on a
Moun
awfor
c
isation, Lo
SW Austra
, NSW, Au
? Long-range transport of Na and aeolian
dust impacts this remote inland site.
ent of PM
2.5
(particulate matter with aerodynamic diameters less than 2.5 ?m) at
obile and smoke sources
zardreductionburns,re-
al-?red power stations,
s,andagedseasaltfrom
, asele-
overall
Science of the Total Environment 630 (2018) 432?443
Contents lists available at ScienceDirect
Science of the Total Environment
journal homepage: www.elsevier.com/locate/scitotenv
the SouthernOcean to the remotealpinestudy site. The impact of recent climate change wasrecognised
vated smoke and windblown soil events correlated with drought and El Ni?o periods. Finally, the
0.4)% were the ?ve PM
2.5
source types, each with its own distinctive trends. The autom
wereascribedtoasigni?cantlocalin?uencefromtheroadnetworkandbush?reandha
spectively. Long-range transport are the dominant sources for secondary sulfate from co
windblownsoilfromtheinlandsalineregionsoftheLakeEyreandMurray-DarlingBasin
Long-range transport
Snowy Mountains
El Ni?o
Aeolian dust
Accepted 19 February 2018
Available online 24 February 2018
Editor: Jianmen Chen
Yarrangobilly,intheSnowyMountains,SEAustraliaareexaminedandquanti?ed.Anewaerosolmonitoringnet-
work was deployed in June 2013 and aerosol samples collected during the period July 2013 to July 2017 were
analysed for 22 trace elements and black carbon by ion beam analysis techniques. Positive matrix factorisation
and back trajectory analysis and trajectory clustering methods were employed for source apportionment and
to isolate source areas and air mass travel pathways, respectively. This study identi?ed the mean atmospheric
PM
2.5
mass concentration for the study period was (3.3 ? 2.5) ?gm
?3
. It is shown that automobile (44.9 ?
0.8)%, secondary sulfate (21.4 ? 0.9)%, smoke (12.3 ? 0.6)%, soil (11.3 ? 0.5)% and aged sea salt (10.1 ?
Keywords:
PM
2.5
Positive matrix factorisation
Received in revised form 19 February 2018
sition and source apportionm
Article history:
Received 3 January 2018
Characterisation of atmospheric aerosols is of major importance for: climate, the hydrological cycle, human
health and policymaking, biogeochemical and palaeo-climatological studies. In this study, the chemical compo-
article info
hancedsmokeandsoilaerosolloadings.
? Drought and El Ni?o conditions en-
? Corresponding author at: Australian Nuclear Science
E-mail address: Carol.Tadros@ansto.gov.au (C.V. Tadro
https://doi.org/10.1016/j.scitotenv.2018.02.231
0048-9697/Crown Copyright ? 2018 Published by Elsevie
abstract
arysulfate,smoke, soilandagedsea salt
Australia
? Sources of PM2.5: automobile, second-
? FirstPM2.5dataset(2013?2017)forthe
Snowy Mountains alpine region, SE
Stuart Hankin ,ReginaRoach
a
Australian Nuclear Science and Technology Organ
b
Connected Waters Initiative Research Centre, UN
c
NSW National Parks and Wildlife Service, Sydney
HIGHLIGHTS
nd source identi?cation of atmospheric
tains, south-eastern Australia
d
a
,PaulineC.Treble
a,b
,AndyBaker
b
,DavidD.Cohen
a
,ArmandJ.Atanacio
a
,
cked Bag 2001, Kirrawee DC, NSW 2232, Australia
lia, Sydney, NSW, Australia
stralia
GRAPHICAL ABSTRACT
Chemical characterisati
aerosols in the Snowy
CarolV.Tadros
a,b,
?,JagodaCr
a
and Technology Organisation, Locked Bag 2001, Kirrawee DC, NSW 2232, Australia.
s).
r B.V. All rights reserved.
tial
ng-
ata
and McTainsh, 2003; Rutlidge et al., 2014). In
of trace metals from aerosols in natural a
palaeoclimate studies (Sigl et al., 2015). The con
deposited in soils, sediments, peat, speleot
been used as palaeo proxy indicators of clim
Ocean and therefore with little impact from regional anthropogenic in-
433C.V. Tadros et al. / Science of the Total Environment 630 (2018) 432?443
addition the deposition
rchives is important in
centration of aerosols
hems and glaciers have
ate and environmental
anthropogenic origin and from local and long-range transported
sources in the Snowy Mountains. In essence, this work will provide a
benchmark to understand the local and regional dynamics of aerosol
sources,providingthemostextensiveaerosolbaselinedatasetforappli-
cation in other studies regionally.
2007). Depositing aerosols are also known to
chemical properties of soils and sediments (Si
alter the physical and
monson, 1995; Hesse
?uences.Thecurrent datasetis important for achievinga better under-
standing of the composition of atmospheric aerosols from natural and
2010) or results in ocean acidi?cation in coastal regions (Doney et al.,
implications including poten
knowledge, this isthe ?rst lo
gion and provides a crucial d
1. Introduction
The main aerosol sources which in?uence the atmosphere of the
Australian continent include windblown soil-dust, sea-salt, biomass
burning, and biogenic secondary organic aerosols from volatile organic
aerosols(Rotstaynetal.,2009).Aerosolsaresolidand/orliquidparticles
suspendedinair(P?schl,2005).Theseparticlesmaybeemitteddirectly
into the atmosphere from i) natural sources, such as: biogenic volatile
organic compounds emitted from vegetation, sea-salt, soil-dust, volca-
noes, biomass burning; or ii) anthropogenic sources, such as: coal-
burning, metal smelting, vehicle emissions (Colbeck and Lazaridis,
2014).Onceintheatmosphere,theaerosolsaretransporteddownwind
of their source area, during which time chemical reactions can occur,
resulting in the formation of secondary aerosols from precursor gasses
such as SO
2
and NO
x
(e.g. Squizzato et al., 2013; Suni et al., 2008).
Thus aerosols contain chemically distinct concentrations of trace
metals, characteristic of their supply source, air mass history and geo-
graphical proximity to coastal, desert, rural, industrial and polar emis-
sion sources.
Aerosols suspended in the atmosphere are known to impact the cli-
mate system. Aerosol particles have a direct radiative forcing effect on
the Earth's surface and atmosphere, because they can absorb or scatter
incoming solar radiation. When solar radiation is re?ected back to
space, a smaller amount of solar energy reaches the ground and this
has a cooling effect on the regional and global climate, whereas some
aerosol particles can absorb solar radiation and this warms the atmo-
spheric layer. Aerosols also have indirect radiative forcing effects on
the climate through modifying the formation and microphysical cloud
properties, leading to a cooler climate. Acting as condensation nuclei,
solubleaerosolscanincreasethecondensationmoistureincloudsbyin-
creasing the droplet number concentration and therefore cloud albedo,
or the re?ection of solar radiation to space. Aerosols also decrease the
droplet size in clouds, suppressing precipitation (and affecting the hy-
drological cycle), which may also lead to an increase in cloudiness and
re?ection of solar radiation. These direct and indirect aerosol effects
on the climate are determined by aerosol particle size, structure and
chemical composition (P?schl, 2005).
Air quality is the most visible effect of aerosols in the environment,
affecting visibility, aviation and road traf?csafetyaswellashuman
health. Fine particulate matter with an aerodynamic diameter smaller
than 2.5 ?g(PM
2.5
) can foster acute and chronic diseases (Pope III
et al., 2009), hence to reduce air pollution, policy interventions,
targetingconcentrationlimitsandairqualityguidelines,areestablished
based on monitoring data (WHO, 2006). Atmospheric deposition of
aerosols is also of speci?c interest in ecosystem biogeochemistry
(Mahowald, 2011). Aerosols provide trace metals that are essential for
productivity in terrestrial (Chadwick et al., 1999) and marine ecosys-
tems (Jickells, 1995). Alternatively they can have a negative impact,
for example acid rain, due to atmospheric sulfate or nitrate deposition,
by enhancing the leaching of nutrients from land ecosystems (Likens,
aerosol derived proxies for interpreting palaeo-archives are discussed. To our
term detailed temporal and spatial characterisation of PM
2.5
aerosols for the re-
set for a range of multidisciplinary research.
Crown Copyright ? 2018 Published by Elsevier B.V. All rights reserved.
changes (Rea, 1994; Kohfeld and Harrison, 2001; Frisia et al., 2005;
Frappier, 2006; Marx et al., 2011; Allan et al., 2015; Ridley et al.,
2015).Thus anunderstandingof the composition,source and transport
of atmospheric aerosols are highly relevant for studies in various disci-
plines including climate processes, human health and policies, terres-
trial and marine biogeochemical cycling and palaeo-climatology.
Our study site is located in Kosciuszko National Park, a high
elevation-alpine site in mainland Australia (1059 m above sea level).
Thelocationofthestudysiteisasensitivemonitorforpasthydroclimate
and environmental change as it is a region prone to ash input from ?re
activity and wind-blown dust deposition, as it is located within the
south-eastern pathway of dust transport from active emission sources
in central Australia (Shao et al., 2011). The site is also strategically
unique in Australia, being a remote inland site; it is conducive to study
the impact of long-range atmospheric transport from regional and nat-
ural emission sources.
The impetus for the current study is to de?ne the chemical charac-
teristicsand inputsources and gain a better understandingof transport
processes of atmospheric particulate matter to karst in the Snowy
Mountains alpine region, Yarrangobilly, SE Australia. This study was
conducted as part of a wider project to reconstruct past environmental
changefromcavedeposits(speleothems)inordertobetterunderstand
past variability in climate, ?re history and environmentalchange in the
SnowyMountainsregion.Traceelementsareoneofthecommonlyused
proxies for providing information on the hydro-climate regime sur-
rounding the depositional conditions in sediments and carbonate rock
archives. As atmospheric aerosols are a direct source of trace elements,
constraining the sources of trace elements may be highly relevant for
thereconstructionofpastenvironmentalchangefromspeleothems.Al-
though the atmosphere may supply a signi?cant source of elements to
thesoil and hence drip-waters, less attention has beenplaced on quan-
tifying the atmospheric input from various sources above the cave.
Dredge et al., 2013 highlighted that aerosols brought into the cave by
air currents are a potential source of elements for speleothem deposi-
tion, however neither the source of aerosols nor the atmospheric pro-
cesses associated with them were investigated. Hence there is a
current knowledge gap in our understanding of the holistic role of at-
mosphericinput,particularlyaerosols,asapotentialsourceofelements
thathavebeentransportedinthecaveviain?ltratingdrip-waterandin-
corporated into speleothem calcite.
This paper presents the?rst high resolution four-year PM
2.5
dataset
for the region. Ion beam analysis (Cohen et al., 1996) was used to iden-
tify the elemental composition of the PM
2.5
samples, following which
positive matrix factorisation (PMF; Paatero and Tapper, 1994) was ap-
plied to identify the contributing sources. Additionally, long-range
source contribution regions to this remote inland site were identi?ed
using backward air mass trajectory calculations. To compliment the
study, results were alsocompared to oneof theglobal ?baseline? atmo-
sphericmeasurementsitesatCapeGriminTasmaniaAustralia,aremote
coastal site that receives air?ow predominately from the Southern
2. Methods
2.1. Study site and area description
An aerosol PM
2.5
monitoring station was deployed on 26 June 2013
above Jillabenan Cave, NSW, Australia (35? 43?S, 148? 29?E; Fig. 1a),
2.5 m from an existing weather station and 170 m from Harrie Wood
Cave, which is the location of the cave drip water monitoring sites
(Tadros et al., 2016). The aerosol sampling site is located within the
Yarrangobilly Caves karst system and is situated in the northern part
of Kosciuszko National Park, in the Snowy Mountains region of south-
eastern New South Wales. Access to the Yarrangobilly Caves is via the
Snowy Mountains Highway (Fig. 1a). At an elevation of 1287 m above
sea level (a.s.l.) this road, and the cave site, receives snowfall during
the winter months. The sampler is at an altitude of 1059 m a.s.l and is
on thewestern sideof theGreatDividingRange(Fig. 1b). The Great Di-
viding Range, a series of plateaus, extends from Cape York Peninsula in
northern Queensland southward to Victoria. It ranges in altitude from
300 m to 2228 m a.s.l; the height of Mt. Kosciusko and the highest
point on the Australian mainland. The remainder of the continent has
low topographical relief, where the average elevation is less than
300 m. Hence the location of the study site is important as it lies
es,w
rling
2.7?E
S13
San
ng
of
434 C.V. Tadros et al. / Science of the Total Environment 630 (2018) 432?443
Fig.1.(a)Topographyandlocationoftheaerosolstudysiteinproximitytosurroundingcav
outline maps of the main dust source regions in Australia (e.g. Lake Eyre Basin, Murray Da
length of the eastern seaboard of Australia is shown as a single grey line. The numbers indicate
(2011)and correspond to: Lake Mungo (1; 33.8?S144.1?E); Lake Windaunka (2; 31.0?S14
SaltPans(5;29.1?S133.6?E);LakeEyreNorth(6;28.0?S137.5?E);SimpsonDesert(7;24.0?
131.0?E);GibsonDesert(10;25.0?S126.0?E);LittleSandyDesert(11;24.0?S122.0?E);Great
Desert (14; 21.0?S131.0?E);Riverina agricultural (15; 34.7?S146.5?E).(c)Wind roseshowi
26/02/2016. The wind rose was compiled using available hourly data measured at a height
Cave. The mean wind direction is shown as a red vector.
eatherstationsandroadways.(b)MapofAustraliashowinglocationofthestudysiteand
Basin and Australia's deserts). The Great Dividing Range which extends along the entire
the location of signi?cant dust source regions in Australia, reported by Cohen et al.
); East Flinders ranges (3;29.9?S 139.6?E); Olympic Dam (4;31.0?S 136.5?E); Emu Fields
7.4?E);GreatVictoriaDesertWest(8;28.0?S124.5?E);GreatVictoriaDesertEast(9;28.0?S
dyDesertWest(12;21.0?S124.0?E);GreatSandyDesertEast(13;24.5?S130.5?E);Tanami
the frequency, wind speedand direction using data collected over the period6/11/2012to
0.2 m above ground level, from the automated weather station located outside Jillabenan
435C.V. Tadros et al. / Science of the Total Environment 630 (2018) 432?443
south-east of the Lake Eyre drainage basin and Murray Darling basin, a
large area of persistent dust activity (Prospero et al., 2002) as well as
great deserts in the western plateau (Fig. 1b).
The wind rose for our site, based on hourly wind data (Fig. 1c),
shows that on an annual basis approximately 30% of the winds is from
thenorthwest quadrant (WNW,NWandNNW).Themeanwinddirec-
tionis290?andfor21%ofthetime,windblowsfromtheWNW.During
summer, onshore winds from the south-southeast are dominant, while
during winter, stronger inland winds from the north-northwest direc-
tion dominate.
2.1.1. Sampling
AstandardIMPROVEPM
2.5
cyclonebasedaerosolsamplingunit,oper-
atingat22lmin
?1
andataheightof2.0mabovegroundlevel(agl;Cohen
et al., 1996), was used to collect two aerosol samples per week. Between
03/07/2013 to 02/07/2017, PM
2.5
samples were collected on Sunday and
Wednesday from 24:00 to 24:00 Australian Eastern Standard Time the
next day. Bulk aerosol samples were collected on 25 mm diameter
stretchedTe?on?lters,approximately250 ?gcm
?2
thickbeforeexposure
(pore size: 3.0 ?m; PALL life sciences). A data gap between January and
mid-March 2015 was due to ?eld access restrictions.
2.1.2. Chemical analysis
The collected aerosol samples were analysed on a 2 MV STAR accel-
erator using ion beam analysis techniques: Proton Induced X-ray Emis-
sion Analysis (PIXE), Proton Induced Gamma-ray Emission Analysis
(PIGE), Proton Elastic Scattering Analysis (PESA) and Rutherford Back
Scattering Analysis (RBS) at ANSTO (Cohen et al., 1996; Cohen, 1998;
Cohenetal.,2004a;Cohenetal.,2004b).PIXEprovidesdataforselected
elements between Al and Pb while PIGME, PESA and RBS were used to
provide information on elements lighter than Al. The four techniques
were applied simultaneously using an 8 mm diameter beam of
2.6 MeV protons and the concentrations (ng m
?3
) of: H, N, Na, Al, Si,
P,S,Cl,K,Ca,Ti,V,Cr,Mn,Fe,Co,Ni,Cu,Zn,Se,BrandPb,inthe?nepar-
ticles sampled were determined. Elements including Mg and Ba could
not be measured analytically in the aerosol samples due to a low
signal-noise ratio (Bird and Williams, 1989; Bird, 1990). Additionally,
black carbon (BC) concentrations were determined using a laser HeNe
absorption system, assuming a mass adsorption coef?cient of 7 m
2
g
?
1
(Taha et al., 2007).
2.2. Data analysis
2.2.1. Chemical mass closure
Following Malm et al., 1994, the degree of mass closure, ratio of the
reconstructed chemical mass compared to the gravitational mass for
each ?lter was evaluated, by employingthe following overall equation:
Reconstructed mass RCM???Salt ? Ammonium sulfate ? Soil
? Smoke ? Organics? BC ?1?
Estimates of each source component are de?nedbyEqs.(2)?(6)
below:
Salt ? 2:54 Na?C138 ?2?
Ammonium sulfate ? 4:125 S?C138 ?3?
Soil ? 2:20 Al?C138?2:49 Si?C138?1:63 Ca?C138?1:94 Ti?C138?2:42 Fe?C138 ?4?
Smoke ? K?C138?0:6Fe?C138 ?5?
Organics ? 11 H?C138?0:25 S?C138?? ?6?
where: [ ] represents the elemental concentration, and a reconstructed
vs. measured PM
2.5
mass concentration correlation (r
2
) greater than
0.7 is required for accurate source apportionments made by PMF.
2.2.2. Source apportionment model: PMF
Elemental source ?ngerprints and their contributions to the total
PM
2.5
atthesamplingsitewereresolvedfromthemeasurementsbyap-
plying PMF (Paatero and Tapper, 1994). PMF is a receptor modeling
technique; whereby a multivariate statistical technique was applied to
the complete dataset. The receptor model is an independent technique
usingthemonitored aerosol chemical data only i.e.a priori information
aboutthesourcesfromthereconstructedchemicalmassinSection2.2.1
isnotrequired(Vianaetal.,2008).Itisbasedontheprinciplethatmass
and species conservation can be assumed (Hopke et al., 2006), and
wherekeyelementsineachfactorareusedtodeterminesourcepro?les
represented by that factor. In this application, the PMF program was
used (PaateroandTapper,1994).PMFsolvesthestandard bi-linearfac-
tor analysis model which can be speci?ed as (Paatero, 2010):
X ? GF ? E ?7?
whereXisamatrixofmeasuredelements,GandFarefactormatricesto
bedetermined,andEisamatrixofresiduals.Thiscanalsobewrittenas:
x
i;j
?
X
p
k?1
g
i;k
f
k;j
? e
i;j
?8?
Assuming n observations are available of m elements, then X is an n
by m matrix, i.e. x
i,j
represents the concentration of element j in the ith
sample. PMF then determined the two factor matrices G and F; if p
sourcesarecontributiontothemeasurements,Gisannbypmatrixcon-
tainingthecontributionfromeachsourcetoeachsample,andFisapby
m matrix containing the source ?ngerprints.
The matrices G and F are determined using an optimisation process
whichminimises thefunction Q,while the resolved factor elements re-
main non-negative:
Q ?
X
n
i?1
X
m
j?1
e
2
i;j
s
2
i;j
?9?
where s
i,j
is a speci?ed error of the form (Cohen et al., 2010):
s
i;j
? MDL
i;j
? Error
i;j
max x
i;j
C12
C12
C12
C12
; y
i;j
C12
C12
C12
C12
C16C17
?10?
where MDL
i,j
is the minimum detectable limit, Error
i,j
is the statistical
error, and y
i,j
is the ?tted value i.e. Y = GF.
2.2.3. Back trajectory analysis
The contribution of dust from inland Australia transported to the
monitoring site by west to north-westerly prevailing winds was re-
solved by applying the method outlined in Cohen et al. (2011). Fifteen
possible dust source regions were selected and each region was repre-
sented as a rectangular grid cell which best approximated the extent
ofthedesertregion.Thenameandpositionofeachgridcellisindicated
in Fig. 1b. and the grid size used to represent the desert regions about
their midpoints, is listed in Table 3 of Cohen et al. (2011). Next, for
days with outlier soil concentrations, 10-day hourly back trajectory
analyses were calculated using HYSPLIT (HYbrid Single-Particle La-
grangian Integrated Trajectory; Draxler et al., 2016; Stein et al., 2015)
to generate density maps and identify the number of grid intersection
points i.e. the relative impact of each source region to the dust ?nger-
print. Additionally, for all other sources, 10-day hourly back trajectory
analyses were conducted for outlier days, speci?cally where the source
concentrationexceededthemeanbytwostandarddeviations,andden-
sity maps were produced to identify source regions. Back trajectory
density maps were generated, for which the horizontal position of the
back trajectory (or trajectory endpoint) was determined every 30 min.
The region of interest was sub-divided into grid cells of 0.5? by 0.5? di-
mension, and if a back trajectory endpoint landed in the grid cell, a
3. Results
3.1. PM
2.5
chemical composition and mass closure
A statistical summary of the species measured in the atmospheric
PM
2.5
samplesandatmosphericmassconcentrationsduringthesampling
periodareprovidedinTable1.Atotalof360daysweresampled,however
threeoutliereventswereexcludedfromfurtheranalysisastheyexceeded
2.5 ?gm
?3
; the national 24-h air quality standard for PM
2.5
. The average
mass concentration of PM
2.5
is (3.3 ? 2.5) ?gm
?3
. This concentration is
relatively low however similar to the total mass concentration at the
?baseline? station at Cape Grim (5.7 ? 3.0) ?gm
?3
,re?ecting the study
site's remoteness. The dominant species of PM
2.5
were BC, sulfur, sodium
and hydrogen. A comparison of the concentrations of the remaining data
to those at Cape Grim (Crawford et al., 2017; Table 1), demonstrate that
concentrations of the soil related elements (Al, Si, Fe) and BC are higher
by a factor of 3.6, 2.9, 1.8 and 1.3, respectively, than the long term trends
of regional background atmospheric PM pollution.
ThePM reconstructedmassconcentrationswerecalculatedbyap-
Table 1
Mean, standard deviation (SD), maximum values and MDLs for elemental species
(ngm
?3
)andPM
2.5
(?gm
?3
)determinedoverJillabenanCavebetweenJuly2013andJuly
2017 compared with those at Cape Grim (from April 1998 to June 2016).
Species Jillabenan Cave (n = 357) Cape Grim (n = 1858)
Mean
(ng
m
?3
)
SD max MDLs
(ng
m
?3
)
Mean
(ng
m
?3
)
SD Max MDLs
(ng
m
?3
)
H 137 116 3614 1.97 85 94 1101 1.97
N 132 179 3237 31.46 144 236 2366 23.42
Na 178 88 942 27.75 1009 710 4866 7.21
Al 14.0 9.7 139.5 1.31 3.9 7.5 219 1.84
Si 38.6 27.6 398.9 0.77 13.0 23.3 598 0.96
P 1.2 0.4 8.2 0.64 3.4 77.0 2455 1.05
S 179 83.8 779 0.57 233 159 1401 0.80
Cl 54.0 48.6 1201 0.57 1520 1086 7251 0.98
K 26.8 24.7 791 0.44 39.8 30.7 836 0.43
Ca 6.5 3.8 35.1 0.42 36.5 24.4 236 0.49
Ti 0.90 0.67 10.3 0.31 0.66 1.64 61.2 0.28
V 0.12 0.06 0.6 0.25 0.89 2.21 46.2 0.39
Cr 0.20 0.14 5.46 0.18 0.14 0.31 4.69 0.33
436 C.V. Tadros et al. / Science of the Total Environment 630 (2018) 432?443
Mn 0.33 0.14 2.38 0.12 0.56 0.75 12.1 0.14
Fe 9.0 6.6 81.9 0.10 4.93 9.18 147 0.15
Co 0.14 0.06 1.03 0.12 0.16 0.23 5.53 0.07
Ni 0.30 0.36 5.95 0.08 0.36 1.15 30.0 0.12
Cu 0.29 0.08 1.27 0.11 0.20 0.42 13.7 0.12
Zn 0.59 0.30 7.33 0.14 0.88 12.4 531 0.09
Se 0.11 0.07 1.09 0.48 0.24 0.40 3.24 0.30
Br 1.74 0.73 10.5 0.34 2.56 2.48 20.4 0.26
Pb 0.43 0.27 3.37 0.85 0.61 1.07 10.8 0.49
BC 309 150 2510 19.07 241 167 2147 20.06
PM
2.5
mass concentrations (?gm
?3
)
Gravitational
mass
3.33 2.47 70.2 0.16 5.67 3.04 23.4 0.16
RCM 2.70 1.71 43.3 0.029 ?? ? ? ?
RCM (%) 90 11 133 ???
grid cell counter was incremented. Where results for cluster analysis are
presented, the back trajectories corresponding to the highest 20 samples
in summer and winter were clustered to investigate seasonal change in
source regions. For cluster formation we used the PAM (Partitioning
Around Medoids; Kaufman and Rousseeuw, 2005)program.
Todeterminethesensitivityoftheresultstothestartingbacktrajec-
tory height, two starting heights of 300 m and 500 m agl were com-
pared. The starting height resulted in minor differences to the back
trajectorypath;thereforetheresultsfromthe300maglstartingheight
are presented. Further, these heights were chosen to reduce topo-
graphiceffectsandtoensurethattheairmasseswerewithinthebound-
ary layer for a signi?cant proportion of time.
Table 2
Summary of PMF factors, percentage contribution of each factor to the total measured PM
2.5
ma
pro?les are summarised in Fig. A.2.
Factor % Mass
contribution
?
error
Dominant species Source; comments
Factor
1
44.9 ? 0.83 H, N, BC, P, Cu, Cr, Pb, Br, Zn,
V, Mn, Co and S
Automobile; This factor is composed
et al., 2002; Fujita et al., 2007), and
and P from diesel engine oil, Br and
wear and V in lubricating oils.
Factor
2
21.4 ? 0.87 S, Ni, P, Na, Cr, V, Co, N, Cu, Pb
and Zn
Secondary Sulfate;SO
2
from the combustion
?lters were fully neutralised and occurred
Also contains tracers from industrial
Factor
3
12.3 ? 0.58 K, Zn, Se, Br, BC Smoke; The main contributor to this
reduction burns (Cachier et al., 1991
(Zn, Se).
Factor
4
11.25 ? 0.45 Al, Fe, Ti, Si, Ca, Mn, Ni and Na Soil; contains the ?ve main elements
slightly above the range for alumina-silicates
Factor
5
10.14 ? 0.38 Cl, Na, Ca Sea Aged; [Na/Cl] ratio is 1.13, higher
continent (M?ller, 1990).
2.5
plying the mass equations in Section 2.2.1. A comparison between the
reconstructed and gravimetric PM
2.5
mass concentrations, shows good
agreement with a correlation coef?cient of r
2
=0.98(Fig. A.1). The
mean reconstructed mass yielded 90% of the measured PM
2.5
mass; y
= 0.711 ? 0.006 (Fig. A.1; Table 1) and is consistent with the ?ndings
of previous studies (Cohen et al., 2010). This mass de?cit is attributed
to the measurement process, since particle-bound water on the
Te?on-membrane ?lter and components related to nitrates were not
measured (Chow et al., 2015).
3.2. Identi?ed sources and source regions
Five main factors were identi?ed in the PMF analysis and thus ?ve
atmosphericsourcetypes weredeterminedbasedonthekeyelemental
compositionineachfactor.TheresultsaresummarisedinTable2.Based
on airborne PM
2.5
concentrations, the main sources of atmospheric
emissions detected at Yarrangobilly are automobiles (45%), secondary
sulfate (21%), smoke (12%), windblown soil (11%) and aged sea salt
(10%). Time series of the ?ngerprints are presented in Fig. 2,andFig. 3
shows the percentage of measured elements allocated to each ?nger-
print. A seasonal contribution from automobiles, secondary sulfate and
soil is evident but not for aged sea salt. A clear trend in smoke was ob-
served,withhigherconcentrationsinMarch,AprilandOctober.(Fig.2).
With the largest contribution to PM
2.5
, the automobile factor was
dominated by H and BC as well as Pb, S, P, Zn, Cu, V and Cr (factor 1:
Table 2; Fig. 3).The automobile aerosol timeseriesis strongly seasonal,
peaking in austral summer (Fig. 2a). Back trajectory cluster analyses
ss, dominant species within each source ?ngerprint and source name. Identi?ed source
mainly of H, a dominant BC component; indicative of vehicular emissions (Zhu
the elements: Pb from resuspension of historic leaded petrol (Kristensen, 2015), S
Mn from fuel additives, Cu and Cr from clutch and brake lining wear, Zn from tyre
of fossil fuels, e.g. coal-?red power stations. Most of the sulfate on the
as (NH
4
)
2
SO
4
, not as ammonium bisulfate (NH
4
)HSO
4
or sulfuric acid H
2
SO
4
.
heavy metals.
factor is K; a tracer for biomass burning, produced by bush?res and hazard
; Chow et al., 2015). It also includes the contribution of plant and/or soil elements
associated with soil dust, in addition to Mn, Ni and Na. The [Al/Si] ratio is 0.37,
(0.25?0.35; Cohen et al., 2010).
than seawater (0.86), suggesting a loss of Cl during transport from sea to
smo
437C.V. Tadros et al. / Science of the Total Environment 630 (2018) 432?443
Fig.2.Timeseriesofthecontribution(ngm
?3
)of:(a)automobile,(b)secondarysulfate,(c)
identi?ed the higher proportion of automobile aerosols in summer
compared to winter and this is attributed to increased visitor numbers
in this tourist area and different directions of the fetch region (Fig. 4a).
Secondary sulfate was the second largest contributor to PM
2.5
(21%;
factor 2: Table 2) with maximum concentrations in summer (Fig. 2b).
The back trajectory density map for the top 23 secondary sulfate mea-
surements show some of the back trajectories have passed over New
South Wales and Victorian urban regions (Fig. 4b).Furthermore,twoof
the three trajectory clusters for summer and winter had passed over
power stations located in NSW and Victoria (Fig. 4b) indicating the
in each source indicates the threshold (mean + 2?) for outlier events. The 2.5 month gap betw
Fig. 3. Percentage allocation of each
ke,(d)soil,and(e)agedseatothetotalairbornePM concentrations.Thereddashedline
dominance of secondary sulfate from coal derived pollution. However,
one trajectory cluster in summer and winter is localised and considering
a number of trace elements were present in the secondary sulfate ?nger-
print (Fig. 3; Table 2), this indicates a smaller contribution of secondary
sulfate from local industrial release (discussed further in Section 4.1).
Smoke (factor 3: Table 2) contributes 12% to the PM
2.5
mass. Potas-
sium is the major component of the smoke factor, 59% of the K was allo-
cated to this factor, with signi?cant allocations from Zn (35.4%), Se (27%),
Br (25%) and BC (20%; Fig. 3). Outlier events in the smoke time series
(Fig. 2c) correspond to signi?cant increases in regional ?re emissions
2.5
een January and mid-March 2015 indicates no available data.
element's mass to a source.
from bush?res, controlled hazard reduction burns and domestic wood
stove heating (labelled I-III respectively) and brought to site from north-
ern air masses (Fig. 4c), which we examine in further detail.
The highest concentration of smoke occurred on the 17 December
2014. Several ?res were ignited by lightning on 15 December 2014 in
central and north-eastern Victoria, inland of the Great Dividing Range
(BoM, 2017a). The four most signi?cant ?res were large scale (120 ha
to 6800 ha), between 154 and 286 km southwest of the study site. The
closeproximityofthese?resandintensityofthesmokesignalrecorded
at the study site, suggests the likely impact of these ?res.
Spring (SON) 2013, 2014 and 2015 were Australia's three warmest
springs on record (BoM, 2017b) characterised by above average tem-
peratures and coincided with below average winter?spring rainfall.
These conditions reduced soil (and vegetation) moisture which pro-
moted signi?cant bush?res in New South Wales, south Queensland
and Victoria, respectively. Elevated concentrations of smoke were re-
cordedfortheseperiods(Fig.2c),suggestingJillabenanCaveswasin?u-
enced by smoke from these events.
Based on ?re incident data obtained from the National Parks local
?re service, local hazard reduction burns correlated with the observed
elevated concentrations of smoke in May 2014, April 2015 and 2016
and March to May 2017. Elevated smoke levels were also recorded for
April2014 andMarch2015andarelikelytobeduetohazardreduction
burns further a?eld in NSW and Victoria; as autumn (MAM) is the
prime hazard reduction period for NSW and Victoria. The winter July
2014 outlier event is attributed to wood ?re smoke from domestic
heating in the local vicinity; due to the passage of an unusual vigorous
cold front producing snow as low as 600 m across the NSW tablelands
(BoM, 2017a). Overall, higher concentration trends in March and April
are due to hazard reduction burns, whereas higher concentrations in
spring are attributed to bush?res.
Soil dust (11%; factor 4: Table 2) contains the ?ve key crustal ele-
mentsAl,Si,Ca,TiandFe,aswellasNaandNi(Table2;Fig.3).Theden-
sitymapofoutliereventsshowsthatthehighercellcrossingsofthesoil
factor are associated with winds mainly from the north-west quadrant
(Fig. 4d). Similar to Cohen et al. (2011) long-range dust transport
from 21 outlier events to the study site from 14 desert regions and the
Riverina area were investigated (Fig. 5). As indicated by the dots in
Fig. 5, the main dust source regions with the highest intersections
weretheRiverina(box15)andLakeMungo(box1)intheMurrayDar-
lingBasinandLakeWindaunka(box2),EastFlindersranges(box3)and
Olympic Dam(box4)within theLake EyreBasin.Soilispicked upfrom
eco
and
438 C.V. Tadros et al. / Science of the Total Environment 630 (2018) 432?443
Fig.4.Backtrajectorydensitymapsatthestudylocationforoutlierevents:(a)vehicle(b)s
sulfate,themeansummerandwintertrajectoryclusteringresultsareindicatedinredcircles
by back trajectories and the study site location is denoted as a solid triangle. The marginal grap
ndarysulfate,(c)smoke,(d)soil,and(e)seasaltemissions.Forautomobileandsecondary
whitesquares,respectively.Ineachmap,thepixelsrepresentthenumberofcellcrossings
hics of each image show the latitude and longitude.
the
439C.V. Tadros et al. / Science of the Total Environment 630 (2018) 432?443
these regions and crosses over the Riverina before arriving at our site.
This region is one of the major sources of Australian dust and the loca-
tion of dry terrestrial salt lakes (Prospero et al., 2002). The observed
Na in soil dust is attributed to terrestrial salt input from this source re-
gion. The time series (Fig. 2d) shows a seasonal cycle with a late-
spring to summer maximum, however there was a particularly high
concentration of soil-derived particles recorded in October 2015, coin-
ciding with Australia's warmest October on record, which was driven
by a strong El Ni?o in the Paci?c. Across the entire southern half of
Australia, including the SE and arid interior of Australia, strong wind
conditions, above average temperatures and dry conditions; below-
Fig. 5. Map of Australia showing back trajectory crossings over the dust source regions for
longitude of each rectangular dust source region.
average rainfall and soil moisture, were recorded (BoM, 2017c). The
presence of Mn and Ni is attributed to entrainment from local industry
and agricultural regions in the Riverina, respectively, evidenced by
back trajectory air mass passage over these locations.
Aged sea salt (factor 5; Table 2) has the lowest contribution to the
PM
2.5
mass at 10%, this is expected due to the remote inland location.
This is in contrast to the dominate source contribution of 57% at Cape
Grim,due tothecoastallocation.Timeseriesof theaged sea saltsource
is variable with no obvious seasonal trend (Fig. 2e). The density map of
outlier events shows a dominant westerly component (Fig. 4e) consis-
tent with the mean westerly wind direction for the study site (Fig. 1c).
Moreover, for these outlier events the back trajectory density map
shows that winds originating from the Southern Ocean had passed
over inland saline regions of Lake Eyre before arriving at our site
(Fig. 4e). Although the sea salt source is predominantly comprised of
Cl, with 100% allocated to this source, 22% of the Ca and 33% of the Na
was also allocated to this source. Moreover 15% of the Na originates
from soil (Fig. 3), and as identi?ed in the soil source, while over land,
Ca and Na from inland salt lakes are incorporated in the air mass.
4. Discussion
4.1. Aerosol sources
By applying the PMF model with the aerosol elemental data set as
input, ?ve main atmospheric sources were identi?ed at Yarrangobilly:
traf?c emissions from automobiles, secondary sulfate from coal com-
bustion, smoke from biomass burning, mineral dust from windblown
soil and aged sea salt. The distinctive chemical characteristics of each
source pro?le were evaluated and the percentage of each element allo-
cated to the different sources was estimated. Seasonality in the sources
wasidenti?ed;ahighercontributionofsecondarysulfatesandautomo-
bile emissions in summer, and dust and smoke aerosols in spring dom-
inated. Moreover, in addition to local contributions, long-range
contributions from distant sources were also determined by examining
extreme events using back trajectories.
The predominant source of PM
2.5
is traf?c-generated emissions (45%;
Table 2). Based on a quantitative meta-analysis of published literature,
Zhou and Levy, 2007 concluded maximum concentrations of ?ne and ul-
high soil events (n = 21; outlier events in Fig. 2d). See Fig. 1b for the name, latitude and
tra?ne particles from automobile emissions are observed between 100
and400mfromtheroadway.Asouraerosolmonitoringsamplerisbe-
tween 30 and 500 m from the local road network (Fig. 1a), the observed
vehicle emissions are attributed to contributions from these access roads.
The Yarrangobilly Caves entry and exit roads along the Snowy Mountains
Highway are situated ~2300 m and at an elevation of 1287 m a.s.l. from
the aerosol unit and the impact from vehicle emissionsfrom the highway
are considered to be limited. The highest contributions of automobile
emissionsoccurredinsummer(Fig.2a),consistentwithseasonalsummer
andwintervisitorvolumedata(2013?2017);wheremorevehiclesaccess
the tourist caves in summer than in winter. The Snowy Mountains High-
way and access roads to the caves are located north-east and south-east
to the sampling site (Fig. 1a). Based on trajectory clustering results
(Fig. 4a); in summer, easterly winds dominate, which pick up vehicle
emissions from the road network and transport them toward the sam-
plingsite,whereasthewinddirectionchangestoamorewesterlycompo-
nentinwinterwhichmayalsocontributetothedropinvehicleemissions
due to the lower density of roads from this direction.
Accountingfor21%,secondarysulfates isthe secondlargestcontrib-
utor of the total PM
2.5
. The two main sources of secondary aerosols are
ascribed to coal combustion burning and local industry. During the pe-
riod of this study, ?ve coal ?red power stations were operational in
New South Wales: Vales Point and Eraring situated on the southern
shores of Lake Macquarie, Bayswater and Liddell located in the Upper
HunterregionofNSWandMt.PiperlocatednearLithgowintheCentral
Westof New South Wales.In thestate of Victoria, four coal ?red power
stations: Loy Yang A, Loy Yang B, Yallourn and Hazelwood power sta-
tions located in the Latrobe Valley of Victoria, were also operational.
440 C.V. Tadros et al. / Science of the Total Environment 630 (2018) 432?443
The back trajectory density map for secondary sulfate measurements
show that some of the back trajectories had passed over the coal-?red
power plants in New South Wales and some (but a lower number)
over the coal-?red power plants in Victoria (Fig. 4b).
Furthermore, results obtained from the trajectory clustering also
show air masses bringing secondary sulfate from the coal-?red power
plants in NSW and Victoria in both summer (Fig. 4b; red circles) and
winter (Fig. 4b; white squares). This con?rms secondary sulfate is pri-
marily coal-derived pollution; however a number of trace elements
were associated with the secondary sulfate ?ngerprint (Fig. 3;
Table 2). Coal burning does produce a number of trace metals (e.g.
Mn,P,Ti,Zn,etc.;Nalbandian, 2012), yet observations from the trajec-
tory clusters are consistent with local industry contributing secondary
sulfate,andmetalemissions.BasedonadatabasesearchoftheNational
Pollution Inventory (NPI, 2017), Hanson Construction Materials at
Williamsdale, located at about 65 km to the north east of the site and
Tumbarumba Structural Softwood Sawmill, located 48 km south west
of the site would have contributed to the SO
2
and metal emissions
(e.g. Pb, Ni, Zn). Secondary aerosol contributions varied signi?cantly
seasonally and maximum concentrations were observed in summer
(Fig. 2b). This seasonality is supported by the fact that in summer, the
combination of pollutants and high levels of ultraviolet (UV) light
causestrongphotochemical reactions. For example, ammonium sulfate
((NH
4
)
2
SO
4
) aerosol is formed when gaseous sulfur dioxide (SO
2
)is
oxidised to sulfuric acid (H
2
SO
4
) by UV light and then neutralised by
ammonia (NH
3
) in the atmosphere (Arrowsmith and Hedley, 1975).
Smoke emissions contributed 12% to the total PM
2.5
mass (Table 2).
Compared to the reported results from the remote site at Cape Grim,
this source contribution is similar in magnitude; where smoke accounts
for 13% of the total PM
2.5
(Crawford et al., 2017). At Yarrangobilly,
smoke has higher contributions in autumn and October (Fig. 2c), consis-
tent with local emission sources from hazard reduction burns and bush-
?res, respectively. The back trajectory density map for smoke (Fig. 4c)
showed elevated concentrations of smoke were also attributed to long-
range transport from intense bush?re events in NSW, south Queensland
and Victoria. This was similar to the observations of Crawford et al.,
2017, which showed smoke transported to Cape Grim originated from
both local and long-range contribution from the Australian mainland.
Due to the remote inland location in a temperate forest, these results
are in accord with expectations for the mid-latitudes, since most of the
Australian biomass burning emissions are from the savannah regions of
tropical Australia (Kasischke and Penner, 2004).
The SnowyMountainsinsouth-easternAustraliareceives terrestrial
soil-derivedaerosolsinspringfromlong-rangetransportfromtheMur-
rayDarlingBasin and LakeEyre (Section 3.2). This is supportedbyback
trajectory analysis of outlier days that showed that air masses travelled
oversaltsourcesintheregionbeforedeliverytosite(Fig.5).Thecontri-
bution of Na to the source ?ngerprint (Table 2; Fig. 3)providedfurther
evidencethatterrestrialNafromthezoneofsaltlakesinthenorth-west
is co-transported with terrestrial soil from this region. The input of ter-
restrial salt from this source region has also been detected at other re-
mote inland areas, for example Wagga Wagga situated 118 km NW of
ourstudysiteandCobar(Shiga et al., 2011).Inaddition,localemissions
of metals e.g. Ni and Mn can be entrained and then adhered to aerosols
as the air masses pass over agriculture and industrial regions, as has
been inferred in previous studies (Marx et al., 2008).
AustraliaisthedominantsourceofdustintheSouthernHemisphere
with substantial dust mobilised from Lake Eyre and the Great Artesian
Basinstartinginspringandreachingamaximumintheaustralsummer
(Prospero et al., 2002; Mitchell et al., 2010; Shao et al., 2011). The high
soil events were evident in the four-year time series as a late spring to
summer maximum, the impact of which generated a strong deposi-
tionalsignalduringthe2015?16ElNi?othatcauseddroughtconditions
(Fig.2d).Thisseasonalityisconsistentwithmaximumdustactivitydoc-
umented from the region. There is clear evidence that drought condi-
tions can alter the landscape and precipitate the transition to an
erosional regime; enhancing the production and mobilisation of large
quantities of dust from desert surfaces (Prospero and Lamb, 2003), be-
cause of this synergy, the increased dust measurements during the
2015?16 El Ni?o indicates the region is sensitive temporally and spa-
tially to the impact of a changing climate.
The contribution of sea salt aerosols (10%) is comparable to other
natural aerosols in the Snowy Mountains. The results indicated marine
aerosols traversed a signi?cant distance from the Southern Ocean to
the high altitude interior part of south-eastern Australia (Fig. 4e). The
sourceofClisseasaltaerosols(Fig. 3)andchloridedepletionduringin-
land transport was identi?ed (Table 2). The process of chloride deple-
tion is a common atmospheric reaction of far-travelled marine
aerosols, where Cl in the sea salt aerosols reacts with an anthropogenic
source of nitric and sulfuric acid, forming gaseous HCl (Ten Harkel,
1997). Considering the study site has high concentrations of N and S
mainly from anthropogenic sources, when sea-salt aerosols are
transported to site, chloride depletion reactions explain the removal of
Cl in the atmosphere.
4.2. Overall implications
Other than the work of Suni et al., 2008 and Paton-Walsh et al., 2014,
there is: sparsity of data on atmospheric emissions deposited in an
Australian alpine forest, and no previous high-resolution data from the
Snowy Mountains. To ?ll this gap in current knowledge, the average per-
centage PM
2.5
load to the atmosphere and the average elemental compo-
sition in the remote high alpine air over a four-year timeframe are
quanti?edforthe?rsttime.Suchresultsmaybeinformativeinevaluating
the extent to which atmospheric aerosols are impacting the natural re-
gional climate (Rosenfeld, 2000; Nicholls, 2005) and subsequent impacts
on the distribution and diversity of species in these alpine environments
(Green and Pickering, 2002; Pickering et al., 2008; Slatyer, 2010).
Automobileandsecondarysulfateemissionsrepresent70%oftheat-
mospheric loading and therefore the primary mode of metal contami-
nation from the atmosphere. A study from the Snowy Mountains
demonstrated that theregion's peat bogs haveaccumulated substantial
quantities of anthropogenic metals since industrialisation, approxi-
mately 150 years ago (Marx et al., 2010).The sources were ascribed to
mining and smelting, coal combustion and agriculture. The monitoring
data from this study quanti?es the anthropogenic PM
2.5
source from
the Snowy Mountains, and are in agreement with the postulated ?nd-
ings of Marx et al., 2010.
This study has demonstrated that longrange transport of anthropo-
genic tracemetalpollutants can be transported toremote regionsasfar
as south-eastern Australia, and has also been demonstrated to be
transported to Cape Grim, Tasmania (Crawford et al., 2017) and as far
as New Zealand (Hesse, 1994; Marx et al., 2008). Furthermore, this
study provides evidence that long range transport of loess and salt
from theinterior to thealpinezone of south-easternAustralia isdepos-
ited annually during the Austral autumn-summer, with increased
amounts during the drought and El Ni?o periods. The soil properties
of the alpine environment in south-eastern Australia are known to be
in?uenced by aeoliandust deposits (Costin,1954; Chen et al., 2002)al-
though at a lower rate compared to earlier times. For example, during
glacial times the rates of dust deposition was concluded to be 1.5 to 3
timesthoseatpresent(Hesse,1994),whilethecurrentratesofdustde-
positionareestimated tobe 1?7mmperthousandyears onthebasis of
dust extracted from drift wood in the Snow Mountains (Walker and
Costin, 1971). Aeolian dust have high clay content (Butler, 1956),
hence dust accession may modify the soil pro?le by imparting proper-
ties different to the underlying parent material.
4.3. Implications for palaeo studies
From a palaeo context, the potential importance of natural aerosol
deposition in a pre-automobile and pre-industrial period is apparent.
of dustentrainment in Australiaduring late (austral) springto summer
caused maximum soil emissions at the study site. No trends were ob-
served for aged sea salt.
It was concluded that long-distance PM
2.5
transport was an impor-
tant mechanism for particle transport from pointsources tothe remote
high alpine site. Speci?cally, the second major source of PM
2.5
,second-
ary aerosols from coal combustion was transported from regional
NSWandVictorianpowerstations.Also,smokefrombush?resoriginat-
ing in the major cities was transported to site. Prevailing westerly air
currents carried aeolian Na and soil from the inland saline regions of
Lake Eyre and Murray Darling basins and aged sea salt from the South-
ern Ocean to the south-east. Furthermore, due to the long transport
time over land, we identi?ed chloride depletion in sea salt aerosols. In
addition, we identi?ed that trace metals are entrained by the wind
and transported to the study site, as evidenced by metal enrichment
of the secondary sulfate and soil factor through the passage over local
industry as well as the agricultural region in the Riverina.
Moreover, we identi?ed that the current climate in particular
drought and El Ni?o periods increase the smoke and soil aerosol load-
ing, suggesting they may serve as proxies for palaeo-studies. We also
postulated that the introduction of water-soluble salts and aeolian
441C.V. Tadros et al. / Science of the Total Environment 630 (2018) 432?443
Basedonmeasurementspresentedhere,theSnowyMountainsreceives
?neparticlesofsoilandsmokethataretransportedbyaeolianprocesses
fromdistantregionsandtohighaltitudes.SimilartoMitchelletal.,2010
we identi?ed that during a dry El Ni?o event a signi?cant increase in
aerosol loading within the dust season from Lake Eyre was observed.
Additionally enhanced smoke concentrations from bush?re activity
driven by increased spring temperatures was also observed
(Section 3.2) and demonstrated to be transported long distances to re-
gionsasfarassouth-easternAustralia.Inthisregard,traceelementsde-
rived from wind-blown soil and smoke aerosols could serve as proxy
data to reconstruct past episodes of aridity in sedimentary records.
Bush?res yield smoke emissions of K, Zn, Se, Br and BC (Fig. 3)and
these aerosol particles are water-soluble (Yamasoe et al., 2000). Fine
windblown soil is composed mainly of Al, Si, Ca, Ti and Fe, as well as
Mn,NiandNa(Fig.3)andarederivedfromamixtureofmineralsofvary-
ing solubilities e.g. aluminosilicate clays, metal oxides, carbonates, gyp-
sum and halite salts (Costin et al., 1952; Costin, 1955). However
heterogeneous reactions of soil dust particles with reactive gasses includ-
ing nitric, hydrochloric, and sulfuric acids can increase their hygroscopic-
ity through the conversion to more soluble compounds (Sullivan et al.,
2009). The deposited water soluble trace element fraction of the smoke
and soil dust in the soil zone are therefore labile and available for biolog-
ical uptake, can undergo chemical processes in the soil zone, or
transported by percolating rainwater through the subsoil and bedrock
to cave systems, or discharged to lakes, reservoirs and river systems.
The notion that dust and smoke can in?uence thephysical and geochem-
icalpropertiesofsoilsandsedimentsisimportantasitmaycomplicatein-
terpretation in proxy records particularly if aerosol deposition is the
dominant source of trace element input rather than the local bedrock
(Chadwick et al., 1999), or when there is an increased rate of aeolian de-
position due to an increased frequency of events. This is particularly rele-
vant in regions like the Snowy Mountains where changes in the climate
are forecasted to decrease precipitation and increase the frequency of
?res (Nicholls, 2005).
Constraining atmospheric aerosol sources provides a unique oppor-
tunity to accurately shape our understanding of climate and environ-
mental conditions controlling aerosol inputs to soil and water bodies,
enabling accurate palaeo-environmental interpretations to be made
from trace elements preserved in natural archives regionally. Analysis
of aerosols highlights that there are a range of potential processes that
may affect the palaeo-environmental interpretation of trace elements
in sedimentary archives such as lake sediments, peat deposits and
speleothems.Anunderstandingofthesubsequentprocessesandmech-
anism(s) leading to the delivery and preservation of trace elements to
the sedimentary archive in the palaeo record is also required. In the
caseofspeleothemarchiveswewillbeaddressingthisinaforthcoming
paper, where the atmospheric dataset from this study is applied via a
mass balance, to quantify the contribution of aerosols to a cave drip
water dataset that was collected concurrently at the same site.
5. Conclusions
This paper presents the ?rst long term data set at Yarrangobilly in
the Snowy Mountains, SE Australia, providing a detailed quantitative
analysisof PM
2.5
duringthefour-yearmonitoringperiod.This includes:
a statistical summary of mean PM
2.5
mass concentrations for aerosol
species, temporal variations and elemental percentage allocation to
each measured source. Five sources, representing: automobile, second-
ary sulfate, smoke, soil and aged sea salt, describe the contributors to
?ne particles in the alpine air of the Snowy Mountains region.
Distinctiveseasonalityoccurredforthedifferentsources.Weidenti-
?edautomobileemissionsincreasedoversummerduetotheincreasein
visitors, where access points around the remote site is by vehicles. The
secondary sulfate maxima during summer is caused by photochemical
reactionsandincreasedsmokeemissionsinautumnandspringisattrib-
utedtotheseasonforhazardreductionburnsandbush?res.Seasonality
clayto the local soil may gradually alter thesoil physicaland geochem-
ical pro?le. Hence an understanding of ?ne particle measurements and
soilprocessesisrequiredforaccuratepalaeo-environmentalinterpreta-
tions in natural archives.
6. Acknowledgements
TheauthorsgratefullyacknowledgethesupportofthestaffattheCen-
tre for Accelerator Science, ANSTO for access to the ion beam analysis fa-
cilities. The NOAA Air Resources Laboratory (ARL) made available the
HYSPLIT transport and dispersion model and the relevant input ?les for
generation of back trajectories used in this paper. George Bradford and
thestaffatYarrangobillyCavesandNSWNPWSarethankedfortheirded-
ication and on-going ?eld support and access permission. We also thank
NPWS Southern Ranges senior ?re ranger Andrew Grant for providing
?re and controlled burn data. We thank Jianmen Chen and two anony-
mous reviewers for their constructive reviews that improved the
manuscript.
Appendix A. Appendices
Fig. A.1. Linear regression analysis between reconstructed vs. gravimetric mass for 357
samples collected at Jillabenan Cave, NSW, Australia, between 2013 and 2017.
442 C.V. Tadros et al. / Science of the Total Environment 630 (2018) 432?443
References
Allan, M., Fagel, N., Van Rampelbergh, M., Baldini, J., Riotte, J., Cheng, H., Edwards, R.L.,
Gillikin, D., Quinif, Y., Verheyden, S., 2015. Lead concentrations and isotope ratios in
speleothems as proxies for atmospheric metal pollution since the industrial revolu-
tion. Chem. Geol. 401, 140?150.
Arrowsmith, A., Hedley, A.B., 1975. The formation of ammonium sulphate particles by at-
mospheric reactions. Rev. Pert. Quim 17, 26.
Bird, J.R., 1990. Total analysis by IBA. Nucl. Instrum. Methods Phys. Res., Sect. B 45 (1?4),
514?518.
Bird, J.R., Williams, J.S., 1989. Ion Beams for Material Analysis. Academic Press, p. 719.
BoM, 2017a. http://www.bom.gov.au/climate/mwr/ (last access: 30 November 2017).
BoM, 2017b. http://www.bom.gov.au/climate/current/annual/aus/ (last access: 30 No-
vember 2017).
BoM, 2017c. http://www.bom.gov.au/climate/current/statements/scs52.pdf (last access:
30 November 2017).
Butler, B.E., 1956. Parna-an aeolian clay. Aust. J. Sci. 18 (5), 145?151.
Cachier, H., Ducret, J., Bremond, M.P., Yoboue, V., Lacaux, J.P., Gaudichet, A., Baudet, J.,
1991. Biomass burning aerosols in a savanna region of the Ivory Coast. In: Levine, J.
(Ed.), Global Biomass Burning: Atmospheric, Climatic and Biospheric Implications.
MIT Press, Cambridge, MA, pp. 174?180.
Chadwick, O.A., Derry, L.A., Vitousek, P.M., Huebert, B.J., Hedin, L.O., 1999. Changing
sources of nutrients during four million years of ecosystem development. Nature
397 (6719), 491?497.
Chen, X.Y., Spooner, N.A., Olley, J.M., Questiaux, D.G., 2002. Addition of aeolian dusts to
soils in southeastern Australia: red silty clay trapped in dunes bordering Murrum-
bidgee River in the Wagga Wagga region. Catena 47 (1):1?27. https://doi.org/
10.1016/S0341-8162(01)00176-X.
Fig. A.2. Source pro?les for the identi?
Chow, J.C., Lowenthal, D.H., Chen, L.W.A., A, L.-W., Wang, X., Watson, J.G., 2015. Mass re-
construction methods for PM
2.5
: a review. Air Qual. Atmos. Health 8 (3), 243?263.
Cohen, D.D., 1998. Characterisation of atmospheric ?ne particles using IBA techniques.
Nucl. Instrum. Methods Phys. Res., Sect. B 136:14?22. https://doi.org/10.1016/
S0168-583X(97)00658-7.
Cohen, D.D., Bailey, G.M., Kondepudi, R., 1996. Elemental analysis by PIXE and other IBA
techniques and their application to source ?ngerprinting of atmospheric ?ne particle
pollution. Nucl. Instrum. Methods Phys. Res., Sect. B 109, 218?226.
Cohen, D.D., Stelcer, E., Hawas, O., Garton, D., 2004a. IBA methods for characterisation of
?ne particulate atmospheric pollution: a local, regional and global research problem.
Nucl. Instrum. Methods Phys. Res., Sect. B 219, 145?152.
Cohen, D.D., Garton, D., Stelcer, E., Hawas, O., 2004b. Accelerator based studies of atmo-
spheric pollution processes. Radiat. Phys. Chem. 71 (3?4), 759?767.
Cohen, D.D., Crawford, J., Stelcer, E., Bac, V.T., 2010. Characterization and source appor-
tionment of ?ne particulate sources at Hanoi from 2001 to 2008. Atmos. Environ.
44 (3), 320?328.
Cohen,D.D., Stelcer,E.,Garton,D., Crawford,J.,2011.Fine particlecharacterization,source
apportionment and long-range dust transport into the Sydney Basin: a long term
study between 1998 and 2009. Atmos. Pollut. Res. 2 (2), 182?189.
Colbeck, I., Lazaridis, M.,2014.AerosolScience:TechnologyandApplication.JohnWiley&
Sons Ltd., West Sussex, UK, p. 496.
Costin, A.B., 1954. A Study of the Ecosystems of the Monaro Region of NSW. Government
Printer, Sydney.
Costin, A.B., 1955. Alpine soils in Australia with reference to conditions in Europe and
New Zealand. Eur. J. Soil Sci. 6 (1):35?50. https://doi.org/10.1111/j.1365-2389.1955.
tb00828.x.
Costin, A.B., Hallsworth,E.G., Woof,M., 1952.Studies inpedogenesisinNewSouthWales.
III. The alpine humus soils. Eur. J. Soil Sci. 3 (2):190?218. https://doi.org/10.1111/
j.1365-2389.1952.tb00643.x.
ed sources from the PMF model.
443tal Environment 630 (2018) 432?443
Crawford, J., Cohen, D.D., Stelcer, E., Atanacio, A.J., 2017. Long term ?ne aerosols at the
Cape Grim global baseline station: 1998 to 2016. Atmos. Environ. 166, 34?46.
Doney, S.C., Mahowald, N., Lima, I., Feely, R.A., Mackenzie, F.T., Lamarque, J.-F., Rasch, P.J.,
2007. Impact of anthropogenic atmospheric nitrogen and sulfur deposition on ocean
acidi?cation and the inorganic carbon system. Proc. Natl. Acad. Sci. U. S. A. 104 (37),
14580?14585.
Draxler, R., Stunder, B., Rolph, G., Stein, A., Taylor, A., 2016. HYSPLIT4 USER's GUIDE Ver-
sion 4 - Last Revision:February 2016. Retrievedfrom. https://www.arl.noaa.gov/doc-
uments/reports/hysplit_user_guide.pdf.
Dredge,J.,Fairchild,I.J.,Harrison,R.M.,Fernandez-Cortes,A.,Sanchez-Moral,S.,Jurado,V.,
Gunn, J., Smith, A., Sp?tl, C., Mattey, D., Wynn, P.M., Grassineau, N., 2013. Cave aerosols:
distribution and contribution to speleothemgeochemistry.Quat.Sci.Rev.63,23?41.
Frappier, A.B., 2006. Empirical Orthogonal Function Analysis of Multivariate Stalagmite
TraceElementData:Detectingthe1982ElChich?nVolcanicEruption,ArchivesofCli-
mate Change in Karst. 10. Karst Waters Institute Special Publication, pp. 113?115.
Frisia, S., Borsato, A., Fairchild, I.J., Susini, J., 2005. Variations in atmospheric sulphate re-
corded in stalagmites by synchrotron micro-XRF and XANES analyses. Earth Planet.
Sci. Lett. 235 (3?4), 729?740.
Fujita, E.M., Campbell, D.E., Arnott, W.P., Chow, J.C., Zielinska, B., 2007. Evaluations of the
chemical mass balance method for determining contributions of gasoline and diesel
exhaust to ambient carbonaceous aerosols. J. Air Waste Manage. Assoc. 57 (6),
721?740.
Green, K., Pickering, C., 2002. A Scenario for mammal and bird diversity in the Snowy
Mountains of Australia in relation to climate change. Mountain Biodversity: A Global
Assessment. Parthenon, New York and London, pp. 239?247.
Hesse, P.P., 1994. The record of continental dust from Australia in Tasman Sea sediments.
Quat. Sci. Rev. 13 (3), 257?272.
Hesse, P.P., McTainsh, G.H., 2003. Australian dust deposits: modern processes and the
Quaternary record. Quat. Sci. Rev. 22 (18?19), 2007?2035.
Hopke, P.K., Ito, K., Mar, T., Christensen, W.F., Eatough, D.J., Henry, R.C., Kim, E., Laden, F.,
Lall, R., Larson, T.V., Liu, H., Neas, L., Pinto, J., St?lzel, M., Suh, H., Pattero, P., Thurston,
G.D.,2006.PMsourceapportionmentandhealtheffects:1.Intercomparisonofsource
apportionment results. Journal of Exposure Science and Environmental Epidemiol-
ogy. 16 (3), 275?286.
Jickells, T., 1995. Atmospheric inputs of metals and nutrients to the oceans: their magni-
tude and effects. Mar. Chem. 48 (3?4), 199?214.
Kasischke, E.S., Penner, J.E., 2004. Improving global estimates of atmospheric emissions
from biomass burning. J. Geophys. Res. Atmos. 109:D14. https://doi.org/10.1029/
2004JD004972.
Kaufman, L., Rousseeuw, P.J., 2005. Finding Groups in Data: an Introduction to Cluster
Analysis. John Wiley & Sons, Inc., New Jersey, p. 368.
Kohfeld, K.E., Harrison, S.P., 2001. DIRTMAP: the geological record of dust. Earth Sci. Rev.
54 (1?3), 81?114.
Kristensen, L.J., 2015. Quanti?cation of atmospheric lead emissions from 70 years of
leaded petrol consumption in Australia. Atmos. Environ. 111, 195?201.
Likens, G.E., 2010. The role of science in decision making: does evidence-based science
drive environmental policy? Front. Ecol. Environ. 8 (6):e1?e9. https://doi.org/
10.1890/090132.
Mahowald,N.,2011.Aerosolindirecteffectonbiogeochemicalcyclesandclimate.Science
334 (6057), 794?796.
Malm, W.C., Sisler, J.F., Huffman, D., Eldred, R.A., Cahill, T.A., 1994. Spatial and seasonal
trends in particle concentration and optical extinction in the United States.
J. Geophys. Res.-Atmos. 99 (D1), 1347?1370.
Marx, S.K., Kamber, B.S., McGowan, H.A., 2008. Scavenging of atmospheric trace metal
pollutants by mineral dusts: inter-regional transport of Australian trace metal pollu-
tion to New Zealand. Atmos. Environ. 42 (10), 2460?2478.
Marx,S.K.,Kamber,B.S.,McGowan,H.A.,Zawadzki,A.,2010.Atmosphericpollutantsinal-
pine peat bogs record a detailed chronology of industrial and agricultural develop-
ment on the Australian continent. Environ. Pollut. 158 (5):1615?1628. https://doi.
org/10.1016/j.envpol.2009.12.009.
Marx, S.K., Kamber, B.S., McGowan, H.A., Denholm, J., 2011. Holocene dust deposition
rates in Australia's Murray-Darling Basin record the interplay between aridity and
the position of the mid-latitude westerlies. Quat. Sci. Rev. 30 (23?24):3290?3305.
https://doi.org/10.1016/j.quascirev.2011.07.015.
Mitchell, R.M., Campbell, S.K., Qin, Y., 2010. Recent increase in aerosol loading over the
Australian arid zone. Atmos. Chem. Phys. 10 (4):1689?1699. https://doi.org/
10.5194/acp-10-1689-2010.
M?ller, D., 1990. The Na/Cl ratio in rainwater and the seasalt chloride cycle. Tellus B. 42
(3), 254?262.
Nalbandian, H., 2012. Trace Element Emissions From Coal. Report CCC/203. IEA Coal Re-
search, London, UK 89 pp.
Nicholls, N., 2005. Climate variability, climate change and the Australian snow season.
Aust. Meteorol. Mag. 54 (3), 177?185.
NPI, 2017. http://www.npi.gov.au (last access: October 2017).
Paatero,P.,2010.User'sGuideforPositiveMatrixFactorizationProgramsPMF2andPMF3,
Part 1: Tutorial. University of Helsinki, Finland.
Paatero, P., Tapper, U., 1994. Positive matrix factorization: a non-negative factor model
with optimal utilization of error estimates of data values. Environmetrics 5 (2):
111?126. https://doi.org/10.1002/env.3170050203.
Paton-Walsh, C., Smith, T.E.L., Young, E.L., Grif?th, D.W., Gu?rette, ?.-A., 2014. New emission
factors for Australian vegetation ?res measured using open-path Fourier transform in-
frared spectroscopy ? part 1: methods and Australian temperate forest ?res. Atmos.
Chem. Phys. 14 (20):11313?11333. https://doi.org/10.5194/acp-14-11313-2014.
Pickering, C., Hill, W., Green, K., 2008. Vascular plant diversity and climate change in the
alpine zone of the Snowy Mountains, Australia. Biodivers. Conserv. 17 (7),
1627?1644.
Pope III, C.A., Ezzati,M., Dockery, D.W.,2009.Fine-particulate air pollution and life expec-
tancy in the United States. N. Engl. J. Med. 360 (4), 376?386.
P?schl, U., 2005. Atmospheric aerosols: composition, transformation, climate and health
effects. Angew. Chem. Int. Ed. 44 (46):7520?7540. https://doi.org/10.1002/
anie.200501122.
Prospero, J.M., Lamb, P.J., 2003. African droughts and dust transport to the Caribbean: cli-
mate change implications. Science 302 (5647), 1024?1027.
Prospero,J.M.,Ginoux,P.,Torres,O.,Nicholson,S.E.,Gill,T.E.,2002.Environmentalcharac-
terization of global sources of atmospheric soil dust identi?ed with the Nimbus 7
Total Ozone Mapping Spectrometer (TOMS) absorbing aerosol product. Rev.
Geophys. 40 (1):1002. https://doi.org/10.1029/2000RG000095.
Rea, D.K., 1994. The paleoclimatic record provided by eolian deposition in the deep sea:
the geologic history of wind. Rev. Geophys. 32 (2), 159?195.
Ridley, H.E., Asmerom, Y., Baldini, J.U.L., Breitenbach, S.F.M., Aquino, V.V., Prufer, K.M.,
Culleton, B.J., Polyak, V., Lechleitner, F.A., Kennett, D.J., Zhang, M., Marwan, N.,
Macpherson, C.G., Baldini, L.M., Xiao, T., Peterkin, J.L., Awe, J., Haug, G.H., 2015. Aero-
sol forcing of the position of the intertropical convergence zone since AD 1550. Nat.
Geosci. 8 (3):195?200. https://doi.org/10.1038/ngeo2353.
Rosenfeld, D., 2000. Suppression of rain and snow by urban and industrial air pollution.
Science 287 (5459):1793?1796. https://doi.org/10.1126/science.287.5459.1793.
Rotstayn,L.D., Keywood,M.D.,Forgan,B.W., Gabric,A.J., Galbally,I.E., Gras, J.L.,Luhar, A.K.,
McTainsh, G.H., Mitchell, R.M., Young, S.A., 2009. Possible impacts of anthropogenic
and naturalaerosols on Australian climate: a review. Int. J. Climatol. 29 (4), 461?479.
Rutlidge,H., Baker, A., Marjo,C.E., Andersen, M.S., Graham, P.W.,Cuthbert, M.O., Rau, G.C.,
Roshan, H., Markowska, M., Mariethoz, G., Jex, C.N., 2014. Dripwater organic matter
and trace element geochemistry in a semi-arid karst environment: implications for
speleothem paleoclimatology. Geochim. Cosmochim. Ac. 135, 217?230.
Shao, Y., Wyrwoll, K.H., Chappell, A., Huang,J., Lin, Z., McTainsh,G.H., Mikami,M., Tanaka,
T.Y.,Wang,X.,Yoon,S.,2011.Dustcycle:anemergingcorethemeinearthsystemsci-
ence. Aeolian Res. 2 (4), 181?204.
Shiga, Y., Greene, R.S.B., Scott, K.M., Stelcer, E., 2011. Recognising terrestrially-derived salt
(NaCl) in SE Australian dust. Aeolian Res. 2 (4):215?220. https://doi.org/10.1016/j.
aeolia.2011.02.003.
Sigl, M., Winstrup, M., McConnell, J.R., Welten, K.C., Plunkett, G., Ludlow, F., B?ntgen, U.,
Caffee, M., Chellman, N., Dahl-Jensen, D., Fischer, H., Kipfstuhl, S., Kostick, C.,
Maselli, O.J., Mekhaldi, F., Mulvaney, R., Muscheler, R., Pasteris, D.R., Pilcher, J.R.,
Salzer, M., Sch?pbach, S., Steffensen, J.P., Vinther, B.M., Woodruff, T.E., 2015. Timing
and climate forcing of volcanic eruptions for the past 2500 years. Nature 523
(7562):543?549. https://doi.org/10.1038/nature14565.
Simonson, R.W., 1995. Airborne dust and its signi?cance to soils. Geoderma 65, 1?2),
1?43.
Slatyer, R., 2010. Climate change impacts on Australia's alpine ecosytems. The ANU Un-
dergraduate Research Journal. 2, 17.
Squizzato, S., Masiol, M., Brunelli, A., Pistollato, S., Tarabotti, E., Rampazzo, G., Pavoni, B.,
2013. Factors determining the formation of secondary inorganic aerosol: a case
study in the Po Valley (Italy). Atmos. Chem. Phys. 13 (4), 1927?1939.
Stein, A.F., Draxler, R.R., Rolph, G.D., Stunder, B.J.B., Cohen, M.D., Ngan, F., 2015. NOAA's
HYSPLIT atmospheric transport and dispersion modeling system. Bull. Am. Meteorol.
Soc. 96 (12):2059?2077. https://doi.org/10.1175/BAMS-D-14-00110.1.
Sullivan, R.C., Moore, M.J.K., Petters, M.D., Kreidenweis, S.M., Roberts, G.C., Prather, K.A.,
2009. Effect of chemical mixing state on the hygroscopicity and cloud nucleation
properties of calcium mineral dust particles. Atmos. Chem. Phys. 9 (10):3303?3316.
https://doi.org/10.5194/acp-9-3303-2009.
Suni,T.,Kulmala,M.,Hirsikko,A.,Bergman,T.,Laakso,L.,Aalto,P.P.,Leuning,R.,Cleugh,H.,
Zegelin, S., Hughes, D., van Gorsel, E., Kitchen, M., Vana, M., H?rrak, U., Mirme, S.,
Mirme, A., Sevanto, S., Twining, J., Tadros, C., 2008. Formation and characteristics of
ions and charged aerosol particles in a native Australian Eucalypt forest. Atmos.
Chem. Phys. 8 (1):129?139. https://doi.org/10.5194/acp-8-129-2008.
Tadros, C.V., Treble, P.C., Baker, A., Fairchild, I., Hankin, S., Roach, R., Markowska, M.,
McDonald, J., 2016. ENSO?cave drip water hydrochemical relationship: a 7-year
dataset from south-eastern Australia. Hydrol. Earth Syst. Sci. 20 (11):4625?4640.
https://doi.org/10.5194/hess-20-4625-2016.
Taha,G.,Box,G.P.,Cohen,D.D.,Stelcer, E.,2007.Blackcarbonmeasurementusinglaserin-
tegrated plate method. Aerosol Sci. Technol. 41 (3), 266?276.
Ten Harkel, M.J., 1997. The effects of particle-size distribution and chloride depletion of
sea-salt aerosols on estimating atmospheric deposition at a coastal site. Atmos. Envi-
ron. 31 (3), 417?427.
Viana, M., Kuhlbusch, T.A.J., Querol, X., Alastuey, A., Harrison, R.M., Hopke, P.K.,
Winiwarter, W., Vallius, M., Szidat, S., Pr?v?t, A.S.H., Hueglin, C., Bloemen, H.,
Wahlin, P., Vecchi, R., Miranda, A.I., Kasper-Giebl, A., Maenhaut, W., Hitzenberger,
R., 2008. Source apportionment of particulate matter in Europe: a review of methods
and results. J. Aerosol Sci. 39 (10), 827?849.
Walker, P.H., Costin, A.B., 1971. Atmospheric dust accession in South-Eastern Australia.
Aust. J. Soil Res. 9 (1), 1?5.
WHO, 2006. http://www.euro.who.int/document/e90038.pdf (last access: 27 November
2017).
Yamasoe,M.A., Artaxo, P., Miguel,A.H., Allen, A.G., 2000.Chemicalcomposition of aerosol
particlesfromdirectemissionsofvegetation?resintheAmazonBasin:water-soluble
species and trace elements. Atmos. Environ. 34 (10):1641?1653. https://doi.org/
10.1016/S1352-2310(99)00329-5.
Zhou, Y., Levy, J.I., 2007. Factors in?uencing the spatial extent of mobile source air pollu-
tion impacts: a meta-analysis. BMC Public Health 7 (89):1?11. https://doi.org/
10.1186/1471-2458-7-89.
Zhu, Y., Hinds,W.C., Kim,S., Sioutas,C., 2002.Concentration and size distribution of ultra-
?ne particles near a major highway. J. Air Waste Manage. Assoc. 52 (9):1032?1042.
https://doi.org/10.1080/10473289.2002.10470842.
C.V. Tadros et al. / Science of the To