JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 106, NO.

Dll, PAGES 12,067-12,097, JUNE 16, 2001

An emerging ground-based aerosol climatology:

Aerosol optical depth from AERONET
B. N. Holben, • D. Tanr•, 2 A. Smirnov,•,3 T. F. Eck, 1,4,sI. Slutsker,•,3
N. Abuhassan,•,3 W. W. Newcomb,•,3 J. S. Schafer,1,4,6B. Chatenet, 7 F. Lavenu,s
Y. J. Kaufman,9 J. Vande Castle,•ø A. Setzer,TMB. Markham, • D. Clark, TM
R. Frouin, •3 R. Halthore,•4,•s A. Karneli, •6 N. T. O'Neill, •7 C. Pietras,•s
R. T. Pinker, •9 K. Voss,2øand G. Zibordi 2•

Abstract. Long-term measurementsby the AERONET program of spectralaerosol

optical depth, precipitablewater, and derivedAngstrom exponentwere analyzedand
compiledinto an aerosolopticalpropertiesclimatology.Quality assuredmonthly means
are presentedand describedfor 9 primary sitesand 21 additional multiyear siteswith
distinctaerosolregimesrepresentingtropical biomassburning,boreal forests,midlatitude
humid climates,midlatitude dry climates,oceanicsites,desert sites,and backgroundsites.
Seasonaltrends for each of these nine sitesare discussedand climatic averagespresented.

1. Introduction total aerosol loading; thus we have no definitive measure of

changefor future assessment[Andreae,1996]. Regardlessof
Man is altering the aerosolenvironmentthrough land cover currentconditions,the extentof local aerosolperturbationson
change, combustionof fossil fuels, and the introduction of a globalscaleis the subjectof extensivegroundlevel, airborne
particulateand gasspeciesto the atmosphere.Each perturba- and satelliteresearch[Kaufmanet al., 1997;King et al., 1999].
tion has some impact on the local aerosolenvironment.How Investigationshave been initiated by concernsranging from
much aerosolman is contributingto the atmosphereis not radiativeforcingby aerosols,long-termimpactson climate and
known. Even more fundamental, we do not know the current
public health, aestheticand ecologicalimpacts,as well as the
future of sea level habitationsand political entities. The re-
•BiosphericSciencesBranch,NASA GoddardSpaceFlightCenter, sourcesput into suchinvestigations havelargelybeen local and
Greenbelt, Maryland. contributed to an enormous,yet mostly uncoordinateddata-
2Laboratoired'OptiqueAtmospherique,CNRS,Villeneuved'Ascq,
France. base that makes global assessment of our past and present
3ScienceSystemsandApplications,Inc., Lanham,Maryland. aerosol disposition difficult. Coordination between surface-
4Raytheon STX Corporation,Lanham,Maryland. based network observations and satellite measurements will be
SNowat GoddardEarthSciences andTechnology Center,Univer- required to develop a long-term monitoring systemof the
sity of Maryland, Baltimore County, Baltimore, Maryland. Earth's aerosol environment.
6Nowat ScienceSystemsandApplications,
Inc.,Lanham,Maryland.
7Laboratoire
Interuniversitaire
desSyst•mesAtmosph6riques,
Uni- The simplest,and,in principle,the mostaccurateand easyto
versit6s Paris VII et Paris XII, CNRS, Cr6teil, France. maintain monitoringsystemsare groundbased.Aerosol opti-
SCESBIOCentreNationalde la Recherche
Scientifique,
Toulouse, cal depthis the singlemostcomprehensive variableto remotely
France.
assessthe aerosolburden in the atmospherefrom ground-
9Climateand RadiationBranch,NASA GSFC, Greenbelt,Mary-
land. basedinstruments.This variable is used in local investigations
•øUniversity
of NewMexico,Albuquerque,
NewMexico. to characterize aerosols,assessatmosphericpollution, and
•qnstitutodePesquisas
Espaciais,
SaoJosedosCampos,
SanPaolo, make atmosphericcorrections to satellite remotely sensed
Brazil.
data. It is for these reasonsthat a record of aerosol optical
•2National
OceanicandAtmospheric
Administration,
SilverSpring,
Maryland. depth spanningmost of the twentieth centuryhas been mea-
•3Scripps
Instituteof Oceanography,
La Jolla,California. sured from Sun photometers.The vast majority are site spe-
•4Department of AppliedScience,Brookhaven NationalLabora- cific, short-term investigationswith little relevance for sea-
tory, Upton, New York. sonal, annual, or long-term trend analysis,however a few
•SNowat NavalResearchLaboratory,Washington,D.C.
•6BenGurionUniversityof the Negev,SedeBoker,Israel. multiyear spatial studieshave contributedto our knowledge
•7CARTEL, Universit6de Sherbrooke,Sherbrooke,Quebec,Can- and experience(Table 1). The followingsectionreviewsthese
ada.
investigations, past and present,which significantlyaddressed
•sSAIC-GSC,Beltsville,Maryland,and NASA GSFC, Greenbelt, long-term measurementsover widely distributedlocationsor
Maryland.
•9Department
of Meteorology,
Universityof Maryland,College provided a significantcontributionthat allowed development
Park, Maryland. of a network for long-termphotometricaerosolobservations.
2øDepartment
of Physics,
University of Miami,Miami,Florida. The earliest systematicresultscome from the Smithsonian
2•Space
Applications
Institute,JointResearchCentreof the Euro- Institution solar observatories.Roosene! al. [1973] computed
pean Commission,Ispra, Italy.
extinctioncoefficientsfrom 13 widely separatedsites during
Copyright2001 by the American GeophysicalUnion. the first half of the twentiethcenturyusingspectrobolometer
Paper number 2001JD900014. observations by the AstrophysicalObservatoryof the Smithso-
0148-0227/01/2001JD 900014509.00 nian Institution. They concludedthe aerosol burden did not

12,067
12,068 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY

Table 1. Summaryof ExistingAerosol Optical Depth MeasurementsFrom Network-RelatedSources

Coverage
Reference (Number of Sites) Temporal Comment

Forganet al. [1993] Global (95) 1971-1986 unverifiable

Ben Mohamedet al. [1992] Niger (7) 1986-1987 accuracy-0.05 est.
DMlmeida [1987] West Africa (11) 1980-1984
Dutton et al. [1994] background(4) 1977-1992 pyrheliometer,accuracy-0.04
Flowerset al. [1969] USA (43) 1960-1966 500 nm, accuracy-0.04
Forgan[2000] Australia(7) 1990to present accuracy
Gushchin[1988] former SovietUnion (33) 1968 to 1984 accuracy-0.03
Holbenet al. [1991] West Africa (15) 1985-1986 accuracy-0.05
Holbenet al. [1998] global(-100) from 1993to present QA, accuracy-0.015
Michalskyet al. [1994] eastUSA (11) from 1992 to present QA, accuracy-0.015
Herberet al. [1993] Antarctica(5) 1956to 1992 various inst.
Roosenet al. [1973] North and SouthAmerica,Africa (13) 1905-1922 Smithsonian observatories
Smirnovet al. [1995] globaloceans 1967 to 1994 56 references
Smirnovet al. [1996] Canada(4) 1986to 1992 500 nm, accuracy-0.02
SKYNET (Takamura) eastAsia (8) 1996 to present
Bigelowet al. [1998] USA (40) 1995 to present unverifiable
Jinhuanand Liquan [2000] China (5 cities) 1980-1994 pyrheliometer,broad-band
Shaw [1982] polar regions(5) 1912-1982 varioussites,techniques
Volz [1969] centralEurope (30) 1961-1965 500 nm, accuracy-0.04

detectiblychange at remote high-altitudesitesfrom 1905 to discernabletrendsare observablesuggesting no anthropogenic

1950. Seasonaland volcaniceruptionswere evident,however induced trends.
long-term trends were not. Gushchin[1988] reported extensivemeasurementsin the
Volz [1959] developedan analogSun photometerwith a 500 former Soviet Union from 1972 to 1984 as part of the
nm interference filter that became the basis for extensive ob- BAPMoN program.Observationswere taken within the spec-
servationalnetworksin Europe [Volz, 1969] and the United tral range of 340-627 nm at more than 30 sites,which char-
States.Flowerset al. [1969] describedthe U.S. Volz Sun pho- acterize rural, urban/industrial, maritime/continental, and
tometer network consistingof 29 stationsacrossthe United desert aerosol. Most observations were not continuous, but
Statesfrom 1961 to 1966. They showedthe base 10 turbidity multiyearand monthlyaveragesof aerosoloptical depth and
parameterto vary asa functionof localityand time of year and its spectraldependencealongwith the maximaland minimum
thuswere the first to developan aerosolclimatologybasedon valueswere presentedfor somesites.
light extinction.The highestvalueswere reported for eastern West Africa and desert dust aerosol has been the focus of
U.S. stations in July and August, the lowest values in the three networkSun photometricinvestigations for much of the
intermountainbasin of the western United States.Although 1980s. The African Turbidity Network from 1980 to 1984
the recordis only availablefrom graphspresentedin the above [DMlmeida,1987],the NASA 15 siteSaheliannetwork[Holben
citation, valuesare generallyconsistentwith current measure- et al., 1991], and the Niger network from 1986 to 1987 [Ben
ments.An accuracyassessment of the Flowerset al. [1969]data Mohamedet al., 1992]. Clearly, thesenetworksdemonstrated
is not presented. The European network operated 30 Sun the temporal and spatial variability and overall high aerosol
photometersfrom 1963 to 1967 [Volz, 1969].Volz reported a loadingin West Africa, duringthis decadeof extremeSahelian
casestudyfrom central Europe, but no apparent attempt was drought.
made to put the databaseinto a climatologicalcontext. No Smirnovet al. [1995] compiledand summarizedaerosolop-
accuracyassessments were made for this network. The data- tical depth observationsfrom over 50 publishedsourcesfor
base is available from the National Climatic Data Center in marine observationsfrom cruisesspanning30 years of obser-
Ashville, North Carolina, United States. vations worldwide. These results showed a wide scatter of
Shaw [1982] reported 800 measurementsof atmospheric opticalparametersin generaland somesignificantscatterbe-
aerosolopticaldepth (500 nm) polewardof 65ø latitudeprior tween optical data for certain areasin particular. Coastaldata
to 1982. He suggestedthat the southern polar regions are are greatly influencedby continentalsourcesand accordingly
pollution free with backgroundconditionsrangingfrom 0.025 the aerosoloptical depth valuesare as a rule greater than in
at McMurdo to 0.012 at the South Pole. Similar background the remote ocean air. Unfortunately, numerousexperiments
conditionsat McMurdo were reported for Barrow and Fair- were sporadic,employedonlya fewwavelengths,and measure-
banks, Alaska, in summer, but values increase to 0.135 and ment accuracysometimeswas unknown.However, it was evi-
0.110, respectively,in March and April with the onsetof arctic dent that in a clean maritime atmospherequasi-neutralspec-
haze. He also convertedVolz [1968] turbidity measurements tral behaviorof aerosoloptical depth predominates.
from 1912to 1922to AOD at Uppsala(60øN)for March means Aerosol optical depth has been retrieved usingbroadband
and noted that the backgroundvaluesare -0.06 lower than the (0.3-4/•m) directnormalobservations madefrom pyrheliom-
contemporarymeasurements, suggesting that arctichaze is a eters,which are more commonlydeployedat meteorological
recent phenomenonof anthropogenicorigin. observatories. Jinhuan and Liquan [2000] reported observa-
Herber et al. [1993] reported several long-term measure- tionsfrom a 12 stationnetwork operatingin China'snorthern
ments from coastal Antarctica and satellite observations. interior from 1980 to 1994. Effective Optical depth retrievals
Recordsdatingto 1956 clearlyshowedthe influenceof volca- were made at 750 nm with an estimatedaccuracyof 10% due
nic eruptionson stratosphericloading;however,no long-term to the retrieval method, but no mention is made of errors due
HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY 12,069
12,070 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY

to calibrationuncertainty.High optical depths,for the colder Africa,AEROCAN in Canada)since1993.The resulthasbeen

interior citiesin winter due to emissionsfrom coalheatingand a long-termquality-assuredrecordfor a largediversityof aero-
springtimepulsesfrom dustsources,were routinelyreported. sol types,mixtures,and sourceand transportregions.An un-
In contrast,Beijing reported a summermaximum.None of affiliated network basedin Japan (Skynet)providessimilar
thesesitesnecessarilyrepresentrural conditions. measurements.

Dutton [1994] presentedbroadbandaerosolopticaldepths This paper presentsthe first of a multipaper studyof the
from pyrheliometers(0.3-3/•m) asmonthlyaverageanomalies climatology of aerosol optical properties retrieved from
from measurementstaken at four NOAA/CMDL background AERONET measurements.We present here the monthly
stations(Barrow, Alaska, Mauna Loa, Hawaii, Samoa,U.S. aerosolopticaldepthsmeasuredor interpolatedto 500nm, the
Territories,and SouthPole) from 1977to 1992.All sitesbeing multispectralAngstromparameter(or wavelengthexponent),
far removedfrom anthropogenicsourcesof aerosolsshowed and the retrievedprecipitablewater at nine selectedsitesrep-
no significantlong-term trends after the effect of volcanic resentingaerosolsfrom biomassburning,desertdust,biogenic/
eruptionswere removed.Slight seasonalitywas observedat background,and anthropogenic/urban sources.In most cases
Mauna Loa associatedwith Asian dustfrom March to May and the monthly record is continuous,however several sites are
Arctic haze during the sametime period in Barrow. limited to summerand/or dry seasonssummariesdue to the
Historically,the most ambitiousattempt to monitor back- difficultyof maintainingyear-longmeasurements. To providea
groundaerosolopticaldepth levelswas that organizedunder more completeassessment of the programresults,we provide
the auspicesof the WMO BAPMoN programwhichofficially Appendix A of monthly and yearly averageaerosoloptical
operatedfrom 1972to 1992.The networkwascomposed of 95 depth, Angstromexponent,and precipitablewater for addi-
sitesoperatedby member countriesthat providedprocessed tional AERONET siteswith over 2 years of quality-assured
data to the National Climate Data Archive in Ashville, North data.
Carolina. The diversity of instruments,expertise, analysis
1.1. Instrumentation and Methods
methods,and quality controlled a World MeteorologicalOr-
ganization(WMO) evaluationcommitteeto recommendaban- All of the measurements reportedin this paper were made
donment of the network and declare the data archive in its with CIMEL Sun/skyradiometers,which are a part of the
presentform as unsuitable"... for scientificanalysisof either AERONET global network. These instrumentsare described
short or long term changesin global aerosoloptical depth." in detail by Holbenet al. [1998];however,a brief description
Experiencefrom BAPMoN was used in formulatingand es- will be given here. The automatic-trackingSun- and sky-
tablishingthe follow-on network Global AtmosphericWatch scanningradiometersmake direct Sun measurementswith a
(GAW) in 1996. The GAW network consistsof ---12 back- 1.2øfull field of view at leastevery15 min at 340, 380, 440, 500,
groundlocationswith identicalinstrumentsprovidedand sup- 675, 870, 940, and 1020nm (nominalwavelengths). The direct
portedby the Swissgovernment.Resultsfrom thisnetworkare Sun measurementstake 8 s to scanall eightwavelengths,with
not publishedbut informationis obtainedthroughthe GAW a motor-drivenfilter wheel positioningeach filter in front of
website.Subsetsof the BAPMoN programhavebeen shownto the detector.A sequenceof three measurements(termed a
be successful in Canada [Smirnovet al., 1996] or developed triplet) taken 30 s apart are made resultingin three measure-
spin-offnetworksasin Australia [B. W. Forgan,personalcom- mentsat eachwavelengthwithin a 1 min period. These solar
munication, Bureau of Meteorology, Melbourne, Victoria, extinctionmeasurementsare then used to compute aerosol
Australia,2000). opticaldepth at eachwavelength(ra(X)) exceptfor the 940
Michalskyet al. [1994]reportedon a networkof 11 rotating nm channel,whichis usedto retrieveprecipitablewater (PW)
shadow band radiometers established in the eastern United in centimeters. The filters utilized in these instruments are
States.The network is notablefor its high temporalresolution interferencefilterswith a band pass(full width at half maxi-
and in situcalibrationapproach.Data collectionbeganin 1992 mum) of the 340 nm channelat 2 nm and the 380 nm filter at
and continuesto the presentand is availablethroughthe In- 4 rim, while the band passof all other channelsis 10 rim.
ternet. This instrumentapproachhasrapidlyexpandedto nu- The data, whichwe analyzedfor the GoddardSpaceFlight
merousresearchfacilitiesand programs[Bigelowet al., 1998] Center (GSFC), in Maryland, utilized only ra(X) measure-
but has not coalesced into a coordinated network. mentsfrom Mauna Loa Observatory (MLO) calibratedinstru-
A developingnetwork of inexpensivelight-emittingdiode ments.These referenceinstrumentsare typicallyrecalibrated
handheldSun photometersis being supportedby the NASA at MLO every -3 monthsusingthe Langleyplot technique
Global Learningand Observationsto Benefitthe Environment with morningdata only. The zero air massvoltages(V o, in-
(GLOBE) programwith the goalof involvingstudentscientists strumentvoltagefor direct normal solarflux extrapolatedto
to take scientificqualitymeasurements [Mims,1999].The po- the top of the atmosphere[Shaw,1983]) are inferredwith an
tential to make regular observationsat internationallydis- uncertaintyof -0.2 to 0.5% for the MLO-calibratedreference
persedsecondaryschoolswith a commondatabasecouldbe of instruments[Holbenet al., 1998].Thereforethe uncertaintyin
great benefiteducationallyand scientifically.No resultsare yet r, becauseof the uncertaintyin zero air massvoltages(com-
available. puted as the standarddeviation/meanof the Vo valuesfrom
The AERONET program [Holbenet al., 1998] is an auto- MLO) for the referenceinstrumentsis better than 0.002 to
matic robotic Sun- and sky-scanningmeasurementprogram 0.005.
that has grownrapidly to over 100 sitesworldwide(Plate 1). The stabilityin time of the Vo values derived from MLO
The program provides the satellite remote sensing,aerosol, Langleyanalyses for one of our referenceinstruments(101) is
land and ocean communitiesquality-assuredaerosoloptical shownin Figure 1. The data in thisfigurecoverthe time period
propertiesto assessand validate satelliteretrievals.The real- of September30, 1997,to September11, 1999,a nearly2 year
time globallydistributednetwork has grownthroughinterna- interval. The filters in use for this instrument and all others in
tional federation (PHOTON, primarily in Europe and West the AERONET network from 1997 onward were ion-assisted
 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY 12,071

nm agreedto within 0.015 (rms), which is similarto our esti-

mated level of uncertaintyin ra retrieval for field instruments.
The techniqueof Brueggeet al. [1992] is used to retrieve
precipitablewater. Given the currentdiscussion for the opti-
mal methodfor PW retrievalsfrom Sunphotometryobserva-
tions[Schmidet al., 2000] and the relativelylargeuncertainty
in the modified Langley Vo, we conservatively estimateour
uncertaintyto be _+10%. Informal comparisons to retrievalsof
PW from Ramanlidar, microwaveradiometer,otherSunpho-
tometer methodsand radiosondessupportthis estimate.
The Angstromwavelengthexponenta valuespresentedin
thispaperwere computedasthe slopeof the linear regression
of In % versusIn X usingthe 440, 500, 675, and 870 nm filter
• i•-&-•
675nm data. Prior to 1995 there were no 500 nm filter data, therefore
0 -g + 500nm the a valuesin 1993-1994were computedfrom the 440, 675,
and870 nm dataonly.The % dataat 500 nm whichwe present
-iO 3"4.:'tJ
':.:.ci?
for 1993-1994are computedfrom interpolationof the 440 and
0 I00 200 300 400 500 600 700 675 nm data on a In % versusIn A scale.It is recognizedthat
there is often significantspectralvariation of a for aerosolsize
Days from Initial Calibration
distributionswith an accumulationmode [Eck et al., 1999;
Figure 1. Time seriesof the zero air massvoltages(i/o) of O'Neill et al., 2000].In thispaperwe presentonlythe 440-870
AERONET referenceinstrument101. I/o valueswere deter- nm linear fit determinationof a as a first-orderparameter
mined from the Langley techniquefrom stable morningsat indicativeof the general size distributionand the relative dom-
MLO.
inanceof fine versuscoarsemode particles.
The AERONET %(A) data in this paper were cloud
screenedfollowingthe methodologyof Smirnovet al. [2000a],
depositioninterference filters. We computed the average and here we presentjust a brief outlineof the procedure.The
yearlychangein l/o from a linear regressionof l/o versustime principal filters used for the cloud screeningare based on
for the entire 711 day data record.As depictedin Figure 1, temporalvariabilityof the % (X), with the assumption being
there are varyingratesof changefor the differentwavelengths, that greatertemporalvariancein % is due to the presenceof
rangingfrom 0.24%/yr for 870 nm to -3.90%/yr for 675 nm. clouds.The first filter is a checkof the variabilityof the three
The changein l/o with time showsin general, a linear ten- % valuesmeasuredwithin a 1 min period. If the difference
dency,with the exceptionof the 940 nm filter. This is due to a betweenminimumand maximum%(A) within this 1 min in-
much larger uncertaintyin l/o determinationfor the 940 nm terval is greaterthan 0.02 (for % lessthan 0.667) or 0.03 %
channelas a result of water vaporvariabilityat MLO as con- (for % greaterthan 0.667) then the measurement is identified
trastedwith a verystableaerosolenvironment.The repeatabil- as cloudcontaminated.Then the time seriesof the remaining
ity of morningLangley-derivedvaluesof l/o for aerosolchan- ra(X) are analyzedfor the presence of rapidchanges or spikes
nelsat MLO, from typically5 to 15 mornings,is excellentwith in the data. A filter based on the second derivative of the
a coefficientof variability(standarddeviation/mean)averaging logarithmof % (X) asa functionof time isemployedto identify
only ---0.3-0.5%, while the value for the 940 nm channel av- rapid variations which are then eliminated as observations
erages2-4 %. affectedby cloud.Other secondary-order cloudscreeningand
The Sun/skyradiometersat sitesother than GSFC utilized in data quality checksare also made, and these are describedin
this study were intercalibrated against a MLO-calibrated detailby Smirnovet al. [2000].This cloud-screening technique
AERONET reference instrument at GSFC both before de- has not beenvalidatedon a broad scalealthoughthe proce-
ploymentin the field and postdeployment. The period of time dure was favorablytested on experimentaldata obtained in
betweencalibrationsfor field instrumentstypicallyvariesfrom different geographicaland optical conditions[Smirnovet al.,
6 to 12 months.A linear rate of changein time of the zero air 2000]. For a variety of sites, our cloud-screening algorithm
massvoltagesis then assumedin the processingof the data eliminated from 10% to 50% of the initial data.
from field sites. Our analysissuggeststhat this results in an
uncertaintyof ---0.01-0.02 in r a(A) (wavelengthdependent) 1.2. Mauna Loa, Hawaii
due to calibrationuncertaintyfor the field instruments. The Cimel Sun photometer on Mauna Lea, Hawaii, is le-
Eck et al. [1999]havecomputedthe combinationof calibra- cated at the Mauna Lea Observatory (MLO) (19ø53'N,
tion uncertainties(l/o) and uncertaintyin ozone(due to sea- 155ø57'W)whichis 3400 m abovesealevelon the north sideof
sonality and atmosphericdynamics)and Rayleigh optical the gentlyslopingshieldvolcano(summitis 4170 m). The
depth(dueto variabilityin air pressure),for opticalair massof nearbysurfaceconsistsof bare lavarockwith no vegetationor
1, in the mannerefRussellet al. [1993].The resultingestimated soil and therefore minimal local productionof aerosol. In
total uncertaintyis ---0.010-0.021in computedra(X ) for field addition,other factorsthat contributeto the very low aerosol
instruments(which is spectrallydependentwith the higher loadingat MLO are its mid-Pacificlocation(>3500 km from
errors in the UV) and ---0.002to 0.009 for reference instru- the nearestcontinent)and its heightabovethe marinebound-
ments.Schmide! al. [1999] compared'ra(A) valuesderived ary layer. MLO is consideredto be the best location for cali-
from four differentsolarradiometers(onewasan AERONET brationof directSun-observing instrumentation by the Langley
Sun-skyradiometer), operatingsimultaneously together in a method appliedto morningdata (air mass> 2), due to the
field campaignand found that the 'ra valuesfrom 380 to 1020 extremelylow and stableaerosoloptical depth [Shaw,1983].
12,072 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY

a) o o o•,

0,•0•
0,04 004

I•'002
002

o,o•
00!
O, OO

: : i ' n-964

2.0

•,20
1•0

o 0.0
Jan Jan Jan Jan Jan Jan 0o00 0.02 0.04 0.06 0.08 0,00 0.02 0.04 0.06 0.08 O. 10
1994 1995 1996 1997 1998 1999 aerosol optical depth(SO0 rim) aerosolopticaldepth(500 nm)

f) 20 ,:•-.-c.,- .. .....

o.si.............

o,• •½
..........
:•....;.'t.....••...,:.,.•.:....,:•.•
...........
%.,.:..........
:•.•..,....:...:.•
........ o {0

o.4
•
o.2
•
0
Jan Jan Jan •n Jan Jan 0. I 0,5 o.g I •: 1.7 2.1 2.5 2•9
1994 1995 19• 1997 1998 1999 Angstrom parameter c•

Figure 2. MaunaLoa, Hawaii.Meanmonthlyvaluesof aerosolopticaldepthat thewavelength500nm for

thewholeperiodof measurements
(a) (thebarsindicateplusorminusonestandarddeviation),
meanmonthly
(b) andmeandailyvalues(c) of aerosol
opticaldepthat 500nm,meandailyvaluesof precipitablewater(d),
frequency
of occurrence
of aerosolopticaldepth(e) andAngstrom parameter(f), scattergram
of Angstrom
parameter
versusaerosolopticaldepth(g),andscattergram of aerosol
opticaldepthversusprecipitable
water
(h).

However,typicallyafter 0900-1000 LT air flowingup the to The aerosolopticaldepthdata,whichwe presenthere, are

the observatoryaltitudefrom the breakupof the marineinver- daily averagesrepresentativeof the completediurnal cycle
sion layer often resultsin rapid temporalvariationand an (Figure2, Table2). Sincecloudcoveristypicallygreaterin the
increasein aerosol concentrations[Perryet al., 1999; Shaw, afternoon,due to the arrival of moisture-richmarine boundary
1979].Emissionof gasesand the formationof aerosolsas a layerair, there are moremorningcloudless observations, and
result of volcanicactivity on the island of Hawaii occurpri- thereforethe dailyaveragesof aerosolopticalthickness at 500
marilybelowthe MLO altitudebut are sometimes transported nm (ra•00)are moreheavilyweightedby morningdatawhen
to the observatoryaltitude and aboveduringthe daytimeby tabOOis alsolower.As a resultof frequentcalibrations
fromthe
upslopeflow causedby heatingof the mountainslopeand morningLangleyanalysis,our estimateduncertaintyin instan-
growth of the marine mixed layer [Luria et al., 1992;Ryan, taneousvaluesof tabOO is approximately_+0.003.The seasonal
1997]. variationof the monthlyaverageaerosolopticaldepth(Figure
 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY 12,073

Table 2. DatabaseSummaryfor MeasuringPeriod for Mauna Loa Observatory,Hawaii (Lat 19ø32'NLong 155ø34'W,
Elevation3397 m), 1994-1999a
'ra500 O' O• tr. PW tr N Month

Jan. 0.015 0.006 1.77 0.41 0.22 0.17 57 3

Feb. 0.017 0.009 1.68 0.42 0.23 0.18 62 3
March 0.033 0.022 1.39 0.51 0.21 0.11 98 5
April 0.033 0.018 1.32 0.41 0.20 0.11 90 4
May 0.030 0.022 1.14 0.45 0.32 0.14 109 5
Ju n e 0.020 0.008 1.35 0.48 0.32 0.14 120 6
July 0.014 0.005 1.57 0.51 0.25 0.14 53 4
Aug. 0.013 0.004 1.78 0.42 0.30 0.15 75 4
Sept. 0.014 0.005 1.63 0.38 0.26 0.11 65 4
Oct. 0.017 0.009 1.32 0.62 0.32 0.18 87 5
Nov. 0.017 0.012 1.48 0.59 0.36 0.22 91 5
Dec. 0.015 0.010 1.62 0.54 0.31 0.22 57 4
Year 0.020 0.008 1.50 0.20 0.27 0.05 964 52

aAerosolopticaldepthat 500 nm (ra500),Angstromexponent(a), precipitablewater (PW), associated

standarddeviations((r), numberof days
(N) and monthsin the observationperiods.

2a) showsmaximum values in the spring seasonmonths of occurrencehistogram of raSOO for the entire data record of
March, April, and May. This seasonalpeak in springis due to 1994-1998 (Figure 2e) showsa skeweddistributionwith a peak
the long-rangetransportof primarily Asian aerosolsto MLO from 0.01 to 0.02, with lesserfrequenciestrailing off at higher
[Perryet al., 1999; Shaw, 1980]. Perry et al. [1999] have mea- Ta values.
sured springpeaks in fine soil massand elementsassociated Daily averageprecipitablewater amounts(Figure 2d), in
with fly ash (bromine, zinc, and lead) from coal burning, in centimeters,exhibitlarge dailyvariabilityand someseasonality
additionto an anthropogenicsulfateenhancement.Lidar ob- with higher values in summer and fall. The range of values
servationsat MLO [Barnesand Hoffman, 1997] show only a from lessthan 0.05 cm to almost 1.0 cm is nearly identical to
slight enhancementof stratosphericaerosol backscatterfrom the range of PW valuesmeasuredby Dutton et al. [1985] from
the Mount Pinatubo eruption (June 15, 1991) in 1994 and 1978 to 1983 (morning data only), alsofrom Sun photometric
approachingbackgroundlevels in 1995. Nearly continuous retrievals.It is interestingto note that the dry period in the
monitoringof ra by AERONET commencedin 1996 (Figure winter-springof 1998 (Figure 2d) is very similar in duration
2b), thereforemost of the data presentedhere are not signif- and magnitudeto a dry period observedby Dutton et al. [1985]
icantly affectedby volcanicaerosolsin the stratosphere,since at MLO from December 1982 to March 1983 both of which
there are only 2 monthswith observationsin 1994. occurred during strong E1 Nifio cycles.The relationshipbe-
Daily mean valuesof r a500(Figure 2c) showthe springsea- tweenaerosolopticaldepth and precipitablewater (Figure 2h)
sonal peaks (maximum daily values of-0.09) but also very showsno significantcorrelation, thus suggestingthat part of
large day-to-dayvariability due to variation in air masstrajec-the reason may be that aerosol and water vapor are trans-
tories transportingaerosolsfrom differing sourceregions to ported at different altitudesaboveMLO and/or that there are
MLO [Perryet al., 1999].Daily variabilityis alsodue in part to trajectoriesfrom severalsourceregionswith differingseasonal
variation in the productionand transportof volcanicaerosols combinationsof raSOO and PW.
from the activevolcanoeson the island.It is noted that raSOO in The Angstrom wavelength exponent plotted versus raSOO
the spring months (March-May) of 1999 was significantly (Figure 2g) showsa weak trend of decreasingvaluesof a as
higherthan was measuredfor the springmonthsof 1996-1998 tabOOincreases.This may be the result of some of the highest
(Figures2b and 2c). Thus it appearsthat there is significant observationsof raSOO being associatedwith the transport of
interannualvariabilityin aerosoltransportto MLO. Minimum Asian soil dustcomposedof coarsemode (>1/xm) sizeparti-
valuesthroughoutthe 1995-1998period rangefrom -0.006 to cles.All of the valuesof a > 0.8 for raSOO > 0.06, shownin
0.009, which is similar to the minimum for 1982-1992 observed Figure 2g, are from spring 1999 when there was a higher than
by Dutton et al. [1994] of 0.008 from the winter of 1990-1991 normal level of transportof aerosolto MLO. Theseepisodesof
prior to the Mount Pinatubo eruption. It is noted, however, relativelyhigh a and raSOO in 1999 suggesttransportof indus-
that Dutton et al. [1994] computedraSOO from measurements trial pollution or possiblymixed desert dust and industrial
made in the morning only during the Langley measurement pollution.Part of the reasonfor the wide range in a valuesfor
sequence,and thus their values are lower than would be for raSOO < 0.02 is due to the uncertainty in ra of-0.003 ap-
daily averages,which is what we report here. Dutton et al. proachingthe magnitudeof the ra data, thusresultingin larger
[1994] also determined a stratosphericbackgroundvalue of errors in a computation.The frequencyof occurrenceshisto-
aerosoloptical depth of 0.004 for the winter of 1990-1991. gramof a (Figure 2f) showsa broad peak from - 1.0 to 2.0 with
Shaw[1979]computeda meanvalue of total atmosphericcol- a minimum near zero and a maximum at 3.0. The annual
umn raSOO = 0.019 for MLO measurementsmade from Mar- averagevalue of a computedfrom the monthly meansis 1.48.
ch-August 1976 and January-February 1977. The value we In comparison,Dutton et al. [1994] computedan a of 0.7 for
compute from the multiyear AERONET observationsfrom near-backgroundconditionsin 1990-1991, and Shaw [1979]
monthly means for January-August (the same months as inferred a valuesthat ranged from 1.1 to 3.5 with a mean of
[Shaw,1979]) is similar,0.022. In addition,Shaw [1982] com- 1.63; however,his data from 1976 was influencedby volcanic
puted an annualmean tabOO at MLO for 1980of 0.020,which is aerosolsin the stratospherefrom the Augustinevolcanoerup-
equal to our multiyear annual mean of 0.02. The frequencyof tion. At low r a the stratosphericbackgroundaerosoloptical
12,074 HOLBENET AL.:EMERGINGGROUND-BASED
AEROSOL
CLIMATOLOGY

o 5 o o d o o d o o
'• (UJU
OOS)
q•,d•0
le311do
10soJee
 HOLBEN ET AL.' EMERGING GROUND-BASED AEROSOL CLIMATOLOGY 12,075

g)

Jan Jan Jan Jan Ji•n Jan Jan Jan Jan Jan Jan Jan Jan Jan
1993 1994 1995 1996 1997 1998 1999 1993 1994 1995 1996 1997 1998 1999
i) 25 , J) 30
ß ; :

i n•1325

25
o• 20

o
=20
,.I5
o
o
o ,.

o10
o

o •10

• 5

o
0 0,2 0.4 0.6 0.8 I 1.2 1.4 0.1 0.5 0.9 1.3 1.7 2.1 2.5 2.9
aerosol optical depth (500 nm) Angstrom parameter c•

2.0

E •.•

E
o 1.o

0.0
0 I 2 3 4 5 6
water vapor content (cm)

Figure 3. (continued)
12,076 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY

depth (-0.003 to 0.005) may comprisea significantcontribu- associatedwith the competitionof airflowsfrom northern and
tion to the total column-integratedaerosoloptical properties. southwesternsectors,and frequent air stagnationabove the
Typical nonvolcanicstratosphericoptical depths,as summa- area, is most likely responsiblefor the fall pattern [Bryson,
rizedbyRussell et al. [1993],resultin a varyingfrom 1.0to 1.5 1966;Trewartha,1954].Parallelsetsof graphsof the Angstrom
for the 500 to 1000 nm wavelengthrange. parameterdo not displayany obviousseason-to-season trends.
The daily averagevalues of the total precipitablewater
1.3. GSFC, Greenbelt, Maryland amount(Figure 3h) showtypicallyhighervaluesin the sum-
GoddardSpaceFlight Center (39ø01'N,76ø52'W,50 m ele- mertimewhichrangeto maximumvaluesof 4-5.5 cm,while in
vation), locatedin suburbanWashington,D.C., and -30 km November-February,valuesare typicallylessthan 1 cm. This is
south of industrial Baltimore, aerosol environment is influ-
encedby a synopticscalemeteorology.A southerlyflow due to
the Bermudahigh is a dominantfeature from late springto the Winter, Dd-F
earlyfall months,and a westand northwesterlyflow typifiesthe
other months.Someepisodesof eachtype of flow may occurat
anytime of the year and be regionallymodifiedby coldfronts
with a strong southerlyflow in advanceof the front and a
northwesterlyflow behind. Most heavy industryis located to
the north, and local emissionsare dominatedby automobiles,
owned by the 2.3 million metropolitan area residents.The
landscapeis dominatedby deciduoustreesleafedout from late
April to October.
Figure 3a illustratesthe monthly averagedaerosoloptical 0.0 02 OA 0•6 O•a •0 I Jt 1•4

depthat 500nm for the 7 yearrecord(1993-1999)at Goddard.

Spring,M-A-M
A totalof 1297dailyaverages are analyzed.The aerosoloptical
depthis dominatedby a markedincreasein opticaldepthfrom
Juneto September,whichpeaksduringJulyand August.The
7 year Julymean of %500 is 0.48. In contrast,the aerosoloptical
depth decreasesto a minimum during the winter months,av-
eraging-0.10 from Novemberto January.Standarddeviations
generallyincreasewith the mean valuesof %. Monthly aver-
ages of % and a for the whole period of observationsare
shownin Figures3c and 3d. With respectto the % variations,
a "classical"annualpattern with an increaseto maximumtur-
bidity in the summertime[Flowerset al., 1969;Petersonet al.,
Summer, J-J-A
1981] is apparentin all the years.A winter minimumis always
in evidence.Becauseof the post-Pinatubocontribution,% was
slightlyhigher for the fall and winter of 1993-1994 than the
value expectedfrom the historicaldata. A notabledecreaseof
the Angstromparametera (Figure 3d) in 1993 is alsoassoci-
ated with the post-Pinatuboeffect. Generally,no regular pat-
tern is seen in the mean monthly values of the Angstrom
parameter, althougha late winter and early springminimum
can be noticed. The mean annual values of a are close to
1.6-1.7 for all years(exceptfor 1993).
Daily averagevaluesof %500for all 7 yearsshowvery large
day-to-dayvariationespeciallyduringthe summertime(Figure Fall. S-O-N

3g). The dramaticincreasein summeraerosolloadingoverthe

eastern United States is a dynamic mixture of natural and
anthropogenicsources,processedby convectionwithin humid
air masses.A histogramof Ta and a is shownin Figures3i and
3j. The aerosoloptical depth probabilitydistributionis rather
narrow with the modal value of about 0.1. The probability
distributionof a is relativelybroaderwith the modal value of
about 1.7. Figure 4 showsthe seasonalvariabilityof aerosol
opticaldepth and the Angstromparameter(DJF is winter,
MAM is spring,JJA is summer,and SON is fall). It canbe seen Angstromparameter
from Figure 4 and Table 3 that the atmospherewas typically
Figure 4. Goddard Space Flight Center, Greenbelt, Mary-
more turbid and morevariablein the summer(wider distribu-
land. Seasonalfrequency of occurrenceof aerosol optical
tionsor largerrrvalues).The wintermonthshavethe narrow- depthandAngstromparameterfor (a, b) winter(December•
est probabilitydistributionof %s00with a modal value about JanuarysFebruary),(c, d) spring(March•April•May), (e, f)
0.08. In the springthe maximum shiftstoward greater values summer(June•July•August), and (g, h) autumn (Septem-
(0.18), and in the fall, the distributionwidensashighervalues ber•October•November). The sum of all frequenciesis
of *a appear.The complexnature of atmosphericprocesses, equal to 100% for each season.
 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY 12,077

Table 3. DatabaseSummaryfor MeasuringPeriod for GoddardSpaceFlight Center, Greenbelt,Maryland (Lat 39ø01'N

Long 76ø52'W,Elevation50 m), 1993-1999
zasoo tr a tr• PW tr N Month
Jan. 0.10 0.06 1.70 0.37 0.87 0.54 72 5
Feb. 0.12 0.06 1.48 0.37 0.72 0.29 78 6
March 0.14 0.08 1.50 0.34 0.94 0.57 96 6
April 0.19 0.10 1.41 0.34 1.19 0.63 119 6
May 0.21 0.10 1.49 0.33 1.81 0.74 110 7
June 0.37 0.23 1.65 0.29 2.80 0.99 120 7
July 0.48 0.29 1.75 0.28 3.42 0.94 142 7
Aug. 0.44 0.28 1.76 0.27 3.15 0.91 135 7
Sept. 0.31 0.26 1.76 0.27 2.49 0.89 120 7
Oct. 0.15 0.12 1.65 0.35 1.50 0.59 114 7
Nov. 0.11 0.08 1.70 0.30 1.05 0.59 112 7
Dec. 0.09 0.06 1.78 0.37 0.89 0.57 107 7
Year 0.23 0.14 1.64 0.13 1.73 0.98 1325 79

consistentwith the general synopticpattern for the area in aerosols.The nearestcity, Albuquerque,New Mexico, is 100
which polar and arctic air masses,which are cold and dry, km north, thus no local anthropogenicsourcesare present.
dominatein winter, and moistwarm tropical air massesdom- The aerosoloptical depth record began in May 1994 and
inate in summer. Mean monthly values of the precipitable continuesalmostunbrokento the present(Figure 5, Table 4).
water are presentedin Figure 3e. The relativelycold summer The monthlyaveragedrecord clearlyshowsthe long-termsea-
of 1996was influencedby air advectedfrom northern Canada, sonalvariationsin aerosolopticaldepth.A gradualincreasein
yielding smaller values for the months of July and August. Ta500from the Januarylow of 0.03 to relativelybroad but low
Otherwise,the temporalannualvariabilityof precipitablewa- summerpeak in June, July, August, and Septemberof 0.11 is
ter is quite similar from year to year. Derived mean monthly followedby a gradualdecreasethroughthe fall to the midwin-
values are consistent(within 10%) with the calculationsof ter minimumvalues(Figure5a). The mean annualTa5OO of 0.08
Gueymard [1994](estimates basedon the long-termsurfacelevel representsone of the lowestvaluesin the AERONET network.
temperaturerecord)madefor the Washington,D.C., area. Summermaximumsof % typicallyoccur in July-September
The relationshipbetweendaily averagesof precipitablewa- and minimumsin November-February(Figure 5c). Daily av-
ter and aerosolopticaldepth at 500 nm (Figure 31) showsa erages,as expected,show a large variability within the mean
strong correlation (r2 of0.56).Thisresultisconsistent withthe annualcycle,which typicallyvariesbetween0.02 and -0.30 at
intra-annualvariability of aerosoloptical properties,synoptic 500 nm (Figure5d). Instantaneous measurements occasionally
air masses,and associatedamount of precipitablewater. The far exceeddaily averagedmaximumsduring dust stormsbut
trendmaybe observedin Figure3f wheremeanmonthlyvalues are short-lived.Some dust events,which may be associated
of %500 areplottedversus PWmonthly means, increasing
ther2 with cold fronts and cloudyconditions,are filtered out of the
to 0.79. The exponentialfit is consistentwith the relationship data set.
establishedfor the U.S. Atlantic coastsitesduring the TAR-
Retrievedprecipitablewater closelyfollowsthe aerosoldy-
FOX experiment[Smirnovet al., 2000b].
namicswith peakvaluesin JulythroughSeptember(peak daily
1.4. Sevilleta, New Mexico averagesas high as 3 to 3.5 cm) associatedwith the summer
Sevilleta is located in the arid intermountain basin of the monsoonflow into the desertsouthwestern United States(Fig-
American Southwest, ---1400 km east of the Pacific Ocean ure 5e). Minimum valuesoccur throughout the winter season
(34ø21'N,106ø53'W,elevation1850m). The annualprecipita- with lowest values in November-December (below 0.5 cm). It
tion (240 mm/yr) is characterizedby the dry, cold, winter is noteworthy that the maximum aerosol loading does not
monthsof DecemberthroughFebruary(10 to 15 mm/month) occur during the dry windy springperiod but rather is associ-
with a transitioninto the warmer,windy but still generallydry ated with the wettesttime of year when water vapor may play
springperiod of March-May. Springis followedby a hot, dry a rolein thescattering properties of aerosols(Figure5i) (r2 =
June and then a hot but wetter summer"monsoon"period of 0.32).
July and August and early September(40 to 45 mm/month). On average, Sevilleta is a very low concentrationaerosol
Summer precipitation generally occurs as intense thunder- environment.The percent frequencyof occurrencehistogram
storms often account for half of the annual total. Fall is char- showsa peakfrequencyof 39% (•'aS00: 0.07), whichdeclines
acterizedby moderatetemperatureswith dryingfrom October rapidly (<2% for opticaldepths>0.2) (Figure 5f). The Ang-
to November[Moore,1996].E1Nifio andLa Nifia eventsstrongly strom parameterhistogramsuggestsa variety of aerosoltypes
influencethe nonmonsoon precipitation[Dahmet al., 1994]. (Figure 5g). Approximately5% of the daily averageda values
The sparsevegetation, consistingof annual grassesand have valueslessthan 0.5, indicatingthat fine particlesdomi-
shrubspecies,is typicalof high-altitudeintermountaindeserts. nate the observedscatteringeffects.The most probable a is
Vegetation cover is generallydictated by availablemoisture ---1.3,which is a typical value assumedfor midlatitude rural
and respondsrapidlyto seasonalrainfall events.Thus dry con- conditions. Less than 10% of the observations exceed 2.0 that
ditionsduring high springwindsprobablycontributeto local would likely be causedfrom dominanceof accumulationmode
12,078 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY

Jan Feb Mar" Ap•" May Jun Jul Aug Sep Oct Nov Dec

c) o

o.•o

0.05
0 O0

Jan Feb Mar' Apr' May Jun Jut Aug $ep Oct Nov Dec.

d) o•,•
n•1407•
:2.5
•.................
i •: ::i
o 0.3
o

0.2 0.4 0..0 0J 0,2 0.3 0.4

aerosol optical depth(500 rim) aerosol optical depth(500 nm)

g) 20
n-!4o7.4

i: ..............
...........
.......... o•1O

•5
•.•'• ,:.•: • '•

o
Jan Jan Jan jan .Jan 0.! O,$ O.O t•3 1.7 2. I o•.o 0.5 i.o •.5 zo 2..5 ,.o 3•$
1994 1995 1996 1997 1998 19• Angstrom parameter water vapor content(cm)

Figure 5. Sevilleta,New Mexico. Mean monthlyvaluesof aerosolopticaldepth at the wavelength500 nm

(a) andAngstromparameter(b) for the wholeperiodof measurements (the barsindicateplusor minusone
standarddeviation),mean monthly(c) and mean dailyvalues(d) of aerosolopticaldepthat 500 nm, mean
dailyvaluesof precipitablewater (e), frequencyof occurrenceof aerosolopticaldepth (f) and Angstrom
parameter(g), scattergramof Angstromparameterversusaerosoloptical depth (h), and scattergramof
aerosoloptical depth versusprecipitablewater (i).

aerosolemissionsgeneratedby regionalwildfiresin the nearby estryresearchreserve,is locatedin Oregon'scentral Cascade

mountains. Mountains ---250km east of the PacificOcean.The landscape
is highlydissected
by streamsand smallriversin narrowvalleys
1.5. H.J. Andrews, Oregon with sharp ridgelines.Regionally, the elevation rangesfrom
The H.J. AndrewsExperimentalForest, a U.S. Forest Ser- 450 m to 1600 m. The Cimel site is located on a ridgeline
vice/Long-Term Ecological Research Reserve (LTER) for- (latitude44ø15'N,longitude122ø9'W)at 827 m andis typically
 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY 12,079

Table 4. DatabaseSummaryfor MeasuringPeriod for Sevilleta,New Mexico (Lat 34ø21'NLong 106ø53'W,Elevation

1447m), 1994-1999
ra500 O' O/ O'a PW o- N Month
Jan. 0.03 0.02 1.26 0.54 0.58 0.18 87 4
Feb. 0.04 0.03 1.35 0.54 0.65 0.25 82 4
March 0.06 0.04 1.07 0.36 0.65 0.23 106 4
April 0.08 0.04 0.91 0.35 0.61 0.23 66 4
May 0.09 0.04 0.97 0.30 0.91 0.35 118 6
June 0.10 0.05 1.24 0.28 1.31 0.52 150 6
July 0.11 0.05 1.38 0.28 2.19 0.72 153 6
Aug. 0.12 0.05 1.67 0.32 2.43 0.46 168 6
Sept. 0.10 0.07 1.57 0.37 1.85 0.55 151 6
Oct. 0.07 0.03 1.40 0.45 0.96 0.41 139 6
Nov. 0.05 0.03 1.49 0.52 0.69 0.23 98 5
Dec. 0.05 0.04 1.46 0.37 0.59 0.20 89 5
Year 0.07 0.03 1.31 0.23 1.12 0.67 1407 62

not influencedby local mountain-valleyinversions.The dom- tural burningof grassfieldsin the Willamette valley raisedthe
inant natural vegetationis Douglasfir in a patch work of old instantaneous%500to nearly 1 on severaloccasionsin August
growth,clear-cuts,and regrowth.Outsidethe reserve,little old and September 1999. Daily ra500for all years (Figure 6d)
growthremains,the landscapebeingdominatedby regrowthin showed a small variation from 0.03 to -0.2 because of lack of
variousstagesof development.Regional precipitation is vari- strongregionalaerosolsources.Episodicsmokeeventsdue to
able, being dependenton Pacificstormsmodified by orogra- regionalforest fires raised the %s00above0.2 on only 23 days
phy. Typically, this rangesfrom 2000 to 3000 mm annually in 4 years(Figure 6d).
within the watershed, and the Cimel site averages2290 mm The aerosoloptical depth frequencydistributionclearly il-
[BierlmaierandMcKee,1989].Precipitationfallsprimarilyfrom lustratesthat nearly75% of the %•00observationsare below0.1
Octoberto May with a 3-4 month dry seasonthe remainderof and half of thosebelow0.05 (Figure 6f). The frequencydistri-
the year. Sourcesof aerosolsare expected from local and bution of the Angstromparameter a showsa broad range 0.3
regionalwild and prescribedfiresduringlate summerand fall. to 2.2, indicatinga wide range in particle size; however, the
Biomassburning from agriculturalgrassfields in the fall and dominantrangeis 1.3 to 1.8,whichincludesvaluesexpectedfor
industrial/urbanaerosolsare transportedfrom the Willamette rural backgroundconditions(Figure 6g). The data were par-
valley 100 km upwind and possiblyfrom California'scentral titioned into two meteorologicaltime periods,June,July, and
valley. A long-rangetransport from Asian dust sourceshas August, the driest period, and September,October, and No-
been observed.On a geologicaltimescale,volcanismin the vember, the onset of the wet season. The wet season has a
Cascaderange is likely. No published aerosol studieshave greaterfrequencyof lower opticaldepthsthan the dry season,
been conductedin this region. which would be expectedfrom an increasedflow of clean
The meteorologyis dominatedby a strongwesterlyflow off Pacificair massesover the site. Correspondingly,the central
the PacificOcean -250 km to the west.The flow is particularly tendencyof the Angstromparameteris shiftedapproximately
strongfrom December to March when most of the precipita- 0.1 higher for the dry period becauseof local or regional
tion falls. From June to September, blocking high-pressure aerosolproductionfrom biomassburning.Approximately30%
ridgescan develop,preventingtransportof Pacificorigin aero- of the points in the wet seasonhave an average Angstrom
sol. Low-level thermal lows beginningin California's central parameterlessthan 1 and optical depthslessthan 0.1. We are
valley developnorthwardinto Oregon and the PacificNorth- unable to determine whether this is due to cloud contamina-
west increasingthe influenceof local and regionalsourceson tion, marine aerosol,or dust,however,marine aerosolis likely.
aerosol loading during stagnantconditions.Rare midwinter Approximately10% of the dry seasonobservationsfall into
arctichigh-pressuresystemscan affect the region. this category.
The LTER has maintained a seasonal AERONET site since Precipitablewater retrievalsclearly show a systematicde-
1994 from approximatelyJune to October. Extensivecloudi- clinefrom JulyandAugust(2 to 2.5 cm) to minimumsof -0.5
nessand inaccessibilityin the winter precludesmeasurements cm in November and December despitethe onset of the wet
throughoutthe year. H.J. Andrewsis characterizedby aerosol season(Figure6e). Thisapparentdiscrepancy maybe partially
opticalpropertiestypicalof midlatitudemaritime-influenced explainedby the reductionin air temperatureof -20øC from
background locations(Figure6, Table 5). At 500 nm, monthly Augustto December,reducingsignificantlythe capacityof the
averagedaerosoloptical depth is low for all months;however, atmosphereto hold water. Additionally, measurementsare
the lowest values of 0.04 occur with the onset of the rainy onlymade duringfair weatherconditions,thusno information
seasonand cleanPacificair. Midsummer meansapproach0.10 is availableduring the more frequentprecipitationeventsof
(Figure6a). Episodesof biomass burningemissions are evident the wet season.The aerosolopticaldepth and PW are nearly
duringthe dry season,and indicationsof smallerparticlesare uncorrelated (r2- 0.04), (Figure6i).
shown by slightly higher Angstrom parameters during this
time. The 6-month mean ra5OOis 0.06. 1.6. Cape Verde
Elevatedra5OO
at H.J. Andrewsis highlyvariablefrom year to Sal Island,CapeVerde (16ø45'N,22ø57'W),is located-600
year (Figure 6c), which is dependenton the intensityand km westof Dakar, Senegal,in the outflowarea of Saharandust
durationof the July throughSeptemberdry season.Agricul- from West Africa. The nearest town of -6000 residents is
12,080 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY

a) o.2o

o.oo

Jan Feb Mar Apr May Jun Jul Aug Sip Oct Nov Dec

b) 2.5 c) 0.25

2.0

o•2o
•o•!o

o.o

Jan Fib MaF Apr May Jt•t Jut Aug •p Oct Nov Dec

d) 0.7 2.5

'•'0.6 0 2.0 ....

:.:•..-...?...•..•
...............
•.....•.•.?.:....:.•
....................
•..............................

0.2
0
0. i

0.0 0.0
0.0 0.2 0.4 0.6 o.o o.• o.2 o.• o.• o.5 o.•
aerosol optical depth(500 rim) •rosol opticaldepth(500 nm)

0. I 0.5 0.9 1.3 1.7 2,1 2.5 2.9 0.0 O.S 1.0 1.5 2.0 2.5 3.0

Angstromparameter• water vapor content(cm)

Figure 6. H.J. Andrews,Oregon, as in Figure 5.

about 3 km away from the measurementsite. The dominant observationsare shown in Figures 7a and 7b. The aerosol
easterlywind directionis well known to be influencedby east-
optical depth for this site is high throughoutthe year with
erly waves, Saharan storms, and associateddust outbreaks elevatedvaluesin summer(from May to September)and sec-
[Carlsonand Prospero,1972;Schutz,1980].A 4 year measure- ondarypeaksin winter (January-February). April andOctober
ment record (1994-1999) has been collectedas part of the correspondto the lowest aerosol contents.Monthly means
PHOTON network,with a total of 726 dailyaveragesanalyzed rangebetween0.26 (April) and 0.68 (June) (Figure 7a, Table
(Figure 7). 6). Mean monthlyvaluesof %•00showsignificantinterannual
Monthly averagesof Ta500and a for the whole period of variabilityof aerosolopticaldepth.The Angstromparametera
 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY 12,081

Table 5. DatabaseSummaryfor MeasuringPeriod for H.J. Andrews,Oregon (Lat 44ø14'N Long 122ø13'W,Elevation
830 m), 1994-1999
'ra5OO tr ot tr. PW rr N Month

Feb. ß.......................
March ........................
April ........................
May 0.15 0.09 1.52 0.26 1.12 0.34 10 2
June 0.08 0.05 1.53 0.26 1.41 0.43 61 4
July 0.08 0.06 1.67 0.28 1.61 0.39 129 5
Aug. 0.10 0.09 1.61 0.22 1.65 0.50 112 5
Sept. 0.07 0.06 1.43 0.39 1.32 0.33 101 5
Oct. 0.07 0.07 1.18 0.51 1.07 0.35 72 5
Nov. 0.04 0.03 1.18 0.52 0.89 0.35 19 3
Dec. 0.04 0.02 0.74 0.52 0.82 0.32 10 1
Y ear514 30

is typicallybelow 0.5 (Figure 7b). The high aerosolloadingin
summerwith correspondinglow a indicatesthat dust domi-
natesthe aerosolregimeassociatedwith frequent Saharandust
outbreaks.This annual cycle has been observed by several
authors(for example,the early studyby Jaenickeand Schiitz
[1978]). It is associatedwith dust transportedover long dis-
tancesat an altitude typicallybetween 2 and 5 km. Satellite
data, like TOMS [Chiapelloet al., 1999a] also indicate a sea-
sonalpattern in aerosolloadingwith a maximumin summer-
time. This time evolutionis quite consistentover the 4 yearsof
measurements (Figure 7c). It is notedthat dustconcentrations
at groundlevel havea minimumduringthe sameperiod [Chia-
pello et al., 1995].
The situationis more complexin wintertime. The relative
high aerosol content can still be associatedwith dust trans-
ported at a lower altitude and comingfrom other sourcesas
observedby Chiapelloet al. [1997]. The contributionof the
marine seasalt when the opticalthicknessis low (around 0.2)
may also represent up to 30% of the total optical thickness
[Chiapelloet al., 1999b].
Daily averagevalues of r a5OO for all years showvery large
day-to-dayvariation (Figure 7d). A histogramof tabOO is wide
with a modalvalue of about0.20 (Figure7f).
Unfortunately, there is no obviousmethod to distinguish
sea-saltaerosolsfrom dustin our data sinceboth aerosoltypes
are associated with the low Angstromparameters(Figures7g
and 7h). On the other hand, there is obviouslyin late winter-
early spring, a contributionfrom a different aerosol type as
reported by the higher values of the Angstrom parameter;
monthly averagesin this seasonare around 0.8-0.9 (i.e.,
March-May 1995)with someindividualcaseslargerthan 1.0
(Figure 7g). Chemical analysisfrom samplesperformed at
groundlevel and air masstrajectoryanalysis[Chiapelloet al.,
1999b]suggestthat there is possiblepollution by sulfatescom-
ing from urban and industrialregionsin North Africa. How-
ever, it is difficult to concludeabout the origin and type of
aerosolsduringthis period. For the entire data set the proba-
bility distributionof a is narrow with a modal value of --•0.1-
0.3 (Figure7g).
As expected,the precipitablewater is a maximumduringthe
summermonths(Figure 7e), whichcorrespondsto the north-
ern-most position of the Intertropical Convergence Zone
(ITCZ) (June-September)and a minimum in the dry season
(January-April).There is no correlationbetweenprecipitable
water and aerosolopticaldepth (Figure 7i).
To conclude,the aerosolcontentoverCapeVerde is highall
year. Very high valuesin summercan clearlybe attributed to
desertdustbut the contributionsof other aerosolsourcesmay
be significantduring the other seasons.
1.7. Banizoumbou, Niger
Banizoumbou,Niger (13ø45'N, 02ø39'E), is located in the
Sahelregion,betweenthe Saharadesertto the north and the
Sudanian zone to the south. The aerosol climate is influenced
by the harmattan,an easterlyor northeasterlywind laden with
dusttransportedfrom the Sahara.Prospero[1981],DMlmeida
[1987], and Pye [1987] identify severalaerosolsourceregions
and associatedtransports that contribute to the harmattan.
Banizoumbouis mainly influencedby sourceslocatedin Niger,
south Algeria, Libya, and Chad. N'Tchayi et al. [1997] have
shown that in addition to the sources located in the Sahara the
semiarid Sahelian region is also a major sourceof dust. The
dustloadingin the atmospheredependson the meteorological
and surfaceconditionsin the sourceregion. High winds and
strongconvectiveprocessesare needed for lofting particlesin
the atmospherefor long-rangetransport.The area can alsobe
affected by the presence of biomassburning aerosols.The
savannavegetationis characteristicof the Sudanianzone, and
fire activitiesare important in the zone during December to
February. Biomassburning aerosolshave a size distribution
with a significantfraction in the accumulationmode (a few
tenthsof microns),while dustparticlesgenerallypresentlarger
sizesnear the sourcesthat result in smaller Angstrom expo-
nents (Figure 8 and Table 7).
The climate of the area is characterizedby a single and
usuallyshort rainy seasonand dependson the presenceof the
ITCZ. The ITCZ correspondsto the transition zone between
dry air massescoming from the north and moist air coming
from the equatorial regions.As expected,the PW increases
whenthe ITCZ is movingnorthwardand is a maximum(Figure
8e) when the ITZC reachesthe area, rangingfrom 1.0 cm in
Januaryup to 4.0 cm in June throughSeptember.
Seasonaltrends in r,•00 are not apparent from these data,
however,valuesare highall year (largerthan 0.2) with primary
peaksin October,February,andApril (Figures8a and 8c). The
distribution(Figure 8f) has a peak value at 0.2-0.4, and the
yearly averageis approximately0.48 (Table 7). These maxi-
mumsare associated with very smallAngstromexponents(less
than 0.15), whichindicatesthat dustis the main contributorto
the optical thickness(Figures 8b and 8g). There are several
12,082 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY

O8

Jan Feb Mar Apr May Jun Jut Aug Sep Oct Nov Dec

f) 25•w•••••r••-• . ..?..• h)2.0

ß
............................

0 .0
Jan d•n J• J• Jan Jan 0.05 0.45 0•85 !.25 1.65 2.05
1994 1995 1996 1997 1998 1999 •ros01 optical•th (500 m) aerosol optical depth(SO0 am)

............................................
:...........................
:'4
.................................
...................................
........................
:
.

Jan j• Jan Jan Jan Jan O•t 0 0,50 0.• I .•0 •70 o

1994 1995 1996 1997 1998 1999 Angstrom

para•ter •
Figure 7. Cape Verde, Sal Island, as in Figure 5.

secondarypeaksin January,March, June, and July associated
with either smallor large (relative to the averageconditions)
Angstromexponents,lessthan 0.05 in June and -0.3 in Jan-
uary.The highervaluesof the Angstromparameterin January
correspondto the presenceof biomassburning aerosols.It is
more obvious in December when
(0.45). Augustand Septemberare relativelycleanmonthsdue
to the scavenging
of the atmosphereby precipitation.
The prevailingaerosoltype is clearly dust comingfrom the
Saharan/Sahelianzone and transported over the area. The
largestvaluesof %500are associatedwith smallvaluesof alpha
(Figure 8h) which is characteristicof the presenceof dust.
Over our 2 years of measurements,the turbidity is always
significantlyabovebackgroundlevels.From March to Junethe
a reaches its maximum very low Angstromexponentconfirmsthe presenceof dust.In
Decemberand January,there is clearlya secondaerosoltype
that contributesto the turbidity as confirmedby the higher
values of the Angstrom exponent.At that time, aerosolsre-
 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY 12,083

Table 6. DatabaseSummaryfor Cape Verde (Lat 16ø43'NLong 22ø56'W,Elevation60 m), 1994-1999

q'aSO0 O' Ot O'a PW o' N Month
Jan. 0.36 0.30 0.39 0.27 1.96 0.83 55 5
Feb. 0.44 0.30 0.25 0.12 1.93 0.90 77 5
March 0.27 0.35 0.55 0.32 2.09 0.70 75 4
April 0.26 0.12 0.58 0.29 2.21 0.79 59 3
May 0.40 0.32 0.37 0.31 2.39 0.72 76 4
June 0.68 0.29 0.16 0.09 3.10 0.71 46 3
July 0.60 0.20 0.18 0.08 3.53 0.93 46 3
Aug. 0.41 0.20 0.21 0.11 3.90 0.89 44 3
Sept. 0.48 0.38 0.23 0.15 3.70 0.83 47 3
Oct. 0.28 0.23 0.32 0.21 3.33 0.80 68 4
Nov. 0.33 0.24 0.39 0.27 2.39 0.74 78 4
Dec. 0.30 0.20 0.36 0.32 1.81 0.67 55 3
Year 0.40 0.13 0.33 0.14 2.69 0.76 726 44

suitingfrom biomassburning activitiesare then mixed with
dust dependingon the wind direction. In the rainy season,
from late Julyto early October,in additionto dustbackground
conditions,humidityeffectsmay contributeto the atmospheric
turbidity,but there is no correlationbetweenprecipitablewa-
ter and optical thickness(Figure 8i).
1.8. Mongu
Mongu, Zambia (15ø15'S,23ø09'E, 1107 m elevation),is
located in west central Zambia on the eastern edge of the
Zambezi River floodplain.The local regionalvegetativecover
is grassland,seasonalmarsh, and croplandin the floodplain
and principallymiombowoodlandon the higherground.The
annualvariation in the aerosoloptical depth is dominatedby
the practiceof agriculturalbiomassburning(Figure 9a, Table
8), which occursprimarilyduringthe secondhalf of the dry
seasonand includesthe beginningof the wet season(August-
November).Averagerainfall for the 7 monthsdry seasonof
April-October is lessthan 8% of the mean annual total (969
mm). Compared to the biomassburning seasonin South
America (for example,Cuiaba,Figure 10a) which reachesa
maximumfor --•2 months(August-September),the burning
seasonis longer (3-4 months)in the savannaregion of south
centralAfrica near Mongu.Scholeset al. [1996a]estimatedthe
geographicaldistributionof the amount of biomassburned,
utilizingsatelliteestimatesof burnedarea [Justice
et al., 1996]
in 1989southof the equatorand showedmuchhigheramounts
to the north of Zambia with low amounts south of Zambia.
This isprimarilydue to the north-southgradientin rainfall and
thus vegetationproduction.Scholeset al. [1996b] combined
thesebiomass-burnedestimateswith emissionfactors depen-
dent on fuel type to showa strongN-S gradient in trace gas
productionfor southernAfrica southof the equator.The fre-
quencyof occurrenceof high aerosol loading of absorbing
aerosolsfrom biomassburninghasbeen measuredfrom satel-
lite retrievals made in the UV wavelengthsfrom the TOMS
instrument[Herman et al., 1997]. These data show that the
regionwith the mostprevalentheavysmokeis north of Zam-
bia, correspondingto the regionwith higher biomass.
Garstanget al. [1996] analysisof trajectoriesover southern
Africa showedfive aerosoltransportmodesthat are likely to
occurfrequentlywith transportpossiblein all major directions
from and to westernZambia. They alsonoted that subsidence
from anticycloniccirculation is a dominant feature during
much of the biomassburning seasonwith four stablevertical
layersidentified in the troposphere.The two layersmost im-
portant in controllingaerosolvertical and horizontaltransport
occurredat 1.5 km abovethe surface(top of the diurnalmixing
layer), which is broken every 5-7 days,and a very persistent
layer at 3.5 km abovethe surface,which is subsidence induced.
From the monthlymeansof ra5OO (Figure 9c) it is noted that
there is significantinterannualvariability in the length of the
biomassburning season.For example, the averageNovember
tabOO is nearly 3 times greater in 1996 than in 1995. This vari-
ation is due in large part to the timing of when rainfall starts
increasingat the beginningof the wet seasonbut is probably
also due in part to the predominant circulation modes and
trajectoriesin a givenyear. Daily averagevaluesof raSOO for all
four years (Figure 9d) showvery large day-to-dayvariations,
especiallyin 1996 and 1997 peak burningseasonmonths,thus
showingthe influence of variable air trajectories.The daily
averagevaluesof the precipitablewater (Figure 9e) showthat
they are typically low in June-July at 0.5-1.5 cm, while in
November,the PW valuesare muchhigher,rangingfrom 2.5 to
3.5 cm, due to the southward advance of the Intertropical
ConvergenceZone (ITCZ). The dailyvaluesof PW in August
and Septembershowa large-amplituderange, suggestingtra-
jectoryvariationsin this season.The relationshipbetweenPW
and raSOO (Figure 9i) showsthat over the total 6 monthseason,
there is not a strongrelationship,but it is noted that there is a
lack of very high raSOO cases(>0.8) for PW < 1.0 cm. If only
peak burning seasonmonthsof August-October were shown,
there would be a somewhatstrongerrelationshipbetweenraS00
and PW becausemany of the valuesof low raS00 with high PW
(2-4 cm) occurin November1995 and December1996when
PW is high and rains have already commencedwashingout
someaerosoland suppressing more burning.Thus trajectories
from the north in August-September,which have higher PW
amounts,sometimesadvectair with higher smokeaerosolcon-
centrationsfrom theseregionswith higherbiomassand higher
emissions[Eck et al., 2001]. The frequencyof occurrencehis-
togramof raSOO (Figure 9f) showsa skeweddistributionwith a
peak at 0.1 to 0.2 and a steadilydecreasingfrequencyat higher
optical depths,which results from the smoke from biomass
burning.
The relationship between Angstrom wavelength exponent
(a) and raSOO is shownin Figure 9h. One of the main features
of this plot is high valuesof a at high raS00which is character-
istic of small-particlesmoke aerosolswhich typicallyhave ac-
cumulation modal radius values of 0.13-0.15 •m of the log-
normalvolume sizedistribution[Remeret al., 1998;Reid et al.,
1998]. Comparisonof a versusraSOO measurementsfor smoke
12,084 HOLBEN ET AL.' EMERGING GROUND-BASED AEROSOL CLIMATOLOGY

0•0

Jan Feb Mar Apr May Jun Jul Aug $ep Oct Nov Dec

b)

o?

o•6

• o•

Jan Feb Mar Apr May Jun Jul Aug $ep Oct Nov Dec

0 2O

1,0 L...............
.,*...
.......................................................................................................
:,

2.0 ....

o
1.5

õs

o
Jan Jul •n Jul Jan Jul o.os 0.45 o.os 1.25 •.65 2•o5 2•4s 2.05 0.0 0.5 t 0 i.5 ZO 2.5 3.0
1995 ! 9• 1997 aerosol optical depth(500 nm) aerosoloptical depth(500 rim)

............................
•............
;2..•...•
..............
•:...,..?
..............
...;..
.........
:• •. ..• g • •. •.:...• :.

..Jan Jul Jan Jul Jan Jul -O•t 0.1 0..30•$ 0.7 0•0 i.i 1•3 1.5 0 t 2 3 4 S 6
! 99S 1996 !997 Angstrom
parameter
a watervapor
content
(cm)
Figure 8. Banizoumbou,Niger, as in Figure 5.

from borealforestfiresin Canada[Markhamet al., 1997]shows found to range from 0.84 for aged smoke to 1.42 for fresh
a strikingsimilarityto the Mongu data. For both locationsthe smoke.However, in that West African site it is possiblethat
a value tendsto asymptoteat a value of --•1.8, thus suggesting the relativelylow a valuesmay be influencedby the presence
similar sizesmokeparticlesfrom vastlydifferent environments of Sahelian/Saharan dust as a secondaryaerosoltype.
and thus fuel types.A similar feature is seenfor South_Amer- Someof the lowervaluesof a at low %•00for Mongu maybe
ican biomassburningsmoke(see Cuiaba Figure 10h) with a due to windblownsoil aerosolof much larger size particles
values typicallybetween 1.6 and 2.0 for %s00values over 0.8. either from local soilsor from long-rangetransportfrom dis-
This contrastswith the measurementsof Liousseet al. [1995] tant sourcessuchas dry lake bedslike Etosha Pan and Mak-
for savannasmokeat Lamto, Ivory Coast,where a valueswere gadikgadiPan to the southwestand southof Mongu, respec-
 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY 12,085

Table 7. DatabaseSummaryfor Banizoumbou,Niger (Lat 32ø22'NLong 02ø39'E,Elevation250 m), 1995-1997

z,•oo tr a tr, PW tr N Month

Jan. 0.47 0.50 0.31 0.26 1.15 0.48 25 1

Feb. 0.71 0.50 0.15 0.08 0.96 0.19 41 2
March 0.47 0.26 0.09 0.07 1.64 0.60 17 1
April 0.76 0.33 0.09 0.05 2.22 0.90 21 1
May 0.41 0.20 0.06 0.07 3.34 1.03 21 1
June 0.53 0.33 0.06 0.07 4.19 0.37 32 2
July 0.52 0.24 0.13 0.14 4.30 0.47 42 2
Aug. 0.39 0.18 0.22 0.16 4.68 0.38 25 1
Sept. 0.33 0.12 0.20 0.16 4.38 0.33 23 1
Oct. 0.59 0.46 0.18 0.12 2.87 0.94 57 3
Nov. 0.39 0.36 0.25 0.14 1.24 0.52 55 2
Dec. 0.24 0.14 0.51 0.27 1.24 0.43 57 2
Year 0.48 0.15 0.19 0.13 2.68 1.45 416 19

tively.The multiyearmeansof averagemonthlya valuesshow
verylittle seasonalvariation(Figure9b) from Juneto October,
howeverthe value in Decemberis -0.7 lower, perhapsdue to
episodesof dust transport, or possiblydue in part to cloud
contamination.It is noted that the June monthly mean a is
equalto the Septembermean (Figure 9b), suggesting that the
aerosolin June may be dominatedby smokefrom smallcook-
ing firesprior to the peak biomassburningseason(landscape
fires) of late July-October.The frequencyof occurrencehis-
togram of daily averagea (Figure 9g) at this site showsa
relativelynarrow peak at 1.6-2.0, again due to the dominance
of smoke aerosols for the months of June-November.
1.9. Cuiaba, Brazil
Cuiaba, Brazil, is located in central South America, imme-
diately to the southof the Amazon Basin, (15ø33'S,56ø4'W,
250 m elevation)in a regionthat is cerrado(savanna)vegeta-
tion, which has been largely convertedto agriculturalland.
Annual burning of these cerrado and agriculturallands is a
commonpracticewhich occursprimarily at the end of the dry
seasonin August-Septemberbut whichmay continueinto Oc-
tober-Novemberdependingon the timing of the rainfall. In
additionto biomassburningof cerradovegetationin the region
surroundingCuiaba,there is alsobiomassburningof tropical
forestto the north (-500 km) and burningof grazinggrass-
landsin the Pantanal(the world'slargestseasonallyflooded
wetland) to the south (-100 km). Remeret al. [1998] have
shownfrom trajectoryanalysisthat smokefrom thesedifferent
regionsis advectedover Cuiabaduringthe burningseasonand
that thesediffering trajectoriesmay result in the advectionof
smokeaerosolwith somewhatdifferingsize distributionsand
that total precipitablewater may alsovary. The CIMEL Sun/
sky radiometer site was located -10 km north of the city of
Cuiaba and therefore may also be influencedby some urban/
industrialaerosolproduction.
A detaileddiscussion of the seasonality
of the aerosoloptical
depth,Angstromwavelengthexponent,and precipitablewater
for Cuiaba is givenby Holben et al. [1996] for measurements
made in 1993 and 1994. Here we present an analysisof data
with 1995 added to the previous2 years and emphasizea
discussion of averageseasonalityand interannualvariations.
The 3 year average monthly variation in q'aS00 at Cuiaba
(Figure 10a, Table 9) clearly showsthat the peak in smoke
concentrationsfrom biomassburning aerosolsoccursin Au-
gust and September.However, q'adoes not return to back-
groundlevelsuntil December,thusthere is somelocalburning
and/or advectionof smokefrom other regionsin the monthsof
October-Novemberwhichis the beginningof the rainy season.
During the 5 monthsdry seasonat Cuiaba,May throughSep-
tember, only -9% of the total annual averageprecipitation
(1373 mm) occurs.Decreasesof q'ain October-Novemberwith
the onsetof rains (aerosolwashoutand lessflammablefuels)
vary from year to year during 1993-1995 (Figure 10c). For
example,the October monthly average%500for 1994 is nearly
doublethe value for 1993. The dailyvariabilityof *aS00 during
the burningseasonat Cuiabacan be quite large (Figure 10d).
For example,duringAugust-September1995the rangeis from
-0.3 to 2.4, as a result of air trajectoriescomingfrom burning
regionson somedaysand on other daysthe trajectoriesmaybe
from a directionwith very few fires. The averageYa500 for the
nonburningseasonmonths of June, December, and January
are low and consistent,suggestinga relativelystableand low
backgroundq'aS00 of--0.15. The frequencyof occurrenceshis-
togramof dailyaverageq'aS00 (Figure 10f) is skewedwith peak
frequency occurringbelow 0.25 representingconditionsap-
proachingmonthlyaveragebackgroundconditionsand a large
numberof highervaluesdue primarilyto biomassburningwith
a peak value of-2.3.
The aerosolparticle size distributionhas a significantsea-
sonal change at Cuiaba, as is indicated by strong seasonal
variation in the Angstromwavelengthexponent(a) (Figure
10b).Angstromwavelengthexponentvaluesaverage-0.6-0.7
for nonburning seasonmonths, while for the peak burning
seasonmonths of August-September,the averageis 1.7-1.8.
As shownin Figure 9h, there is a strongrelationshipbetween
Ta500and a at Cuiaba, with the majority of a valuesfor Ta5OO >
0.6 falling in the range 1.5-2.0. This is somewhatsimilarto the
a valuesfor biomassburningsmokein Mongu,Zambia (Figure
9h), thus implying similar aerosol size distributionsfor the
smokein thesetwo vastlydifferentregions.Remeret al. [1998]
has found the accumulationmode particle size of the lognor-
mal volume size distribution for biomassburning smoke in
Cuiaba to be typicallyabout 0.13/am modal radiusbut with a
rangeof 0.12-0.17/am. Kotchenruther and Hobbs[1996]found
the humidificationfactor of South American biomassburning
aerosolsto be rather small (-1.05-1.35), suggesting little in-
fluenceof relative humidity on aerosolsize and optical prop-
erties.This mayexplainin part whythe a valuesof smokefrom
biomassburning are so similar in suchwidely differing envi-
ronmentsas African savanna(see Mongu, Figure 9h), boreal
forestfiresin Canada[Markhamet al., 1997],and SouthAmer-
ican tropical forest and savanna[Holbenet al., 1996].
12,086 HOLBEN ET AL.: EMERGING GROUND-BASEDAEROSOLCLIMATOLOGY

jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

2.0

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

d) 2.0•.••• •• . 02O

$ $ :i S '

õ •o ß :i •

o 0.5

o
Jan •1 J• Jul Jan Jul Jan Jul o. o5 0•45 0.85 1.25 ! .65 0.5 1.0 1.5 ZO
1995 1996 1997 1998 aerosoloptical deptla(500 nm) •ero•ol •fic•l dep• <•

•"•:"•'•
........:.•........• .........
• .......................
•........
o

o
0.! 0.5 0.9 1.3 1.7 2.1 2.5 2.0 I 2 3 4 5
Jan Jul J• •I Jan Jul J• Jul
1995 ! 996 1997 1998 Angstrom parameterec water vapor content(cm)

Figure 9. Mongu,Zambia, as in Figure5.

The a valuesin Cuiaba at lower TaS00 (<0.4) showa wide withpeakfrequencyat 1.6-1.8resulting
frombiomass burning
rangefrom ---0.1to 2.2,thussuggestingdifferingaerosoltypes aerosolsdominatedby accumulationmodeparticles.A signif-
on differentdays.However,in contrastto Mongu,whichshows icant numberof occurrences from 0.8 to 1.4 are representative
monthlyaveragea valuesthatareequalforburningseason and of bimodal size distributionswith varyingrelative concentra-
nonburningseasonalike, there are more casesof lower a tions of accumulation mode versus coarse mode aerosols.
values in Cuiaba in nonburningseasonmonths,resultingin The seasonal variationof precipitablewaterat Cuiaba(Fig-
lower monthlyaveragea valuesindicativeof a predominant ure 10e) ranges from 1-3.5 cm in June-Julyat the middleof
influenceof larger aerosolparticles.The frequencyof occur- the dry seasonto 4.0-5.5 cm in Januaryin the middleto wet
renceshistogram of a (Figure10g)showsa skeweddistribution season.Significant day-to-day variabilityin PW occurs,associ-
 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY 12,087

Table 8. DatabaseSummaryfor Mongu, Zambia (Lat 15ø15'SLong 23ø09'E,Elevation 1107m), 1995-1998

zas00 tr a tr• PW tr N Month

Feb. ß.......................
March ........................
April ........................
May ........................
June 0.17 0.07 1.86 0.14 1.31 0.27 28 3
July 0.23 0.13 1.78 0.23 1.13 0.34 81 4
Aug. 0.44 0.21 1.79 0.18 1.16 0.34 110 4
Sept. 0.60 0.31 1.80 0.14 1.58 0.57 90 4
Oct. 0.49 0.18 1.79 0.22 2.03 0.66 93 4
Nov. 0.33 0.21 1.60 0.34 2.80 0.68 60 4
Dec. 0.13 0.06 1.10 0.71 3.19 0.55 32 2
Year .................. 494 25

ated with air masstrajectoriesfrom different sourceregions. forestfireswith heavyaerosolloading(Figure 11d). In contrast
However,there is very little correlationbetweenra5OOand PW to the years 1994 and 1995, which had numerous fires and
for the entire combinedwet plus dry seasondata set (Figure associatedbiomassburning aerosols,the relativelymoist years
10i). In contrast,however,Remer et al. [1998] found some of 1997-1999(especially1997) had relativelylittle burningand
correlationbetween PW and tabOO for Cuiaba when primarily thereforemuchlower aerosolopticaldepth (Figure 11d). This
burningseasondata are analyzed.This occurssincePW actsas time seriesis in relatively good agreementwith satellite esti-
a tracer of air massorigin and sincethe highestsmoke aerosol matesof the area coveredby heavysmoke,as estimatedby the
concentrationsoriginate in the forest burning regionsto the TOMS sensor[Hsuet al., 1999],for all of Canadafrom 1979 to
north, which also have the highestPW concentrations. 1998. For Canada, TOMS estimatesshow 1997 as having the
1.10. Thompson, Manitoba, Canada least heavysmokecoverageof anyyear, while 1994was a year
of moderateto high coveragewithin the contextof the 19 years
Thompson,Manitoba (55ø47'N,97ø50'W,218 m elevation)
is located near the northern ecotone of the boreal forest zone
of satellitemonitoring.However,TOMS data showthat for all
of Canada, 1998 was a year of extensiveburning while our
of central Canada.The local land coveris dominatedby forest
measurementsin Thompsonshowrelativelylittle smokecom-
of three species:black spruce,jack pine, and aspenwith nu-
pared to 1994 and 1995, thus suggestingthat other regionsof
merouslakes and pondspresent.The climateis characteristic
Canada experiencedweather conditionsmore conduciveto
of a high-latitudemidcontinentalgeographiclocation, with
long severewintersand heavysnowpack.The warm springand fires that year.
summer seasonsexperiencea high degree of variability in Although the seasonaland interannualvariation of aerosol
precipitationamount,which resultsin very large interannual optical depth is dominatedby biomassburning aerosolsfrom
variations in forest fire frequency and total area burned forest fires, there are additional sources of aerosol which are
[Stocks,1991].The seasonalvariationand interannualvariabil- presentalso.There is a large nickel smeltingoperationlocated
ity in aerosol optical depth is dominated by the amount of in the town of Thompson,and the forest itself is a sourceof
biomassburningaerosolsproducedfrom forest fires (Figures biogenicallyproducedaerosols.In addition to the production
11a-11c,Table 10). Data for June 1994 and mostof July 1994 of pollen in the spring, coniferousforests also produce bio-
were measuredat a site ---35 km northwestof Thompsonbe- genic hydrocarbonswhich lead to the formation of atmo-
causethe Thompsonmonitoringsitewasnot establisheduntil sphericparticulates[Kavouraset al., 1998]. Comparisonof op-
late July 1994. It is noted that 1994 was the driest year on tical depth data from winter to early springmonthsfor this site
record for this region, resultingin numerousand widespread in 1996[Markhamet al., 1997]beforebiomassburningtypically

Table 9. DatabaseSummaryfor MeasuringPeriod for Cuiaba,Brazil (Lat 15ø30'SLong 56ø00'W,Elevation250 m), 1993-
1995

%s00 tr a tr= PW tr N Month

Jan. 0.17 0.09 0.62 0.33 4.59 0.32 10 1
Feb. ß.......................
March ........................
April ........................
May ........................
June 0.15 0.05 0.74 0.28 2.47 0.72 13 1
July 0.22 0.12 1.24 0.26 2.50 0.40 38 2
Aug. 0.81 0.54 1.67 0.26 2.52 0.71 73 3
Sept. 1.17 0.40 1.72 0.15 3.46 0.73 71 3
Oct. 0.45 0.20 1.22 0.48 4.32 0.56 34 2
Nov. 0.36 0.13 0.99 0.62 4.26 0.82 25 3
Dec. 0.17 0.10 0.63 0.33 ...... 14 1
Year .................. 273 16
12,088 HOLBEN
ETAL.'EMERGING
GROUND-BASED
AEROSOL
CLIMATOLOGY

a) •
1.6

Jan Feb Mat- Apr May Jun Jul Aug $©p Oct Nov Dec

Jan Feb Mar" Apr M•ay Jun Jul Aug Sep Oct Nov Dec

! i i :i :i :i' ! n-278

2.0•..................................
:i.,•..,--..
...................
i :•'
..................................................
ø
'•. • • ..

. ...........................
.................
i...................................
i............
;
•I.0
1•

<: O.$•-•'..-*:.:.,...:.:*..,i
........................................
•.........................................
i :......................................

0
0.0•*•
'•'•••• '••::
Jan Jul Jan Jul Jan Jul oøo5 0.0 O.$ l.O l.S 2.0 2.5
! 993 1994 ! 995 aerosolopticaldepth(500 nm) •ro•l •ttcal •th (• nm)

g) 30 i) 2.s

5 •........................................
."•;:'
'".'•2
..............................
•.............
";'";
•t 25 2.0 'i i*

•'2 ••-..............................
'•- ......................... •i• •............. •o
i......................................
!-.-•
..........................
i.................................
• .............................
•:.
.........................
•.'• •......... •...,. ..............
............
•<...,..:•....
.........::.................................

0.9 1.3 I •7 2J o i 2 3 4 5 6
Jan Jul Jan Jul Jan Jul
Angstrom parameter water vapor content(cm)
1993 1994 1995

Figure10. Cuiaba,Brazil,asin Figure5.

begins show slightly

highervaluesinspring, whichmaybedue nm).In Figure 11d,weseethatthereisalargedailyvariability
inparttobiogenically produced aerosols. aerosolsin %s00
In addition, forallyears
except1997and1999, whichisduemainly
maybeadvected intotheregion fromdistant source regionsto varying
airmass trajectories
fromforestfireregions (and
having different
airmass characteristics.
Forexample, Smirnovprecipitation
variability)
butalsoto a muchlesser degree, to
etal. [1996] haveshown thattropicalairmasses influencethe long-rangeaerosoltransport
fromothersources. Totalcol-
siteofWynward, Saskatchewan (51ø46'N, 104ø12'W)about 5% umn-integrated PW (Figurelie) alsoshowsignificant daily
of the time andthat theseair masses withhigher variability,
are associated withhighvaluesin midsummer suggestiveof air
aerosol opticaldepthsobserved at thatsite(upto 0.36at 500 massadvectionfromthe southand/orlarge-scale
regionalcon-
 HOLBEN
ETAL.'EMERGING
GROUND-BASED
AEROSOL
CLIMATOLOGY 12,089
a)

•ø•

• o• III$ o

Jan Feb •ar Apr May •n Jul A• Sip Oct

• 2 0 •' : -• •

1994Jan
!995Jan
1996Jan
1997 Jan
t998 Jan
1999 0t 05
aerosol optical depth (500 nm) aerosol optical depth (SO0 rim)
g) 2S
• n•459 •

I• 20

-•• • • " :• --.•:: ..:'•....••o o

2!i••'i¾*
•":•'
:•.l •s

Jan Jen Jan Jan Jan Jan ol o.s o.g Is [7 21 2.5 2.o
1994 1995 1996 1997 1998 1999 Angstromparameter (•

Figure11. Thompson,
Manitoba,
Canada,
asin Figure5.

vergenceincreasing theprecipitable
water.Therelationshipof 0.5therearetypically only1 to 3 days,if any,for each0.1
betweendailyaverage %500andPW(Figure11i)shows thatfor r, s00
intervalbin.Thisis a resultof thehighspatialandtem-
days where precipitable
waterexceeds 2.5cmtherearevery poralvariability offorestfiresmoke, especially
athighaerosol
fewcases of veryhighra500.Thedays withhighprecipitable optical depths.
waterarelikelytohavehada southerlyflowcomponent, there- The relationship betweendailyaverage Angstrom wave-
foresuggestingthatmostof thelargeforestfiresdonotoccur length exponent (a) and%•00 (Figure 11h)shows twoprincipal
to thesouthof Thompson. Thefrequency histogram for the features:awiderange ofa atmoderate tolowaerosol optical
dailyaverager, s00
values
(Figure11f)shows thata largenum- depths (<0.4)andarelatively narrow rangeofa athighoptical
berofoccurrences areobservedbelow 0.2andthatabove depths(>0.5). The widerangeof a associated with the low
12,090 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY

Table 10. DatabaseSummaryfor MeasuringPeriod for Thompson,Canada(Lat 55ø47'NLong 97ø50'W,Elevation218 m),

1994-1999

Ta500 O' a O'c• PW tr N Month

Feb. ß.......................
March ........................
April ........................
May 0.17 0.12 1.41 0.47 1.10 0.47 27 3
June 0.28 0.45 1.54 0.44 1.76 0.60 90 5
July 0.27 0.35 1.86 0.41 2.06 0.56 118 6
Aug. 0.21 0.25 1.73 0.38 2.02 0.60 116 6
Sept. 0.10 0.10 1.61 0.43 1.47 0.48 73 6
Oct. 0.06 0.03 1.39 0.62 0.95 0.33 35 3
Nov. ß .......................
Dec. ß .......................
Year .................. 459 29

optical depth casesmay be due to a wide variety of aerosol
typeswith different associatedsize distributions.The high a
values may possiblybe associatedwith pollution from the
nickel smeltingoperationin town [Markhamet al., 1997] and
low a valuesperhapsassociated with pollen dispersalfrom the
forestsand/or combinationsof other biogenicaerosolsor dust
from unpavedroads during dry time periods.In addition, at
low Ta500 (<0.10), the uncertaintyin Ta measurementof 0.01-
0.02 can result in large errors in the computedvalues of a
[O'Neill et al., 2000]. The a valuesof 1.4-2.1, which occurat
highopticaldepths(,aS00> 0.5), are typicalof biomassburn-
ing values(see the Mongu, Zambia, and Cuiaba,Brazil, sec-
tions). These a values result from accumulation-mode-
dominatedsmokeaerosolswith typicallognormalvolume size
distribution radius values of 0.13 to 0.17 t•m for the accumu-
lation mode. The range in a for biomassburning aerosolsis
dependentin part on the type of fuels burned, the type of
combustion(flamingor smoldering),and the agingprocesses
of the aerosolas it is transported[Reidet al., 1999]. Monthly
averagevaluesof a shownin Figure 11b, reach a maximumin
Julyand Augustdue in part to the presenceof biomassburning
aerosolsin these monthsbut also perhapsdue to lesspollen
and other largeparticleaerosolsproducedin midsummerthan
in May-June when biomassburningaerosolsare alsopresent.
The frequencyof occurrences histogramof a (Figure1lh) also
showsthe wide rangeof valuesobservedat this sitewith a peak
occurringat the 1.6-1.8 a range bin.
2. Conclusion
Monthly statisticsfor aerosolopticaldepth (,as00),precipi-
tablewater and Angstromexponenthavebeencomputedfrom
AERONET direct Sun observationsof 2 or more yearsat sites
representingbiomassburning aerosols,backgroundaerosols,
desert aerosols,and aerosolsgeneratedin urban landscapes.
The resultsclearlyshowseasonaldynamicsin aerosolloading,
type, and precipitablewater.
Backgroundlevelsof aerosols,which we define as *a5øø less
than 0.10, were observedat almost all sitesbut varying fre-
showedonly weak increasesin aerosol loading above back-
groundduringsummermonths.Similarly,the seasonalsitesin
the boreal forest and tropical woodlands largely exhibited
backgroundlevelsprior to dry seasonbiomassburning,pre-
dictablyincreasingmonthlyaveraged*a5øø to 1.2 in the tropics
but with very large interannualvariabilityin the boreal forests.
The greaterinterannualvariabilityin the boreal forestsitewas
causedby the influenceof wildfires,while the tropical sites
were influencedby anthropogenicfires.
The West African sites influenced by desert dust, Cape
Verde, and Banizoumbouwere abovebackgroundlevelsfor all
months.A bimodalaerosolopticaldepth distribution(winter
and summerpeaks) was observedat Cape Verde as it lies
downwindof the extensivedust sourceregions.Dust particles
are the dominant aerosol type at the island site despite its
oceaniclocation.The Ta500 monthlyaveragesrangebetween0.2
and 0.7 year-round,lower than the high drought observations
of the mid 1980s.Banizoumbois influencedby elevatedlevels
of dustand biomassburningduringthe winter seasonand dust
for much of the rest of the year as indicated in the dramatic
changesof the Angstromexponent.Again, thesevalues are
lower than the peak values of the droughtyears of the mid
1980s.
Biomassburningconstitutessignificantincreasesin monthly
loadingand order of magnitudedaily spikesfor tropicalsites.
Becauseof their anthropogenicnature associatedwith land
coverchange,the spikesare a predictablesignalin the aerosol
loading during August through September in both South
America and southernAfrica. The Angstromparametertypi-
cally rangesfrom 1.7 to 2.1, indicatingvery small particles
generatedby the combustionprocesses. Correspondingly, wild-
fire biomassburningin the Northern Hemispheremidlatitudes
typicallyoccursin the summermonthspeakingin Augustand
September;however,the extent is highlyvariable depending
on localmeteorologicaland fuel-loadingconditions.Daily ,aS00
at these sites can range from backgroundconditionsto well
over 1.
All northern midlatitude continental sites showed seasonal

quencies.Mauna Loa Observatory,locatedin the mid-Pacific
Ocean abovethe marine boundarylayer, exhibitedthe lowest
values <0.02, only slightlyperturbed duringthe springAsian
dust seasonplus transportof Asian pollution and infrequent
emissionsfrom localvolcanism.Backgroundlevelsmay alsobe
observedat GSFC mainly duringwinter months.Sevilletaand
H.J. Andrews, representingdry and wet midlatitude regions,
cyclesof warm seasonpeaks and cool seasonlow values of
*aS00.Particularlydramaticwas Goddard SpaceFlight Center
with a July Ta500averageof -0.5. This waswell correlatedwith
precipitablewater, suggestingthat humidificationcontributes
significantlyto the aerosolburden,while polar continentalair
masstrajectoriesbring air with low aerosoloptical depth and
low precipitablewater. Likewise, the midlatitude circulation
 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY 12,091

slowsduring summermonthsallowingaerosolconcentrations anticipate these data will provide part of a synthesisfor a

to build. Additionally more hours of sunlight can enhance
photochemicalsmog. In all our nondust sites, there was a
positivecorrelation between the precipitablewater and the
aerosolopticaldepth.For tropicalbiomassburningsites,thisis
a meteorological-land-use phenomenonlargelyassociated with
the coincidenceof sourceregions.The midlatitude sitesstud-
ied probablyrepresenta significantregionalbiogenicemission
sourcecombinedwith humidificationand swellingof sulfate
and biogenicaerosolsas highertemperaturesincreasethe ca-
pacity of the atmosphereto hold water vapor. U.S. mid-
Atlantic urban aerosolshavebeen shownto be stronglyhygro-
scopic,while biomassburning aerosolsin Brazil are very
weaklyhygroscopic[Kotchenruther et al., 1999].
The resultof theseobservations clearlydocumentsa pulseof
aerosolswith large a at manyof our globallydistributedsitesin
August and September.Although this network providespoint
observations,the diurnal daily and long-term observationswill
yield important details of aerosoloptical properties to aug-
ment the spatial observationsafforded by other aerosolnet-
works and satellite retrievals. Thus in the near future, we
globallybalancedassessment of the impact of aerosolson the
radiationbalanceof the planet.
Appendix A
Approximately100 sitescomprisethe AERONET network
representinga wide diversityof aerosolregimes.We provide
valuesof monthly and annual ra500,Angstromexponent,and
PW for sites with 2 or more years of quality-assureddata
(Table A1). Annual meanswere not computedfor seasonal
sites.An exhaustiveanalysisof these 21 sitesis not intended
but rather a cursory inspectionof these data clearly show
seasonaltrends, annual differencesbetweensites,and mixing
of aerosoltypesthat can be further influencedby water vapor.
The island sites of Lanai, Hawaii, and San Nicholas Island,
California, representclean oceanicconditionswith the mean
annual Ta500--' 0.08 although the Lanai site has a decidedly
lower Angstrom exponentindicativeof marine aerosols.Ber-
muda (ra500= 0.14), Dry Tortugas,Florida (r a500= 0.18),
Kaashidhoo,Maldives (WAS00 = 0.20), and Bahrain (WAS00 =

Table A1. Aerosol Optical Depth at 500 nm (Ta500),AngstromExponent(a), PrecipitableWater (PW), Associated
StandardDeviations(rr), and Number of Days (N) and Months in ObservationPeriods
%500 cr a cr PW o- N Month

Lanai, Hawaii, USA (Lat 20ø49'N Long 156ø59'W,Elevation 80 m), 1995-1999

Jan. 0.08 0.08 0.71 0.41 2.95 0.48 34 2
Feb. 0.08 0.03 0.56 0.40 2.55 0.46 28 2
March 0.12 0.06 0.74 0.36 2.54 0.47 36 2
April 0.11 0.04 0.62 0.27 2.77 0.35 44 3
May 0.11 0.06 0.70 0.22 3.17 0.36 38 2
June 0.06 0.02 0.60 0.23 3.32 0.51 20 2
July 0.06 0.02 0.96 0.25 3.45 0.57 56 2
Aug. 0.06 0.03 0.87 0.31 3.74 0.41 58 3
Sept. 0.06 0.03 0.85 0.35 3.68 0.52 41 3
Oct. 0.06 0.03 0.60 0.32 3.53 0.65 42 2
Nov. 0.06 0.04 0.63 0.33 3.34 0.79 39 3
Dec. 0.08 0.06 0.69 0.46 2.89 0.59 49 3
Year 0.08 0.02 0.71 0.12 3.16 0.42 485 29

San NicolasIsland, California, U&4 (Lat 33ø15'N Long 119ø29'W,Elevation 133 m), 1998-2000
Jan. 0.04 0.03 1.06 0.62 1.03 0.56 35 2
Feb. 0.06 0.03 1.01 0.44 1.02 0.39 29 2
March 0.11 0.06 0.74 0.26 1.22 0.39 31 2
April 0.13 0.09 0.75 0.30 1.30 0.31 41 2
May 0.12 0.06 0.91 0.16 1.67 0.47 26 1
June 0.09 0.04 1.06 0.40 1.72 0.49 36 2
July 0.09 0.04 1.37 0.38 2.34 1.03 33 2
Aug. 0.08 0.04 1.49 0.53 1.94 0.94 47 2
Sept. 0.11 0.06 1.27 0.41 2.10 0.95 37 2
Oct. 0.08 0.04 1.42 0.45 1.27 0.43 40 2
Nov. 0.04 0.02 1.33 0.68 1.21 0.55 34 2
Dec. 0.04 0.02 1.20 0.66 0.88 0.29 56 2
Year 0.08 0.03 1.13 0.25 1.48 0.47 445 23

Dry Tortugas,Florida, USA (Lat 24ø36'NLong 82ø47'W,Elevation 10 m), 1996-1999

Jan. 0.15 0.05 1.29 0.45 3.37 0.88 9 2
Feb. 0.16 0.08 1.13 0.51 2.82 1.06 25 1
March 0.14 0.09 1.29 0.53 2.81 1.02 33 1
April 0.26 0.16 1.54 0.36 2.90 0.94 33 1
May 0.39 0.19 1.64 0.28 3.25 1.07 25 1
June 0.18 0.11 1.03 0.57 4.87 0.52 73 3
July 0.21 0.14 0.73 0.53 4.87 0.74 105 4
Aug. 0.17 0.12 0.97 0.65 5.18 0.78 89 4
Sept. 0.20 0.15 1.07 0.67 4.88 1.10 76 4
Oct. 0.11 0.05 0.97 0.58 4.35 1.18 47 2
Nov. 0.12 0.04 0.99 0.43 3.41 1.00 42 2
Dec. 0.11 0.04 0.75 0.51 3.30 0.98 31 2
Year 0.18 0.08 1.12 0.28 3.83 0.92 588 27
12,092 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY

Table A1. (continued)

Ta500 tr a tr PW tr N Month

CART Site, Oklahoma,USA (Lat 36ø36'NLong 9T24'W, Elevation315 m), 1994-1999

Jan. 0.08 0.06 1.39 0.23 1.00 0.38 8 1
Feb. 0.08 0.06 1.18 0.39 0.87 0.26 18 1
March 0.20 0.10 1.36 0.22 1.15 0.47 10 1
April 0.16 0.10 1.03 0.46 1.43 0.64 59 4
May 0.36 0.22 1.27 0.23 2.54 0.72 27 1
June 0.19 0.08 1.32 0.23 3.55 0.92 19 1
July 0.20 0.06 1.32 0.34 4.61 0.70 15 1
Aug. 0.28 0.14 1.57 0.32 3.82 0.67 27 1
Sept. 0.22 0.16 1.42 0.39 3.34 1.06 50 3
Oct. 0.14 0.11 1.58 0.28 2.38 0.84 23 2
Nov. 0.06 0.06 1.22 0.27 1.19 0.41 7 1
Dec. 0.07 0.06 1.71 0.22 1.08 0.75 12 1
Year 0.17 0.05 1.36 0.19 2.25 1.31 275 18

Bondville,Illinois, USA (Lat 40ø03'NLong 88ø22'W,Elevation212 m), 1996-1999

Jan. 0.06 0.03 1.35 0.42 0.37 0.04 3 1
Feb. 0.08 0.06 1.26 0.22 0.84 0.63 13 1
March 0.14 0.06 1.36 0.25 0.88 0.64 31 2
April 0.19 0.13 1.37 0.37 1.41 0.54 32 2
May 0.31 0.28 1.24 0.33 2.19 0.75 43 2
June 0.31 0.23 1.41 0.45 3.10 0.96 45 3
July 0.29 0.18 1.59 0.35 3.48 0.99 49 3
Aug. 0.33 0.25 1.62 0.32 2.92 0.77 78 4
Sept. 0.30 0.31 1.42 0.47 2.39 0.99 89 4
Oct. 0.16 0.10 1.48 0.48 1.54 0.72 79 4
Nov. 0.09 0.08 1.42 0.39 0.91 0.47 19 2
Dec. 0.06 0.02 1.52 0.22 0.58 0.27 17 1
Year 0.19 0.11 1.42 0.12 1.72 1.07 498 29

Sherbrooke,Quebec,Canada (Lat 45ø22'NLong 71ø55'W,Elevation300 m), 1995, 1998-2000

Jan. 0.07 0.03 1.55 0.73 0.38 0.45 9 2
Feb. 0.07 0.04 1.63 0.95 0.29 0.14 14 2
March 0.11 0.04 1.27 0.34 0.58 0.14 10 2
April 0.09 0.06 1.46 0.45 0.76 0.26 20 1
May 0.13 0.06 1.20 0.35 1.70 0.70 18 1
June 0.23 0.17 1.48 0.39 2.07 0.85 39 2
July 0.22 0.15 1.66 0.38 2.59 0.77 44 2
Aug. 0.15 0.16 1.57 0.40 2.15 0.66 60 3
Sept. 0.10 0.10 1.26 0.54 1.56 0.68 26 3
Oct. 0.09 0.05 1.66 0.46 1.11 0.44 34 3
Nov. 0.06 0.03 1.97 0.63 0.91 0.56 13 3
Dec. 0.06 0.03 2.03 0.43 0.42 0.23 6 1
Year 0.12 0.06 1.56 0.26 1.21 0.78 293 25

Waskesiu,Saskatchewan,
Canada (Lat 53ø55'NLong 106ø04'W,Elevation550 m), 1994-1999

Feb. -..
March --.
April .-.
May 0.19 0.11 1.40 0.35 1.19 0.48 49 4
June 0.23 0.27 1.41 0.39 1.61 0.51 93 4
July 0.20 0.24 1.50 0.43 2.03 0.53 133 5
Aug. 0.18 0.20 1.55 0.41 1.95 0.48 144 6
Sept. 0.07 0.05 1.35 0.48 1.40 0.40 112 5
Oct. 0.06 0.05 1.24 0.50 0.95 0.36 68 5
Nov. 0.04 0.03 1.37 0.54 0.75 0.34 16 1
Dec. 0.03 0.02 1.43 0.43 0.44 0.13 6 1
Year -.. ............... 621 31

Alta Floresta,Brazil (Lat 09ø55'SLong 56ø00'W,Elevation175 m), 1993-1995, 1999

Jan. 0.09 0.03 0.66 0.24 5.41 0.12 4 1
Feb. 0.08 0.04 0.57 0.27 5.22 0.29 13 1
March 0.11 0.12 0.74 0.51 5.10 0.41 8 1
April 0.09 0.04 0.98 0.46 4.82 0.55 17 1
May 0.10 0.06 1.21 0.53 4.61 0.52 23 1
June 0.11 0.04 1.30 0.62 3.69 0.69 29 2
July 0.16 0.12 1.50 0.46 2.94 0.44 55 2
Aug. 1.18 0.65 1.89 0.24 2.84 0.55 81 4
Sept. 1.48 0.73 1.66 0.28 4.19 0.65 80 4
Oct. 0.63 0.24 1.50 0.26 4.76 0.43 31 2
Nov. ßß ß
Dec. ...
Year ... ............... 341 19
 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY 12,093

Table A1. (continued)

'ra5OO cr ot cr PW cr N Month

Brasilia,Brazil (Lat 15ø55'SLong 47ø54'1,V,,

Elevation1100 m), 1993-1995

Feb. 0.11 0.06 1.04 0.23 2.51 0.25 4 1

March ........................
April 0.07 0.03 0.77 0.14 2.99 0.29 7 1
May 0.10 0.03 1.12 0.18 2.60 0.16 12 1
June 0.08 0.04 0.95 0.32 1.74 0.51 63 3
July 0.09 0.07 0.96 0.27 1.70 0.27 71 3
Aug. 0.23 0.23 1.15 0.35 1.56 0.41 76 3
Sept. 0.56 0.36 1.44 0.29 2.10 0.69 52 2
Oct. 0.35 0.21 1.32 0.42 2.60 0.46 27 2
Nov. 0.16 0.08 0.83 0.45 2.89 0.55 19 2
Dec. 0.14 0.04 0.64 0.34 3.21 0.41 6 1
Year .................. 337 19

Los Fierros,Boliva (Lat 14ø33'SLong 60ø55'W,Elevation 170 m), 1996, 1998-1999

Feb. ß.......................
March ........................
April ........................
May 0.08 0.02 1.16 0.25 4.22 1.11 26 2
June 0.14 0.07 1.52 0.21 4.25 0.54 33 2
July 0.19 0.09 1.69 0.17 3.23 0.51 75 3
Aug. 1.07 0.80 1.88 0.16 3.58 0.90 90 3
Sept. 1.16 0.68 1.86 0.11 3.86 0.65 43 2
Oct. 1.02 0.48 1.75 0.23 4.56 0.74 27 2
Nov. 0.24 0.07 1.03 0.35 5.06 0.44 16 1
Dec. ß.......................
Year .................. 310 15

Atica, Chile (Lat 18ø28'S Long 70ø15'IV,,Elevation25 m), 1998-1999

Jan. 0.27 0.14 1.42 0.16 3.28 0.82 30 1
Feb. 0.45 0.20 1.29 0.22 5.04 0.35 24 1
March 0.45 0.17 1.32 0.31 4.69 0.37 31 1
April 0.32 0.10 1.43 0.12 3.01 0.62 24 1
May 0.27 0.09 1.26 0.22 2.15 0.37 46 2
June 0.24 0.07 1.25 0.24 1.97 0.38 49 2
July 0.24 0.07 1.17 0.22 2.13 0.34 52 2
Aug. 0.32 0.13 1.13 0.18 1.88 0.28 47 2
Sept. 0.31 0.09 1.18 0.10 1.84 0.35 29 1
Oct. 0.24 0.07 1.24 0.12 1.98 0.17 26 1
Nov. 0.27 0.08 1.25 0.17 2.50 0.47 30 1
Dec. 0.33 0.14 1.26 0.14 3.07 0.98 28 1
Year 0.31 0.07 1.27 0.09 2.79 1.09 416 16

Bermuda(Lat 32ø22'NLong 64ø41'W,Elevation10 m), 1996-1999

Jan. 0.10 0.06 0.87 0.45 2.13 0.50 16 2
Feb. 0.14 0.06 0.80 0.35 1.84 0.28 20 2
March 0.13 0.05 0.98 0.31 1.89 0.46 44 3
April 0.18 0.07 0.98 0.31 2.02 0.52 59 3
May 0.24 0.16 1.09 0.33 3.05 0.62 59 3
June 0.19 0.12 1.10 0.46 3.56 0.75 48 3
July 0.17 0.12 0.90 0.42 4.45 0.74 63 3
Aug. 0.11 0.06 0.86 0.39 4.91 0.54 75 3
Sept. 0.14 0.08 0.93 0.50 4.20 0.90 52 3
Oct. 0.11 0.06 0.78 0.44 2.93 0.72 55 4
Nov. 0.10 0.04 0.84 0.45 2.43 0.54 63 4
Dec. 0.13 0.08 0.92 0.48 2.35 0.63 28 3
Year 0.14 0.04 0.92 0.10 2.98 1.07 582 36

Ilorin, Nigeria (Lat 08ø19'NLong 04ø20'W,Elevation350 m), 1998-1999

Jan. 0.90 0.34 1.02 0.18 2.85 0.64 30 1
Feb. 0.76 0.19 0.71 0.25 3.14 0.95 24 1
March 0.48 0.19 0.54 0.26 4.24 0.35 23 1
April 0.70 0.41 0.34 0.17 4.51 0.67 32 2
May 0.43 0.22 0.38 0.25 4.95 0.58 50 2
June 0.45 0.19 0.50 0.28 4.89 0.39 40 2
July 0.27 0.16 0.92 0.43 5.06 0.28 19 2
Aug. 0.27 0.05 1.32 0.25 4.78 0.28 8 2
Sept. 0.34 0.19 0.75 0.46 5.02 0.27 11 1
Oct. 0.33 0.27 0.78 0.31 4.53 0.71 24 1
Nov. 0.37 0.12 0.80 0.25 3.35 0.30 28 1
Dec. 0.80 0.41 0.90 0.38 2.60 0.81 25 1
Year 0.51 0.22 0.75 0.28 4.16 0.91 314 17
12,094 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY

Table A1. (continued)

q'a500 O' Ol O' PW o' N Month

Bondoukoui,BurkinaFaso (Lat 11ø50'NLong 03ø45'W,Elevation0 m), 1996-1997

Jan. 0.32 0.20 0.77 0.33 1.47 0.46 24 1
Feb. 0.71 0.60 0.22 0.11 1.05 0.44 46 2
March 0.80 0.55 0.26 0.16 1.95 0.94 38 2
April 0.56 0.33 0.18 0.11 3.02 1.04 29 2
May 0.48 0.52 0.15 0.15 3.73 0.89 49 2
June 0.38 0.21 0.17 0.17 4.17 0.36 45 2
July 0.39 0.16 0.24 0.18 4.32 0.41 32 2
Aug. 0.31 0.19 0.48 0.40 4.68 0.36 27 1
Sept. 0.30 0.16 0.43 0.27 4.53 0.28 19 1
Oct. 0.39 0.25 0.27 0.20 3.81 0.55 28 1
Nov. 0.30 0.17 0.78 0.36 1.95 0.75 39 2
Dec. 0.36 0.34 0.78 0.35 1.24 0.48 59 2
Year 0.44 0.17 0.39 0.25 2.99 1.38 435 20

Bidi-Bahn,BurkinaFaso (Lat 14ø03'NLong 02ø27'W,Elevation0 m), 1996-1997

Jan. 0.43 0.49 0.65 0.42 2.01 0.64 25 1
Feb. 0.93 0.70 0.18 0.07 1.43 0.24 31 2
March 0.60 0.49 0.25 0.13 2.20 0.98 51 2
April 0.41 0.29 0.31 0.27 2.52 1.11 50 2
May 0.91 0.67 0.09 0.08 4.29 1.07 28 1
June 0.63 0.36 0.10 0.09 4.86 0.50 46 2
July 0.59 0.26 0.23 0.19 5.18 0.50 53 2
Aug. 0.55 0.27 0.31 0.24 5.58 0.43 36 2
Sept. 0.52 0.29 0.37 0.15 5.34 0.43 26 2
Oct. 0.52 0.32 0.24 0.08 3.50 1.01 28 1
Nov. 0.34 0.29 0.31 0.14 1.61 0.32 28 1
Dec. 0.20 0.14 0.72 0.30 1.59 0.61 28 1
Year 0.55 0.21 0.31 0.19 3.34 1.63 430 19

SedeBoker,Israel (Lat 30ø31'NLong 34ø28'E,Elevation480 m), 1996, 1998-1999

Jan. 0.19 0.17 1.20 0.62 1.06 0.35 29 2
Feb. 0.21 0.28 1.02 0.69 0.83 0.25 29 3
March 0.20 0.13 0.95 0.63 1.03 0.23 48 3
April 0.27 0.13 0.74 0.44 0.96 0.17 26 2
May 0.25 0.17 0.42 0.23 1.21 0.33 27 2
June 0.17 0.06 0.87 0.37 1.49 0.40 56 3
July 0.23 0.09 1.06 0.32 1.85 0.48 64 3
Aug. 0.23 0.08 1.12 0.31 2.02 0.48 85 3
Sept. 0.22 0.08 0.91 0.40 2.13 0.45 71 3
Oct. 0.22 0.07 1.02 0.36 2.03 0.44 44 2
Nov. 0.19 0.10 0.90 0.39 1.45 0.43 42 3
Dec. 0.11 0.05 1.04 0.37 1.09 0.37 26 2
Year 0.21 0.04 0.94 0.21 1.43 0.47 547 31

Ispra, Italy (Lat 45ø48'NLong 08ø37'E,Elevation235 m), 1997-1998

Jan. 0.25 0.20 1.53 0.34 0.87 0.19 12 1
Feb. 0.34 0.21 1.55 0.15 1.18 0.28 21 1
March 0.39 0.28 1.49 0.31 1.27 0.31 23 1
April 0.39 0.38 1.50 0.36 1.71 0.48 14 1
May 0.46 0.23 1.62 0.15 2.17 0.35 23 1
June 0.52 0.38 1.44 0.24 3.02 0.87 27 1
July 0.37 0.30 1.61 0.13 3.25 0.70 42 2
Aug. 0.40 0.20 1.57 0.16 3.50 0.78 50 2
Sept. 0.44 0.37 1.46 0.23 2.61 0.84 37 2
Oct. 0.35 0.27 1.39 0.30 2.10 0.70 32 2
Nov. 0.22 0.15 1.46 0.28 1.61 0.54 14 1
Dec. 0.21 0.10 1.56 0.24 1.43 0.25 6 1
Year 0.36 0.10 1.52 0.07 2.06 0.87 301 16

Bahrain (Lat 26ø19'NLong 50ø30'W,Elevation0 m), 1998-1999

Jan. 0.23 0.09 1.21 0.21 1.84 0.64 22 1
Feb. 0.29 0.11 1.00 0.27 2.29 0.70 14 1
March 0.27 0.14 0.78 0.54 1.53 0.39 26 1
April 0.33 0.12 0.68 0.21 1.88 0.55 30 1
May 0.34 0.16 0.52 0.18 1.78 0.53 31 1
June 0.32 0.09 0.68 0.18 1.67 0.24 21 1
July 0.48 0.19 0.55 0.28 2.76 0.87 30 2
Aug. 0.45 0.15 1.10 0.23 3.61 0.89 31 1
Sept. 0.38 0.11 1.03 0.22 3.12 0.74 28 1
Oct. 0.25 0.09 1.25 0.29 2.29 0.65 31 1
Nov. 0.24 0.06 1.34 0.25 1.95 0.30 28 1
Dec. 0.21 0.07 1.26 0.19 1.79 0.68 29 1
Year 0.32 0.09 0.95 0.30 2.21 0.64 321 13
 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY 12,095

Table A1. (continued)

Za500 iT Of iT PW tr N Month

Kaashidhoo,Maldives (Lat 04ø57'N Long 73ø27'E,Elevation10 m), 1998-1999

J an. 0.27 0.10 1.09 0.32 4.46 0.81 13 1
Feb. 0.28 0.14 1.24 0.17 4.72 0.81 27 2
March 0.30 0.17 1.24 0.21 4.56 0.61 48 2
April 0.26 0.19 1.01 0.31 5.42 0.66 49 2
May 0.19 0.09 0.62 0.30 5.88 0.56 26 2
June 0.13 0.05 0.43 0.16 5.27 0.73 25 2
July 0.20 0.08 0.30 0.12 4.81 0.65 14 2
Aug. 0.15 0.04 0.52 0.34 4.89 0.52 32 2
Sept. 0.11 0.05 0.62 0.41 5.39 0.38 38 2
Oct. 0.13 0.07 0.75 0.38 5.09 0.49 34 2
Nov. 0.17 0.10 1.07 0.38 4.73 0.76 46 2
Dec. 0.17 0.10 0.97 0.35 5.00 0.88 31 2
Year 0.20 0.07 0.82 0.32 5.02 0.41 383 23

Dalanzadgad,Mongolia (Lat 43ø34'N Long 104ø25'E,Elevation1470 m), 1997-2000

Jan. 0.07 0.04 1.34 0.64 0.22 0.09 29 2
Feb. 0.10 0.05 0.94 0.57 0.25 0.08 31 2
March 0.21 0.25 0.58 0.35 0.27 0.10 58 3
April 0.22 0.23 0.68 0.37 0.50 0.20 56 3
May 0.25 0.19 0.78 0.42 0.76 0.32 52 3
June 0.14 0.09 0.61 0.35 1.02 0.30 34 3
July 0.14 0.12 1.32 0.62 2.08 0.58 31 3
Aug. 0.15 0.15 1.13 0.78 1.44 0.42 36 3
Sept. 0.11 0.09 1.30 0.57 1.07 0.45 46 2
Oct. 0.06 0.04 1.62 0.67 0.62 0.20 38 2
Nov. 0.06 0.06 1.55 0.51 0.43 0.16 32 2
Dec. 0.05 0.03 1.82 0.68 0.23 0.07 16 2
Year 0.13 0.07 1.14 0.42 0.74 0.58 459 30

0.32) clearly showelevatedlevelsthat are seasonallydepen- SteveRyan (MLO), Mukufute M. Mukelabai (Mongu), JoseChagas
dent due to influences of a variety of continental aerosol (Cuiaba), Alfredo Pereira (Brazil network),Bento de Silva (Brasilia),
O. Manga, J. L. Rajot, E. F. SantosSoares,A. Werem (PHOTON
sources. The small desert coastal town of Arica, Chile, has a
sites), Ray Johnson(Thompson), S. Khudulmur and Enkhtuvshin
high mean annual aerosol loading (r,•o•, = 0.31), yet the (Dalanzadgad),Judd Welton and David Bates (Dry Tortugas), Mark
Angstromexponent(1.27) is not typicalof the expectedaero- Yarborough (Lanai), Miguel Rivas (Arica), Jeff Chilton, Naval Air
sol type generatedby either desertlandscapesor marine aero- Warfare Center,Point Mugu (SanNicholas),T. O. Aro (Ilorin), Mike
sols. Desert aerosol-influencedsites such as Ilorin, Nigeria Willis (Bahrain), Paula Palachek (Waskesiu),Mike Snyder (Bond-
ville), Alain Royer and Amadou Bokoye (CARTEL, Universit6 de
(r,•o,, = 0.51), Bondoukoui,Burkina Faso (%500= 0.44), Sherbrooke,Sherbroke,Quebec, Canada) and FrancescoCappellani
Bidi-Bahn, Burkina Faso (r,•oo = 0.55), have mean annual (Ispra). Our thanks to our funding sourcesthat had the foresightto
Angstrom exponentsless than 0.8; however, desert-located supportthis internationalcollaborativeactivity;theseincludeMichael
sitessuchas SedeBoker, Israel (r, s00= 0.21), and Dalanza- King, EOS Senior Project Scientist,Diane Wickland and Robert Cur-
ran of NASA Headquarters for support of AERONET, M. Alain
dgad, Mongolia (r,•00 = 0.13), have Angstrom exponents Podaire (CNES) and Daniel Vidal-Madjar (CNRS) for support of
indicativeof seasonallyvariableparticle sizes(0.94 and 1.14, PHOTON, Division of Environmental Biology (DEB) of NSF for
respectively).Additionally, the rural North American sites supportof the LTER sites,and Philippe Teillet and Gunar Fedosejevs
such as CART Site, Oklahoma (r,•oo : 0.17), Bondville, of the CanadaCentre for Remote Sensingfor supportingAEROCAN
(the CanadianAERONET componentcoordinatedby the CARTEL
Illinois (%,00 = 0.19), and Sherbrooke,Quebec (TaS00= group of the Universit6 de Sherbrooke),the Department of Energy
0.12), demonstratea warm-season peak and a narrowrangeof Atmospheric Radiation Measurement program for support of the
a mean annual Angstrom exponentfrom 1.36, 1.42, and 1.56, Southern Great Plains site and NASA/EOS MODIS contract NAS5-
respectively.Ispra, Italy (0.36), is influencedby a rather con- 31363 for support for the Dry Tortugas site. Finally, our thanks to
stant aerosol sourceyear-round as shownby high r, sooeach Ellsworth Dutton and Russ Schnell of NOAA/CMDL for their long-
term supportfor our calibration effort at Mauna Loa, Hawaii.
month and a smallrangein Angstromexponentfor all months
(meana = 1.52). Lastly,the seasonal valuesfrom the forested
References
boreal site, Waskesiu, Saskatchewan(r,,00 = 0.23, a =
1.41), and tropicalforestedsites,Alta Floresta(?,s,,o= 1.48, Andreae, M. O., Raisingdustin the greenhouse,Nature,380, 389-390,
1996.
• = 1.66), Brazil, and Brasilia, Brazil (r, soo= 0.56, • =
Barnes,J. E., and D. J. Hofmann, Lidar measurementsof stratospheric
1.44), clearly respondto the warm-dry-season influenceof aerosolover Mauna Loa Observatory,Geophys.Res.Lett., 24, 1923-
biomassburning, as describedin detail earlier. 1926, 1997.
Ben Mohamed, A., J.P. Frangi, J. Fontan, and A. Druilhet, Spatialand
temporalvariationsof atmosphericturbidity and related parameters
Acknowledgments. This paper representsconsiderableeffort over in Niger, J. Appl. Meteorol.,31, 1286-1294, 1992.
many yearsby numerouspeoplewho have collectivelydevelopedthe Bierlmaier, F. A., and A. McKee, Climatic summaries and documen-
AERONET programand allowedanalysisof the data presentedhere. tation for the primary meteorologicalstation, H.J. Andrews Exper-
Our thanks to the individual site managerswho have maintained an imental Forest, 1972to 1984. Gen. Tech.Rep.PNW-242, 56 pp., U.S.
activeinterest in the project and resultedin long-term observations, Dep. of Agric.. For. Serv.,PacificNorthwestRes. Station,Portland,
includingFred Bierlmaier (H.J. Andrews),Doug Moore (Sevilleta), Oregon, 1989.
12,096 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY

Bigelow,D. S., J. R. Slusser,A. F. Beaubien,and J. H. Gibson,The parisonof trendsin the troposphericand stratospheric aerosolop-
USDA ultravioletradiation monitoringprogram,Bull. Am. Meteo- tical depthsin the Antarctic,J. Geophys.Res., 98, 18,441-18,447,
rol. Soc., 79, 601-615, 1998. 1993.
Bruegge,C. T., J. E. Conel, R. O. Green, J. S. Margolis,R. G. Holm, Herman, J. R., P. K. Bhartia, O. Torres, C. Hsu, C. Seftor, and E.
and G. Toon, Water vapor column abundanceretrievals during Celarier, Global distributionof UV-absorbingaerosolsfrom Nim-
FIFE, J. Geophys.Res.,97, 18,759-18,768, 1992. bus 7/TOMS data,J. Geophys.Res.,102, 16,911-16,922,1997.
Bryson,R. A., Air massesstreamlines,and the Boreal forest, Geogr. Holben, B. N., T. F. Eck, and R. S. Fraser, Temporal and spatial
Bull., 8(3), 228-269, 1966. variabilityof aerosolopticaldepthin the Sahelregionin relationto
Carlson, T. N., and J. M. Prospero,The large-scalemovementof vegetationremotesensing,Int. J. RemoteSens.,12, 1147-1163,1991.
Saharanair outbreaksover the northernequatorialAtlantic,J.Appl. Holben, B. N., A. Setzer, T. F. Eck, A. Pereira, and I. Slutsker, Effect
Meteorol., 11,283-297, 1972. of dry seasonbiomassburningon Amazon Basinaerosolconcentra-
Chiapello, I., G. Bergametti, L. Gomes, B. Chatenet, F. Dulac, J. tions and opticalproperties,J. Geophys.Res.,101, 19,465-19,481,
Pimenta,and E. SantosSoares,An additionallow layertransportof 1996.
Sahelianand Saharandustover the northeasternTropical Atlantic, Holben, B. N., et al., AERONET--A federated instrument network
Geophys.Res.Lett., 22, 3191-3194, 1995. and data archive for aerosol characterization, Remote Sens.Environ.,
Chiapello,I., G. Bergametti,B. Chatenet,P. Bousquet,F. Dulac, and 66(1), 1-16, 1998.
E. Santos Soares, Origins of African dust transportedover the Hsu, N. C., J. R. Herman, J. F. Gleason, O. Torres, and C. J. Seftor,
northeasterntropicalAtlantic,J. Geophys. Res.,102, 13,701-13,709, Satellite detection of smoke aerosolsover a snow/icesurface by
1997.
TOMS, Geophys.Res.Lett., 26, 1165-1168, 1999.
Chiapello,I., J. M. Prospero,J. Herman,andC. Hsu,Nimbus-7/TOMS Jaenicke,R., and L. Schtitz, Comprehensivestudy of physicaland
detection of mineral aerosols over the North Atlantic Ocean and
chemicalpropertiesof the surfaceaerosolin the CapeVerde Islands
Africa, J. Geophys.Res., 104, 9277-9292, 1999a. region,J. Geophys.Res.,83, 3585-3599, 1978.
Chiapello,I., G. Bergametti,B. Chatenet,F. Dulac, C. Moulin, A. Jinhuan,A., and Y. Liquan, Variation characteristics of atmospheric
Vermeulen, C. Devaux, I. Jankowiak, and E. Santos Soares, Contri- aerosolopticaldepthsand visibilityin North China during 1980-
bution of the different aerosolspeciesto the aerosoloptical depth 1994, Atmos. Environ., 34, 603-609, 2000.
over the northeasterntropicalAtlantic,J. Geophys. Res.,104, 4025- Justice,C. O., J. D. Kendall, P. R. Dowtry, and R. J. Scholes,Satellite
4033, 1999b. remotesensingof fires duringthe SAFARI campaignusingNOAA
Dahm, C. N., and D. I. Moore, The E1-Nifio/Southern oscillation advancedveryhighresolutionradiometerdata,J. Geophys. Res.,101,
phenomenonandthe Sevilletalong-termecological researchsite,in 23,851-23,863, 1996.
L TER Report,edited by D. Greenland,LTER Clim. Comm.,LTER Kaufman, Y. J., and R. S. Fraser, Light extinctionby aerosolsduring
Publ. 18, pp. 12-20, 1994. summerair pollution,J. Clim.Appl. Meteorol.,22, 1694-1706, 1983.
D'Almeida, G. A., On the variabilityof desertaerosolradiativechar- Kaufman,Y. J., D. Tanre, H. R. Gordon, T. Nakajima,J. Lenoble,R.
acteristics,J. Geophys.Res., 92, 3017-3026, 1987. Frouin, H. Grassl,B. M. Herman, M.D. King, and P.M. Teillet,
D'Almeida, G. A., A model for Saharandusttransport,J. Clim.Appl. Passiveremote sensingof troposphericaerosol and atmospheric
Meteorol., 25, 903-916, 1996.
correctionfor the aerosol effect, J. Geophys.Res., 102, 16,815-
Dutton, E.G., Aerosolopticaldepthmeasurements from four NOAA/ 16,830, 1997.
CMDL monitoringsites,in Trends'93:A Compendiumof Data on
Kavouras,I. G., N. Mihalopoulos,and E. Stephanou,Formation of
Global Change,editedby T. A. Boden,D. P. Kaiser,R. J. Sepanski,
atmosphericparticlesfrom organicacidsproducedby forests,Na-
andF. W. Stoss,pp. 484-494, CarbonDioxideInf. Anal. Cent.,Oak ture, 395, 683-686, 1998.
Ridge Natl. Lab., Oak Ridge, Tenn., 1994.
King, M.D., D. M. Byrne,J. A. Reagan,and B. M. Herman, Spectral
Dutton, E.G., J. J. DeLuisi, and A. P. Austring, Interpretation of
variationof opticaldepthat Tucson,ArizonabetweenAugust1975
Mauna Loa atmospherictransmissionrelative to aerosols,using
and December 1977,J. Appl. Meteorol.,19, 723-732, 1980.
photometricprecipitablewateramounts, J.Atmos.Chem.,3, 53-68,
1985. King, M.D., Y. J. Kaufman, D. Tanre, and T. Nakajima, Remote
Dutton, E.G., P. Reddy, S. Ryan, and J. J. DeLuisi, Features and sensingof troposphericaerosolsfrom space:Past presentand fu-
ture, Bull. Am. Meteorol. Soc., 80, 2229-2259, 1999.
effects of aerosoloptical depth observedat Mauna Loa, Hawaii:
Kotchenruther, R. A., and P. V. Hobbs, Humidification factors of
1982-1992,J. Geophys.Res.,99, 8295-8306, 1994.
Eck, T. F., B. N. Holben, J. S. Reid, O. Dubovik, A. Smirnov, N. T. aerosolsfrom biomassburning in Brazil, in Smoke, Clouds,and
O'Neill, I. Slutsker,and S. Kinne, The wavelengthdependenceof Radiation-Brazil,edited by V. W. J. H. Kirchhoff, pp. 117-120,
the opticaldepth of biomassburning,urban and desertdustaero- Trans Tech, Clausthal-Zellerfeld,Germany, 1996.
sols,J. Geophys.Res., 104, 31,333-31,350,1999. Kotchenruther,R. A., P. V. Hobbs, and D. A. Hegg, Humidification
Eck, T. F., B. N. Holben, D. E. Ward, O. Dubovik, J. S. Reid, A. factorsfor atmosphericaerosolsoff the mid-Atlanticcoastof the
Smirnov,M. M. Mukalabai,N. C. Hsu, N. T. O'Neill, andi. Slutsker, United States,J. Geophys.Res., 104, 2239-2251, 1999.
Characterizationof the opticalpropertiesof biomassburningaero- Liousse,C., C. Devaux, F. Dulac, and H. Cachier,Aging of savanna
solsin Zambia during the 1997 ZIBBEE experiment,J. Geophys. biomassburningaerosols:Consequences on their opticalproperties,
Res., 106, 3425-3448, 2001. J. Atmos. Chem., 22, 1-17, 1995.
Flowers, E. C., R. A. McCormick, and K. R. Kurfis, Atmospheric Luria, M., J. F. Boatman,J. Harris, J. Ray, T. Straube,J. Chin, R. L.
turbidity over the United States,1961-1966,J. Appl. Meteorol.,8, Gunter, G. Herbert, T. M. Gerlach, and C. C. Van Valin, Atmo-
955-962, 1969. sphericsulfurdioxideat Mauna Loa, Hawaii, J. Geophys.
Res.,97,
Forgan,B. W., J. J. DeLuisi, B. B. Hicks, and E. N. Rusina,Report on 6011-6022, 1992.
the measurementsof atmosphericturbidity in BAPMoN, WMO Markham, B. L., J. S. Schafer, B. N. Holben, and R. N. Halthore,
Rep., 94, 77 pp., World Meteorol. Organ., Geneva, Switzerland, Atmosphericaerosol and water vapor characteristics over north
1993. central Canada during BOREAS, J. Geophys.Res., 102, 29,737-
Garstang,M., P. D. Tyson,R. Swap,M. Edwards,P. Kallberg,andJ. A. 29,745, 1997.
Lindesay,Horizontal and vertical transportof air over southern Michalsky,J. J., J. A. Schlemmer,N. R. Larson,L. C. Harrison, W. E.
Africa, J. Geophys.Res.,101, 23,721-23,736,1996. Berkheiser III, and N. S. Laulainen, Measurement of the seasonal
Greenland, D., et al., A climate analysisof Long-Term Ecological and annualvariability of total column aerosolin a northeasternUS
Researchsites,in Climateof Sevilleta,editedby D. I. Moore, chapter network,in Proceedings of the InternationalSpecialtyConferenceon
17, http://lternet'edu/im/climate/climdes/'
2000. Aerosolsand AtmosphericOptics:RadiativeBalance and VisualAir
Gueymard,C., Analysisof monthlyaverageatmosphericprecipitable Quality,vol. A, pp. 247-258, Air and Waste Manage.Assoc.,Pitts-
water and turbidity in Canada and northern United States, Sol. burgh, Penn., 1994.
Energy,53, 57-71, 1994. Mims, F. M., III, An internationalhaze-monitoringnetwork for stu-
Gushchin, G. P., The Methods,Instrumentationand Resultsof Atmo- dents, Bull. Am. Meteorol. Soc., 80, 1421-1431, 1999.
sphericSpectralMeasurements (in Russian),200 pp., Gidrometeoiz- Moore, D., SevilletaNational Wildlife Refuge, in A ClimaticAnalysis
dat, St. Petersburg,Russia, 1988. of Long-TermEcologicalResearchSites,edited by D. Greenland, T.
Herber, A., L. W. Thomason, V. F. Radionov, and U. Leiterer, Com- Kittel, B. P. Hayden, and D. S. Schimel,and LTER Clim. Comm.
 HOLBEN ET AL.: EMERGING GROUND-BASED AEROSOL CLIMATOLOGY 12,097

Members, chap. 17, 1996. (Available as http://www.lternet.edu/ Shaw,G. E., Solarspectralirradianceand atmospherictransmission at
documents/Publications/climdes/) Mauna Loa Observatory,Appl. Opt., 21, 2006-2011, 1982b.
N'Tchayi Mbourou, G., J. J. Bertrand,and S. E. Nicholson,The diur- Shaw,G. E., Sunphotometry,Bull.Am. Meteorol.Soc.,64, 4-10, 1983.
nal and seasonalcyclesof wind-bornedust over Africa North of the Smirnov,A., O. Yershov,and Y. Villevalde, Aerosol optical depth in
Equator,J. Appl. Meteorol.,36, 868-882, 1997. the Atlantic Ocean and Mediterranean Sea, Proc. SPIE, 2582, 203-
O'Neill, N. T., T. F. Eck, B. N. Holben, A. Smirnov, O. Dubovik, and 214, 1995.
A. Royer, Bimodal size distributioninfluenceson the variation of Smirnov,A., N. T. O'Neill, A. Royer, A. Tarussov,and B. McArthur,
Angstromderivativesin spectralandopticaldepthspace,J. Geophys. Aerosol optical depth over Canada and the link with synopticair
Res., in press,2000. masstypes,J. Geophys.Res.,101, 19,299-19,318, 1996.
Perry, K. D., T. A. Cahill, R. C. Schnell,and J. M. Harris, Long-range Smirnov, A., B. N. Holben, T. F. Eck, O. Dubovik, and I. Slutsker,
transportof anthropogenicaerosolsto the National Oceanic and Cloud screeningand qualitycontrolalgorithmsfor the AERONET
AtmosphericAdministrationbaselinestationat Mauna Loa Obser- database,RemoteSens.Environ.,73(3), 337-349, 2000a.
vatory,Hawaii, J. Geophys.Res.,104, 18,521-18,533,1999. Smirnov, A., B. N. Holben, O. Dubovik, N. T. O'Neill, L. A. Remer,
Peterson,J. T., E. C. Flowers,G. J. Berri, C. L. Reynolds,and J. H. T. F. Eck, I. Slutsker,and D. Savoie,Measurementof atmospheric
Rudisill,AtmosphericturbidityovercentralNorth Carolina,J.Appl. opticalparameterson U.S. Atlantic coastsites,ships,and Bermuda
Meteorol., 20, 229-241, 1981. during TARFOX, J. Geophys.Res., 105, 9887-9901, 2000b.
Prospero,J. M., Aeolian transportto the world ocean,in The Sea,vol. Stocks,B. J., The extent and impact of forest fires in northern cir-
7, The OceanicLithosphere,edited by C. Emiliani, pp. 801-874, cumpolar countries,in Global BiomassBurning.'Atmospheric,Cli-
Wiley Intersci., New York, 1981. matic, and BiosphericImplications,edited by J. S. Levine, pp. 197-
Pye,K.,AeolianDustandDustDeposits,334 pp., Academic,SanDiego. 202, MIT Press,Cambridge,Mass., 1991.
Calif., 1987. Trewartha, G. T., An Introductionto Climate,369 pp., McGraw-Hill,
Reid, J. S., P. V. Hobbs, R. J. Ferek, D. R. Blake, J. V. Martins, M. R. New York, 1954.
Dunlap, and C. Liousse,Physical,chemicaland opticalpropertiesof Volz, F. E., Turbidity at Uppsala from 1909 to 1922 from Sjostroms
regionalhazesdominatedby smokein Brazil, J. Geophys.Res.,103, solarradiationmeasurements, Sued.Meteorol.Hydrol.Inst. Rep. 28,
32,059-32,080, 1998. pp. 100-104, Stockholm,Sweden, 1968.
Reid, J. S., T. F. Eck, S. A. Christopher, P. V. Hobbs, and B. N. Volz, F. E., Some resultsof turbidity networks,Tellus,21, 625-630,
1969.
Holben, Use of the Angstromexponentto estimatethe variabilityof
opticaland physicalpropertiesof agingsmokeparticlesin Brazil, J.
N. Abuhassan, T. F. Eck, B. N. Holben, B. Markham, W. W. New-
Geophys.Res.,104, 27,473-27,489,1999.
Remer, L. A., Y. Kaufman,B. N. Holben, A.M. Thompson,and D. P. comb,J. S. Schafer,I. Slutsker,and A. Stairnov,BiosphericSciences
Branch, NASA GSFC, Code 923, Greenbelt, MD 20771.
McNamara, Biomassburning aerosolsize distributionand modeled
opticalproperties,J. Geophys.Res.,103, 31,879-31,892, 1998. (brent@aeronet.gsfc.nasa.gov)
D. Tanr6, Laboratoire d'Optique Atmospherique, CNRS, Ville-
Roosen, R. G., R. J. Angione, and C. H. Clemcke, Worldwide varia-
neuve d'Ascq,France.
tionsin atmospherictransmission,1, Baselineresultsfrom Smithso-
B. Chatenet, Laboratoire Interuniversitaire des Syst•mes Atmo-
nian observations,Bull. Am. Meteorol. Soc., 54, 307-316, 1973.
sph6riques,Universit6sParis VII et Paris XII, CNRS, 94010--Cr6teil
Russell,P. B., et al., Pinatuboand pre-Pinatuboopticaldepthspectra: Cedex, France.
Mauna Loa measurements,comparisons,inferred particle size dis- F. Lavenu, CESBIO/CNRS, Toulouse, France.
tributions,radiative effects,and relationshipto lidar data, J. Geo- Y. J. Kaufman, Climate and Radiation Branch, NASA GSFC,
phys.Res.,98, 22,969-22,985, 1993. Greenbelt, MD 20771.
Ryan, S., The wind field around Mauna Loa derived from surfaceand
J. Vande Castle, University of New Mexico, Albuquerque, NM
balloon observations, J. Geophys.Res.,102, 10,711-10,725,1997. 87106.
Schmid,B., J. Michalsky,R. ttalthore, M. Beauharnois,L. Harrison, J. A. Setzer, Instituto de PesquisasEspaciais,Sao Jose dos Campos,
Livingston,P. Russell,B. Holben, T. Eck, and A. Stairnov,Com- San Paolo, Brazil.
parisonof aerosolopticaldepth from four solar radiometersduring D. Clark, NOAA, Silver Spring,MD 20746.
the fall 1997ARM intensiveobservationperiod, Geophys.Res.Lett., R. Frouin, ScrippsInstitute of Oceanography,La Jolla, CA 92093.
26, 2725-2728, 1999. A. Karneli, Ben Gurion University of the Negev, Sede Boker Cam-
Schmid, B., J. J. Michaslky, D. W. Slater, J. C. Barnard, R. N. pus 84990, Israel.
Halthore, J. C. Liljegren, B. N. Holben, T. F. Eck, J. M. Livingston, N. T. O'Neill, CARTEL, Universite de Sherbrooke, Sherbrooke,
P. B. Russell,T. Ingold, and I. Slutsker,Comparisonof columnar Quebec, Canada.
water vapor measurementsduring the fall 1997 ARM Intensive C. Pietras, SAIC-GSC, Beltsville, MD, and NASA/GSFC, Green-
ObservationPeriod: solar transmittancemethods,Appl. Opt., 40, belt, MD 20771.
1886-1896, 2001. R. T. Pinker, Department of Meteorology,Universityof Maryland,
Scholes,C. O., J. D. Kendall, and C. O. Justice,The quantity of College Park, MD 20742.
biomassburned in southernAfrica, J. Geophys.Res., 101, 23,667- K. Voss, Department of Physics,University of Miami, Miami, FL
23,676, 1996a. 33124.
Scholes,C. O., D. E. Ward, and C. O. Justice,Emissionsof tracegases G. Zibordi, SpaceApplicationsInstitute, Joint ResearchCentre of
and aerosolparticlesdue to vegetationburning in SouthernHemi- the European Commission,Ispra, Italy.
sphereAfrica, J. Geophys.Res.,101, 23,677-23,682, 1996b. T. F. Eck, Goddard Earth Sciencesand TechnologyCenter, Uni-
Schutz,L., Long range transportof desertdustwith specialemphasis versity of Maryland, Baltimore County, Baltimore, MD 21250.
on the Sahara, Ann. N.Y. Acad. Sci., 338, 515-531, 1980. J. S. Schafer,ScienceSystemsand Applications,Inc., Lanham, MD
Shaw,G. E., Aerosolsat Mauna Loa: Opticalproperties,J.Atmos.Sci., 20706.
36, 862-869, 1979. R. Halthore, Naval ResearchLaboratory,Code 7228, Washington,
Shaw,G. E., Transportof Asian desertdustto the Hawaiian Islands, D.C. 20375.
J. Appl. Meteorol.,19, 1254-1259, 1980.
Shaw,G. E., Atmosphericturbidity in the polar regions,J. Appl. Me- (ReceivedMay 31, 2000; revisedNovember 27, 2000;
teorol., 21, 1080-1088, 1982a. acceptedDecember 7, 2000.)