Source-receptor relationships for airborne measurements of CO2, CO and O3 above Siberia

17 Pages·2010·3.32 MB·English

Atmos. Chem. Phys.,10,1671–1687,2010
Atmospheric
www.atmos-chem-phys.net/10/1671/2010/
Chemistry
©Author(s)2010. Thisworkisdistributedunder
theCreativeCommonsAttribution3.0License. and Physics
Source-receptor relationships for airborne measurements of CO ,
2
CO and O above Siberia: a cluster-based approach
3
J.-D.Paris1,A.Stohl2,P.Ciais1,P.Ne´de´lec3,B.D.Belan4,M.Yu.Arshinov4,andM.Ramonet1
1LaboratoiredesSciencesduClimatetdel’Environnement/IPSL,CEA-CNRS-UVSQ,UMR1572,GifsurYvette,France
2NorwegianInstituteforAirResearch(NILU),Kjeller,Norway
3Laboratoired’Ae´rologie,ObservatoireMidiPyre´ne´es,CNRS-UPS,Toulouse,France
4ZuevInstituteofAtmosphericsOptics,SBRAS,Tomsk,Russia
Received: 22January2009–PublishedinAtmos. Chem. Phys.Discuss.: 9March2009
Revised: 14January2010–Accepted: 7February2010–Published: 15February2010
Abstract. We analysed results of three intensive aircraft to lower mixing ratios in spring, when the airmass aerosol
campaignsaboveSiberia(AprilandSeptember2006,August burden is important. In the lower troposphere, large-scale
2007)withatotalof∼70hofcontinuousCO , COandO deposition processes in the boreal and sub-arctic boundary
2 3
measurements. The flight route consists of consecutive as- layer is a large O sink, resulting in a ∼20ppb difference
3
cents and descents between Novosibirsk (55◦N, 82◦E) and with overlaying air masses. Lagrangian footprint clustering
Yakutsk(62◦N,129◦E).Weperformedretroplumecalcula- isverypromisingandcouldalsobeadvantageouslyapplied
tions with the Lagrangian particle dispersion model FLEX- to the interpretation of ground based measurements includ-
PART for many short segments along the flight tracks. To ingcalculationoftracers’sourcesandsinks.
reducetheextremelyrichinformationonsourceregionspro-
vided by the model calculation into a small number of dis-
tinct cases, we performed a statistical clustering – to our
1 Introduction
knowledgeforthefirsttime–intopotentialsourceregionsof
thefootprintemissionsensitivitiesobtainedfromthemodel Atmospheric transport of pollutants can occur both at low
calculations.Thistechniquenotonlyworkedwelltoseparate altitude in the boundary layer and in the free troposphere,
source region influences but also resulted in clearly distinct where it can be much faster. Export of North-East Asian
tracerconcentrationsforthevariousclustersobtained. High and North American emissions is associated with strong
COandO3 concentrationswerefoundassociatedwithagri- uplift in the warm conveyor belts of mid-latitude cyclones
culturalfireplumesoriginatingfromKazakhstaninSeptem- (e.g. Cooper et al., 2001; Liang et al., 2004; Owen et al.,
ber 2006. A statistical analysis indicates that summer up- 2006).Ontheopposite,Europeanemissionshaveatendency
take of CO2 is largely explained (∼50% of variance) by air to remain in the lower troposphere and are most frequently
mass exposure to uptake by Siberian and sub-arctic ecosys- channelled to the Arctic or to Siberia (Wild et al., 2004;
tems.Thisresultedinanaverage5to10ppmdifferencewith Stohl et al., 2002, 2007a; Duncan and Bey, 2004; Law and
overlayingairmasses. Stratosphere-troposphereexchangeis Stohl,2007;Liuetal.,2002). Fasterzonaladvectionoccurs
foundtostronglyinfluencetheobservedO3 mixingratiosin in winter and early spring in association with the Siberian
spring, but not in summer. European emissions contributed High(NewellandEvans,2000;Wildetal.,2004;Liuetal.,
to high O3 concentrations above Siberia in April 2006 and 2003). In April, European pollutants can contribute signifi-
August2007,whileemissionsfromNorthEasternChinaalso cantly to tracers’ concentrations in the Asian outflow to the
contributed to higher O3 mixing ratios in summer, but tend Pacific(Liuetal.,2003;Liangetal.,2004). Withrespectto
ozone production, European export increase O concentra-
3
tions over Siberia by 2 to 4ppb in spring and 2 to 6ppb in
Correspondenceto: J.-D.Paris summer(Wildetal.,2004;DuncanandBey,2004).
([email protected])
PublishedbyCopernicusPublicationsonbehalfoftheEuropeanGeosciencesUnion.
1672 J.-D.Parisetal.: Source-receptorrelationshipsforairbornemeasurementsaboveSiberia
Pollutant export to or across Siberia, however, has CO, O and other tracers have been implemented in vari-
3
been studied almost exclusively through modelling stud- ousparts ofSiberiausingaircraft campaigns equippedwith
ies and lacks measurement-based assessment. Pochanart et flask sampler (Nakazawa et al., 1997), aircraft vertical pro-
al. (2003) has conducted ground based measurements sup- files (Lloyd et al., 2002; Levin et al., 2002; Ramonet et al.,
ported by transport model analysis, showing that European 2002),theTranssiberianrailroad(Oberlanderetal.,2002)or
emissions lead to a 1 to 4ppb increase in O near Baikal talltowers(Kozlovaetal.,2008).
3
Lake. Enerothetal.(2003), basedonbacktrajectoriesanal- Three airborne campaigns were made across Siberia in
ysis of transport to the West-Siberian site of Zotino have April 2006, September 2006 and August 2007 respectively.
shownthatstagnantflowconditionswereresponsibleofres- Numerous vertical profiles of continuous measurements of
piration build-up leading to high CO2 concentrations (and CO,CO2, O3 andaerosolswerecollected, extensivelysam-
ruled out European emissions). But measurements of CO2 plingtheregionaltroposphereatdifferenttimesoftheyear.
and pollutants’ distribution are still required over Siberia, Afterdescribingthedataset,weanalysethetracers’concen-
alongwiththepossibilitytorelatethemtoremoteemission trationsusingaclusteringtechniquebasedonthepartitioning
sourcesi.e. source-receptorrelationships(SRR). offootprintsobtainedwithaLPDM.
Contrarily to single back-trajectories, the SRR obtained More specifically, we focus on the extent to which clus-
from a Lagrangian particle dispersion model (LPDM) ac- ters(groupswithsimilardistributions)ofpotentialemission
countsfortheatmosphericturbulenceandconvection(Stohl sensitivity(PES)mapsforreceptorpositionsrelatetotheair
et al., 1998). Han et al. (2005) compared the result of re- masses chemical composition at these receptors. The PES
gional source apportionment using single backtrajectories isexpressedhereasaresidencetime. Theclustersareused
andbackwarddispersion. Theirbackwarddispersionmodel togeneralizeSRRthroughouttheaircrafttrajectorythrough
addedindividualturbulence-relatedstochasticcomponentto the Siberian air shed. We try to answer the following ques-
alargenumberofHYSPLITbacktrajectories. Theydemon- tions: (1)canweidentifyinourdatathecontributiontoCO
2
stratedthebetterabilityofLPDMtoidentifyregionalpoint andCOconcentrationsofEuropeanorotherremoteanthro-
sources of reactive mercury. Use of LPDM has been pro- pogenic sources predicted by models? (2) Is it possible to
posed (Gerbig et al., 2003; Lin et al., 2003) and eventually identifycontributionsfromforestfireorfromothertypesof
demonstrated (Lauvaux et al., 2008) for meso- to regional- biomass burning? (3) To what extent can we explain the
scaleinversionofCO2surfacefluxes. seasonal variability in CO2, CO and O3 concentrations by
Clustering of backtrajectories has been widely used for variations in atmospheric transport patterns through inten-
the analysis of atmospheric composition measurements at sive campaigns? (4) Is there a consistent signal about the
fixed observatories and has been found to be an efficient regional carbon source/sink distribution emerging through
method for separating air masses with different properties modelanalysisofCO concentrations?
2
(Moody and Galloway, 1988, Dorling et al., 1992, Sirois Section2describesthecampaigns,instruments,modeland
and Bottenheim, 1995, Eneroth et al., 2003). Moody and statistical tools. Section 3 describes the variability of CO ,
2
Galloway(1988)werethefirsttoattemptclusteringofback- CO and O observed above Siberia during the three cam-
3
trajectoriesinordertoassessthewetdepositionofacidcom- paigns. In Sect. 4 we discuss, using cluster analysis, the
poundsovertheBermuda.Thesametechniquehasbeenused tracerconcentrationsobservedinselectedflights. Section5
atthecontinentalscaletoderiveSRRsforO3concentrations investigatesthe“seasonal”(inter-campaign)variationofthe
in central Siberia and supported the conclusion of elevated connectionbetweensourceregionsandtracegasconcentra-
COinairmassesinfluencedbyEuropeanemissions(Pocha- tionsoverSiberia.
nartetal.,2003). Traubetal.(2003)appliedasimpleback-
trajectories’ partitioning technique to the extensive analysis
of measurements obtained from an aircraft. No attempt has 2 Dataandmethod
been made yet, to our knowledge, to use a clustering tech-
niquebasedonLPDMfootprints. 2.1 Campaignsoverviewandinstruments
Aircraft measurements only provide a snapshot of the at-
mosphere at a particular time, but they have the potential The airborne platform is an Antonov-30 dubbed “Optik-E”
toexploretransportprocessesandimpactacrosstransported charteredbyTomsk’sInstituteofAtmosphericOptics.Itwas
plumes(Takegawaetal.,2004;Fehsenfeldetal.,2006;Stohl equippedincollaborationbetweenFrenchandRussianlabo-
etal.,2007b;Methvenetal.,2006;Lelieveldetal.,2002)or ratories for the measurement of CO , CO, O , aerosols and
2 3
in the outflow of massively emitting regions (Jacob et al., meteorological parameters. Intensive campaigns took place
2003). Here we investigate data from three recent YAK- over Siberia between 11 and 14 April 2006, between 7 and
AEROSIB (Airborne Extensive Regional Observations in 10 September 2006 and between 17 and 20 August 2007.
Siberia)intensivecampaignsthatsampledtheSiberiantropo- The flight route consists in a large, continental-scale loop
sphereatdifferenttimesoftheyear(Parisetal.,2008).Mea- from Novosibirsk in central Siberia to Yakutsk in eastern
surement programmes of different types dedicated to CO , Siberia (Fig. 1) that is done in four days. Each campaign
2
Atmos. Chem. Phys.,10,1671–1687,2010 www.atmos-chem-phys.net/10/1671/2010/
J.-D.Parisetal.: Source-receptorrelationshipsforairbornemeasurementsaboveSiberia 1673
Fig.1. Horizontal(toppanel)and3-D(bottompanel)flighttracks. Eachcampaignfollowedasimilarpath. Flightsarenumberedinthe
bottompanel.Thedifferentflightsofeachcampaign(foureach,typicallyoneperday)areshownindifferentcolors.
consistsof4flights(Fig.1)lastingupto8h(limitedbyre- absorptiongascorrelationwithanaccuracyof5ppbor5%.
fuelling needs and airports availability). These flights are The instrument is based on a commercial infrared absorp-
numbered 1 to 4, 5 to 8 and 9 to 12 for the April 2006, tion correlation gas analyser (Model 48C, TEI Thermo En-
September 2006 and August 2007 campaigns respectively. vironment Instruments, USA; Ne´de´lec et al., 2003). O is
3
Theflightrouteisrepeatedwithminormodificationsforeach measuredbyamodifiedUVcommercialfastresponseozone
campaign. Flightswereconductedinallweatherconditions, analyser(ThermoInstrumentsModel49)withaprecisionof
usuallybetween09:00and17:00LT(Parisetal.,2008).Pro- 2ppb,2%foranintegrationtimeof4s.
filesarecollectedasoftenaspossibleastheflighttrackscon-
sistmainlyofascentsupto7kmaltitudeanddescents.Thick 2.2 Atmosphericbackwardtransportmodel
cloud decks were flown over instead of conducting normal
ascentanddescentthroughthem. Atmospheric transport was investigated using the FLEX-
TheAprilcampaignwasdominatedbystagnantflowcon- PART v6.2 LPDM. FLEXPART calculates the trajectories
ditionsandlowsurfacetemperatures(Parisetal.,2008),with of tracer particles using the mean winds interpolated from
mostofthegroundcoveredbysnow.Incontrast,theSeptem- the analysis fields plus random motions representing turbu-
ber2006andAugust2007campaignswerecharacterizedby lence (Stohl and Thomson, 1999). Results presented here
westerlyflow,warmerandhighlyvariabletemperatures,and use ECMWF analysis fields although both ECMWF and
significantfrontalactivityintheflightarea. GFS(NOAA/NCEP)fieldswereusedforapriorassessment
Detailed instrument descriptions can be found in Paris oftransporterror(seehttp://zardoz.nilu.no/∼andreas/YAK/).
et al. (2008); we give here only a short description. CO Formoistconvectivetransport,FLEXPARTusesthescheme
2
is measured by a modified Non Dispersive Infrared Anal- ofEmanuelandZivkovic-Rothman(1999),asdescribedand
yserbasedonacommercialLi-Cor6262,withaccuracyand testedbyForsteretal.(2007). Abackwardsimulationmode
precision of 0.15ppm, obtained by periodical in-flight cal- is available, described in more detail by Stohl et al. (2005)
ibration against WMO-referenced reference gases bracket- and Seibert and Frank (2004). Here the backward method
ing atmospheric concentrations, and regulation of pressure, isusedtoanalysetransportpathwaysfrompotentialfluxre-
temperature and flow in the cells. CO is measured by IR gions to the receptor position. Each simulation consists of
www.atmos-chem-phys.net/10/1671/2010/ Atmos. Chem. Phys.,10,1671–1687,2010
1674 J.-D.Parisetal.: Source-receptorrelationshipsforairbornemeasurementsaboveSiberia
Fig. 2. Regions chosen for data reduction prior to the clustering analysis. The black line shows the fight track during the campaigns.
AnthropogenicCOemissionsfromtheEDGARdatabasearealsoshowninlogarithmicgreyscale,ing(CO)m−2yr−1,toillustratedifference
betweenregions.
40000 particles released whenever the aircraft has moved singlecampaignimprovestheseparationcapabilityoftheal-
0.15◦ in latitude or longitude, or 10hPa in altitude. This gorithm,whereasdefiningj overthethreecampaignsallows
corresponds to about one minute of sampling or to a layer inter-campaigncomparison.
∼100m thick during ascent or descent. Released particles The M boxes are identified a priori as regions with spe-
were followed 10 days backward in time. Gridded PESs cific sources or sink for the species relevant to this study.
in three vertical levels (0–300m, 300–3000m, and 3000– Such identification is based either on case studies from the
50000m) are stored at 1◦×1◦ resolution every 24h. The AprilandSeptember2006campaigns(Parisetal.,2008),or
fraction of stratospheric air is recorded as the percentage on the questions to be addressed with this dataset, such as
of all particles originating from the stratosphere (i.e. have thetransportofEuropeanemissionstoSiberia. Theregions
residedinthestratosphereanytimewithinthelast10days). identifiedare:
2.3 Clusteranalysis – Europe: Europeanpollutantemissionsandoutflow;
– WesternRussia: EuropeanRussianpollutantemissions
Airbornecampaignscandeliverdetailedinformationonthe
andoutflow;
atmosphericstateataparticulartime. Togeneralizethisin-
formation and to organize the YAK-AEROSIB data set, a – NE China and Koreas: extratropical cyclones can lift
methodissoughtthatgroupsthedataaccordingtocommon NEAsianpollutantovercentraloreasternSiberia;
transportproperties. Clusteranalysisissuchanexploratory
tool, which sorts multivariate data into groups as dissimilar – Central Asia: Central Asia has a different ecosystem
aspossiblebutwhosepropertiesarenotknownapriori. We and low anthropogenic sources but it is a potential
seektoinvestigatetowhichextentfootprintscanexplainthe source of biomass burning (e.g. van der Werf et al.,
air mass chemical composition in CO , CO and O . Foot- 2006);
2 3
prints are implemented as 10-day summed spatial distribu-
– Arctic: Arctic air from beyond the Arctic front can be
tion of PES obtained from FLEXPART, complemented by
observedoverSiberiaduringcoldairoutbreaks;
10-dayaveragedrelativecontributionsfromthestratosphere,
defined as the region with potential vorticity >2 PVU. To – Localregion: coverstheareadirectlyflownoverbyour
facilitate analysis we have reduced the number of variables aircraft;
from the gridded PES by further summing these PES over
large regions of interest for our study (Fig. 2). A vector – Stratosphere:Anextravariableisaddedwhichaverages
" #T the10-dayproportionofstratosphericparticles.
10 10 10
x = Pr ,Pr ,...,Pr ofdailyPESr inbox
j 1,j 2,j M,j k
d=1 d=1 d=1 The boxes’ geographical extent is the result of a trade-off
m (m=1, 2, ..., M) is associated with each consecutive re- betweenmaximizingdifferencesbetweentransportfromthe
ceptorpositionj. Thistimeseriesx willconstituteourset various boxes, and an insufficient number of particles if
j
ofj realizationsofM variablestocluster. Definingj overa boxes are chosen to be too small. There are large regions
Atmos. Chem. Phys.,10,1671–1687,2010 www.atmos-chem-phys.net/10/1671/2010/
J.-D.Parisetal.: Source-receptorrelationshipsforairbornemeasurementsaboveSiberia 1675
notcoveredbyanybox.Theoretically,thiscouldmeanthata
footprint is located entirely outside all of the boxes. How-
ever, the footprints normally cover quite large regions, so
theyalwaysoverlapwithoneormoreoftheboxes. Bysep-
arating the boxes from each other, differences between the
variousfootprintscanbemaximizedbytheclusteringalgo-
rithm. Choosingboxesthatarecontiguousmakethesepara-
tionmoresensitivetohorizontaltransporterror, andassoci-
atedPESlesscontrasted.
More remote regions are less connected (few or no parti-
cles reaching the region) by transport to the measurements.
As a result, remote regions’ average footprint distributions
are skewed toward zero. To account for this skewness and
different region sizes, normalization is applied to the re-
gions’ average footprint time series x according to x =
N
(x−p )/x¯ wherep isthe5thpercentileofx andwas
0.05 0.05
foundtooptimallyseparateclusters. Therobustnessofdata
reductionintoregionsanddatanormalizationwastestedby Fig.3. Silhouetteindexoftheclusteringalgorithmseparationca-
runningtheclusteringalgorithmforvaryingregionsizes,lo- pability as a function of the number of clusters for all data (thick
line)andseparatedcampaignbycampaign(thinlinewitherrorbars
cations and number of regions (bootstrapping), and various
showing±1stddev).
normalizationfunctions.
The K-means algorithm implemented in the MATLAB
software’sStatisticstoolboxisusedforclustering. K-means
very high confidence in reproducibility of cluster centroids
is a classic partitioning (non-hierarchical) algorithm (see
determination. Thenormalizationandthemetricswerealso
e.g. Wilks, 2006) which attempts to separate observations
found to have a strong impact on the result. In this respect,
into a fixed number of groups. (1) It defines K centroids
clusteranalysisisnotanobjectiveclassificationtechniqueas
withinitialrandompositionvectorofsizeM=7(numberof
thecriteriasetfortheclusteringhavetobedeterminedsub-
regions); then(2)associateseachpointx tothenearest(in
j jectively.
Euclidianmetrics)clustercentroid,(3)movesthecentroidto
the centre of the cluster and (4) repeats steps 2 and 3 until
convergence is achieved. In the process, any empty cluster
3 Campaignobservations
is discarded. In order to maintain the number of clusters, a
pointhavingmaximumdistancetoitscentroidissingledout
3.1 CO measurements
2
anddeclaredcentroidofanewcluster.
Figure3showsthe“silhouette”index(S∈[−1,1]),amea- Figure 4a, d and g shows the observed CO concentra-
2
sureoftheseparationcapabilityoftheclustering,asafunc- tions averaged for each of the 4 flights of each campaign.
tionofthechosennumberofclusters. TheSindexisdefined Among our surveys in 2006 and 2007, the minimum BL
as: CO concentrationshavebeenobservedinAugust2007(av-
2
S= 1 XN dnearest−dcentroid (1) einraggcea3m6p6apigpnmsbitelsohwou1ldkmbednuoritnegdFthliagthatt1m2o).spWhehriilcecCoOmpianr--
N i=1max(dnearest,dcentroid) creasesby∼1.4ppmyr−1duetoglobalanthropogenice2mis-
with N number of points, and where point i has a dis- sions. The flight where this CO minimum was measured
2
tance to the nearest cluster d and distance to its as- occurredmostlyoveraforestedregion, withsparseagricul-
nearest
sociated centroid d . The S index informs about how tural landscapes and industrial centres (Kemerovo, Novosi-
centroid
close a point is to its cluster’s centroid, relative to the birsk) nearby. This CO minimum was coincident with a
2
other nearest cluster centroid (Matlab R2006b documenta- veryhighBLtop(upto∼3.5kma.g.l.),asdeducedfromhu-
tion, Statistics toolbox, http://www.mathworks.com/access/ midityandCO gradients. AttheZotinotalltower,average
2
helpdesk/help/toolbox/stats/silhouette.html). UsingtheS in- daytime CO mixing ratios measured in the boundary layer
2
dex,theminimumnumberk ofclusterswasfoundtobeop- were390ppmand382ppminAprilandSeptember2006,re-
timally set to 6 in the global clustering option, and 4 in the spectively(Kozlovaetal.,2008). Kozlovaetal.’sSeptember
campaignbycampaignoption(Fig.3). valueissignificantlyhigherthanthemixingratioobservedin
As the clustering algorithm is sensitive to a priori (ran- thelowerpartofourverticalprofiles(Fig.4d;valuesbetween
dom) position of the centroids, the process is repeated 20 371and377ppmforFlights5and8,theclosesttotheZotino
timesandthebestresultonly(optimizingtheseparationbe- site),possiblyduetoregionalvariabilityinsourcesandsinks
tweencentroids)isusedintheanalysis. Thisalsoensuresa of CO . In the free troposphere, CO mixing ratios were
2 2
www.atmos-chem-phys.net/10/1671/2010/ Atmos. Chem. Phys.,10,1671–1687,2010
1676 J.-D.Parisetal.: Source-receptorrelationshipsforairbornemeasurementsaboveSiberia
Fig.4.ProfilesofmedianCO2,COandO3mixingratiosforeachflightoverthe3campaigns(toprow:April2006,middle:September2006,
bottom:August2007).Thehorizontallinesjointhe10thand90thpercentilesofeachaltitudelevel.
between 377 and 379ppm in August 2007, comparable (up above the front. Except for local emissions in the lower
totheinterannualtrend)tothe377–378ppmrangemeasured troposphere, the August 2007 campaign has shown low CO
inSeptember2006butmuchlowerthaninApril2006(387– mixing ratios (background 100±5ppb) with little variation,
388ppm), reflecting the hemispheric CO seasonal cycle. although satellite fire detection has shown numerous fires
2
CO mixingratiosinApril2006wereashighas392ppmin in temperate and boreal Eurasia. High ozone values (up to
2
polluted filaments encountered between 5 and 6km altitude 90ppb) were found near the flight ceiling during the Au-
(Parisetal.,2008). gust 2007 campaign (Fig. 4i), and the O vertical gradient
3
wasthesteepestofthethreecampaigns.
3.2 COandO measurements
3
4 Cluster-basedSRRrelationships: casestudies
Figure4bshowstheaverageCOprofilesforeachflightdur-
ingtheApril2006campaign. Throughoutthiscampaignthe
Inthissectionweapplytheclusteringalgorithmanddiscuss
averageCOmixingratiowas175ppb. Anumberofplumes
theresultstoexaminethemeasurementsforaselectedflight
with higher CO mixing ratios were repeatedly observed. A
fromeachofthethreecampaigns(Flight1on11April2006,
plumewithCOupto220ppbwasfoundtobeassociatedto
Flight 5 on 7 September 2006 and Flight 9 on 17 Au-
upliftofapollutedairmassfromNEChinabyawarmcon-
gust 2007). All three flights are going from Novosibirsk to
veyorbelt(Parisetal.,2008). Duringthiscampaign,theO
3
Myrni (green track in Fig. 1). The three flights took off at
profileshowsarelativelyflatverticalprofile(from∼60ppb
12:00, 09:00 and 09:30LT, and landed at 21:00, 19:00 and
at 6km to ∼45ppb at 1km, Fig. 4c) compared to summer
18:15LT respectively. The clustering is applied on a single
campaigns.
campaign-basis. Theclusteringbeingindependentfromthe
During the September 2006 campaign the average CO
observedconcentrations,statisticalseparationoftheconcen-
concentration was about 100ppb, with a large variability
trationsacrossdifferentclustersisinterpretedasavalidation
(Fig.4e),duetothepresenceofairmasseswithverydistinct
oftheSRR.
origins. For example during Flight 5 the aircraft crossed a
warmfrontacrosswhichstrongcompositiongradientswere
found, with CO mixing ratios up to 150ppb immediately
Atmos. Chem. Phys.,10,1671–1687,2010 www.atmos-chem-phys.net/10/1671/2010/
J.-D.Parisetal.: Source-receptorrelationshipsforairbornemeasurementsaboveSiberia 1677
Fig.5. AveragefootprintmapsfordatabelongingtoclustersA–DthroughoutFlight1(11April2006). Logarithmiccolorscaleshowsthe
log10ofresidencetime,i.e.thePESbelow300m.ThenumberofelementsN inaparticularclusterisgivenaboveeachmap.
4.1 Flight1: pollutionfromNorth-easternChina Figure 6 shows the distribution (quartiles and median) of
CO ,CO,O andwatervapourmeasurementsassociatedto
2 3
Figure 5 illustrate the ability of the clustering algorithm to eachcluster. AssociatedaltitudeisgiveninFig.6e,andthe
separatetransportpatterns(andPES)betweenclustersiden- four (A–D) clusters’ centroid (in std dev-normalized coor-
tifiedthroughoutthe11April2006flight. Thepanelsshow dinates) position vector is given in Fig. 6f. Each cluster’s
theaveragefootprint(airmassresidencetimebelow300m) centroidpositionvectorcanberelatedtotheclusteraverage
obtainedfromthepopulationofbackwardplumesassociated footprintmapinFig.5.
to each of the 4 clusters. Cluster A appears to gather foot- The data associated to Cluster D (sensitive to NE China
printsincludingairtransportedfromtheEuropeanboundary emissions) are encountered between 5 and 6.5km altitude
layerand,secondarily,localinfluence. ClusterBisrelatedto (Fig. 6e) and have a median CO concentration of 180ppb
stronglocalinfluence(brownanddarkredinFig.5b)andto (Fig. 6b), well differentiated from other clusters. On the
Arcticairmasses. opposite, CO values (median 388.6ppm; Fig. 6a) are not
2
Footprints classified as Cluster C (Fig. 5c) dominate the distinguishably higher. This cluster identifies the trans-
flight (73%). This cluster is characterized by a strong portofNorthEasternChinaemissionstoeasternSiberiaon
residence time in the free troposphere, and exhibits the 12April2006. TheabilityoftheLPDMtoresolvethinlay-
highest stratospheric influence compared to the other clus- ersassociatedtosynopticfeaturesisausefuladvantageover
ters (Fig. 6f, green line). Potential emission sensitivity is Eulerianmodels.
typically stronger over South East and East Siberia. A Cluster C is ubiquitous throughout Flight 1 and has a
marginalnumber(1%)offootprintsareassociatedtoCluster higherstratosphericinfluencereflectedinitshighO concen-
3
A(Fig.5a),whereairmassesaresensitivetoEuropeanemis- trations(upto85ppb).ClusterBhashighCOconcentrations
sions. ClusterB(20%;Fig.5b)gathersasignificantamount (median174ppb)correspondingtolocalemissionsensitivity
ofobservationsfromairmasseshavingresidedovernorthern ofthelowertroposphere.
SiberiaandtheArctic. FootprintsinClusterD(5%;Fig.5d)
areexposedtosurfaceexchangeinsouthernSiberiaandthe 4.2 Flight5: firesandCO uptake
2
Kazakhstan. The four well-differentiated footprint maps il-
lustratethevalidityofcluster-basedpartitioning. Figure7showstheaveragefootprintforeachclusterforthe
7September2006flight. Asintheprevioussection,cluster-
ingwasperformedonthewholecampaign(here,theSeptem-
ber 2006 campaign). Data within Cluster B (43%; Fig. 7b)
www.atmos-chem-phys.net/10/1671/2010/ Atmos. Chem. Phys.,10,1671–1687,2010
1678 J.-D.Parisetal.: Source-receptorrelationshipsforairbornemeasurementsaboveSiberia
Fig.6. (a)Boxplotsofmedianandinter-quartilerange(IQR)forCO2concentrationsmeasuredduringFlight1(11April2006)associated
toeachcluster(clustersA–Dinabscissa,samedenominationasinFig.5). OutlierswithintheIQRareincludedinthewhiskers; outliers
beyondtheIQRareshownasdots. (b)SameforCO.(c)SameforO3. (d)Sameforwatervapourmixingratio. (e)Sameforaltitude. (f)
Clustercentroidpositionvectorforeachcluster.Graybarsindicatethepopulationofeachcluster(numberofsamples).
were mostly associated to FT zonal flow, with limited sen- TheEuropean(ClusterC)airmasshasamedianCOcon-
sitivitytopotentialsurfaceemissionsoverEuropeanRussia centrationof110ppb(Fig.8b)atamedianaltitudeof5km,
and the Black Sea area. This cluster has the highest strato- tobecomparedto105ppbforthemoreubiquitousClusterB.
spheric signature of the dataset (Fig. 8f). Cluster D is also Althoughthisisonlyasinglecase,itprovidesacaseofeast-
ubiquitous in this flight (39%; Fig. 7d). It gathers footprint wardEuropeanpollutionexportinthemid-troposphere,sug-
with sensitivity highest over Kazakhstan, between Caspian gestingthatthisexportdoesnotalwaysfollowthelowlevel
and Aral Seas. A large number of fires detected by ATSR advection pathway described by modelling studies (Wild
firecountinthisregion,probablyofagriculturalorigin,were et al., 2003; Stohl et al., 2002; Duncan and Bey, 2004).
injectedintheFTastheregionwassweptbyafront(Pariset Other characteristic of this air mass include relatively high
al., 2008). ClusterA(14%; Fig.7a)isclearlyassociatedto CO (median 376ppm) and an O median concentration of
2 3
anArcticairmasschannelledsouthwardintheBL.ClusterC 53ppb.
(4%;Fig.7c)reflectsadvectionofairmassesincontactwith
potentialEuropeanemissions.
Atmos. Chem. Phys.,10,1671–1687,2010 www.atmos-chem-phys.net/10/1671/2010/
J.-D.Parisetal.: Source-receptorrelationshipsforairbornemeasurementsaboveSiberia 1679
Fig.7.SameasFig.5forFlight5on7September2006.
Figure 8b shows that Cluster D has the most elevated as CO oxidation within the plume and CO assimilation by
2
CO concentration, up to 158ppb (median 121ppb). It is intact vegetation, and reflect the difficulty to retrieve fire
also associated to high O (median 55ppb) but CO mix- emissionratiowiththeairmassintegratingsourcesandsinks
3 2
ing ratios are not particularly high (375ppm). High CO upstreamoftheburningarea.
and O3 are likely to reflect the sensitivity of Cluster D Cluster A gathers observations affected by Arctic air
(Fig. 7d) to the biomass burning detected in the Caspian (Fig. 8f) with low CO , CO and O concentrations values
2 3
region (especially in Northern Kazakhstan at ∼50◦N, be- (medianvalues373ppmCO , 103ppbCOand34ppbO ).
2 3
tween 50 and 70◦E). To illustrate our analysis we focus It is typically encountered at low altitude (1km) within the
now on individual biomass burning plumes found through BL and consists of moist air (Fig. 8d). Advection of Arctic
this flight. The plume with the highest CO concentration airinthecampaigndomainsampledduringSeptember2006
(144±14ppb) exhibits a high CO-O3 correlation (R2=0.68, occurredintheboundarylayer. FLEXPARTresultsindicate
regression slope 0.24molmol−1). This plume was deter- thattotalPESovertheArcticwas5daysormoreinthelast
minedtobeabiomassburningplumeusingFLEXPARTand 10 days for the majority of data in Cluster A. The low CO
satellitedetectedfirehotspots,withthefirehotspotslocated andCO concentrationssuggestastrongisolationfrompol-
2
overKazakhstan(http://zardoz.nilu.no/∼andreas/YAK/),and lution sources in the air mass history. During the five days
hence is illustrative of Cluster D. The positive regression ofmeridiantransportfromtheArctic,exposuretouptakeby
slope is usually associated to net O3 production within the highlatitudevegetationwithpermanentdaylightperiodand
plumefromcombustion-generatedprecursorslikeCO(Pfis- tosurfacedepositionmustalsohavecontributedtolowCO
2
teretal.,2006;ValMartinetal.,2006). Inanotherbiomass andlowO valuesrespectively.
3
burningplumetheregressionslopewasfoundtobenegative
(−0.10molmol−1). Both of these plumes are encountered 4.3 Flight9: CO uptake,forestfireandstratospheric
2
at2–2.5kmaltitude. TheO3-COregressionslopesarecom- input
parable to those found in the literature for boreal forest fire
plumes(range0.05–0.30molmol−1;ValMartinetal.,2006; Figure9showstheaveragefootprintforeachofthe4clusters
Realetal.,2007andreferencestherein). forthe14August2007flight. ClusterBindicatesthedomi-
Both of these plumes have a high CO-CO correlation nantfootprintpatternforthisflight(65%; Fig.9b). Ithasa
2
(respectively R2=0.80 and 0.78) with negative CO-CO re- strongstratosphericcomponent(Fig.10f)andtheweaksur-
2
gressionslopesof−0.06and−0.09molmol−1 respectively. facefootprintdensityisspreadoverWesternSiberia. Cluster
Thesenegativeslopesreflectcompetinglossprocessessuch C(17%;Fig.9c)hasahighsensitivitynorthofLakeBaikal.
www.atmos-chem-phys.net/10/1671/2010/ Atmos. Chem. Phys.,10,1671–1687,2010
1680 J.-D.Parisetal.: Source-receptorrelationshipsforairbornemeasurementsaboveSiberia
Fig.8.SameasFig.6forFlight5on7September2006.
ClusterA(9%;Fig.9a)iscentredovernorthEuropeanRus- August 2007 vs. 34ppb in September 2006. In Siberia FT
sia and Scandinavia while Cluster D (9%; Fig. 9d) reflects CO concentrations are minimum in August (Ramonet et
2
airofArcticorigin,comparabletoClusterAintheSeptem- al., 2002), and September concentrations increase by about
ber2006case. 2ppmintheFTand5ppmintheBLrelativetoAugust.
Figure10b–cshowsthatClusterChasthehighestCOval- Cluster B shows the highest stratospheric component for
ues(median122ppb,upperquartile191ppb). SuchhighCO thisflight.ThehighestO3values(upto91ppb)associatedto
valuespointtoregionalfireinfluence, asseenonATSRfire itareconsistentwiththe“stratospheric”classification. High
atlasfortheperiodofthecampaign(notshown).Thealtitude O3 concentrations (up to 90ppb; Fig. 11) are associated to
rangeisverylarge(1–6km). ThehighestCOconcentrations more than 50% of freshly (2 days) exported stratospheric
within this cluster (300ppb) were observed in two distinct particles.
layersinthelowerFT(seeFig.11).
Comparing CO concentrations in the low-altitude Arc-
2
ticclustersofSeptember2006(ClusterA)andAugust2007
(Cluster D), the latter (median 369ppm) is lower by
∼4ppm (despite the anthropogenic inter-annual increase of
∼1.4ppmyr−1). The same is found for O with 25ppb in
3
Atmos. Chem. Phys.,10,1671–1687,2010 www.atmos-chem-phys.net/10/1671/2010/