Jan 20, 1995 - the North Atlantic Ocean were determined as part of a program designed to .... are collected from aluminum, walk-up type, sampling towers.
JOURNAL
OF GEOPHYSICAL
RESEARCH,
VOL. 100, NO. D1, PAGES 1199-1213, JANUARY
20, 1995
Trace elements in the atmosphere over the North Atlantic R. Arimoto, R. A. Duce,• B. J. Ray, W. G. Ellis Jr., J. D. Cullen, and J. T. Merrill GraduateSchoolof Oceanography,Universityof RhodeIsland, Narragansett
Abstract. The concentrations of traceelementsin aerosolparticlesfrom the atmosphereover the North AtlanticOceanweredeterminedaspartof a programdesignedto characterize the chemicalclimatologyof theregion. For thesestudies,whichwerepartof the AtmosphereOceanChemistryExperiment(AEROCE), 2 yearsof sampleswerecollectedat TudorHill, Bermuda(BTT), andat RaggedPoint,Barbados(BAT); and 1 yearof sampleswascollectedat Mace Head, Ireland(MHT) andat the Izafia Observatory,Tenerife,CanaryIslands(IZT). One major componentof the aerosolwas atmosphericdust,andthe rankingfor the medianmineral dustconcentrations asrepresentedby aluminumwas BAT > IZT > BTT > MHT. The A1 concentrations at BAT, IZT, andBTT rangedover 4 ordersof magnitude,i.e., from 0.001 to 10 gg m-3. At MHT the maximumdustconcentrations were abouta factorof 10 lower than at the other sites,but the lower end of the rangein dustconcentrations was similar at all sites. The mineraldustconcentrations generallywere highestin summer,andthe flux of atmosphericdust was dominatedby sourcesin North Africa. The elementsshowingclearenrichmentsover the concentrations expectedfrom seasalt or crustalsourceswere I, Sb, Se, V, andZn. At Izafia, which is in the free troposphere(elevation-2360 m), the concentrations of Se andI were much lower than at the boundarylayer sites;this differencebetweensitesmostlikely resultsfrom the marineemissionsof theseelements. The impactof pollutionsourceson traceelement concentrations
was evident at all sites but varied with season and location.
The concentrations
of
elementsoriginatingfrom pollutionsourcesgenerallywere low at Barbados. Analysesof trace elementratiosindicatethatthereare large-scaledifferencesin the pollutionemissionsfrom North America versusthosefrom EuropeandAfrica. Emissionsfrom pyrometallurgicalindustries, steelandiron manufacturing,andpossiblybiomassburningare moreevidentin the atmospheric samplesinfluencedby transportfrom EuropeandAfrica.
1. Introduction
The Atmosphere-Ocean Chemistry Experiment (AEROCE) is a multidisciplinary program focusing on the atmospheric and marine chemistry of the North Atlantic Ocean region. The trace
element
studies
described
here
were
conducted
for
AEROCE in part as a result of concernsover perturbationsto biogeochemical cycles caused by human activities. In addition, the chemistry and transport of mineral aerosol is considered in some detail because the magnitude of the dust flux from Africa to the North Atlantic is important from a geochemical standpoint [e.g., Duce et al., 1991]. In this paper we present results from the four AEROCE stations (Figure 1): Ragged Point, Barbados (BAT, 13.18øN, 59.43øW); Tudor Hill, Bermuda (BTT, 32.24øN, 64.87øW); Mace Head, Ireland (MHT, 53.43øN, 9.73øW); and the Izafia Observatory,Tenerife (IZT, 28.30øN, 16.48øW). AEROCE is providing a unique record of long-term trends and climatological averageconcentrationsof a variety of trace substancesin the atmosphere over the North Atlantic. A
major focus of climatology of North Atlantic designed to (1)
our program is to characterize the chemical trace elements in aerosol particles over the Ocean. More specifically, our studies are characterize the geographical and temporal variability in the concentrationsof mineral aerosol, sea salt, and selected enriched trace elements in the atmosphere; (2) assess how this variability in concentrations is related to physical, chemical, and biological processesoccurringin the atmosphere and the oceans; and (3) investigate the mechanisms involved in the air-sea exchange of trace elementsand improve estimatesof the fluxes of trace elements across the air-sea interface.
Studies
of the trace
element
fluxes
related
to the third
objective will be reported elsewhere. Our studies of the climatology of trace elements also make it possible to delineate trends in concentrations, and they bring to light some of the potential impactsof changingsourcesand source distributions for trace elements in the atmosphere.
2. Experiment 1Departments of Meteorology andOceanography, Collegeof Geosciences and Maritime Studies,TexasA&M University,College Station.
Copyfight1995 by the AmericanGeophysical Union. Paper number94JD02618. 0148-0227/95/94JD-02610505.00 1199
2.1.
Sampling
andSample
Handling
Aerosol particle samples were collected for 2 years from Bermuda and Barbados starting in June and August 1988, respectively. One year of data are available from the Izafia Observatory, Tenerife (Canary Islands), and from Mace Head, Ireland, where sampling commenced in late April 1989 and August 1989, respectively. At BTY, BAT, and MHT, samples
1200
ARIMOTO
ET AL.: TRACE ELEMENTS
IN THE ATMOSPI-IE•
day and blanks are collected every third day. The daily sampling interval was used to facilitate comparisionsamong all of the aerosol speciesstudiedfor AEROCE, i.e., sulfate,
methanesulfonate, nitrate, 2tøPb,and?Be. The sampling interval was not optimal for all elements,but it was suitable for sea salt and mineral dust which are two componentsof major interest. The sampling interval also was selected to make it possible to interpret our data using air mass trajectories. The aerosol samples are sealed in plastic bags, and each week they are mailed to the University of Miami and then to our laboratory at the University of Rhode Island (URI). For the trace element studies, the aerosol particle samples are selectedfor analysisbased on the volume of air pumped,i.e.,
50øN
30øN
all of the sampleswith volumesgreaterthan 100 m3are
10øN
Figure 1. Sampling stations for the Atmosphere-Ocean Chemistry Program(AEROCE).
are collected from aluminum, walk-up type, sampling towers (-20 m tall), while at IZT, samplingis conductedfrom the roof of a Spanish Meteorological Observatory. The Izafia Observatoryis the only AEROCE site not on a coast,and it is located at an elevation of-2360 m. At IZT the samplersonly operate during the night when downslope conditionsenable us to collect samplesfrom the free troposphere. At the AEROCE primary sites of Bermuda and Mace Head, custom-made computer-basedsystemsare used to control the aerosol particle sampling with respectto wind direction, wind speed,condensationnuclei (CN) counts,relative humidity, and the occurrence of rain. This system, based on a design by Schwartze et al. [1988], was developed to minimize the contamination of samples from local sources. A simpler system was used at the secondarysites of Barbadosand Izafia; at those sites, CN counts, humidity, and rain were not used as control parameters. The wind sectors chosen for sample collection
at the various
sites were
based on the location
of
the sampling platform, the local topography, etc. The wind sectors, expressed as degrees true, were as follows: BAT = 335 ø to 130 ø, BTT = 180 ø to 330 ø, MHT = 180ø to 300 ø, IZT = 270 ø to 45 ø.
Bulk aerosol particles were collected with high-volume
analyzed. This volume of air is required to obtain a sufficient and representative mass of the key trace species, and it is equivalentto a minimum pumpingtime of-10% per day. Differences in the volumes of air sampled at the different sites introduce possible biases in the data; this is a potential problem not only when comparing the data among the sites but also with previous studies. The potential problem with experimental design is more than offset, however, by the ability to intercomparethe AEROCE aerosoland precipitation data sets from different groups. This is made possible by employing a daily sampling interval for all of the studies. Furthermore, the ability to use air mass trajectories in interpreting the daily aerosol data, which is also made possible by the daily sampling protocol, is a decided advantage. The arithmetic mean volumes (+ standard deviation) of the air sampledwere BAT = 956 + 240 (numberof samples, n = 344) and 1016 + 215 (n = 347) for the first and secondyears of sampling,respectively;BTT = 682 + 452 (n = 183) and 853 + 547 (n = 196) for the first and secondyear; IZT = 365 + 170 (n = 173) ßand MHT = 1182 + 1425 (n = 143). For the URI trace element studies, the aerosol-laden filters
and blank filters are preparedfor chemical analysisin a class 100 clean room. The filters are cut using clean custom-made plastic scissorsand plastic forceps. All manipulationsof the filters are done by personswearingclean room garb, including jumpsuits, face masks, hats, and shoulder-length plastic gloves. Quarters of the sample and blank filters are individually pressedinto pellets at a pressureof-150 to 200
kg cm-2 usinga stainlesssteelpresswith titaniumend-caps and a plasticliner. Separatealiquotsof the filters are usedfor
samplersoperatingat approximately 45 m3 h-1. The samplers the two irradiation schemes described below, and the remainder are made of noncontaminatingmaterials, and to the greatest extent possible, standardized aerosol sample handling protocolswere implementedat each of the stations. Samples were collected on single, unwashed, 8 in x 10 in Whatman 41© filters (Whatman International Limited, Maidstone, England). For each of the stationsthe volumesof air sampled were calculated from flow tubes with calibrated orifice plates. Errors in the volumes of air sampled are estimatedto be 5 to 10%. The cutoff characteristicsof the samplersare such that particles smaller than 10 [tm are captured efficiently (J. M.
Prospero, University of Miami, personal communication, 1989), and therefore the samplers' collection characteristics should not significantly bias the results. One sampleand one field blank are collectedeachday from the primary stations,but at Mace Head, inclement weather occasionallymade it impossibleto changesampleson a daily basis. At the secondarystations,samplesare collectedevery
is frozen and storedin a samplearchive. 2.2.
ChemicalAnalyses
The AEROCE samplesare analyzedby instrumentalneutron activation analysis(INAA) using the 2 MW researchreactor
(neutronflux = 4 x 10•2 neutronscm-2 s-t) operatedby the Rhode Island Nuclear Science Center [Duce et al., 1983]. These methodsare nondestructiveand providedata for up to 26
elementsthrough two independentanalytical procedures. A "short" (1 to 2 min) irradiation scheme is used for the determinationof A1, Br, Ca, C1, Cu, I, Mg, Mn, Na, and V. A "long" (-28 hours) INAA procedure is used for the determination of Ba, Co, Cr, Cs, Eu, Fe, Hf, Rb, Sb, Sc, Se, Ta, Tb, Th, Yb, and Zn. The concentrations
of the trace elements were determined
throughthe useof flux monitors(shortINAA) or flux monitors
ARIMOTO
ET AL.: TRACE ELEMENTS
and internal standards(long INAA). Peak searchesand peak analyses initially were performed using a Prime 550 minicomputer, but more recently a VAX/VMS-based GENIE workstation multichannel analyzer system (Canberra Nuclear, Meriden, Connecticut) was used for this purpose. The uncertainties
in
the
concentrations
of
most
elements
IN THE ATMOSPHE•
1201
seasonalchangesin transportpathways. Finally, and perhaps most important, the multiple-colinearity imposed on the chemical data by meteorological processestends to obscure the information
that can be discerned
relative
to sources and
source regions.
are
estimatedto be -10%, mainly due to the uncertaintiesin the 3.1. Mineral Aerosol flow measurementsystem and to blank corrections. For Cu, The long-rangetransportof mineral dust and its flux across Mn, Sb, and Zn the uncertaintiesare higher, 10 to 50% owing the air-sea interface links the biogeochemicalcycles of the to the lower relative sensitivityof INAA for these elementsas well as larger blank corrections. Blank correctionsare made continents, atmosphere, and oceans. Dust over the North by subtractingthe concentrationsof the elements in the field Atlantic has been the subjectof scientificreportsdating from the late 1700s [e.g., Dobson, 1781], and the occurrence of blanks from those in the samples. Nine different Ge(Li) detectors were used for the sample analysis, and the dust storms has long been of interest to mariners and operational detection limits were based on counting statistics, navigators [e.g., Darwin, 1846]. An African source for the with a lower limit nominally set at 100 countsper elementper mineral dust over the tropical North Atlantic is well established. For example, geochemicalstudiesindicate that sample. Quality control and quality assurance (QC/QA) for the dust from the Sahara is the parent material for soils found on chemical analyses are documented and maintained through various islands in the Caribbean and in the western Atlantic [Muhs et al., 1990]. The quantity of mineral matter involved interlaboratory calibrations, including the U.S. EPA QC/QA program and through the analysis of standard reference is considerable. Junge [1979] estimatedthat 60 to 200 Tg of Saharan dust are generatedeach year. More recently the materials. The reference materials routinely analyzed for QC/QA include National Instituteof Standardsand Technology atmospheric input of mineral dust to the North Atlantic has standardreference materials suchas coal fly ash (SRM 1633 or been estimatedto be -220 Tg yr-• [Duce et al., 1991]. A connectionbetween the mineralogyof eolian particles 1633a), urban particulate matter (SRM 1648), and ashless and deep-sea sedimentshas beenrecognizedfor decades[e.g., blank filter (SRM 2681). Rex and Goldberg,1958; Griffin and Goldberg,1968; Chester 3. Results
et al., 1972; Windom, 1975]. These and more recent studies
and Discussion
[e.g., Blank et al., 1985] demonstratethat the long-range
A set of summary statistics for the trace element concentrations
at each of the AEROCE
sites is not included
here but is available from the American Geophysical Union
on microfiche. • Includedin the summarytableare 2 yearsof data for BAT and BTT and one year of data for IZT and MHT. The different periods of coverage reflect the times when the stations first became operational. The raw data also can be obtained by contactingthe authors. In the discussion that follows,
the trace elements
are
classified into groups on the basis of their presumptive sources: these are mineral aerosol,atmosphericsea salt, and a group of enriched elements (EEs). Sources for the trace elements are then assessedthrough analysesof interelemental relationships and through a consideration of transport pathways and temporal variability in the concentrations. Multivariate statisticaltechniqueshave become widely used for assessing the sources for trace substances in the atmosphere,but after extensivediscussionswith our AEROCE colleagues, we opted not to use this approach for several reasons. First, the numberof EEs is small given the numberof sources that are potentially important. Second, in a substantial number of samples the concentration of one or more of the EEs was below detectionlimits, and thesemissing values complicate the analyses. Third, some of the anthropogenicsourceemissionsare not well characterizedand
atmospheric transport of weathered crustal material from the continents supplies a major portion of the nonbiogenic sedimentsthat accumulateon the seafloor. The importanceof informationconcerningthe fluxes of mineral aerosolis further enhancedby the application of such information to studiesof paleoclimate and to analysesof geographicalpatternsin the composition of deep-sea sediments [e.g., Rea et al., 1985; Sarnthein et al., 1982; Sirocko and Sarnthein, 1989]. Oceanographicstudies have shown that the dissolution of certain trace metals, suchas Fe and A1, following the air-to-sea transfer of mineral aerosol is an important and sometimes dominant
source for these elements
in surface waters of the
open ocean [Duce, 1986; Orians and Bruland, 1986; Maring and Duce, 1987; Prospero et al., 1987; Martin and Fitzwater, 1988; Zhuang et al., 1990; Duce and Tindale, 1991]. Much of the recentattentionhas focusedon Fe becausephytoplankton productionapparentlycan be limited by the supply of eolian iron [Martin and Fitzwater, 1988].
The results concerning mineral aerosol presented here address(1) the elementalcompositionof atmosphericdust and (2) the temporal and spatial variability in dust concentrations, particularly in relation to atmospheric transport pathways over the North Atlantic. The elemental composition of mineral
dust from
Barbados,
Bermuda,
and Izafia
was
investigated by focusing on the ratios of the trace elements in other cases are not distinctive. Furthermore, as shown relative to aluminum, which is often used as a representative below, the pollution sourcesvary in importancewith respect crustal element and makes up -8% of the upper continental to sourceregion, and their impacts are stronglyinfluencedby crust [e.g., Uematsu et al., 1983; Duce et al., 1983]. Elemental ratios relative to A1 were calculated for samples collected under high-dust conditions, i.e., mineral aerosol
•Set of supplementarystatisticsis available with entire
concentrations of > 2 •g m-3for BAT andBTT and> 5 •g m-3at
article on microfiche. Order from American Geophysical Union, 2000 Florida Avenue, N.W., Washington,DC 20009.
IZT. The elemental ratios were calculated by a linear fitting procedure known as reduced major axis (RMA) regression analyses (also known as the calculation of lines of organic
DocumentD94-001; $2.50. Paymentmustaccompany order.
1202
ARIMOTO
ET AL.:
TRACE
ELEMENTS
IN THE ATMOSPHERE
correlation),as describedby Hirsch and Gilroy [1984]. These the NorthAtlantic[e.g.,Prosperoand Carlson,1972;Lepple, authorsconcludedthat RMA is preferredover ordinary least 1975; Schiitzet al., 1981; Morales, 1986]. For example, squaresbecauseit more accuratelyreflects the relationships Prospero et al. [1970] traced the origin of a dust event at between variables such as the trace element concentrations Barbadosto West Africa, with a transporttime of-5 days. presentedhere. Related studiesby Carlson and Prospero [1972] showedthat The ratios of eighteenelements(Ba, Ca, Co, Cr, Cs, Eu, Fe, the mineral aerosolfrom Africa is transportedat altitudesof Hf, Mg, Mn, Na, Rb, Sc, Ta, Tb, Th, V, Yb) calculatedrelative 1.5 to 4.5 km in large anti-cycloniceddies. to A1 for the high-dust events are similar in the Barbados, Dust transportfrom Africa occursthroughoutthe year, but Bermuda, and Izafia data sets (Table 1). The observedelement
to A1 ratios for the high-dustsamplesalso are comparableto thoseof the upper continentalcrustbasedon the compilation of Taylor and McLennan [1985]. This similarity is further evidencethat many of the trace elementswe studyoriginate from a crustal source. During the high-dustevents at BAT, BTT and IZT, mineral aerosolis by far the dominantinorganic
the sourceregions and main transportpathwayschangein concert with movementsof the Inter-Tropical Convergence Zone (ITCZ) [seeD'Almeida, 1986]. A pronouncedseasonal
cycle in mineral aerosol concentrationsis evident at each of the four stations (Figure 2), presumablydue to shifts in the major wind fields and the resultant changes in transport pathways [see Schiitz, 1980]. The dust fluxes to the North aerosol constituent on a mass per unit volume basis. In Atlantic during the summer generally are higher than those in contrast, at Mace Head, the maximum dust concentrationswere winter. It is noteworthy, however, that the single highestdust never so high that the pollutant contributions to the trace concentration observed in 2 years of sampling at Barbados elements could be ignored. Under high-dust conditions at occurredduring the winter of 1988-1989. Bergametti et al. [1989] used elemental ratios (Si/A1 and BAT, BTT, and IZT, someelementssuchas Ca and Mg whose concentrationsare more commonly controlledby sea salt can Fe/A1) to distinguish between Sahelian versus Moroccan be dominatedby atmosphericdust (Table 1). Similarly, the source regions for dust particles collected over the Canary concentrationof Sb was at times stronglyinfluencedby dust, Islands. These authors found that the S i/A1 and Fe/A1 ratios but in general, Sb was substantiallyenriched. For example, both were higher in dust originating from a Sahelian source Ellis et al. [1993] concludedthat at Bermuda,only 6% of the comparedwith dust from the vicinity of Morocco. More than Sb was crustal. 20 years earlier, Delany et al. [1967] observed seasonal The short-term variability of the dust concentrationsis variations in the magnetic properties of dust collected at clearly linked to synoptic meteorology, and during sporadic Barbados. Based on analysesof weather maps, these authors events,which generally last 2 to 4 days, the atmosphericdust concludedthat desert soils from Morocco were a likely source concentrations at BAT, BTT, andIZT can exceed100gg m-3. for the dust during the summer, while more southerly Isentropic air trajectories indicate that the dust flux is influences, i.e., vegetatedregions to the south of Dakar, were dominatedby sourcesin North Africa [Arimoto et al., 1992], implicated in the winter. Similarly, meteorologicalanalyses and this is fully consistentwith previous studiesof dust over by Morales [1986] indicate a seasonalnorth-to-southshift in dust stormactivity in Africa from summerto winter. A threedimensional analysis of the atmospheric cycling of dust Table 1. Mass Ratios of Crustal Elements to Aluminum particles by Joussaume [1990] also suggested seasonal changesin the African origins of mineral aerosol. for High-DustEventsat Barbados,Bermuda,andIzafia Silicon is not determinedby our methods,but we were able Observed to evaluatepossiblechangesin sourceregionsby considering the Fe/A1 ratios for 2 years of data from Barbados. Based on the studies of Bergametti et al. [1989] and Delany et al., Barbados Bermuda Izafia Average [1967], one would expect higher Fe/A1 ratios in the winter (BAT) (BTT) (IZT) CrustalRocka owing to more southerly source regions with a higher magnetic content. We tested this hypothesisby considering Ba 6.2 x 10.3 1.2 x 10.2 9.8 x 10.3 6.8 x 10-3 the samplesfrom Barbadoswith the 5% highest and the 5% Ca 2.9x10 -• 3.3x10 -• 3.6x10 -• 3.7x10 -• Co 2.4 x 10-4 2.4 x 10-4 3.0 x 10-4 1.2 x 10-4 lowest Fe/A1 ratios. The median Fe/A1 ratio for the complete Cr 1.1 x 10-3 1.9 x 10-3 1.5 x 10-3 4.4 x 10.4 sample set from Barbadoswas 0.58; this is comparableto a Cs 4.7x10 -5 6.2x10 '5 7.1x10 -5 4.6x10 -5 crustal average of 0.44 [Taylor and McLennan, 1985]. The Eu 2.2 x 10-5 2.3 x 10-5 2.9 x 10-5 1.1 x 10-5 median Fe/A1 ratio for the uppermost5% of the sampleswas Fe 5.1 x 10-• 6.1 x 10-• 7.0 x 10-• 4.4 x 10-• 2.2, while the averageratio for the lowest 5% of the samples Hf
5.2 x 10-5
5.8 x 10-5
8.1 x 10-5
7.2 x 10-5
Mg
3.7x10. •
3.2x10 -•
3.0x10 -•
1.6x10 -•
Mn Na Rb Sb Sc Ta Tb Th V Yb
1.1 x 10-2 1.1x10 ø 1.1 x 10-3 1.2 x 10.5 1.7 x 10-4 2.1 x 10-5 1.6 x 10-5 1.6 x 10.4 1.5 x 10-3 4.2 x 10-5
9.5 x 10-3 3.3x10 '• 1.7 x 10-3 1.3 x 10.5 2.0 x 10-4 2.1 x 10-5 1.5 x 10.5 2.0 x 10.4 3.5 x 10-3 5.0 x 10-5
1.2 x 10-2 1.1x10. • 1.6 x 10-3 1.4 x 10.5 2.3 x 10-4 2.8 x 10-5 1.9 x 10.5 2.0 x 10-4 1.6 x 10-3 5.2 x 10-5
7.5 x 10-3 3.6x10 -• 1.4 X 10-3 2.5 X 10.6 1.4 x 10-4 2.7 x 10-5 8.0 X 10-6 1.3 x 10.4 7.5 X 10-4 2.7 x 10-5
aTay•.orandMcLennan,1985.
was 0.31.
Many of the samples (18 of 29 = 62%) exhibiting the highestFe/A1 ratios were collectedfrom Novemberto January, while most (66%) of the sampleswith lowest 5% Fe/A1 ratios were collected between August and November. The Fe/A1 ratios are highly variable from day to day, however, and both high and low Fe/A1 ratios can occur within a given month. While semiquantitative at best, these results support the hypothesized seasonal differences in the African source regionsfor dustover the North Atlantic. The geographical variability in the dust concentrations over the North Atlantic is pronounced, with the median concentrations of dust as represented by A1 as follows:
ARIMOTO ET AL.: TRACE ELEMENTS 1N THE ATMOSPHERE
1203
z ¸
z
¸
z ¸
[-m $ri 'uo!lmluoauo D
[-m $ri 'uo!lmluoauo D
