns are observed for remote continental regions in Australia or the Sahara desert. The inferred troposphericNO2 columnsfor the remoteoceansare closeto zero, mainly becausefor this area the total atmosphericNO2 column was assumedto be mostly locatedin the stratosphere (see above). In Januarythe maximum NO2 columnsfor the easternUnited States,westernEurope, and easternChina are
twiceas largeas in the othermonths; i.e., 33-43x l015 moleculescm-2 versus 14-20- x l0 Is moleculescm-2 The largerwintervalueslikely resultfrom a decreased lossof NO2 by reactionwith OH (the major NOx lossprocessin the lower troposphere),confinementin the boundary layer, reduced vertical transportfrom the boundarylayer to the free troposphere,and slightlyincreasedsourcestrengths. ß
3. Description of the Models 3.1.
IMAGES
The IMAGES model [M•iller and Brasseur, 1995; Pham et al., 1995; Granier et al., 1996, 1999, 2000a, 2000b] is a
three-dimensional global chemistry transport model that calculatesmonthly averageddistributionsof some60 species. It has a horizontal
resolution
of 5 ø x 5 ø and includes
25 levels
in the verticalextendingfrom the surfaceto 50 hPa (-22 km). The dynamicsis governed by prescribedmonthly averaged transportand temperaturefields. The distributionof clouds and precipitation rates are also prescribed according to monthly averagedclimatologicalvalues.The rapid turbulent exchangesin the planetaryboundarylayer are parameterized using a diffusion coefficientdependingon the vertical gradient of potentialtemperature.The coefficientsrepresenting this diffusion coefficient are calculatedusing monthly mean temperaturedata. The chemicalschemein the model is adaptedfrom M•iller and Brasseur [1995], and its current version is given by
12,650
VELDERS ET AL.: GLOBAL TROPOSPHERIC NO2 COLUMN DISTRI]3UTIONS
Granier et al. [2000b]. The chemistry includes an explicit descriptionof nonmethanehydrocarbons chemistryincluding the formation of peroxy-acetylnitrate (PAN), importantfor the transportof reactivenitrogenfrom sourceregionsto the free troposphere[Moxim et al., 1996; Horowitz and Jacob,
averageconcentrationsof NO2 were stored, calculatedfrom the mixing ratios at every 20 min time step.The intermediate datawere not savedseparately.
1999]. The reaction rates are from DeMore et al. [1997]. The
4. Modeled NO2 Column
anthropogenic emissionsare from the EDGAR emissiondatabase [Olivier et al., 1996], which is developed in close In the Plates 3 and 4 the modeled troposphericNO2 cooperationwith the Global EmissionsInventory Activity columns (vertical column densities) calculatedby the (GEIA) of the International Global Atmospheric IMAGES and MOZART models (respectively)are shown. Chemistry/International Geosphere-BiosphereProgramme Plates3 and 4 can be comparedwith Plate 2 showingthe IGAC/IGBP [see also Granier et al., 2000b]. These emissions GOME NO2 columns.To obtaina clearerpicture,the lowest are specifiedon a 1ø x 1ø globalresolutionfor CH4,NOx, CO, levels (0-0.5or0-1x 10•5molecules cm-2)areleftblank.The and nonmethanehydrocarbons. They includeemissionsfrom modeled troposphericNO2 columns are obtained from the fossil fuel use (industry, transport,fuel production,and calculatedNO2 mixing ratios by integratingfrom the surface transmission),biofuel combustion,industrial processesand to the tropopause. solventuse, waste treatment,and agriculturalwasteburning. GOME is in a Sun synchronousorbit. It passesover the A seasonalvariationis imposedon the anthropogenic emis- equatorat 1030 local time (LT) and coversall longitudesat sionsaccordingto Miiller [ 1992]. Biomassburningemissions the equatorin 3 days.To agreeas closelyas possiblewith the are taken from Granier et al. [2000a], basedon inventoriesby local overpasstime of GOME, the NO2 columnsas calculated Hao and Liu [ 1994]. The productionof nitric oxide by light- by IMAGES correspondwith a local time between1000 and ning discharges is distributed according to the flash 1040 LT. They are derivedby calculatingthe ratio of the NO2 frequenciesobtainedfrom satellitemeasurements as reported concentrationat -1030 LT and the diurnal averageconcentraby Turman and Edgar [1982]. Total emissionof NO from tions in the full diurnal cycle of the model. The diurnal lightningis assumedto be 5 Tg N/yr. average concentrationin the middle of the month is then IMAGES calculatesa full diurnal cycle in the first 3 days multipliedby this ratio. Details about the differencebetween of the month,with time stepsof 1 hourfor the first 2 daysand the columnnear 1030 LT and the diurnal averageare discusa time stepof 0.5 hour for the third day. For the rest of the sed in section 5 and Plate 7. For the MOZART calculations month,diurnal averagevaluesare calculatedand a time step only diurnal averageconcentrationsof NO2 were stored,no of 6 hours is used. 3.2.
hourly data. To calculate the NO2 column amounts from MOZART at the GOME overpasstime, we used the results
MOZART
from the IMAGES
The MOZART-2
model is a new version of the model
describedby Brasseuret al. [1998] and Hauglustaineet al. [1998]. It is a three-dimensionalglobal chemistrytransport model which can calculate the distributionof-60 species, with a 20 min time step.It has a horizontalresolutionof 2.8ø x 2.8ø and 34 levels in the vertical extendingfrom the surface to 4 hPa (-36 km). The meteorologicalparametersneededto calculatethe advectivetransport,smaller-scale exchanges, and wet scavengingof long lived chemicalspeciesare prescribed and taken from the outputof the middle atmosphereversion of the National Center of Atmospheric Research (NCAR) communityclimate model CCM3 and are updatedevery 6 hours. The parameterizationof exchangesin the boundary layer is based on Holtslag and Boville [1993] and uses a vertical eddy flux proportional to an eddy diffusion coefficient.The latter dependson a turbulentvelocity scale and on the boundary layer height (Richardsonnumber
model to calculate the ratio between the
tropospheric NO2 columnsat 1030 LT andthe diurnalaverage columns.This ratio was then used, on a monthly basis, to calculate
the
columns
at
1030
LT
for
the
MOZART
calculations.
The NOx emissionsin the modelcalculationsare monthly or yearly averages. Diurnal variations are not taken into account.About 30% of the global NOx emissionscome from roadtransport.Especiallyin and aroundthe biggercitiesroad transport exhibits a strong diurnal variation. During the morning rush hours, larger emissionsof NO• are expected which may affect the NO2 column at the GOME overpass time (1030 LT). We have performed a calculation with IMAGES using a diurnal variationin the NO• emissionsfrom road transportaccordingto Dreher and Harley [1998]. The effect of a diurnal variation in NO2 emissions (versus a constant emission) on the troposphericNO2 column was found to be small, with very local increasesaround Los
dependent).A countergradientterm representingnonlocal Angeles of-3%, in the eastern United States 1-5%, transportassociatedwith dry boundarylayer convectionis northwesternEurope 1-4%, and Japan- 1% (annualaverages). also taken into account.
The chemical scheme in the model is the same as the one
used in IMAGES [Granier et al., 2000b], with reaction rates from DeMore et al. [ 1997]. The emissionsof CH4, CO, NOx,
4.1. Grid
Sizes and Method
of Calculation
Before comparing the modeled NO2 columns with the GOME measurements,the effect of the different grid sizes section 3.1).Theproduction of NO by lightning is dfstributedhas to be considered. In Plates 5a and 5b the GOME NO2 asa functionof spaceandseasonaccording to the locationof columnsare shownmappedonto the IMAGES and MOZART convective cloudsas providedby the general•irculation grids. This smoothesthe plots and removesthe small very model, and the height of the top of convective clouds, as localized spots with high NO2 columns over industrialized areas.In Plate 5 the correspondingannualaveragedmodeled formulated by Price and Rind [1992]. As for the IMAGES model, the global production of NO from lightning is NO2 columnsare also shown.In the middle panelsthe tropoassumedto be 5 Tg N/yr. For our calculationsonly diurnal sphericcolumnsare calculatedby integratingfrom the surface
andhydrocarbons are the sameasthoseusedin IMAGES (see
VELDERS ET AL.: GLOBAL TROPOSPHERICNO2 COLUMN DISTRIBUTIONS to the tropopause,while in the lower panelsa similar method as for the GOME columnsis applied;i.e., subtractingthe NO2 column over the Pacific (around 180øW) from the NO2 columnat all longitudes.The latterprocedureis performedon a monthlybasisand retainsthe latitudinalvariation.Comparing both procedures(Plate 5c versus5e and 5d versus5f), it can be seenthat there is virtually no differencefor industrialized areaswith largetroposphericNO2 columnsas well as for the more remote continental areas (South America, Africa, and Australia). There are some differences over the oceans, but they are small in an absolutesense.The troposphericNO2 columns over the oceans and away from NOx sourcesare
12,651
mayalsoaffectthe modelvaluesin theseregions.The troposphericNO2 columnsin IMAGES are about a factor of 2-5 smaller than GOME for South America and Africa. The values in MOZART are -1-2 times smaller for South America and -3 times for Africa. The modeled columns over
theSaharan andAustralian deserts aresmall(0.2-0.5x l0 is molecules cm-2). The following points have to be mentionedin relation to
the comparison: (1) The emissions usedherecorrespond with a 1990inventoryandareprobablytoo low for 1997,especially for China.Kato andAkimoto[1992] foundin increasein NO• emissionsin easternAsia of 4%/yr in the 1970s and 1980s.
smaller than0.5 x 1015 molecules cm-2(middle panels). This Extrapolatingthis trendgivesan in increasein NO• emissions is the amount that is subtractedfrom both the GOME (Plates for thisareaof 32% between1990and1997.Thiscanexplain only part of the difference between GOME and MOZART. A 5a and 5b) and modeled (Plates 5e and 5f) in the derivation of the troposphericNO2 columnsbut is part of the tropospheric largerincreasein emissions wouldbe neededto explainthe column. whole difference. (2) Differences for South America and 4.2. Comparing Models With GOME
Africa can alsobe causedby errorsin the parameterization of convectiveprocesses whichtransportNO2 from the boundary layerto higheraltitudes.(3) In thiscomparison we onlyused oneyearof GOME data.The NO2 columnsprobablydisplay some year-to-yearvariability, but these real differencesare likely to be smallerthanthe largeuncertainties in the GOME
Plates 3 (IMAGES) and 4 (MOZART) can now be compared with Plate 2 (GOME). We must be cautious in making a quantitative comparison,consideringthe large uncertaintiesin the retrieval of the troposphericNO2 columns from the GOME measurementsas discussedabove (sections 2.3 and 2.4); these are likely to be as large as 50% for individuallocations.A qualitativecomparisonof GOME with
tropospheric column amounts and uncertainties in the
the model calculations is, however, useful, since this uncer-
comparedto model calculations.
tainty is likely to affect the NO2 columns everywherein a
The differences between the measured and modeled columns can also be seen in Plate 6 where the ratios between
similar manner. Note that Plates 2, 3, and 4 have different
scalesto showthe full rangeof the NO2 columns.The figures also differ in the grid sizes (grid size of the GOME data is 0.45ø x 0.45ø). The magnitude of the NO2 columns varies greatlyin different regions.Becauseof the color scalechosen, the low NO2 columnsover remoteareasare not easilyvisible in theseplates.In Plate 5 the columnsfrom GOME are plotted on the IMAGES and MOZART grids, and the color scaleis nonlinear
modeledrepresentation of the boundarylayer,suggesting that a betterunderstandingof thosefactorswill be neededbefore
inter annualfluctuationscan be meaningfullymeasuredand
the NO2 column amountsare plotted, i.e., GOME/IMAGES and GOME/MOZART. For calculating these ratios the GOME
data have been converted to the IMAGES
or the
MOZART grid. No ratio is calculatedif either the measured
or modeledNO2 column was smallerthan 0.1 x 10•5 molecules cm-2.Plate6a showsthatfor mostareastheGOME NO2 columnis 2-3 timeslargerthanthe columncalculatedby
to accentuate the areas with small columns.
IMAGES. Comparing GOME with MOZART (Plate 6b) lower ratios (i.e., better correspondence) are found for the polluted areas;a factor of 0.5-2 for easternUnited Statesand columns over industrialized areas both in the Northern and westernEurope. The columnsmeasuredby GOME over the SouthernHemispheres,low values over remote continental Saharaand Australiaare 1.5-2 timeslargerthan modeledby areas, and very low values over the oceans.Larger tropo- MOZART. The high ratios around 60øN over westernRussia sphericNO2 columnsare found in winter than in summer,the are the result of large NO2 columnsin GOME, probably
The overall picture of the measuredtroposphericNO2 columnsis well reproducedby the model calculations:large
same as in the GOME
observations.
The absolute
values of
the troposphericNO2 columnsas calculatedby MOZART are close to those derived from the GOME measurements, while
resultingfrom an inadequatesubtraction of the stratospheric NO2 column from the total column for these areas.
From thesecomparisonsbetweenthe GOME and modeled
the column amountsfrom IMAGES are significantlysmaller.
tropospheric NO2 columnswe concludethat thereis a good
The maximum
qualitativeagreementbetween the column amountsretrieved
values in IMAGES
are about a factor of 2-3
smaller than in GOME for the United States, -1.5-4 smaller
from the GOME
for Europe, and -3-5 for Asia, using the samegrid (that of
lations. Quantitatively,the GOME columnsare of the same
IMAGES) for both. The maximum values in MOZART are similar to those of GOME for the United States(a factor of 0.8-1.5 smaller) and Europe (0.7-2), but a factor of 1.5-3.5
industrialized
smallerfor Asia (all on the MOZART grid). If averagedover a larger area (subcontinental),the high maximum values in
measurements and both sets of model calcu-
order as the onescalculatedby MOZART for the polluted areas.
The resultsof theIMAGES modeldiffer substantially from those of the MOZART
model. This difference is caused
mainlyby the differentrepresentations of the boundarylayer andresultingdifferentNO2 profile in the lowestkilometerof
very localized areas in MOZART and in GOME are somewhat reduced.The agreementbetweenthe model calculations the modeledtropospherein the two models.The MOZART and the GOME
measurements
is somewhat worse for South
America and Africa, areaswith lower NO2 columns in which biomassburning, savannahfires, and soil emissionsare the dominant NOx sources. Uncertainties in the structure of the
boundarylayer due to largerconvectiveactivityin the tropics
model has a physicallymore realisticboundarylayer with high and more or lesshomogeneous NO2 concentrations. In IMAGES the boundarylayer cannotmaintaina high NO2 concentration throughoutthe whole layer. Emittedpollutants are removed too fast from the boundarylayer by vertical
12,652
VELDERS ET AL.' GLOBAL TROPOSPHERICNO2COLUMN DISTRIBU•ONS
diffusion. The boundary layer, which contributesstronglyto specieswith the largesttroposphericcolumn is HNO3, with areasfor the total troposphericNO2 columnin pollutedareas,therefore -10-18 x 10•5moleculescm-2overthe industrialized contains less NO2 in IMAGES than in MOZART (see also bothIMAGES and MOZART. The NOdNOyratio overthe section 6). industrializedareasin the United Statesand Europeis -2040%
5.
Contributions
to the NO• Column
Now that we have consideredthe modeled tropospheric NO2 columnsin section4, we probe the build up of the NO2 column, some aspectsof the chemicalpartitioning,and the contributionsfrom different vertical layers based on model calculationsonly.
As mentioned above, GOME passesover the equator at 1030 LT, so it performsits measurements at variouslatitudes between-10 and -11 LT. In Plate 7a the differencein tropospheric NO2 column at -1030 LT and the diurnal average column is given as a ratio as calculatedby IMAGES. Over areas in the eastern United
and 30-70%
in MOZART.
The observed
NOdNOy ratio [Parrish et al., 1993] rangesfrom -30 to -50% at daytimeandfrom-50 to -70% diurnallyaveragedin easternNorth America in August. The MOZART values are in better general agreementwith these observationsthan the IMAGES
values. The models differ on their treatment of the
nitrogen speciesresulting in different NOx/HNO3 ratios. This
differenceis mainly causedby a differentwet depositionof HNO3 and does not affect the NOr concentrationin the model
5.1. NOx and NOy
industrial
in IMAGES
States and western
very much [l/elders and Granier, 2001]. With respectto the column,the middle and upper tropospheres contributemore to the HNO3 troposphericcolumnthan in the caseof NO2, but the lower troposphere and boundary layer dominate the columnfor both speciesover industrializedareas.The large
NOycolumnanduncertainty in it showsthatthemodeledNO2
columns could be substantiallyaffected by errors in this Europe the troposphericNO2 column at the GOME overpass partitioning(see section6). time is -80% of the diurnal average column. Over South America and Africa it is slightly less, 50-70%. Any errorsin 5.2. Contribution of Different Layers the diurnal partitioningof NO: are thereforelikely to be too In section 5.1 and in section 4 we have considered the small to substantiallyaffect the comparisonbetween GOME and the model calculations. whole troposphericcolumn of NO2. Now we will considerthe In the atmosphere,NO and NO• exhibit a clear diurnal contributionsof different layers to the total tropospheric variation:during the day NO: is convertedto NO by photoly- column, focusing in particular to how sensitive the total sis, while during the night NO is convertedback into NO: troposphericcolumn is to the boundarylayer. In Plate 8 the mainly by the reactionwith ozone.The sum of NO and NO2 NO2 column in shown integratedup to the tropopausebut (=NO0 remainsmore or less constantduring this cycle. The startingat layer 1 (surface),layer 2 (bottom at 0.17 km), layer annualaveragelifetime of NOx in the IMAGES troposphereis 3 (0.45 km), and layer 4 (0.94 km) as calculated by the --12 hours, while it is -7 hours in the lowest 5 km of the MOZART model. From these plots it can be seen that over Northern Hemispherebetween20øN and 70øN. It is interest- the more polluted continental areas the first three layers ing to see how large the tropospheric NO• column is individually contribute more to the total troposphericNO2 comparedwith the troposphericNOr column.An error in the columnthan all otherlayerscombined.Over the industrialized modeled NO•/NO• ratio could alter the comparison with areasin the easternUnited Statesand westernEuropelayer 1 GOME measurements.This ratio, yearly averaged,is shown makesup 25-40% of the total tropospheric NO2 column,layer in Plate 7b and 7c. In IMAGES (Plate 7b) the NO: column is 2 another 25-40%, layer 3 -15-25%, and the other layers 60-70% of the NOx column over the continents(diurnal and combinedyield 10-20%. Over the remotecontinentalareasof annual average) with maximum values of-80% for the SouthAmerica and Africa the first layer contributes15-25%, easternUnited States,westernEurope,and easternChina. The the secondlayer 10-20%, the third 5-15%, while the other NO2/NOx ratio for MOZART (Plate 7c) is overall slightly layerscontribute40-70%. Over the remoteoceans,i.e., away higher. Over the most polluted areas ratios of-60% are from the continents,the layersabove 1 km altitude contribute found. These lower ratios are most pronouncedin the winter morethan 97% to the total troposphericcolumn.Closerto the months with values as low as 30%, while in summer, there is continentsand in the North Atlantic the layers above 1 km contribute 80-90% to the total column. no differenceover the United Statesand Europe(ratio -80%). The lowest few layers,representingthe boundarylayer, in The lower NO2 and higherNO valuesin winter are probably the resultof directNO emissions,slowerphotolysisof NO2 to the MOZART model clearly contributemost to the tropoNO, and reduced conversion of NO to NO2 becauseof lower spheric NO2 column (the same holds for IMAGES). The ozone concentrations. This effect is not observable in the boundary layer is difficult to model properly in a global IMAGES model, probably becauseof the diffusive nature of atmospheric model, since the chemical and dynamical the boundarylayer, which transportsNO andNO2 upwardout processes in the boundarylayer occuron spatialand temporal off the boundarylayer too rapidly. An error in the NO2/NOx scales that are too small for the relatively coarse spatial ratio could affect the agreementwith the GOME measure- resolution and time step of global models. Therefore, on a ments, although the modeled ratio is close to the observed local scaleof tens of kilometersmuch higherNO2 concentraratio of-80-90% in the boundary layer in eastern North tionsmight be found in the boundarylayer than calculatedby America in August [Parrish et al., 1993]. global models, which can increase the troposphericNO2 Assuming the total NO,• emissionsare relatively well column significantly. As has been shown in section 2.3, known, the troposphericNO2 column in the model calcula- GOME's sensitivity to NO2 in the boundary layer is a tions can also be affected by a different distribution of complex function of albedo and wavelength,resulting in a nitrogenspeciesbetweenNO2 (or NOr) and NOy.The NOy largeuncertaintyin the derivedtroposphericcolumns.
VELDERS ET AL.' GLOBAL TROPOSPHERIC NO2 COLUMN DISTRIBUTIONS
12,653
b) GOME: NO2 column' MOZART
a) GOME' NO2 column' IMAGES grid
J
MaxUSA:13.5 MaxEurop•: 12.8 0.0
0.005
0.01
0.02
0.05
0.1
0.2
0.5
1.0
MaxAsia:8.3 2.0
5.0
>10,0
c) IMAGES NO2 column
ß
•
.
Max USA: 15.0
0.0
0.005
0.01
0.02
0.05
Max Europe: 14.8
0.1
0.2
0.5
1.0
Max Asia: 11.7
2.0
5.0
>10.0
d) MOZART: NO2 column
"•.... i
'"
: :
'Max liSA: •.1 0.0
0.005
0.01
0.02
0.05
Max •=urop;: 6.2 0.1
0.2
0.5
1.0
Ma;Asia: 2.6 2.0
5.0
Max USA: 15.4
Max Europe: 12.8
Max Asia: 6.5 __
0.0
>10.0
0.005
0.01
MOZART:
e) IMAGES: NO2 column: 180W subtracted
002
0.05
0.1
0.2
NO2 column'
'• ,
0.5
1.0
2.0
180-165W
5.0
>10.0
subtracted
• '.• •
:
.
ß
::......':.............!' :
.!
: ........ •........................
: :
Max USA: 6.0
Max Europe:6.1
'•' :.._,_..._ ......
::
:
! ....
i.... • :
i'
: .......
;......... ; ........ : ........
i
,
'
:
: ::
,.
MaxUSA:.-15.1 ' MaxEurope: 12.5
Max Asia:2.5
'•'
: ........ .:.........
'Max Asia: 6.1
t
0.0
0.005
0.01
0.02
0.05
0.1
0.2
0.5
1.0
2.0
5.0 >100
NO2 ( 1.0e+ 15 molecules/cm2)
00
0.005
0.01
0.02
0.05
0.1
02
0.5
1.0
2.0
5.0
>10.0
NO2 (1.0e+ 15 molecules/cm2)
Plate 5. Yearlyaveraged tropospheric NO2columnsof GOME (a, b), IMAGES (c, e), andMOZART (d, f): all plottedon (left) the IMAGES grid of 5ø x 5ø and(right)the MOZART grid of 2.8ø x 2.8ø. (c) and (d) the tropospheric columnis calculated by a summation fromthesurfaceto thetropopause, whilein (e) and(f) the tropospheric columnis calculatedby subtracting a longitudebandoverthe Pacificfrom all longitudes. The lattermethodis similarto what is usedto derivethe tropospheric columnfrom the total columnin the GOME data.The columnscorrespond to -1030 LT. Note a nonlinearcolorscaleis used.
12,654
VELDERS ET AL.: GLOBAL TROPOSPHERIC NO2 COLUMN DISTRIBUTIONS
b) NO2 column'GOME/MOZART
a) NO2 column'GOME/IMAGES
: .
.
.
ß
' 00
05
• ' • ''*' I 1.0
1.5
2.0
2.5
3.0
3.5
4.0
-I
4.5
50
i
ß
0,0
>5.5
"l
"
0,5
1.0
1.5
2.0
2.5
3,0
3.5
.,
4.0
4.5
5.0 >5.5
NO2 GOME/MOZART
NO2 GOME/IMAGES
Plate 6. Ratios of the NO2 columnsfrom GOME relative to the IMAGES and MOZART columns,i.e., (a) GOME/IMAGES and (b) GOME/MOZART. Shown are (as a fraction) the yearly averaged columns correspondingwith 1030 LT. The ratio is only plottedif both the GOME and modelderivedNO2 columnsare
larger than0.1x 10Ismolecules cm-2.
a) IMAGES: NO2 column'10:30amvs diurnalav
b) IMAGES: NO2/NOx column
I
Max0SA'78% MaxEurop•. 80•o 0
10
20
30
4.0
50
60
70
Max Asia 68%
80
90
NO2(10:30)/NO2 (percent)
•--'---'--M•"6S•'79% 'MaxEUrOl•: 81% 0
10
20
30
4.0
50
60
70
Ma•Asia'74% 80
90
NO2/NOx (percent)
c) MOZART:NO2/NOx column
ß •"¾' •'• ß
I•.x USA86% MaxE•;ope•3% 0
10
20
30
40
50
60
MaxAsia.8•% 70
80
90
NO2/NOx (percent) Plate 7. Ratios between (a) the troposphericNO2 column at 1030 LT and the diurnal averagecolumn (IMAGES), andbetweenthe diurnalaYeragetropospheric NO2 columnandthe NO• columncalculatedby (b) IMAGES and (c) MOZART. Yearly av½,'aged columnsare shown.
VELDERS ET AL.' GLOBAL TROPOSPHERIC NO2 COLUMN DISTRIBUTIONS
b) MOZART: NO2 column:0.17 km - tropopause
a' MOZART: NO2 column: 0- tropopause :
:
:
:
•:.___..._•. •'
:
.
:
:
:
.
.
:
12,655
. • ß
:
.
MaxUSA. 15.4 Max •urope: 12.8 0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
MaxUSA: 10.7 MaxEurop•: 8.8
Ma•Asia: 65 9.0
MaxAsia: 4.1
00 0.8 1.6 2.4 3.2 4.0 4.8 5.6 •.4 72 8.0 >8.8
10.0 >11.0
c) MOZART:NO2 column:0.45 km - tropopause d) MOZART:NO2 column:0.94 km - tropopause
i..
...
:'..... :
.........
!' .
Max USA: 5.5 0.0
0.4
0.8
1.2
16
Max Europ•: 4.3' 2.0
2.4
28
3.2
' :
--Max
Ma•Asia: 2.6 36
4.0
00
>44
02
USA. 2.1 0.4
0.6
0.8
MaxEurope: 20 1.0
1.2
1.4
16
MaxAsia' 1.5 1.8
20
>2.2
NO2 (1.0e+ 15 molecules/cm2)
NO2 (1 0e+l 5 molecules/cm2)
Plate 8. Different verticalintegrationsfor calculatingthe yearlyaveragedtropospheric NO2 columnas calculatedby MOZART (1030 LT). The integrations are performedup to the tropopause but startingat, (a) the surface,(b) 0.17 km, (c) 0.45 km, and (d) 0.94 km.
RadiativeforcingtroposphericNO2
Max USA: 0.126
0.00
0.01
0.02
0.03
0.04
Max Europe:0.104
0.05
006
0.07
Max Asia:0.055
0.08
0.09
0.10
>0.11
NO2 (W/m2) Plate9. Annualaverage radiative forcing (W m-2)of tropospheric NO2calculated withMOZART.Theradiative forcingsare diurnalaveragesfor cloudfree conditions,calculatedusingthe daytimeaveragetropospheric NO2 columns.
12,656
VELDERS ET AL.: GLOBAL TROPOSPHERIC NO2 COLUMN DISTRIBUTIONS
5.3. HeterogeneousConversionof HNO3 on Soot
Lary et al. [1997] proposedthat a heterogeneous conversionof HNO3 to NO might take placeon soot.This reaction can increasethe NO and therebyNO2 concentrations in the troposphere. It has its largestaffectover and downwindof
pollutedareaswherelargeconcentrations of bothHNO3and sootoccur,i.e., the eastof China, easternEurope,and Siberia, with smaller effects in the easternUnited Statesand during thebiomassburningseasonin the tropicsoverSouthAmerica and Africa and downwind from there. Hauglustaineet al. [ 1996] showedthat this reactioncouldreducethe HNO3/NOx
ratio in the free troposphere at Hawaii, renderingmodeled valuesin betteragreementwith measurements. The effect of this reactionon the tropospheric NO2 column was studiedusing IMAGES. Since soot is not explicitly
yearly averageNO2 levels at urban sites of-24 ppbv, suburbansitesof -20 ppbv and at rural sitesof -8 ppbv for the 1990s. Their chemiluminescencemeasuring technique is sensitivenot only to NO2 but alsoto othernitrogen-containing compoundssuchas PAN. This cancausean overestimation of the reportedNO2 levels,especiallyin rural areas[EPA, 1998]. Thesedata are thereforelessusefulfor our study,but they do showthat very high mixing ratiosare found in pollutedareas, in agreementwith very large local NO2 columnsover industrialized areas. Measurementsin the free troposphere[e.g., Emrnonset al., 2000] are alsolessrelevantfor this studysince the free tropospherecontributesonly a small amount to the troposphericNO2 columns. Measurementsof reactivenitrogenat rural groundsitesin easternNorth Americawerereportedby Parrish et al. [1993]. In late summerof 1988 they measuredseveraltropospheric
modeledby IMAGES, it was crudelyrepresented as sulfate includingNO, NO2,andNOy at sevensurface (SOn)in the calculations.This is intendedonly to test how tracespecies, locations representativefor rural areas in the populated sensitivethe NO2 columns are to a possibleheterogeneous regions of easternNorth America. Mixing ratios over rural conversionof HNO3. Soot and sulfate have similar sources areasin industrializedregionsof the United StatesandEurope and distributionsand the reactionrate is adjustedto account are best suitedfor validatingthe GOME and modeleddata, for the difference in absolute values between soot and sulfate. considering the grid sizesof GOME and of the modelsused A reactionrate of 0.3 x 10-•5 cm'3s -• was used,whichis here. Parrish et al. [1993] found that the daytime (1000-
equivalent to a conversion lifetimeof 2 daysin theNorthern 1800)medianNOy levelsfor a fully convective well mixed Hemisphere, 0-5 km altitude.Thisresultsin tropospheric NO2 layerwere2 to 5 ppbv.TheNOylevelsvaryat each columnsin the easternUnited Statesand westernEuropeof boundary or more.The daytimeNOJNOy -10 x 10•5 molecules cm-2, an increase of 20-30%.Larger siteby an orderof magnitude ratiowas30-50%.The observed daytimeNO/NOyratiowas increasesoccurover westernRussia(80-120%), Siberia(more -5%. From this we derive NO2 daytimevaluesrangingfrom than 80%), and easternChina (60-80%). This heterogeneous 0.5 to 2.2 ppbv.The diurnalaverageNO2 valuesare -1.1-2.8 conversion of HNO3 has little effect in the Southern ppbvand-1.3-3.2 ppbvat the GOME overpass time of 1030 Hemisphere. Thus reactions on soot may significantly LT. increase the modeled NO2 column amounts.
In Figure2 the modeleddiurnalaverageNO2 mixingratios in the boundarylayer are shown for the United Statesin August.Thesemixingratiosbroadlyagreewith othermodel 6. Validation of TroposphericNO2 calculations[Levy et al., 1999; Horowitz and Jacob, 1999] In the previoussectionswe have comparedthe tropo- which show boundarylayer values above 1 ppbv over the
sphericNO2 columnsfrom GOME with columnamounts industrializedareas.The NO2 measurements[Parrish et al., for a well-mixedboundarylayer.The fromchemistry transport modelsandfounddifferences in the 1993] arerepresentative absolute values of the columns. In this section the focus is on lowestlayersin the modelcalculations do not showa uniform the validationof the GOME NO2 values.Obviously,the most mixing ratio but have high valuescloseto the surfaceand straightforward wayto validateNO2columns fromGOMEis decreasingtowardsthe top of the boundarylayer. For the the modeledmixingratiosare averagedoverthe to comparethemwith columnmeasurements madefromthe comparison ground.A largenumberof NO2columnmeasurements from lowesteight model layersfor IMAGES (0-1.4 km) and the groundbasedstationsis availablebut theymostlyfocuson lowest four model layers for MOZART (0-1.6 km). The stratospheric NO2andarein remotelocations or onmountain calculatedaveragevaluesfor the easternUnited Statesare tops(e.g., datafrom the Networkfor Detectionof Strato- 0.7-1.0 ppbvfor IMAGES and 1-3 ppbvfor MOZART. The spheric Change, NDSC).Thefocusin thisstudyis onNO2in latter values agree well with the measuredvalues (0.5-2.2 the pollutedareasof UnitedStates,Europe,andAsia with ppbv),but thoseof IMAGES aretoo low. In the lowestlayer only a few (or noneat all) systematic NO2 columnmeasure- in the models,NO2 valuesare as high as 4 ppbv (IMAGES) mentsat theselocations.We thereforecomparemodeledNO2 and 15 ppbv(MOZART). In winter,with a smallerboundary mixing ratioswith measurements in the boundarylayer to layer,valuesmorethantwiceas largeare foundbothfor the assessindirectlythe qualityof the modeledNO2 columns. averagein theboundarylayerandin thelowestlayer. A mixingratioof 1 ppbvin a well mixedboundary layerof Usingthecomparisons betweenGOME andthemodelcalculationsin the sections4 and 5 we can then get an idea of the 1 km contributes2.69 x 10•5 moleculescm-2 to the total qualityof the GOME measurements. We alsocompare tropo- column. The observed 1.3-3.2 ppbv of NO2 at -1030 LT yields -3.5-8.6x 10•smolecules cm-2foraboundary sphericNO2 columnsfrom GOME and the modelswith therefore columnsinferred from mixing ratio measurements in the layer of 1 km. Adding to this an NO2 columnin the free boundarylayer. Local measurementsof NO2 in the boundary layer are
troposphere of-1.5 x 10•smolecules cm-2(August datafrom MOZART model), we obtain a total troposphericcolumnof
cm-2.TheNO2columns retrieved performedby many institutesall over the world. They 5.0-10.1x 10•smolecules generallytakeplacein citiesfor localair qualitystudies. The from GOME in the easternUnited Statesin August are -7-10 cm-2,withmaximum values upto 16x 10•5 EPA [ 1998]performs measurements of NO2at varioussurface x 10•5molecules locationsin the United Stateson a daily basis.They report
molecules cm-2 in verylocalized areas.Thisis withinthe
VELDERS ET AL.' GLOBAL TROPOSPHERIC NO2 COLUMN DISTRIBUTIONS IMAGES:
20øN 130øW
120øW
110øW
100øW
NO2
90øW
MOZART:
80øW
70øW
60øW
130øW
120øW
110øW
12,657
NO2
i
i
I
i
100øW
90øW
80øW
70øW
•
60øW
Figure 2. Diurnal averageNO2 mixing ratios (ppbv) in the boundarylayer for August calculatedwith (a) IMAGES and (b) MOZART. The mixing ratios are averagedover the lowest 1.5 km of the modeled atmosphere.
rangeof the columnsinferredfrom the measuredNO2 mixing ratiosand slightlylargerthan the columnscalculatedwith the
cause a local instantaneousabsorption of radiation in the
troposphere of up to -2-10 W m-2,likelyresulting fromair
MOZARTmodel(4-8x 10•5molecules cm-2at --1030LT in pollution. Over polluted areas in the easternUnited States, August, witha localmaximum of 13x 10•5molecules cm-2). northwesternEurope, and the east of China a larger local The valuesfrom IMAGES are2.5-4.2 x 1015moleculescm-2. radiative absorption can be expected. Absorption of solar in thetroposphere by NO2of up to 30 W m-2has The following qualificationsconcerningthis comparison radiation need to be taken into account:
1. The boundary layer thicknessis not known exactly, affecting the comparisonof the modeled columnswith the surfacemixing ratio data.The thicknessof 1 km we assumed hereis probablyon the low end,resultingin a low estimateof the columninferred from the surfacemixing ratios. 2. The boundary layer in the model calculationsdoes not have a uniform mixing ratio but exhibits high surfacevalues decreasingtoward the top of the boundary layer. For this comparison,the averageof the mixingratiosin the lowest1.5 km has been used.This is not necessarilythe sameas a wellmixed boundary layer becauseof possible nonlinear NOx chemistry. 3. The relatively large grid size of the models affects the NOx chemistrysince the NO emissionsare emitted into these boxesand diluted more than in reality, resultingin lower local
been estimatedto take place in clouds during thunderstorms [Solomonet al., 1999]. Note that the local radiative absorption valuesgiven by Solomonet al. [1999] correspondto the net absorptionoccurringin the troposphericair column. This is different from the radiative climate forcing (i.e., the change
in flux at the tropopause)sinceit includesabsorptionwhich would have occurredat the ground anyway.The differenceis largest for clear skies and cloudy skies where the absorberis locatedbelow the cloud (for optically thick cloudswhere the absorber
is located
in or above the cloud the difference
is
relatively small). In this section we use the modeled NO2 columnsfroin IMAGES and MOZART to make a global map of the radiative forcing by NO2 in the tropospherefor clear sky conditions.
A line-by-lineradiativetransfercalculationwhich incorporates the DISORT [Stamneset al., 1988] scatteringcode has been used in this study.This is the samebasic ]nodel as used concentrationsand a different conversion to NO2. 4. High NO2 mixing ratios and columns occur on a local in Solomonet al. [1999]. A perturbationprofile of NO2 with scale in urban and suburbanareas(see large mixing ratio in all of the NO2 in the lowest 1 km is assumedsince the models urban and suburbanareas as measuredby the EPA [1998]), showthat most of the NO2 is in the boundarylayer in polluted which cannot be modeled in a global model and probably regions. Overlap with 03 and 04 is included although their difficult to observe by GOME, but can affect the NO2 influence is small except for O.• absorptionin the UV. A columns from both. ground albedo of 0.05 is assumed,which is reasonablefor 5. The data (GOME, model, and surface NO2 ]nixing continentalurban areasin the blue regionof the spectrum.An ratios) all show considerable spatial variability, which aerosollayer is included in the boundarylayer with an optical depth of 0.1, a single scatteringalbedo 0.85, and an asymmecomplicatesthe comparison. These considerationsdo not changethe overall picturethat try factor of 0.7, which are very con-servativeestimatesfor the tropospheric NO2 columns from GOME over eastern polluted regions. The diurnal cycle is integratedover for the North America are of the same order as those derived from 15th day of each month as a function of latitude for clear sky both the observed in situ NO2 mixing ratios and those from conditions.This yields a local radiativeforcing efficiency per the model calculations
with MOZART.
NO2molecule (W m-2/cm -2)whichcanbe multiplied by the
NO2 column to obtain the local radiative forcing. Plate 9 showsthe diurnal annual averageradiative forcing 7. Radiative Forcing of Tropospheric NO2 of troposphericNO2 as calculatedusingthe daytimeaveraged In addition to playing a role in the formation of ozone in troposphericNO2 columnsfrom MOZART. Sincethe daytime the troposphere,NO2 also absorbssolarradiationand thereby NO2 columnswere not directly available from the MOZART may contributeto the radiative forcing of climate. Solomonet calculations, we used the results from the IMAGES model to al. [1999] showed that over Boulder (Colorado) NO2 can calculatethe ratio betweenthe daytime averagetropospheric
12,658
VELDERS ET AL.: GLOBAL TROPOSPHERIC NO2 COLUMN DISTRIBUTIONS
NO2 columnsand the diurnal averagecolumns.This ratio was measuredtroposphericcolumnamountsinferredfrom GOME then usedon a monthlybasisto calculatethe daytimeaverage are of the sameorder as thosecalculatedby MOZART over columns
for
the
MOZART
calculations.
The
maximum
monthly values attained by the local radiative forcing of
tropospheric NO2are0.1-0.15W rn-2.Muchlarger values can
industrialized
areas. The maximum values in MOZART
are
similar to thoseof GOME for the United States(a factor of 0.8-1.5 smaller) and Europe (0.7-2), but a factor of 1.5-3.5 smallerfor Asia (on the samegrid).If averagedovera larger area (subcontinental)the high maximum values in very
be expectedin cities where very high NO2 concentrations are observed;events that are not well representedin the global and in GOME are somewhat model calculation presentedhere. These local "hot spots" localized areas in MOZART over industrializedareascould be importantfor local heating reduced.The agreementbetweenthe model calculationsand in urbanareas.On a more continentalscalea radiativeforcing the GOME measurements is somewhat worse for South of 0.04-0.1W m-2is calculated overtheeastern UnitedStates, America and Africa, areaswith lower NO2 columnsin which western Europe, and eastern Russia. The calculation of biomassburning, savannahfires, and soil emissionsare the radiative forcing illustrates the regional distribution and dominantNOx sources.The maximum valuesof IMAGES are magnitude,at least to a factor of 2 accuracy.The calculated significantlysmallerthan thoseof GOME: abouta factorof 2values are of the same order as the radiative forcing of the 3 for the United States,- 1.5-4 for Europe,and-3-5 for Asia, well-mixed greenhouse gasN20 (0.14W m-2) but smaller usingthe samegrid for both.
thantheradiative forcing ofthehalocarbons (0.25W rn-2)and Over pollutedareasthe largestcontributionto the troposphericNO2 columnoriginatesfrom the boundarylayer.The of tropospheric ozone(0.2-0.6W rn-2) [Intergovernmental Panel on Climate Change, 1996; Kiehl et al., 1999; Worm boundarylayer is difficult to model properlyin a global Meteorological Organization, 1999]. Globally averaged,the atmosphericmodel, becausethe chemical and dynamical in the boundarylayeroccuron spatialandtemporal estimatedradiative forcing of troposphericNO2 is relatively processes
smallhaving a valueof-0.005 W m-2.Theradiative forcing scalesthat are too small for the relativelycoarsespatial of NO2 showsa smallseasonalcycle,with -15% largervalues than average in winter/spring and-15% lower values in autumn. The seasonalcycle is not very large becausethe largerforcing in the summercomparedwith the winter, due to the zenith angle dependence,is compensatedby the smaller NO2 columns.Assuminga larger groundalbedoof 0.15 in the calculation increasesthe estimated radiative forcing by a factor of 2. Forcing maps derived from the IMAGES calculations
resolutionand time step of global models.Thereforemuch higherNO2 concentrations might be foundon smallerscales (severalkilometers)in the boundarylayer than thosecalculated by global models, which can locally increasethe tropospheric NO2 columnsignificantly. The effect of a possibleheterogeneous conversion(on soot) of HNO3 to NO on the modeled NO2 columns in the tropospherehas been studieswith the IMAGES model. We
are about half those derived from MOZART.
foundthatthisreactioncanincreasethemodeledtropospheric The radiative forcing of troposphericNO2 as estimated NO2 columns with 20-30% in the eastern United States and here is based on cloud free conditions. Since the bulk of the westernEurope and around 100% in westernRussia, Siberia, NO2 residesin the boundarylayer below the clouds,clouds and eastern of China. The modeled ratios NO2/NOx and effectively shield the solar radiation from the NO2. Even NO•XlOyhavebeenanalyzedandfoundto be closeto obserclouds with relatively small optical depths will reduce the vationsover easternNorth America, addingconfidenceto globally averagedradiativeforcing.For example,a cloudwith model calculationsof the NO2 columns. The modeledtroposphericNO2 columnsfrom MOZART as an optical depth of 10 will reduce the radiative forcing by 65%. Thus, as a lower limit we can assume that clouds will well as the columnsmeasuredby GOME agreewith NO2 reduce the estimated local radiative forcing by nearly the columnsderivedfrom observed NO2 mixingratiosin the fraction of local cloud cover. Assuming an average cloud boundarylayerin easternNorth America:the NO2 columns cm-2,fromMOZART cover of 50% the maxima in the local radiativeforcing over fromGOMEare7-10x 10•5molecules the industrializedareasbecomes-0.05-0.08 W m-2. are4-8 x 10•5molecules cm-2,andthosederivedfromthe 8.
surfacemixingratiosare 5-10 x 10•5 molecules cm-2. In IMAGES,NO2is transported outof theboundary layertoo fastresulting in lowerNO2columns (2.5-4x 10•5molecules
Conclusions
TroposphericNO2 columnsderivedfrom GOME datahave
cm-2). Qualitatively, thepattern oftropospheric NO2columns
been comparedwith that from model calculationsby two asmeasured by GOMEagreewellwiththeonescalculated by global three-dimensional chemistry transport models, MOZART andIMAGES.A quantitative comparison is more IMAGES andMOZART. The globalpatternsin NO2 columns difficult, but indicatesthat the GOME NO2 columns,those are well reproducedby the models:large troposphericcalculated by MOZART, andthosederivedsurfacemixing
columns over the industrialized
areas of the eastern United
States,Europe, and Asia, as well as over cities in Africa and South America; increasedNO2 columns are found over the biomassburning areas of Africa and South America. The variation through the seasonsis also similar in both the
GOME andthe modeldatasets,havinglargervaluesin winter and smalleronesin summer.The quantitativecomparisonis limited by the uncertaintyin the tropospheric NO2 columns derivedfrom GOME causedby the uncertaintyin the air mass factorsandcloudcoveranddueto the difficultyof modeling the dynamicsof theboundarylayerproperlyin globalmodels. A tentativecomparisonof the absolutevaluesshowsthat the
ratios are all of the same order for eastern North America.
An estimateof theradiativeforcingof NO2hasbeenmade using the calculated troposphericNO2 columns from MOZART. The local maxima over the eastern United States
andwesternEuropein the radiativeforcingof tropospheric NO2is 0.1-0.15W m-2,whilevaluesof 0.04-0.1W m-2are estimated on a continental scale(cloudfreeconditions). The all sky estimatesare about a factor of 2 smallerthan the clear sky values.Thesevaluesare of the sameorderas the radiative forcing of N20 but smallerthan that of the halocarbonsand
tropospheric ozone.Thegloballyaveraged radiative forcingis negligible, -0.005W m-2.
VELDERS ET AL.: GLOBAL TROPOSPHERIC NO2 COLUMN DISTRIBUTIONS Acknowledgments. We would like to thank M. Buchwitz, R. Harley, D. Parrish, V. Rozanov, S. Solomon, and M. Trainer for usefuldiscussionsand supportin the work and J.-F. Miiller for his help with theIMAGES model.Part of the work was performedwhile one of the authors (K. Pfeilsticker) held a National Research
12,659
Emmons, L. K., et al., Climatologies of NOxandNOy:A comparison of dataandmodels,Atmos.Environ.,31, 1851-1904, 1997. Emmons,L. K., D. A. Hauglustaine,J.-F. Mailer, M. A. Carroll, G. P. Brasseur, D. Brunner, J. Staehelin, V. Thouret, and A. Marenco, Data composites of airborne observations of troposphericozone and its precursors, ,/. Geophys.Res., 105,
Council-NOAA ResearchAssociateship. J. Burrowsand A. Richter 20,497-20,538, 2000. acknowledgefunding by the Universit3 of Bremen, the German Ministry for Educationand Research(BMBF) and the European EnvironmentalProtectionAgency, National Air Quality and Commission.Part of this work has also been supportedby the Emissions TrendsReport1997,U.S. Environ.Prot.Agency,Off. EuropeanCommissionunder the POET project (contractEVK2of Air QualityPlann.andStand.,Res.TrianglePark,N. C., 1998. 1999-00011).Somesimulations havebeenperformedusingcomputer Granier,C., J.-F. Mtiller, S. Madronich,and G. P. Brasseur,Possible time providedby the National Center for AtmosphericResearch causes for the 1990-1993decrease in theglobaltropospheric CO (NCAR), which is sponsoredby the National ScienceFoundation. abundance;A three-dimensionalstudy, Atmos. Environ., 30, 1673-1682, 1996.
Granier, C., J.-F. Mtiller, G. P6tron, and G. Brasseur,A three-
References
dimensional studyof theglobal;CO budget,Chemosphere Global
BovesmannH., J.P. Burrows, M. Buchwitz, J. Frerick, S. Noel, V. V. Rozanov, K. V. Chance, and A. P. H. Goede, SCIAMACHY Mission objectives and measurementmodes, J. Atmos. Sci., 56, 126-150, 1999. Brasscur,G. P., D. A. Hauglustaine,S. Walters, P. J. Rasch, J.-F.
Mtiller, C. Granier, and X.-X. Tie, MOZART: A global chemical transportmodel for ozone and related chemical tracers, 1, Model description,J. Geophys.Res. 103, 28,265-28,289, 1998. Burrows J.P.,
W. Schneider, and K. V. Chance, 'GOME
and
ChangeSci., 1,255-261, 1999.
Granier,C., J.-F. Mtiller, and G. Brasseur,The impactof biomass burningon the global budgetof ozoneand ozoneprecursors, in BiomassBurning and its Inter-Relationshipwith the Climate System, editedby J. L. Innes,M. Beniston, andM. Verstrate, pp. 69-85, Kluwer Acad., Norwell, Mass., 2000a.
Granier,C., G. P6tron,J.-F. Mtiller, andG. Brasseur, The impactof natural and anthropogenichydrocarbonson the tropospheric budget of carbon monoxide,Atmos. Environ., 34, 5255-5270, 2000b.
SCIAMACHY: Remotesensingof stratospheric and tropospheric Hao, W. M., and M. H. Liu, Spatial and temporaldistributionof trace gases,in CEC Air Pollution Research Report 34: Polar tropicalbiomassburning,Global Biogeochem.Cycles,8, 495Stratospheric Ozone Proceedings of the First European 503, 1994. Workshop,3-5 October 1990 Schliersee,Baveria Germany, Hauglustaine,D. A., B. A. Ridley, S. Solomon,P. G. Hess, and S. editedby J. A. Pyle and N. R. P. Harris, pp. 99-102, Office for Madronich,HNO3/NOxratio in the remotetroposphere during Publ. EuropeanCommunities,Luxembourg,1991. MLOPEX 2: Evidence for nitrix acid reduction on carbonaceous Burrows,J.P., et al., Global Ozone Monitoring Experiment,Interim aerosols?,Geophys.Res.Lett., 23, 2609-2612, 1996. ScienceReport, Eur. Space Agency Spec. Publ., ESA SP-1151, Hauglustaine, D. A., G. P. Brasseur,S. Walters,P. J. Rasch,J.-F. 1993. Burrows, J.P., A. Dehn, B. Deters, S. Himmelmann, A. Richter, S. Voigt, and J. Orphal, Atmosphericremote sensingreferencedata
from GOME, 1, Temperaturedependentabsorptioncrosssection of NO2 in the 231-794 nm range,J. Quant. Spectr.Radiat. Trans., 60, 1025-1031, 1998a. Burrows J.P., T. Kurosu, R. Guzzi, E. Cattani, M. Cattani, M. Cervino, Ch. Levoni, and F. Torricella, GOME cloud and aerosol
data products algorithm development, final report contract 11572/95/NL/CN, Eur. Space Agency, Noordwijk, Netherlands, 1998b.
Burrows, J.P., M. Buchwitz, M. Eisinger, V. Rozanov, M. Weber, A. Richter, and A. Ladst•itter-Wei13enmayer, The Global Ozone Monitoring Experiment (GOME), mission, instrumentconcept, and first scientific results, in Proceedings of the Third ERS
Mtiller,L. K. Emmons, andM. A. Carroll,MOZART:A global chemicaltransport modelfor ozoneandrelatedchemicaltracers, 2, Model resultsand evaluation,d. Geophys.Res.,103, 28,29128,335, 1998.
Hegels, E., D. Perner, P. J. Crutzen, J.P. Burrows,A. Ladst•itterWei13enmayer, M. Eisinger, J. Callies, and W. Balzer, Global OzoneMonitoringExperiment:first BrO measurements, in XVII1
Quadrennial OzoneSympoisum 1996L'AquilaItalyAtmospheric Ozone:Proceedings of the Quadrennial OzoneSymposium 1996 Italy 12-21 September 1996, editedby R. D. Bojkovand G. Visconti, pp. 293-296, Parco Scientificoe Technologico d'Abruzzio,L'Aquila,Italy, 1998.
Holtslag,A., andB. Boville,Localversusnonlocal boundary-layer diffusionin a globalclimatemodel,d. Clim.,6, 1825-1842,1993.
Symposium Florence18th-23 ra of March1997,Spaceat the Hoogen,R., V. V. Rozanov,andJ.P. Burrows,Ozoneprofilesfrom
Serviceof our Environment,Eur. SpaceAgencySpec.Publ., ESA SP 414, 585-590, 1998c.
Burrows, J.P., et al., The Global Ozone Monitoring Experiment (GOME): Mission conceptand first scientific results,J. Atmos.
GOME satellitedata:algorithmdescription andfirst validation,d. Geophys.Res., 104, 8263-8280, 1999.
Horowitz,L. W., and D. J. Jacob,Global impactof fossil fuel combustion on atmospheric NOx,d. Geophys.Res.,104, 23,823-
23,840, 1999. Sci., 56, 151-175, 1999. Panelon ClimateChange,ClimateChange1995, Bussemer, M., Der Ring-Effekt: Ursachen und Einflu13 auf die Intergovernmental The Scienceof ClimateChange,CambridgeUniv. Press,New spektroskopische Messung stratosph•irischer Spurenstoffe, York, 1996. Diploma thesis,Univ. of Heidelberg,Heidelberg,Germany,1993. B., Signalprocessing andpattern recognition, in Handbook of Chameides, W. L., et al., Ozone precursor relationshipsin the J•ihne, Computer VisionandApplications, vol.2, editedby B. J•ihne, H. ambientatmosphere, J. Geophys.Res.,97, 6037-6055, 1992. Hauss•icker, andP.Gei131er, Academic,SanDiego,Calif., 1999. DeMore, W. B., D. M. Golden, R. F. Hampson,C. J. Howard, C. E. Kato,N., andH. Akimoto,Anthropogenic emissions of SO2andNO• Kolb, M. J. Kurylo, M. J. Molina, A. R. Ravishankara,and S. P. in Asia:Emissioninventories, Atmos.Environ.,Part A, 26, 2997Sander, Chemical kinetics and photochemicaldata for use in 3017, 1992. stratosphericmodeling, Data evaluation 12, JPL Publ., 97-4, 1997. Kiehl, J. T., T. L. Schneider,R. W. Portmann,and S. Solomon, Climateforcingdue to tropospheric and stratospheric ozone,d. Dreher, D. B., and R. A. Harley, A fuel-basedinventoryfor heavyGeophys.Res., 104, 31,239-31,254, 1999. duty diesel truck emissions,Air WasteManage. Assoc.,48, 352358, 1998.
Eisinger, M., and J.P. Burrows, Tropospheric sulfur dioxide observedby the ERS-2 GOME instrument,Geophys.Res. Lett., 25, 4177-4180, 1998.
Eisinger,M., J.P. Burrows, and A. Richter, Studiesof the precision of GOME irradiance and radiance products and GOME measurements of OC10
and BrO over Antarctica, in ESA
Publication Division (WPP-109), GOME, Global Ozone Monitoring Experiment, Geophysical Validation Campaign, edited by F. Bednarz, pp. 93-105, Eur. SpaceRes. and Technol. Cent., Frascati, Italy, 1996.
Kraus,A. B., F. Rohrer,E. S. Grobler,and D. Ehhalt,The global tropospheric distributionof NO• estimatedby a three-dimensional chemicaltracer model, d. Geophys.Res., 101, 18,587-18,604, 1996.
Kurosu, T., V. V. Rozanov, and J.P. Burrows, Parameterization schemes for terrestrial water clouds in the radiative transfer model
GOMETRAN, d. Geophys.Res., 102, 21,809-21,824, 1997. Lamarque,J.-F., G. P. Brasseur,and P. G. Hess, Three-dimensional
studyof therelativecontributions of thedifferentnitrogensources in the troposphere, d. Geophys.Res.,101, 22,955-22,968,1996. Lary, D. J., R. Toumi,A.M. Lee, M. Newchurch,M. Pirre, andJ. B.
12,660
VELDERS ET AL.: GLOBAL TROPOSPHERIC NO2 COLUMN DISTRIBUTIONS
Renard, Carbon aerosols and atmospheric photochemistry,J. Geophys.Res., 102, 3671-3682, 1997. Leue, C., Quantitative Analysis of NO,• emissionsfrom GOME satellite image sequences,Ph.D. thesis, Univ. of Heidelberg, Heidelberg, Germany, 1999. Leue, C.. M. Wenig, and U. Platt, Retrieval of troposphericNO2
Rozanov, V., D. Diebel, R. J. D. Spurr, and J.P. Burrows, GOMETRAN: A radiative transfermodel for the satelliteproject GOME- the plane parallel version, d. Geophys.Res., 102,
Res., 100, 16,445-16,490, 1995. Noel, S., M. Buchwitz, H. Bovensmann, R. Hoogen, and J. P.Burrows, Atmospheric water vapour amountsretrieved from GOME satellitedata, Geophys.Res.Lett., 26, 1841-1844, 1999. Olivier, J. G. J., A. F. Bouwman, C. W. M. van der Maas, J. J. M. Berdowski,C. Veldt, J.P. J. Bloos, A. J. H. Visschedijk,P. Y. J. Zandveld, and J. L. Haverslag, Description of EDGAR version 2.0: A set of global emissioninventoriesof greenhousegasesand ozone-depletingsubstances for all anthropogenicand mostnatural sourceson a per country basis and on lxl grid, RIVM Rep. 771060002, Natl. Inst. of Public Health and the Environ.,
and Technol. Cent., Noordwijk, Netherlands,1999. Wang, Y., J. A. Logan, and D. J. Jacob, Global simulation of troposphericO3-NOx-hydrocarbon chemistry,2, Model evaluation and global ozonebudget,d. Geophys.Res., 103, 10,727-10,755,
16,683-16,695, 1997. Solomon, S., R. W. Portmann, R. W. Sanders, J. S. Daniel, W.
Madsen, B. Bartram, and E.G. Dutton, On the role of nitrogen dioxidein the absorptionof solarradiation,J. Geophys.Res.,104, concentration from multispectralimage sequences, in Handbook 12,047-12,058, 1999. of ComputerVisionand Applications,vol. 3, chap.37, editedby B. J•ihne,H. Hauss•icker,and P. Gei131er, pp. 738-805, Academic, Stamnes, K., S.C. Tsay, W. Wiscombe, and K. Jayaweera,A numerically stable algorithm for discrete-ordinate-method San Diego, Calif. 1999. radiativetransferin scatteringand emittinglayeredmedia,Appl. Leue, C., M. Wenig, T. Wagner, O. Klimm, U. Platt, and B. J•ihne, Opt., 27, 2502-2509, 1988. Quantitative analysis of NO,• emissionsfrom GOME satellite Turman, B. N., and B.C. Edgar, Global lightning distributionsat imagesequences, J. Geophys.Res.,106, 5493-5505,2001. dawnanddusk,J. Geophys.Res.,87, 1991-1206, 1982. Levy II, H., W. J. Moxim, A. A. Klonecki, and P.S. Kasibhatla, SimulatedtroposphericNO,•: Its evaluation,global distribution Valks, P. J. M., and G. J. M. Velders, The present-dayand future and individual source contributions, d. Geophys. Res., 104, impactof NOx emissionsfrom subsonicaircrafton the atmosphere 26,279-26,306, 1999. in relationto the impact of NO,• surfacesources,Ann. Geophys., 17, 1064-1079, 1999. Logan, J. A., M. J. Parther, S.C. Wofsy, and M. B. McElroy, Tropospheric chemistry:A globalperspective, d. Geophys.Res., Velders, G. J. M., and C. Granier, Sensitivity of washout on 86, 7210-7254, 1981. HNO3/NOx ratio in atmosphericchemistrymodels,d. Geophys. Res., 106, 3125-3132, 2001. Marquard,L. C., T. Wagner,and U. Platt, Improvedair massfactor conceptsfor scatteredradiationdifferentialopticalabsorption Vountas,M., V. V. Rozanov,J.P. Burrows,Ring effect: Impact of rotational Raman scattering on radiative transfer in earth's spectroscopy of atmosphericspecies,d. Geophys.Res., 105, !315-1327, 2000. atmosphere, J. Quant.Spectr.Radiat. Trans., 60, 943-961, 1998. Moxim, W. J., H. Levy II, and P.S. Kasibhatla,Simulatedglobal Wagner, T., Satellite observationsof atmospherichalogen oxides, Ph.D. thesis,Univ. of Heidelberg,Heidelberg,Germany, 1999. tropospheric PAN: Its transportandimpacton NO,, J. Geophys. Res., 101, 12,621-12,638, 1996. Wagner, T., and U. Platt, Satellite mapping of enhanced BrO concentrations in the troposphere, Nature, 395, 486-490, 1998. Mtiller, J.-F., Geographical distribution and seasonalvariation of surfaceemissionsand depositionvelocities of atmospherictrace Wagner, T., C. Leue, K. Pfeilsticker, and U. Platt, Measurementof atmosphericBrO and OC10 by GOME, in Proceedingsof the gases,J. Geophys.Res.,97, 3787-3804, 1992. EuropeanSymposiumon AtmosphericMeasurements from Space Mtiller, J.-F., and G. Brasseur, IMAGES: A three-dimensional (ESAMS 99), 18-22 January1999, pp. 415-419, Eur. SpaceRes. chemicaltransportmodel of the global troposphere, J. Geophys.
Bilthoven, Netherlands, 1996.
Parrish, D. D., et al., The total reactive oxidized nitrogenlevels and
1998.
Wenig, M., C. Leue, B. J•ihne, and U. Platt, Cloud classification using image sequencesof GOME data, in Proceedingsof the
EuropeanSymposium on AtmosphericMeasurements from Space (ESAMS 99), 18-22 January1999, pp. 420-424, Eur. SpaceRes. and Technol. Cent., Noordwijk, Netherlands,1999. World MeteorologicalOrganization,Scientificassessment of ozone depletion: 1998, edited by D. Albritton et al., Rep. 44, Global Ozone Res. and Monit. Proj., Geneva, 1999.
thepartitioningbetweenthe individualspeciesat six rural sitesin easternNorth America,J. Geophys.Res.,98, 2927-2939, 1993. Penner, J. E., D. J. Bergmann, J. J. Walton, D. Kinnison, M. J. J.P. Burrowsand A. Richter,Instituteof EnvironmentalPhysics, Prather, D. Rotman, C. Price, K. E. Pickering, and S. L. Universityof Bremen,P.O. Box 330440,28334 Bremen,Germany. Baughcum, An evaluation of upper troposphericNO,• with two C. Granier, Aeronomy Laboratory, NOAA, 325 Broadway, models,./. Geophys.Res., 103, 22,097-22,113, 1998. Boulder, CO 80303. Pham, M., J.-F. Mtiller, G. Brasseur,C. Granier, and G. Megie, A K. Pfeilsticker,U. Platt, T. Wagner, and M. Wenig, Institut ftir three-dimensional study of the tropospheric sulfur cycle, J. Umweltphysik,Universityof Heidelberg,Im NeuenheimerFeld 229, Geophys.Res., 100, 20,061-20,092, 1995. 69120 Heidelberg,Germany. Platt, U., Differential Optical AbsorptionSpectroscopy(DOAS), in R. W. Portmann,AeronomyLaboratory,NOAA, 325 Broadway, Air Monitoring by SpectroscopicTechniques,chap. 127, edited Boulder, CO 80303. by M. W. Sigrist,pp. 27-84, JohnWiley, New York, 1994. G. J. M. Velders, Air ResearchLaboratory, National Institute of Price, C., and D. Rind, A simple lightning parameterizationfor Public Health and the Environment (RIVM), P.O. Box 1, 3720 BA calculatingglobal lightning distributions,J. Geophys.Res., 97, Bilthoven,Netherlands.(guus.velders@rivm. nl) 9919-9933, 1992.
Richter, A., M. Eisinger, F. Wittrock, and J.P. Burrows, BrO observed from GOME in the Northern Hemisphere in 1997, Geophys.Res.Lett. 25, 2683-2686, 1998. Richter A. and J.P. Burrows, Tropospheric NO2 from GOME measurements, Adv. SpaceRes.,in press,2000.
(ReceivedMay 22, 2000; revisedOctober19, 2000; acceptedNovember21, 2000.)