NASA Technical Reports Server (NTRS) 20120010247: Effect of Wind Speed on Aerosol Optical Depth over Remote Oceans, Based on Data from the Maritime Aerosol Network

2.5 MB·English

by  NASA Technical Reports Server (NTRS)

Atmos. Meas. Tech.,5,377–388,2012
Atmospheric
www.atmos-meas-tech.net/5/377/2012/
Measurement
doi:10.5194/amt-5-377-2012
©Author(s)2012. CCAttribution3.0License. Techniques
Effect of wind speed on aerosol optical depth over remote oceans,
based on data from the Maritime Aerosol Network
A.Smirnov1,2,A.M.Sayer2,3,B.N.Holben2,N.C.Hsu2,S.M.Sakerin4,A.Macke5,N.B.Nelson6,Y.Courcoux7,
T.J.Smyth8,P.Croot9,P.K.Quinn10,J.Sciare11,S.K.Gulev12,S.Piketh13,R.Losno14,S.Kinne15,and
V.F.Radionov16
1SigmaSpaceCorporation,Lanham,Maryland,USA
2NASAGoddardSpaceFlightCenter,Greenbelt,Maryland,USA
3GoddardEarthScienceTechnologyandResearch–GESTAR,Columbia,Maryland,USA
4InstituteofAtmosphericOptics,Tomsk,Russia
5LeibnizInstituteforTroposphericResearch,Leipzig,Germany
6UniversityofCalifornia,SantaBarbara,California,USA
7Universite´ delaRe´union,SaintDenisdelaRe´union,France
8PlymouthMarineLaboratory,Plymouth,UK
9LeibnizInstituteofMarineSciences,Kiel,Germany
10NOAAPMEL,Seattle,Washington,USA
11LaboratoiredesSciencesduClimatetdel’Environnement,Gif-sur-Yvette,France
12P.P.ShirshovInstituteofOceanology,Moscow,Russia
13UniversityoftheWitwatersrand,Johannesburg,SouthAfrica
14Universite´ deParis7etUniversite´ deParis12,Creteil,France
15InstituteforMeteorology,UniversityofHamburg,Hamburg,Germany
16ArcticandAntarcticResearchInstitute,SaintPetersburg,Russia
Correspondenceto: A.Smirnov([email protected])
Received: 22November2011–PublishedinAtmos. Meas. Tech.Discuss.: 5December2011
Revised: 3February2012–Accepted: 6February2012–Published: 17February2012
Abstract. TheMaritimeAerosolNetwork(MAN)hasbeen satellite-basedAODmeasurements. Thebasicrelationships
collecting data over the oceans since November 2006. The are similar for all the wind speed sources considered; how-
MANarchiveprovidesavaluableresourceforaerosolstud- ever,thegradientoftherelationshipvariesbyaroundafactor
iesinmaritimeenvironments. Inthecurrentpaperweinves- oftwodependingonthewinddataused.
tigatecorrelationsbetweenship-borneaerosolopticaldepth
(AOD) and near-surface wind speed, either measured (on-
board or from satellite) or modeled (NCEP). According to 1 Introduction
ouranalysis,windspeedinfluencescolumnaraerosoloptical
depth, although the slope of the linear regression between The World Ocean is the largest source of natural aerosol.
AODandwindspeedisnotsteep(∼0.004–0.005), evenfor Accurateestimationofsea-sprayaerosolproduction, evolu-
strong winds over 10ms−1. The relationships show signif- tion and removal processes is important for understanding
icant scatter (correlation coefficients typically in the range theEarth’sradiationbudget, aerosol-cloudinteractions, and
0.3–0.5); the majority of this scatter can be explained by visibilitychanges(LathamandSmith,1990;O’Dowdetal.,
the uncertainty on the input data. The various wind speed 1999; Haywood et al., 1999; de Leeuw et al., 2000). The
sources considered yield similar patterns. Results are in windspeedisthemajordriverbehindtheproductionofnat-
good agreement with the majority of previously published ural marine aerosol (Lewis and Schwartz, 2004). The ma-
relationshipsbetweensurfacewindspeedandship-basedor rine aerosol concentration and size distribution are strongly
PublishedbyCopernicusPublicationsonbehalfoftheEuropeanGeosciencesUnion.
378 A.Smirnovetal.: Effectofwindspeedonaerosolopticaldepthoverremoteoceans
Fig.1. LatitudinaldependenceofAODseriesacquiredatleast200kmfromthenearestlandmass(a),andAODdependenceonship-based
windspeed(b)duringtheFebruary–April2008cruiseoftheR/VPolarstern.
dependent on wind speed (Blanchard and Woodcock, 1980; partly,byuncertaintiesinfoamformationanditslatitudinal
Gathman, 1982; Lovett, 1978), however, the dependence of distribution(AnguelovaandWebster,2006),byaprocessof
columnaraerosolopticaldepth(AOD)onwindspeedismore quality control that excludes some residual cloud contami-
difficult to detect and quantify, because of scores of differ- nation(ZhangandReid,2010),bytheaccuracyofradiative
ent factors influencing AOD (Smirnov et al., 1995). Estab- transfer models used (Melin et al., 2010), and by more ac-
lishingcorrectrelationshipsbetweenAODandnear-surface curate accounting for surface reflectance effects in satellite
wind speed will help tune global aerosol transport models retrievals(Sayeretal.,2010).
(Jaegleetal.,2011;Madryetal.,2011;FanandToon,2011), Therefore it is useful to utilize the available archive
atmosphericcorrectioninocean-colorstudies(Zibordietal., of ship-based AOD measurements over the oceans ac-
2011),validateAODsretrievedfromsatellitemeasurements quired within the framework of Maritime Aerosol Network
(Kahn et al., 2010; Kleidman et al., 2012), and understand (Smirnov et al., 2009), and analyze AOD in conjunction
biogeochemicalcycles(MeskhidzeandNenes,2010). with information on near-surface wind speed from vari-
Recentlyanincreasedinterestinaerosolopticaldepthover ous sources: measured onboard, simulated by the National
the oceans and its dependence on wind speed manifested CenterforEnvironmentalPrediction(NCEP),andestimated
itself in a number of publications. Satellite-derived and frommeasurementstakenbytheAdvancedMicrowaveScan-
coast or island acquired AODs have been studied by Mulc- ningRadiometer(AMSR-E)aboardAquasatellite.
ahyetal.(2008), Glantzetal.(2009), Lehahnetal.(2010),
Huang et al. (2010), O’Dowd et al. (2010), Kiliyanpilakkil
and Meskhidze (2011), Grandey et al. (2011), Adames et
al.(2011),andSayeretal.(2012).Power-lawandlinearrela- 2 Datasetsusedinthisstudy
tionshipsbetweenAODandwindspeedwereestablishedal-
thoughsamplingissues,uncertaintiesinretrievalalgorithms, MANaccumulatedmorethan2500daysofship-basedAOD
and/orinfluenceofthechosenislandlocationsgaveanindi- measurementsoveraperiodofseveralyears(Smirnovetal.,
cationthattheproblemisfarfrombeingsolvedandthereis 2009, 2011). MAN deploys hand held Microtops II sun-
notyetconsensus. photometers(Morysetal.,2001)andutilizescalibrationand
Satellite-based measurements are undoubtedly the only dataprocessingprocedurestraceabletoAERONET(Holben
tool(atleastatpresent)forglobalaerosolopticaldepthcov- et al., 1998, 2001). The estimated uncertainty of the opti-
erage. Howeverbecauseofexistingsatelliteretrievalbiases cal depth in each channel does not exceed ±0.02 (Knobel-
(Smirnovetal.,2006,2011)theground(ocean)-basedtruth spiesseetal.,2004),primarilyduetointer-calibrationagainst
is needed to correct or constrain them. For example, in the AERONET reference CIMEL instruments that are accurate
southern latitudes (south of 40◦) the sunphotometer AODs to within ∼0.01 in the visible and near-infrared (Eck et al.,
are low compared with satellite retrievals (Smirnov et al., 1999). ThusMANprovideshigh-qualityAODswithknown
2006, 2011). This discrepancy can be explained, at least uncertainty. A public domain web-based archive dedicated
Atmos. Meas. Tech.,5,377–388,2012 www.atmos-meas-tech.net/5/377/2012/
A.Smirnovetal.: Effectofwindspeedonaerosolopticaldepthoverremoteoceans 379
Fig.2. LatitudinaldependenceofAODdailyaveragesusedinthisstudy(a),andlatitudinaldependenceofcorrespondingdailyaveraged
ship-basedwindspeed(b).
Fig. 3. Differences between NCEP and ship-based wind speed as a function of ship-based wind speed (a) and wind speed frequency of
occurrences(%)foreachsubset(b).
to the network activity can be found at: http://aeronet.gsfc. thisprocedurebackwardsintime,weadditionallycomputed
nasa.gov/new web/maritime aerosol network.html. windspeedsaveragedoverthe24hperiodpriortoeachAOD
The ship-based meteorological data collected onboard measurement.
were provided by the cruise PIs. Meteorological measure- The Advanced Microwave Scanning Radiometer-Earth
mentsweremadewiththestandardequipmentatleasthourly. Observing System (AMSR-E) instrument on the NASA
Then linear interpolation was applied to match in time sun- EarthObservingSystem(EOS)Aquasatelliteprovidesnear-
photometermeasurements. surface wind speed (Wentz and Meissner, 2007). These
The National Center for Environmental Prediction data are provided at a spatial resolution of 25km, sepa-
(NCEP)windspeeddatausedwas1◦ by1◦ horizontalreso- rately for daytime and nighttime overpasses. In this study,
lutionoutputevery6h(Derberetal.,1991). Inouranalysis, thedatapointwhichtheMANmeasurementlieswithinwas
foreachmeasuredAODtheNCEPnear-surface(10m)wind used. Because the AMSR-E sampling is spatially incom-
speed data points were linearly interpolated in space and plete, some MAN data lacked a corresponding AMSR-E
time to provide the “instantaneous” wind speed. Repeating windspeedretrieval.
www.atmos-meas-tech.net/5/377/2012/ Atmos. Meas. Tech.,5,377–388,2012
380 A.Smirnovetal.: Effectofwindspeedonaerosolopticaldepthoverremoteoceans
Fig.4.ScatterdensityhistogramsofAODat500nm(series)(a–e)andAngstromparameter(f)versusthesurfacewindspeed.
TheinfluenceofwindspeedonAODinthewholeatmo- all meteorological parameters are simply the same over the
spheric column is a very difficult problem. A link between range of wind speeds considered. Discriminating between
opticalturbidityandparticlegenerationbywindisnoteasy air masses permits a more rigorous analysis of the link be-
todetect, sinceitcanbemaskedbythebackgroundaerosol tween wind speed and optical depth (Smirnov et al., 1995).
(ofcontinentaloriginincoastalareas,forexample). Accord- ThecorrelationsbetweenAODandwindspeedinmaritime
ingly, surface generation effects can be clearly noticed only tropicalairmasseswerefoundtobesignificantlylargerthan
whenmeasurementsaretakeninareasonablytransparentat- those obtained in a study of the same Pacific Ocean data
mosphere. Ideally a relationship between spectral aerosol (Villevalde et al., 1994), where no air mass discrimination
optical depth and wind speed needs to be ascertained in the was made. This means that the correlation coefficient in-
same air mass in order to minimize the influence of other creased when the data were characterized by more uniform
meteorological parameters on optical properties, or when atmosphericconditions.
Atmos. Meas. Tech.,5,377–388,2012 www.atmos-meas-tech.net/5/377/2012/
A.Smirnovetal.: Effectofwindspeedonaerosolopticaldepthoverremoteoceans 381
Fig.5.ScattergramsofdailyaveragedAODat500nm(a–e)andAngstromparameter(f)versusthesurfacewindspeed.
www.atmos-meas-tech.net/5/377/2012/ Atmos. Meas. Tech.,5,377–388,2012
382 A.Smirnovetal.: Effectofwindspeedonaerosolopticaldepthoverremoteoceans
Table1.Listofcruises,cruiseareas,andnumberofmeasurementdaysusedinouranalysis.
Cruisename Cruisearea Timeperiod N ofdays PI
R/VAkademikFedorov2005-2006 SouthAO Dec2005 1 B.HolbenandS.Sakerin
R/VAkademikFedorov2006–2007 SouthAO Dec2006 5 B.HolbenandS.Sakerin
R/VAkademikFedorov2007–2008 SouthAO,SouthIO Dec2007–Apr2008 17 B.HolbenandS.Sakerin
R/VAkademikFedorov2008–2009 SouthAO Dec2008 1 B.HolbenandS.Sakerin
R/VAkademikFedorov2009–2010 SouthernO,SouthAO Dec2009;Feb2010 2 B.HolbenandS.Sakerin
R/VAkademikIoffe2009 SouthAO Nov2009 9 B.HolbenandS.Gulev
R/VAkademikIoffe2010 NorthAO Sep2010 4 B.HolbenandS.Gulev
R/VAkademikSergeyVavilov SouthAO Nov–Dec2004 17 B.HolbenandS.Sakerin
RRSJamesClarkRoss2008 SouthAO Oct–Nov2008 8 B.HolbenandT.Smyth
RRSJamesCook2009 SouthAO Nov2009 10 B.HolbenandT.Smyth
RRSJamesCook2010 SouthAO Nov2010 6 B.HolbenandT.Smyth
R/VKnorr2008 NorthAO Mar–Apr2008 5 P.Quinn
R/VMarionDufresne2007 SouthIO Nov–Dec2007 14 B.HolbenandJ.Sciare
R/VMarionDufresne2008 SouthIO Nov–Dec2008 11 B.HolbenandR.Losno
R/VMarionDufresne2009 SouthIO Nov–Dec2009 8 Y.Courcoux
R/VMarionDufresne2010 SouthIO Jan,Aug–Sep2010 18 Y.Courcoux
R/VMelville2009–2010 SouthPO Jan–Feb2010 8 B.HolbenandN.Nelson
R/VPolarstern2008 SouthernO,SouthAO Feb–Apr2008 16 B.HolbenandP.Croot
R/VPolarstern2008 SouthAO Apr–May2008 9 B.HolbenandA.Macke
R/VPolarstern2009 SouthAO Apr–May,Nov2009 27 B.HolbenandA.Macke
R/VPolarstern2010 SouthAO Apr2010 10 B.HolbenandS.Kinne
R/VRonaldH.Brown2007–2008 NorthandSouthPO Dec2007–Feb2008 26 B.HolbenandN.Nelson
R/VRonaldH.Brown2008 SouthPO Oct–Nov2008 10 P.Quinn
M/VSAAgulhas SouthernO,SouthAO Dec2007–Jan2008 13 B.HolbenandS.Piketh
In other words, the relationship between AOD and wind
speed depends on many factors we simply do not know or
cannotfullyaccountfor(atleastempirically). Agoodexam-
pleispresentedinFig.1aandb. TheR/VPolarsterncruise
considered took place in the winter of 2008 in the South
Atlantic and Southern Ocean (Dr. Peter Croot was a PI for
AODmeasurements). Figure1ashowsthelatitudinaldepen-
dence of AOD series (a series can have one or more mea-
surementspoints,typicallyfiveormore,madewithagapof
under2min;seeSmirnovetal.,2009fordetails)acquiredat
least200kmfromthenearestlandmass,andFig.1bpresents
adependenceonship-basedwindspeed. Itisclearthatthere
isnoobviousrelationshipbetweenAODandwindspeedfor
the subset considered. AODs are quite low while the wind
speed ranges from 3 to 14ms−1. Additional consideration
ofthesubsetacquiredwithin39◦–65◦Sdidnotproduceany
correlationeither.
Thereforeinouranalysiswedecidedtodeploythefollow-
ingstrategy. Becauseallfactorsinfluencingthedependence
of AOD on wind speed cannot be accounted for, we sim-
ply considered only data presumably not influenced by ur- Fig.6.Maritimeaerosolopticaldepthasafunctionofwindspeed.
ban/industrial continental sources, dust outbreaks, biomass
burning, or glaciers and pack ice. In the Northern Atlantic
welimitedtheareatothelatitudinalbeltbetween40◦–60◦N;
in the Southern Atlantic we considered data acquired to the
South of 10◦S; in the Indian Ocean data set included only
Atmos. Meas. Tech.,5,377–388,2012 www.atmos-meas-tech.net/5/377/2012/
A.Smirnovetal.: Effectofwindspeedonaerosolopticaldepthoverremoteoceans 383
Fig.7.Scatterdensityhistogrambetweensimulatednoise-freewindspeedandnoise-freeAOD(a),noisyAODandnoise-freewindspeed(b),
noisywindspeedandnoise-freeAOD(c)andnoisywindspeedsandnoisyAODat500nm(d).Pointsweregeneratedassumingthesteady-
staterelationshipinTable3,τa(500nm)=0.0047×w+0.034(showninblack),andthenaddingGaussiannoiseofamplitude2ms−1tothe
windsand0.02totheAOD.
cruises South of 9◦S. An additional restriction imposed on ispresentedinFig.3a. TherelativenegativeoffsetofNCEP
thedatasetwasexclusionofpointstakencloserthantwode- is evident, although it is not critical for our study. About
greesfromthenearestlandmass.Amongtheselectedcruises, 66%ofthedifferencesarewithin2ms−1. The“series”and
we excluded one (presented in Fig. 1), which showed no “daily” wind speed differences are comparable. Figure 3b
relationship between AOD and wind speed. For any other shows histograms of wind speeds used in our further anal-
individual cruise considered, the slope of the AOD scatter- ysis. High winds (greater than 10ms−1) account for over
plotversuswindspeedwasfoundtobeatleast0.002sm−1. 20%ineachsubsetconsidered.
This“cherry-picking”isjustifiedbytheultimategoaloffind-
ing the most robust possible dependence of AOD on wind
speedovertheoceans. Table1presentsfinaldatasetusedfor 3 Results
ouranalysis. Overallweconsidered239measurementdays.
Figure 2a shows AOD daily averages as a function of lati- Figures 4 and 5, and Table 2, illustrate regressions between
tude, and Fig. 2b presents corresponding daily averages of aerosol optical depth, Angstrom parameter (negative of the
theship-basedwindspeed. logarithmicgradientofAODwithwavelength,overthevis-
The NCEP wind speed data were interpolated in space iblespectrum)andwindspeed. Morethan1100seriesfrom
and time to match the AOD measurement series. In addi- 239daysofaerosolopticaldepthmeasurementscontributed
tiontothe“instantaneous”windspeeds(windspeedsatthe tothestatisticspresented. Overallwecanconcludethatthe
timesmatchingtheAODseries),weusedwindspeedsaver- relationshipbetweenAODandwindspeedislinear,butcor-
agedoverthe24hpriortoeachAODmeasurement,andalso relations are not strong (non-linear relationships were con-
the subset of “steady-state” wind speeds (defined similar to sidered,butdidnotresultinstrongercorrelations). Evenfor
Madryetal.,2011,i.e.,standarddeviationforthedailyaver- thecaseof“steady-state”winds,correlationscoefficientsdo
agedwindspeedshouldnotexceed2ms−1 forwindspeeds not increase significantly. These values, although not high,
less than 10ms−1, or 3ms−1 for wind speeds greater than are statistically significant at a 99% confidence level. Re-
10ms−1). NCEP data were compared with the ship-based sultsobtainedforthe“daily”and“series”datasetsarecom-
meteorological information for cruises considered, and this parable. Averaging AOD over a day removes some noise,
www.atmos-meas-tech.net/5/377/2012/ Atmos. Meas. Tech.,5,377–388,2012
384 A.Smirnovetal.: Effectofwindspeedonaerosolopticaldepthoverremoteoceans
Table2.Regressionstatisticsofspectralopticalparametersversuswindspeed.
Relationship Datasource a b R∗ Datasource a b R∗
τa(440nm)=a×w(ship)+b series 0.0020 0.062 0.23 dailyaverage 0.0022 0.061 0.29
τa(500nm)=a×w(ship)+b 0.0023 0.052 0.28 0.0024 0.052 0.33
τa(675nm)=a×w(ship)+b 0.0029 0.038 0.38 0.0031 0.035 0.47
τa(870nm)=a×w(ship)+b 0.0027 0.039 0.37 0.0030 0.037 0.46
τa(440nm)=a×w(NCEP)+b series 0.0034 0.052 0.31 dailyaverage 0.0035 0.052 0.35
τa(500nm)=a×w(NCEP)+b 0.0037 0.044 0.35 0.0036 0.045 0.39
τa(675nm)=a×w(NCEP)+b 0.0043 0.029 0.44 0.0043 0.028 0.50
τa(870nm)=a×w(NCEP)+b 0.0042 0.030 0.45 0.0043 0.028 0.52
τa(440nm)=a×w(<24h>,NCEP)+b series 0.0042 0.044 0.35 dailyaverage 0.0040 0.047 0.38
τa(500nm)=a×w(<24h>,NCEP)+b 0.0043 0.037 0.39 0.0041 0.039 0.41
τa(675nm)=a×w(<24h>,NCEP)+b 0.0049 0.023 0.47 0.0048 0.022 0.53
τa(870nm)=a×w(<24h>,NCEP)+b 0.0049 0.023 0.48 0.0049 0.022 0.55
τa(440nm)=a×w(steady−state,NCEP)+b series 0.0045 0041 0.40 dailyaverage 0.0044 0.042 0.44
τa(500nm)=a×w(steady−state,NCEP)+b 0.0047 0.034 0.43 0.0045 0.035 0.47
τa(675nm)=a×w(steady−state,NCEP)+b 0.0052 0.021 0.50 0.0051 0.019 0.56
τa(870nm)=a×w(steady−state,NCEP)+b 0.0051 0.021 0.51 0.0052 0.019 0.58
τa(440nm)=a×w(AMSR)+b series 0.0033 0.056 0.28 dailyaverage 0.0040 0.050 0.38
τa(500nm)=a×w(AMSR)+b 0.0036 0.047 0.33 0.0040 0.043 0.40
τa(675nm)=a×w(AMSR)+b 0.0041 0.034 0.40 0.0046 0.028 0.50
τa(870nm)=a×w(AMSR)+b 0.0036 0.038 0.37 0.0045 0.029 0.50
α=a×w(ship)+b series −0.036 0.706 0.35 dailyaverage −0.035 0.732 0.36
α=a×w(NCEP)+b −0.051 0.789 0.39 −0.048 0.813 0.38
α=a×w(<24h>,NCEP)+b −0.059 0.867 0.41 −0.054 0.875 0.41
α=a×w(steady−state,NCEP)+b −0.059 0.863 0.42 −0.053 0.864 0.40
α=a×w(AMSR)+b −0.038 0.648 0.31 −0.042 0.747 0.34
∗R–isalinearcorrelationcoefficient.
associated in part with uncertainties in the AOD and wind In Fig. 6 the relationship between AOD and NCEP wind
speed, andinpartwithnaturalvariability, andmakescorre- averagedwithinprevious24h(“currentstudy”)iscompared
lationcoefficientsslightlyhigher(bylessthan0.1). Various to other studies. The diversity between different relation-
winddatasourcesandwindspeedsubsetsyieldedverysimi- ships established in the literature is evident. However, over
larresults. AsexpectedAngstromparameterdecreaseswith the range 0–10ms−1, the typical change in AOD is simi-
wind speed. An influx of large particles is responsible, at larinmostparameterizations(∼0.04at500nm),andconsis-
leastinpart,forthisanticorrelation. tentwiththeship-bornemeasurementsfromthisstudy. The
The slope of the linear regression of AOD versus main differences between studies are linked to the baseline
wind speed lies in the range 0.002–0.005 for the various AOD for low-wind conditions, and some nonlinearities at
wavelengths, cruises, and wind datasets considered. As ex- high wind speeds. In the former case, this may be partially
pected (because the wind speed’s history is important) the explained by local aerosol sources or satellite retrieval bi-
dataset that uses wind speed averaged within previous 24h ases,specifictoeachindividualstudy’sdataset. Inthelatter
periodand“steady-state”winddatasetyieldedhigherslopes. case, this may often be linked to a paucity of data for high
Table 3 presents regression statistics compiled from vari- wind speeds, such that the determination of the form of the
ous publications. Our results are consistent with the major- relationshipislesswell-defined(althoughasmentionedpre-
ity of previously reported results for ship-based and island- viously, over 20% of the MAN AODs are for wind speeds
based measurements, although being different from Mulc- of10ms−1 orgreater). Further,datafromcoastalsitesmay
ahy et al. (2008). We would like to note that additional be more strongly affected by enhanced foam from breaking
considerationofstricter“steady-state”windconditions(with wavesathighwindspeeds,andsatellitebiases(Sayeretal.,
standard deviation less than 1ms−1 within previous 24h) 2010; Smirnov et al., 2006, 2011) may be more extreme in
did not change the slope at 500nm, but slightly increased suchcases. Theseeffectswouldnotbeexpectedtoinfluence
it to 0.0058 at 870nm. Some of the satellite-derived AODs theMANAODsinthesameway.
yieldedsteeperslopes,althoughwebelievethesetobeanar- ToinvestigatetheextenttowhichuncertaintiesintheAOD
tifact of the satellite-derived AOD overestimation (Smirnov andwindspeedcontributetothelowcorrelations, anumer-
etal.,2006,2011). ical simulation was performed, based on the “steady-state”
Atmos. Meas. Tech.,5,377–388,2012 www.atmos-meas-tech.net/5/377/2012/
A.Smirnovetal.: Effectofwindspeedonaerosolopticaldepthoverremoteoceans 385
Table3.Regressionstatisticsofaerosolopticaldepthversuswindspeed.
Reference AOD Wind Region Relationship a b
source source
Currentpaper SP Ground Global τa(500nm)=a×w(ship)+b 0.0023 0.052
Model τa(500nm)=a×w(instantaneous,NCEP)+b 0.0037 0.044
Model τa(500nm)=a×w(<24h>,NCEP)+b 0.0043 0.037
Model τa(500nm)=a×w(steady−state,NCEP)+b 0.0047 0.034
Satellite τa(500nm)=a×w(AMSR)+b 0.0036 0.047
PlattandPatterson(1986) SP Ground CapeGrim τa(500nm)=a×w+b 0.0028 0.046
Villevaldeetal.(1994) SP Ground Pacific τa(500nm)=a×w+b 0.0033 0.101
Smirnovetal.(1995) SP Ground Pacific τa(500nm)=a×w+b 0.0036 0.123
WilsonandForgan(2002) SP Ground CapeGrim τa(500nm)=a×w+b 0.0035 −0.006
Smirnovetal.(2003) SP Ground Midway τa(500nm)=a×w<24h>+b 0.0068 0.056
Shinozukaetal.(2004) SP Ground Pacific τa(500nm)=b+4.9×10−5w3−3.7×10−5w2 0.017
Muclahyetal.(2008) SP Ground MaceHead τa(500nm)=b+5.5×10−4w2.195 0.060
Lehahnetal.(2010) SP Satellite Global,islandsites τac(500nm)=a×w+b 0.0070 0.015
Adamesetal.(2011) SP Ground Atlantic τa(500nm)=a×w+b 0.0066 0.027
Sayeretal.(2012) SP Model Global,islandsites τa(500nm)=a×w+b 0.0031 0.070
Glantzetal.(2009) SeaWIFS Model Pacific τa(500nm)=b+0.00016×w2.3 0.036
Huangetal.(2010) AATSR Model Global τa(550nm)=a×w+b 0.004 0.085
Lehahnetal.(2010) MODIS Satellite Global τa(500nm)=a×(w−4)+b 0.013 0.080
O’Dowdetal.(2010) MODIS Satellite Pacific τa(550nm)=b+0.00022×w2.47 0.114
τa(550nm)=b+0.033×w0.72 −0.004
Indian τa(550nm)=b+0.0097×w1.09 0.042
τa(550nm)=b+0.011×w1.04 0.040
KiliyanpilakkilandMeskhidze(2011) CALIPSO Satellite Global τa(532nm)=0.15/(1+6.7×e−0.17∗w)
Grandleyetal.(2011) MODIS Model NAtlantic τa(550nm)=a×w+b 0.0097 0.050
SAtlantic τa(550nm)=a×w+b 0.0111 0.041
AATSR Model NAtlantic τa(550nm)=a×w+b 0.0089 0.099
SAtlantic τa(550nm)=a×w+b 0.0034 0.081
relationship in Table 3, τ (500nm)=0.0047×w+0.034. This last number is similar to the correlations observed in
a
First,a50000-memberGaussiandistributionofwindspeeds this study, suggesting the principle factor decreasing corre-
with mean 7.93ms−1 and standard deviation 2.96ms−1 lation from unity for the MAN and wind data studied here
(corresponding to the “steady-state” wind distribution in areuncertaintiesintheinputdata(ratherthanothermeteoro-
Fig. 3), with any resulting negative wind speeds removed, logicaleffects),andthatthenoiseontheAODismorecriti-
wasgenerated. Thesamplesizegivesstatisticsrobusttotwo calforthispurpose. HalvingthemagnitudeoftheGaussian
significantfigures. ThiswasthenusedtocalculatetheAOD, noiseusedfortheperturbationsincreasesthislastcorrelation
assuming the aforementioned wind speed/AOD relationship to 0.76, indicating the effect of noise on the correlations
wasaperfectpredictor. couldbemuchreducedwithmorepreciseinputdata.
Next, the wind and AOD distributions were perturbed by Performingalinearleast-squaresfittotheperturbednoisy
adding Gaussian noise (zero mean in both cases, standard datagivestherelationshipτa(500nm)=0.0031×w+0.047,
deviation2ms−1 forwindspeed,and0.02forAOD).These i.e., an increase of the intercept and suppression of the gra-
uncertaintiesarereasonablyrepresentativeoftheuncertainty dient as compared to the true underlying relationship. This
intheinputdata(e.g.,Knobelspiesseetal.,2004,Wallcraftet resultisconsistentwiththeobservationfromthisstudythat
al.,2009;Sayeretal.,2012).Perturbednegativewindspeeds with increasing levels of temporal averaging to decrease
orAODswerethensettozero,aswouldlikelybeusedasthe noise(instantaneoustodailyorsteady-stateNCEPdata),gra-
minimumvaluetoreportinsuchanAODorwinddataset. dientsbecomestrongerandinterceptssmaller, andsuggests
thatuseofonlyinstantaneousdataforsuchanalyseswillre-
The resulting distributions of data are shown in Fig. 7.
sult in an overestimate of the baseline maritime AOD, and
Correlatingthenoisywindspeedswiththe“true”simulated
underestimateoftheresponsetochangesinthewindspeed.
AODs give R=0.82 (Fig. 7c); correlating true wind speeds
with noisy AODs gives R=0.56 (Fig. 7b); and correlating
noisywindspeedswithnoisyAODsgivesR=0.47(Fig.7d).
www.atmos-meas-tech.net/5/377/2012/ Atmos. Meas. Tech.,5,377–388,2012
386 A.Smirnovetal.: Effectofwindspeedonaerosolopticaldepthoverremoteoceans
4 Conclusions produced by Remote Sensing Systems and sponsored by the
NASA Earth Science MEaSUREs DISCOVER Project and the
OuranalysisoftheMaritimeAerosolNetworkdatashowed AMSR-E Science Team. Data are available at www.remss.com.
a linear relationship between aerosol optical depth over the Gert Konig-Langlo from Alfred Wegener Institute for Polar and
oceans and wind speed for a wind speed range 0–15ms−1. Marine Research (Bremerhaven, Germany) is acknowledged for
There is no indication of a non-linear power-law or expo- providing the meteorological data from R/V Polarstern. British
OceanographicDataCentreisacknowledgedforprovidingAMT19
nential relationship between those quantities for any of the
and 20 meteorological data (the data were supplied to BODC by
wind datasets (ship-based, NCEP, satellite-based) consid-
ChrisBarnard,GarethKnightsandJonSeddon).
ered. However, the gradient of the relationship varies by
around a factor of two depending on the wind data used.
Editedby:A.Kokhanovsky
This highlights that the derivation of such relationships is
sensitivetonotonlytheAODdatasource,butalsothewind
datasource,whichmayexplainsomeofthevariationshown References
withintheliterature.
Variouswindspeedsubsets,instantaneousanddailyaver- Adames,A.F.,Reynolds,M.,Smirnov,A.,Covert,D.S.,andAck-
erman, T. P.: Comparison of MODIS ocean aerosol retrievals
agedAODsyieldedsimilarregressionstatisticswhichproves
withship-basedsunphotometermeasurementsfromthe“Around
therobustnessofourconclusions. Itisnoteworthythat,un-
the America’s” expedition, J. Geophys. Res., 116, D16303,
like in Smirnov et al. (2003) the wind speed range consid-
doi:10.1029/2010JD015440,2011.
eredherewassignificantlywider–upto15ms−1forNCEP
Anguelova, M. D. and Webster, F.: Whitecap coverage from
andAMSR-E,andupto20ms−1forwindsmeasuredonthe
satellite measurements: A first step toward modeling the vari-
ship. ability of oceanic whitecaps, J. Geophys. Res., 111, C03017,
Ourfindingsareconsistentwiththepreviouslyreportedre- doi:10.1029/2005JC003158,2006.
sults,onlydifferingsignificantlyfromMulcahyetal.(2008) Blanchard,D.C.andWoodcock,A.H.:Theproduction,concentra-
andO’Dowdetal.(2010)forhighwindspeeds(>10ms−1). tionandverticaldistributionofthesea-slataerosol, Ann.N.Y.
However, we expect that the future release of the MODIS Acad.Sci.,338,330–347,1980.
Collection 6, which takes near-surface wind speed into de Leeuw, G., Neele, F. P., Hill, M., Smith, M. H., and Vignati,
E.:Productionofseasprayaerosolinthesurfzone,J.Geophys.
account when determining ocean surface reflectance, will
Res.,105,29397–29409,doi:10.1029/2000JD900549,2000.
change the conclusions reported by O’Dowd et al. (2010)
Derber, J. C., Parrish, D. F., and Lord, S.: The new global oper-
in terms of reducing the wind-speed dependence in the re-
ational analysis system at the National Meteorological Center,
trieved AOD (Mishchenko and Geogdzhayev, 2007; Kleid-
WeatherForecast.,6,538–547,1991.
manetal.,2012). TherelationshipbyMulcahyetal.(2008)
Eck, T. F., Holben, B. N., Reid, J. S., Dubovik, O., Smirnov, A.,
overestimates aerosol optical depth, predicting AOD∼0.27 O’Neill, N. T., Slutsker, I., and Kinne, S.: Wavelength depen-
atwindspeed15ms−1, possiblyduetothebreakingwaves denceoftheopticaldepthofbiomassburning,urban,anddesert
atthecoastalsiteandonly14measurementdayscontributing dustaerosol,J.Geophys.Res.,104,31333–31350,1999.
totheoverallstatistics. Fan, T. and Toon, O. B.: Modeling sea-salt aerosol in a coupled
As found in previous studies, there is considerable scat- climateandsectionalmicrophysicalmodel: mass,opticaldepth
terinplotscomparingAODandwindspeed,leadingtocor- andnumberconcentration,Atmos.Chem.Phys.,11,4587–4610,
doi:10.5194/acp-11-4587-2011,2011.
relations typically of order 0.3–0.5. Our results show that
Gathman, S. G.: Optical properties of the marine aerosol as pre-
the known uncertainties in the AOD and wind data used
dictedbytheNavyaerosolmodel,Opt.Eng.,22,57–62,1983.
would be sufficient to degrade the observed correlation be-
Glantz,P.,Nilsson,E.D.,andvonHoyningen-Huene,W.:Estimat-
tweenvariables,whichwereperfectlycorrelatedintruth,to
ingarelationshipbetweenaerosolopticalthicknessandsurface
around0.5.Thisnoisealsoaffectedthecoefficientsoffit,de-
windspeedovertheocean,Atmos.Res.,92,58–68,2009.
creasingthegradientascomparedwiththe“true”case.Thus, Grandey, B. S., Stier, P., Wagner, T. M., Grainger, R. G., and
itisplausiblethat,overtheremoteocean,thetruestrengthof Hodges, K.I.: Theeffectofextratropicalcyclonesonsatellite-
correlationbetweenmaritimeAODandwindspeed,andthe retrievedaerosolpropertiesoverocean,Geophys.Res.Lett.,38,
magnitude of the response, could be significantly stronger L13805,doi:10.1029/2011GL047703,2011.
thanobservedinthesestudies. Haywood,J.,Ramaswamy,V.,andSoden,B.:Troposphericaerosol
climateforcinginclear-skysatelliteobservationovertheoceans,
Acknowledgements. The authors thank Hal Maring (NASA Science,283,1299–1303,1999.
Headquarters) for his support of AERONET. The authors thank Holben B. N., Eck, T. F., Slutsker, I., Tanre, D., Buis, J. P., Set-
Kirk Knobelspiesse (Columbia University and NASA GISS) and zer,A.,Vermote,E.,Reagan,J.A.,Kaufman,Y.,Nakajima,T.,
anonymous reviewer for constructive comments. Measurements Lavenu,F.,Jankowiak,I.,andSmirnov,A.:AERONET–Afed-
onboard R/V Marion-Dufresne were supported by the French eratedinstrumentnetworkanddataarchiveforaerosolcharacter-
Polar Institute (IPEV). NCEP Reanalysis data provided by the ization,RemoteSens.Environ.,66,1–16,1998.
NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, from their
Web site at http://www.esrl.noaa.gov/psd/. AMSR-E data are
Atmos. Meas. Tech.,5,377–388,2012 www.atmos-meas-tech.net/5/377/2012/