GEOPHYSICAL RESEARCH LETTERS, VOL. 28, NO.
17, PAGES 3251-3254, SEPTEMBER 1, 2001
Baseline maritime aerosol: methodology to derive the optical
thicknessand scattering properties
YoramJ.Kaufman
•, Alexander
Smirnov
2,BrentN. Holben
•, andOlegDubovik
2
Abstract. Satellite measurements of the global distributionof
aerosolandtheir effect on climateshouldbe viewed in respectto
a baselineaerosol.In this concept,concentrationof fine mode
aerosol particles is elevated above the baseline by man-made
activities (smoke or urban pollution), while dust or sea-spray
elevatesthe coarsemode.Using 1-3 yearsof measurements in l0
stations of the Aerosol Robotic network (AERONET)
develop a methodology and derive the optical thicknessand
propertiesof this baselineaerosolfor the Pacific and Atlantic
Oceans. Defined as the median for periods of stable optical
thickness (standard deviation < 0.02) during 2-6 days, the
medianbaselineaerosolopticalthicknessover the Pacific Ocean
is 0.052at500nmwith•ngstr6mexponent
of 0.77,and0.071
and 1.1 respectively,over the Atlantic Ocean.
1. Introduction
The humaninducedchangesin the aerosolconcentrationand
properties, or the aerosol responseto climate change (e.g.
droughtsproducingfires and dust) shouldbe measuredrelative
to a "baseline aerosol" with minimal impact of man-made
pollutionor smokeaerosolor dustthat is partiallyin responseto
land use changeand climatedriven droughts.How to define this
baseline aerosol, so that it is both measurable and useful? Over
the oceanaerosolis not necessaryfrom maritimeorigin.There is
a needto separateit from the baselineaerosolof marine origin,
e.g. aerosol from oxidation of DMS and sea spray. Some
residual, long lived, continental aerosol may also
contribute/contaminate this value. Therefore the baseline aerosol
is very broadlyspread,resultingin smallertemporalvariability
thanpollution,smokeaerosolor dustthat originatefrom specific
locations (e.g. city or dry lakebed) and are emitted and
transported to the oceanic measuring site in a specific
combination of meteorological conditions. Therefore, the
baseline aerosol can be defined as the median aerosol for
conditionswith small temporalvariations.Measurementsof the
baseline aerosol should help quantifying the contribution of
natural and man-made sources to satellite or ground based
measurementsof aerosol [Jankowiak and Tanrd, 1992; Husar et
al., 1997;Kaufmanet al., 1997;Mishchenkoet al., 1999;Holben
et al., 2001]. Model assessments
of the aerosolforcingof climate
can also usetheseresults[e.g. Kiehl and Briegleb, 1993; Tegen
et al., 1996; Hansen et al., 1997]. The baselineaerosolis derived
from AERONET maritime measurements taken in the last 1-3
years. The AERONET network is describedby Holben et al.
[1998,2001],climatology
ofmaritime
aerosol
bySrnirnov
eial.
[2001].
• NASA GoddardSpaceFlightCenter,Greenbelt,
Maryland.
2 Science
Systems
andApplications,
Inc.,Lanham,Maryland.
Copyright2001 by the AmericanGeophysical
Union.
Papernumber2001GL013312.
0094-8276/01/2001GL013312505.00
2. Analysis
The baselineaerosol is determined using time sequences
measuredin the AERONET sites.Examplefor two sitesis given
in Fig. 1. One site is in usuallycleanconditionsin the middle of
the Pacific Ocean - Lanai, and the second in the Atlantic Ocean,
we affectedby pollutionfrom North Americaand dustfrom Africa -
Bermuda. The time series of the aerosol optical thicknessat
500nm,X.•oo,
andthe•ngstr6mexponent,
ct- variation
between
opticalthicknessat 440 and 870 nm: ct=ln(x440/xg70)/ln(870/440)
is plotted.The measurements are taken every 15 minutes,daily.
They are cloud screenedand quality assuredfor calibrationand
instrumentalerrors assumingthat no other errors have affected
the measurements[Srnirnovet al., 2000]. As expected,Fig. 1
shows that the conditions in the remote site of Lanai are stable
while in Bermudaepisodesof intrusionof pollutionand dust are
observed most of the time.
To define the baselineaerosolin a consistentway for these
diverse sites, we first compute the median of N consequent
measurements. We repeatthe calculationsshifting each time by
one measurement point (a runningmedian).Two valuesof N are
usedfor comparison:N=50, that corresponds in averagecloudy
conditionsto 2-3 days of measurements and N=100 (4-5 days).
Most high polluted or dusty conditions correspondto data
sequenceswith high standarddeviationof the optical thickness,
c• > 0.02, and are excluded from the baseline statistics. Over the
clean Pacific Ocean the variability of the optical thicknesswas
found to be with standard deviation of c•--0.02 [S•nirnov et al.,
2001]. Thereforewe adoptedthis value for the analysisof the
baseline optical thickness. A median is computed for the
ensembleof the remainingmedians.The use of mediansrather
than average excludes those rare cases of pollution or dust
episodesthat are stableduringthe N measurements and survived
the first screening.This methodresultsin a singleannualvalue,
with no seasonalor wind speeddependence. Table 1 summarizes
the results for N=100 for the 10 oceanic sites. The baseline
aerosolover the PacificOceanis characterizedby x.•0o=0.052
and
ct=0.77 (calculatedas weighted averages).Over the Atlantic
Oceanthe valuesare higher:x.•0o=0.071andct=l.1. Changingthe
number of elements in the running median from 100 to 50
increasedthe medianoptical thicknessonly by X.•oo---0.002.
This
small changein the baselineoptical thicknessshowsvery little
dependenceon the choice of the sample N. 14,000 and 3,000
mediansrespectivelyfor the Pacific and Atlantic oceanspassed
the criteria for N=100 and were used in the baseline statistics
(20,000 and 5,000 respectively for N=50). Increasing the
thresholdof standarddeviationfor the stabilityof the baseline
from c•=0.02 to 0.03 increasesthe derived baselineoptical
thicknessby x.•---0.004.If we would useaveragesinsteadof the
medians, the baseline values would increase between 0.002 to
0.008 for the 10 sitesusedin the study.
The differences between the baseline aerosol in the Pacific,
Atlantic and the one location in the Indian Oceans are
significant.They can be due to differencesin the meteorological
3251
3252 KAUFMAN ET AL.' BASELINE MARITIME AEROSOL

:. ' i , ..• it "- :. '• .

016
c: •:.. =.033+O.016exp(b.
OO27W,' )
'. o 0 12 .--'• : ....•: .....L-4
..........•.,- ',-;....
O.12 .......
o i .,I! '!
........ :..•..........•. •.•.•;..:.:•.............
! l: ? 1/;{
l: l

.z
'• 0.08 ,!' ', 'I;l
: [ Lanai O

"o 0.06 I I
.

Jun Aug Sep Nov Jan

:' I" .. I'",J,: i •' : k. s•,ca.• I I

0 4 •'- • ..... 1

x < O.Oq.... j............. • ............ j---•

t l

o3 •..........
'.....
L:;................'..•,•. L.:.'
....... o
-.•

0 1 2 3 4 5 6

O 2 Columnprecipitable water vapor (cm)

'
Figure 2. Baselineaerosoloptical thicknessfor all the stations
0 I •.-;.....
;-•.!
......,.•,..... ;?.•
.....
;-•-'•:............ (green),and the averageoptical thicknessfor each station(red)
as a functionof the total precipitablewater vaporindicatingthe
Mar May Jul Sep Nov
climatologyof the station.Pacific Ocean- blue squaresand the
Atlantic stations- white circles. The blue line is a fit to the
Month
baselineaerosolexcludingCape Verde and AscensionIsland,
Figure 1. Time sequenceof the aerosolopticalthickness(red) due to the small number of identified medians (<300). Data from
and•ngstr6m
Exponent
(blue)forLanai(datafrom1998)in the Azores Island do not have yet the final calibrationand are not
Pacific Ocean and Bermuda (1996) in the Atlantic Ocean. Dots includedin the analysis.
are the individual measurements taken in 15 minutes intervals.
Lines connect the medians for standard deviation o < 0.02. The
horizontal lines are the medians of the selected medians for the
whole time series(seetable l). conditionsthat changethe strengthof sourcesof the baseline
aerosol. It cannot be ruled out that some contamination of the
baselinevalues by pollution occurs,particularly in the Indian
Ocean. However the present baseline value is compared
favorablywith an estimateof the naturalaerosolopticalthickens
Table 1. The baselineaerosolopticalthickness at 500onm, 'rs00, of 0.07, usingchemicalanalysisof aerosolin the Indian Ocean
the total precipitable water vapor, WV, and the AngstrOm
exponentct, derivedas the medianof an assembleof mediansof region [Ramanathan et al., 2001]. This value correspondsto
100 consecutivemeasurementswith standarddeviation in 'rs00< centerof the visible spectrumand thus is by 0.01 smaller than
0.02.
our estimate. Possiblecontaminationof the baseline optical
thicknessis discussedin regard to Fig. 2 to 4. The baseline
Location Median WV ct # of
optical thicknessand averageoptical thicknessfor each site are
•s,• medians
plotted in Fig. 2 as a function of the total precipitablewater
vaportbr the baselineconditions.The precipitablewatervaporis
usedas a roughindicatorof the meteorologicalconditionsthat
Ascension Island 8øS 15øW 0.100 3.76 0.71 203
affect the aerosolpropertiesin a given site. The aerosolsources
CapeVerde 17øN23øW 0.063 1.92 0.64 130 and wind speedover the oceanmay changebetweentropicaland
Dry Tortugas25øN 83øW 0.079 3.86 1.11 714 temperatezones, characterizedby different precipitablewater
Bermuda 32øN 65øW 0.067 3.53 1.00 957 vapor.The aerosoloptical thicknessand particlesize was found
to be correlated to total precipitable water vapor due to
Azores 38øN 29øW 0.064 2.82 1.39 761
humidificationof the aerosol[Kaufmanand Fraser, 1983]. Note
Pacific Ocean
that the differencebetweenthe averageoptical thicknessand the
derived baselineis much larger in the Atlantic and Indian sites
Nauru 0øS 167øE 0.058 4.05 0.43 1653
than the Pacific sites.If we excludethe small numberof points
Tahiti 17øS 150øW 0.056 3.61 0.80 2104 in the Cape Verde and AscensionIsland(that haveno significant
Lanai 21øN 157øW 0.049 3.23 0.69 6243 influenceon the weightedaverageusedin Table 1) we get a very
San Nicolas 33øN 119øW 0.051 1.07 1.05 3734 high nonlinearcorrelationof the baselineopticalthicknessto the
precipitablewater vapor. Additionalinsightinto the differences
Indian Ocean between the locations is gained from the volume size
distributionsplottedin Fig. 3. The size distributionsare derived
Kaashidhoo 5øN 73øE 0.100 4.69 1.11 400
from the sky radiancespectral-angularmeasurements[Dubovik
and King, 2000; Dubovik et al., 2000]. All stations are
Averages characterizedby fine and coarsemodes.The volume of the fine
Atlantic Ocean 0.07 ! 3.36 1.10 2765 mode is much larger for the Atlantic Ocean than for the Pacific
Pacific Ocean 0.052 2.80 0.77 13734 Ocean. The parametersof a two lognormal fit to the size
distributionare given in Table 2.
 --•,--Bermuda 0.12 .............. Naru Equator
0.03 Lana•
fi--•
-Tahiti
Nauru .--o.
oAzores
[---
ß.' Tahiti 17'S
Lanai 21øN
--•,-
San
Dry N•colas
c 0.10 '
ß San Nicholas
Ascension
33øN
Island
'-,. Dry Tortugas 25øN
Bermuda 32øN

o.o ..........
................................
..................
........
........ 0.08
.; .

0.06

......... •;..f........
0.04

• •, .,. .
o.oo • ...... • .' • ....... 0.02
o.1

radius
I

(•m)
lO
'i
0.00
Figure 3. A•rosolvolum•siz• distribution (dr/clint) averaged Dec.-Feb. March-May June-Aug. Sept.-Nov.
for conditionsthatcodespond to thebasdineopticalthickness Month

4 Pacificsitesand 3 Atlanticsites.The averagesizedistribution

Figure4. Seasonal
variation
of thebaseline
optica
! thickness
for
for eachoceanis given by the solidli,•s (blu• for thePacific, Pacific and Atlantic sites.Multiyear data setswere used.Data
andorangefor the Atlantic). sets from locations and seasons with less than 70 baseline
mediansare notdisplayed.The higherbaselineoverthe Northern
Pacific is probablydue to contaminationby transportof Asian
3. Fine versus coarse mode aerosol aerosol.

Satellitemeasurements of aerosolcan be usedto distinguish

betweenman-madefine aerosoland naturalor inducedby land
changecoarseparticles.Therefore,the baselinespectraloptical higherfor theAtlantic ('Ca_fine'"0.045)
than for the Pacific (l•a.
nne-0.029)oceans.The contributionsof the coarse modes are
thicknessis also separatedinto fine and coarsemodesusingthe
parametersin Table 2a. The optical thicknesstransitionsfrom similar(Atlantic- 0.026, Pacific- 0.023), exceptof the strong
coarse mode in Nauru.
beingdominatedby fine particlesfor shortwavelengths to being
The baselinesize distributions
can be integratedto calculate
dominatedby the coarseparticles for the long wavelengths
(Table 2b). For )•=500 nm the contributionof the fine mode is the number concentrationof the particles for the baseline
conditions(Table2a). Sincethe maritimeparticlesare expected
to be hygroscopic,and the size distribution was derived fbr
particleslarger than 0.05 tamradius,the numberof particles
Table 2. (a) Size distributionparametersfor the average shouldcloselyresemblethe CCN concentration. The efficiency
baselineopticalthickness overtheAtlanticandPacificOceans. of the aerosolto serveas CCN is expressed by the ratio of the
The sizedistributionis integratedfrom radiusof 0.05 tamto 0.44
tamfor thefine modeand0.44 tamto 15 tamfor thecoarsemode.
Rv is thevolumegeometricmeanradius(tam),Reftis the
effectiveradius(tam),o is thegeometricstandarddeviation,(b)
The contributionof the fine and coarsemodesto the spectral 0.25 dailybackground
baselineopticalthickness usingthesizeinformationin (a) and daily back.+pollutlon
refractive index of 1.35-0.001i. monthlybackground

0.2
monthly
back
+poilulion
(a) size distributionparameters mon,hly
derived
';::-.•' 5•
background N=lO0
• ;;.' •: •.- .'-+ • '";":'•
";•""•

Atlantic Ocean Pacific Ocean

Parameter 0.15
Fine Coarse Fine Coarse

Rv(pm) 0.14 2.49 0.13 2.94 0.1

Reft(pm) 0.13 1.83 0.12 2.23

o 0.42 0.78 0.43 0.74
0.05
Numberof particlesin the 250 0.2 185 0.2 ..,½.. :
column(cm-2)
* 106
Volume(Fm3/cm2)
* 106 1.4 2.7 0.9 2.7 0
0 100 200 300 400 500 600 700
Number/Volume
(pm'3) 167 0.10 200 0.09
Day
(b) spectralopticalthickness
Figure 5. Simulation of the method to derive the baseline
aerosoloptical thickness.A time serieswith a baselineaerosol
Atlantic Ocean Pacific Ocean
Wavelength,nm (blue dots)in the form of an annualsinusoidalpatternand 5-15
days randomsinusoidalvariationsis used:'rb,•din•= 0.02 [2 +
Fine Coarse Fine Coarse
sin(D•r/180)]+ 0.02[l+sin(D'tc/(5+10R^tqD))];
whereD is the day
440 0.062 0.026 0.037 0.021 number.R^NDis a randomnumber(0 to l). Pollutionis addedin
5OO 0.047 0.027 0.027 0.021
a semi-random pattern every 24 days:
Xpon:0.15R,nd[l+sin(•rD/24)]
- red dots.The monthlymedian
860 0.012 0.028 0.007 0.023
total optical thicknessis shownby (e) and the median optical
1650 0.001 0.027 0.001 0.024 thicknessfor 50 (.) and 100 (I) point series with variation
2130 0.001 0.024 0.000 0.022 o<0.02 by. The baselinetrue medianopticalthicknessof 0.060
was retrieved as 0.066.
3254 KAUFMAN ET AL.' BASELINE MARITIME
particle numberto volume. This ratio, rather than the number
itself is easierto comparewith in situ measurements that report
the concentrations per unit volume.The ratio of 167 (Table 2a)
for the Atlantic Ocean is similar to the ratio measured in situ
from aircraft of 200e30 and 185e25 [Hegg and Kaufman, 1998]
and other in situ measurementsby Hegg and Jonsson,[2000]
and Van Dingenen et al. [1999] (168_+70 and 165_+30
respectively).
4. Seasonalcycle
The procedureto derive the baselineaerosolis basedon the
availability of a large data set that can be used to extract the
baseline conditions. Derivation of the seasonal trends is therefore
less accurate. We divided the data set for each site to 3 months
seasonalintervals,and averagethe seasonalbaselinevalue for all
the years for which the data is available. Fig. 4 shows the
baseline value for the Pacific and Atlantic sites that have at least
9 months of data. The Pacific baseline shows a systematic
increase in the Northern Hemisphere spring, associatedwith
transportof Asian dust.The Atlantic data alsopeaksin the same
period. This seasonalchange does not indicate a significant
seasonalcycle in the baselineaerosol.
5. Discussion and Conclusions
Using severalyears of measurements in 10 stationsof the
Aerosol Robotic network (AERONET) we derived the optical
thicknessand propertiesof the baselineaerosol.The baseline
aerosolopticalthicknessover the Pacific Oceanis 0.052 at 500
nm with /kngstr6m
exponentof 0.77, and 0.071 and 1.1
respectivelyover the Atlantic. The baselineaerosolis derived
assuming that pollution or dust is significantly more
heterogeneousthanthe baselineaerosoldue to the restrictedsize
of the source and variation of meteorological conditions
responsiblefor generation and transportof the aerosol. The
median is derived for periods of stable optical thickness
(standarddeviation< 0.02) during 2-6 daysof consecutivecloud
free, 15 minute interval measurements. How well do we derive
the baseline aerosol? Since we do not know the ultimate "true"
baseline,we performeda sensitivitystudy,in which a long-term
variation in the baseline aerosol is combined with short term
variations in pollution optical thickness (Fig. 5). The same
approach to derive the baseline aerosol is applied to the
simulateddata set and the result is comparedwith the "true"
simulated baseline. We assume that the difference between the
simulatedandretrievedbaselineopticalthicknessis a measureof
the error in the derived baseline in this study. The resultsshow
that despite the much stronger pollution sources, the
methodologyallows us to derive the baseline aerosol optical
thickness within an error of 0.003 or 0.008. Therefore, it is
expectedthat the contaminationof the baselineis between0.00
and A'c=+0.01. Using size distribution measurementsthat
correspondto the median conditions,the baselineaerosolis
separatedinto fine andcoarsemodes.Satellitemeasurements of
the aerosolopticalthicknesscorrespondingto the fine andcoarse
modes [e.g. MODIS' Tanrd et al., 1997, 1999, Remer et al.,
2001] can then be used to derive the man-made contribution -
elevatedfine mode. This can be a steptowardsidentifyingour
impactontheaerosol
distribution
anditsforcingof climate.
References
Dubovik, O., and M.D.King, A flexible inversionalgorithmfor retrieval
of aerosol optical properties from sun and sky radiance
measurements, J.Geophys.Res.,105, 20,673-20,696,2000.
AEROSOL
Dubovik, O., A.Smirnov, B.N.Holben et al., Accuracy assessmentsof
aerosol optical propertiesretrieved from AERONET sun and sky
radiancemeasurements, J.Geophys. Res., 105, 9791-9806, 2000.
Hansen,J., M. Sato,and R. Ruedy, Radiativeforcingand climate
response,J. Geophys.Res.,102,6831-6864, 1997.
Hegg, D.A., and Y.J.Kaufman, Measurements of the relationship
between submicron aerosol number and volume concentration, J.
Geophys.Res., 103, 5671-5678, 1998.
Hegg, D.A., and H.Jonsson,Aerosol number-to-volumerelationshipand
relativehumidityin the easternAtlantic, J. Geophys.Res.,105, 1987-
1995, 2000.
Higurashi,A., and T. Nakajima, Developmentof a two channelaerosol
retrieval algorithmon global scaleusing NOAA AVHRR. J. Atmos.
Sci., 56, 924-941, 1999.
Holben, B. N., T. F. Eck, 1. Slutsker, D. Tanr•, et al.,, AERONET-A
federated instrument network and data archive for aerosol
characterization, Rem. Sens. Environ., 66, 1-16, 1998.
Holben, B.N., D. Tanre, A. Smirnov, T.F. Eck et al., An emerging
ground based aerosol climatology' Aerosol optical depth from
AERONET, J. Geoph.Res., 106, 12,067-12,097,2001.
Husar, R. B., J. Prospero, and L.L. Stowe, Patterns of tropospheric
aerosolsover the oceans,J. Geophys.Res. 102, 16,889, 1997.
Jankowiak, 1., and D. Tanr•, Climatology of Saharan dust events
observedfrom Meteosat imagery over Atlantic Ocean. Method and
preliminaryresults,J. Clim., 5, 646-656, 1992.
Kaufman, Y. J., and R. S. Fraser, Light Extinction by Aerosol During
SummerAir Pollution.J. Appl. Meteor., 22, 1694-1706, 1983.
Kaufman, Y. J., D. Tanr6, H. R. Gordon, T. Nakajima, et al., Passive
remotesensingof troposphericaerosoland atmosphericcorrection,J.
Geoph.Res., 102, 16,815-16,830, 1997.
Kiehl, J. T., and B. P. Briegleb, The relative roles of sulfateaerosolsand
greenhousegasesin climate forcing,Science,260, 311-314, 1993.
King, M.D., Y. J. Kaufman,D. Tanif:, and T. Nakajima: Remotesensing
of troposphericaerosolsfrom space:past,presentand future. Bull. of
Meteor. Soc.. 80, 2229-2259, 1999.
Mishchenko, M. I., I. V. Geogdzhayev, B. Cairns, et al., Aerosol
retrievalsover the ocean by use of channels I and 2 AVHRR data:
sensitivityanalysisand preliminaryresults,Appl. Opt. 38, 7325-7341,
1999.
Ramanathan, V., P. J. Crutzen, J. Lelieveld et al., The Indian Ocean
Experiment:An assessment of the climate forcing and effects of the
great lndo-Asian haze, J. Geophys. Res., 106, 2001 (accepted, in
press).
Remer, L. A., D. Tanr•, Y. J. Kaufman, et al,, Validation of MODIS
AerosolRetrievalOver Ocean,Geophys.Res.Lett., 2001 (submitted).
Smimov, A., B.N.Holben, T.F.Eck, et al., Cloud screeningand quality
controlalgorithmsfor the AERONET data base,Rem. Sens.Env., 73,
337-349. 2000.
Smirnov, A., B.N.Holben, Y.J.Kaufman, et al., Optical propertiesof
atmospheric aerosol in maritime environments, J.Atm. Sci., 2001
(accepted,in press).
Tanr•, D., Y. J. Kaufman, M. Herman and S. Mattoo, Remotesensingof
aerosolover oceansfrom EOS-MODIS, J. Geoph.Res., 102, 16,971-
16,988, 1997.
Tanr6, D., L.R. Remer, Y.J. Kaufman, et al.,, Retrieval of Aerosol
Optical Thickness and Size Distribution Over Ocean from the
MODIS Airborne Simulatorduring TARFOX, J. Geophys.Res, 104,
2261-2278, 1999.
Tegen, I., A.A. Lacis and I. Fung, The influenceon climate forcing of
mineral aerosols from disturbed soils, Nature, 380,419-422, 1996.
Van Dingenen,R., F.Raes,J.-Ph.Putaud,et al., Processesdeterminingthe
relationshipbetween aerosol number and non-sea-saltsulfate mass
concentratonsin the clean and perturbedmarine boundarylayer, J.
Geophys.Res.,104, 8027-8038, 1999.
B.N.Holben, Y.J.Kaufman, NASA Goddard Space Flight Center,
Greenbelt, MD 20771.
O.Dubovik,A.Smimov, ScienceSystemsandApplications,Inc.,
Lanham, MD 20706.
(ReceivedApril 19, 2001' revisedJune 1,2001;
acceptedJune 12, 2001.)