Introduction
Urban dust aerosols, comprising both natural and anthropogenic contributions
with complex morphological and physiochemical characteristics, have become a
focus of study in global climate change and regional air pollution (Wilson
et al., 2002). Natural dust is derived primarily from long-range transport
with minor local soil contributions and often causes dust events, including
sandstorms, suspended dust, and blown-sand weather (Sun et al., 2001;
Zhang et al., 2003; Chen et al., 2004; Kan et al., 2007; Baddock et al.,
2013); it
has an adverse effect on local air quality (Wang et al., 2004; Ginoux et
al., 2004). Anthropogenic dust produced by human activities is characterized
by high concentrations of toxic heavy metals (e.g., Pb, Zn, Co, Cr, Ni, and As),
which has a long-lasting and deleterious impact on the local environment and human
health (Zdanowicz et al., 2006; Qiao et al., 2013; Lu et al., 2014; Lee et
al., 2015).
Airborne ultrafine particulate
matter (e.g., PM2.5 and PM1)
can enter the alveolar region and blood circulatory system, leading
to health issues and even death (Brunekreef and Holgate, 2002; Nel et al., 2006;
Pickrell et al., 2009; Maher et al., 2013; Elser et al., 2016). Moreover,
anthropogenic dust is an important medium for the formation of secondary
pollutants and plays a significant role in the formation of haze events
(Hanisch and Crowley, 2001; Li et al., 2001; Lee et al., 2002; Usher et al.,
2002; Finlayson-Pitts et al., 2003; Rubasinghege and Grassian, 2009;
Takeuchi et al., 2010; Wu et al., 2011; Huang et al., 2014). Consequently,
it is important to distinguish the characteristics and contributions of
natural and anthropogenic dust in urban aerosols to formulate effective
policies for dust pollution abatement and improving air quality.
Natural and anthropogenic contributions to urban dust aerosols are usually
assessed quantitatively using geochemical and magnetic methodologies
(Gorden, 1988; Xie et al., 1999; Gomez et al., 2004; Spassov et al., 2004;
Kim et al., 2009; Feng et al., 2012). Geochemical methods typically involve
source apportionment and the contribution assessment of representative heavy
metal elements using statistical methods such as chemical mass balance (CMB)
(Chow et al., 2002; Gupta et al., 2007) and factor analysis (FA) (Harrison
et al., 1997a; Salvador et al., 2004).
Pb, Fe, Zn, Cr, Cd, Ni, Ba, and Sb are
frequently used as marker elements for vehicle emissions (Huang et al.,
1994; Adachi and Tainosho, 2004; Meza-Figueroa et al., 2007), while Hg, Pb, Mn,
Cr, Co, Cu, Cd, and Ni are regarded as indicators of coal combustion (Vouk and
Piver, 1983; Pacyna and Pacyna, 2001; Sushil and Batra, 2006).
Since magnetic measurements are rapid, inexpensive, and nondestructive,
environmental magnetism is increasingly being used as an effective approach
to study urban dust pollution (Maher, 1998; Hoffmann et al., 1999).
By combining magnetic properties with morphological features (Muxworthy et al., 2001;
Urbat et al., 2004; Blaha et al., 2008a) as well as heavy metal (Hunt et al., 1984; de
Miguel et al., 1997; Blaha et al., 2008b; Maher et al., 2008) and back-trajectory characteristics (Wehner et al., 2008; Li et al., 2009; Fleming et
al., 2012), the provenance, transport routes, and spatial distribution of
polluted dust aerosols can be investigated. This multidisciplinary approach
is becoming a popular means of urban pollution monitoring and assessment
(Jordanova et al., 2014; Stein et al., 2015; Yan et al., 2015a; Bourliva et
al., 2016).
Using environmental magnetic techniques to assess pollution levels and
sources, different forms of urban dust aerosols in East Asia have been
studied, including atmospheric dustfall, street dust, leaf dust, inhalable
particulate matter, and surface soil. For example, spatial and temporal
pollution patterns were quantitatively estimated from seasonal fluctuations
of the concentration and grain size of magnetic particles in urban roadside
dust (Kim et al., 2007, 2009). A high correlation between magnetic
parameters (magnetic susceptibility and saturation isothermal remanence,
i.e., χ and SIRM) and heavy metal concentrations in street dust,
polluted farmland soil, and atmospheric dustfall was observed, indicating
that these magnetic parameters can be employed as effective proxies to
assess heavy metal pollution (Zhang et al., 2011, 2012a, b; Qiao et al.,
2013). SIRM characteristics of roadside leaves were shown to reflect spatial
variations of magnetic particles in urban dustfall (Quayle et al., 2010;
Hansard et al., 2011, 2012; Maher et al., 2013; Kardel et al., 2012).
Although morphology, grain size, mineral, and element analyses were utilized
in previous works, there are no studies that systematically compare magnetic
signatures of natural dust, urban dust aerosol, and polluted dust from
source to sink.
This study systematically collected surface sediments from potential dust
sources in East Asia, urban dust aerosols in Xi'an, including atmospheric
dustfall (over five consecutive years) and street dust, and typical
anthropogenic pollutants such as vehicle exhaust and fly ash. Morphology and
elemental compositions of magnetic particles in representative samples were
analyzed to facilitate a thorough source–sink comparison. Our results
indicate that natural and anthropogenic contributions to urban dust aerosols
can be differentiated using a combination of their magnetic, morphological,
and elemental characteristics.
Locations of natural surface sediments in the East Asian sources
(a) and urban dust samples in Xi'an (b). NCD – northern Chinese deserts, MG
– Mongolian Gobi, TD – Taklimakan Desert, and TP – Tibetan Plateau. Black
diamonds are street dust sampling sites; blue dots are samples of
consecutive atmospheric dustfall (XA1 at the Institute of Earth Environment,
Chinese Academy of Sciences; XA2 at the Xinxinjiayuan residential
community); red triangles are typical heavily polluted sites, including the
Bell Tower in an area of high traffic density and the Baqiao thermal power
plant.
Sampling and methods
Sampling
Surface sediments were collected in potential dust source regions of East
Asia, including the northern Chinese deserts (the Badain Juran and
Tengger Desert), the Taklimakan Desert, Mongolian Gobi, and Tibetan Plateau
(Fig. 1a). Fine-grained materials were collected from alluvial fans, dry
riverbeds, lake basins, and drainage depressions within Gobi–sandy deserts
at intervals of 100 to 200 km (Fig. 2a–d). To better understand the
different sedimentary characteristics, 48 samples from the northern Chinese
deserts, 50 samples from the Taklimakan Desert, 23 samples from the
Mongolian Gobi, and 32 samples from the Tibetan Plateau were selected for
magnetic measurements. Locations of the
samples are shown in Fig. 1a. Detailed
descriptions were given in Sun et al. (2013).
Sampling sites of natural surface sediments in a dry riverbed (a),
desert margin (b), drainage depressions within sandy desert (c) and Gobi
desert (d), and atmospheric dustfall at XA1 (e) and XA2 (f).
A total of 68 street dust samples were collected from parks, construction
sites, commercial streets, and residential areas in Xi'an following a
3×4 km grid spanning approximately 30 km from west to east and
20 km from north to south (Fig. 1b). The sampling grid covers a range of
different functional areas in Xi'an, including the Industrial District,
Commercial District, Cultural District, Ecological District, and Han
Chang'an city ruins park. We also collected four typical
anthropogenic pollutant samples in June 2017, including one sample of
exhaust from several vehicles, one sample of
fly ash from a dust bag of electrostatic precipitators at the Baqiao thermal
power plant, one street dust sample from the Bell Tower in downtown Xi'an,
which experiences daily traffic jams, and one street dust sample near the
Baqiao thermal power plant where coal burning is the leading pollution
factor. The locations of these samples are shown in Fig. 1b.
Atmospheric dustfall collectors were placed on the top of a four-story
building at the Institute of Earth Environment, Chinese Academy of Sciences,
∼10 m above the ground surface, and a 15-story building
inside the Xinxinjiayuan residential community, ∼50 m above
the ground surface (Fig. 2e, f). The sampling sites situated in southwest
Xi'an consist primarily of commercial and residential districts. Samples
were collected using the wet-collection method (Qian and Dong, 2004) at time
intervals of 3–5 days in spring and 6–7 days in other seasons. Detailed
sampling procedures were reported by Yan et al. (2015a, b); 733 samples were
collected from March 2009 to March 2014. Dust flux (DF, g m-2 day-1) is calculated as follows:
DF=W/(A×T),
where W is the sample weight in g, A is the area in m2, and T is
sampling duration in days.
Methods
Low- and high-frequency magnetic susceptibilities (χlf and χhf, respectively) are measured using a MFK1-FA Kappabridge at
frequencies of 976 and 15 616 Hz. Frequency-dependent magnetic
susceptibility (χfd) is calculated as (χlf-χhf) / χlf×100 %.
The temperature-dependent susceptibilities (χ-T) are measured in an
argon atmosphere (the flow rate is 50 mL min-1) at a frequency of 976 Hz from
room temperature up to 700∘ and back to room temperature using a
MFK1-FA Kappabridge equipped with a CS-3 high-temperature furnace. The
susceptibility of each sample is corrected for background (furnace tube
correction) using the CUREVAL 8.0 program.
Hysteresis loops and first-order reversal curve (FORC) diagrams are measured
by a vibrating sample magnetometer (VSM3900) to a maximum applied field of
1 T. Hysteresis parameters, including the saturation magnetization (Ms),
saturation remanent magnetization (Mrs), and coercivity (Bc), are
obtained after subtracting the paramagnetic contribution. The remanence
coercivity (Bcr) is obtained by demagnetizing samples from +1 T back
to -1 T. The hysteresis ratios Mrs/Ms vs. Bcr/Bc are
used to construct a Day plot.
The FORC diagrams are measured with the averaging time of 200 ms and
produced using FORCinel software (Harrison and Feinberg, 2008). A total of
18 samples are used for detailed iron oxide analyses, including 2 samples
from each natural dust source with modal χlf values, 4 dustfall
samples and 2 street dust samples with high χlf and low χlf, and 2 samples of vehicle exhaust and fly ash.
The magnetic components of these representative samples are separated from
the bulk samples using a 1 T magnet sealed in a polyethylene bag. To confirm
their mineral, morphological, and elemental characteristics, direct
observations and measurements of the samples and their extracted magnetic
particles are performed using a ZEISS EVO-18 scanning electron microscope
(SEM)
equipped with a Bruker XFlash 6130 energy dispersive spectroscope (EDS). Samples are mounted on
the SEM stub with double-sided carbon tape and
then coated with a thin gold film. The specified resolution of the SEM is
<5 nm. The EDS detector is capable of detecting elements with
atomic numbers ≥5 and the detection sensitivity can reach 0.1 wt%.
Bulk samples and magnetic extracts are characterized by randomly selecting
three
to four fields of view and examining all the particles observed within the
selected fields. All the measurements are made at the Institute of Earth
Environment, Chinese Academy of Sciences, Xi'an.
χ-T heating (red line) and cooling (blue line) curves
(a–f)
and magnetic hysteresis loops (g–l) of representative samples of natural
surface sediments (MD0907), atmospheric dustfall (2013.7.18 and 2010.4.30),
street dust (L5-10), and anthropogenic pollutants: fly ash (DCYH) and
vehicle exhaust (QCWQ).
Results
Magnetic mineralogy
χ-T is used to identify magnetic mineral composition. All the χ-T
heating curves (Fig. 3a–f) are characterized by a major susceptibility
decrease at 580∘, i.e., the Curie temperature of magnetite, which
identifies magnetite as the major contributor to χ. All the samples
are irreversible with cooling paths above heating trajectories due to the
neoformation of magnetite (Jordanova et al., 2004; Kim et al., 2009). The
χ-T heating curve of the vehicle exhaust displays a decreasing χ
between 580 and 700∘ (Fig. 3b), suggesting the presence of
hematite.
All samples have similar slightly wasp-waisted hysteresis loops (Fig. 3g–l).
Magnetic saturation is generally reached at a magnetic field of about
300 mT. This is a clear indication of the predominance of low-coercivity
ferrimagnetic minerals in all samples.
Day plot of the ratios Mrs/Ms
vs. Bcr/Bc (a) and FORC diagrams (b–e) for
representative samples from natural surface sediments (NSS), atmospheric
dustfall (AD), street dust (STD), and anthropogenic pollutant (AP). Domain
boundaries and the single-domain+multi-domain mixing line are according to
Dunlop (2002b). Percentages in the Day plot represent the concentrations of
multi-domain in the single-domain+multi-domain mixture.
Hysteresis properties
The Day plot and FORC diagram are powerful methods to identify the domain
state distribution of magnetic materials (Day et al., 1977; Pike et al.,
1999; Roberts et al., 2000; Dunlop, 2002a, b). All the samples agree well
with single-domain + multi-domain admixture curves in the
pseudo-single-domain range of the Day plot (Fig. 4a). The FORC diagrams for
street dust (Fig. 4d) and anthropogenic pollutants (Fig. 4e) have divergent
contours that are characteristic of multi-domain grains. The FORC diagram
for natural surface sediments (Fig. 4b) seems to be characteristic of
pseudo-single-domain and multi-domain behavior, whose outer contours display
a divergent pattern and inner contours are somewhat less divergent. The FORC
distributions of atmospheric dustfall (Fig. 4c) appear to have a mixed set
of contours. The outer contours have a divergent pattern that would be
expected for multi-domain particles, while the inner distribution with
closed contours represent single-domain grains.
Spatial and temporal variations of χ
The magnetic susceptibilities of all bulk samples were measured to estimate
concentrations of magnetic minerals, which are largely controlled by
concentrations of ferromagnetic minerals (Dunlop and Özdemir, 1997; Evans and Heller,
2003; Liu et al., 2012). χfd is sensitive to the
superparamagnetic component. There are virtually no superparamagnetic grains
when χfd is <2 %, while a mixture of
superparamagnetic and coarser grains is indicated with χfd in the
range of 2–10 % (Dearing, 1994; Dearing et al., 1996). The peaking of the χlf frequency distribution curve indicates the χlf values
are most distributed in this interval. Both χlf and
χfd exhibit a distinctive distribution pattern in different sources.
Frequency distribution of χlf (a, c), bivariate plots
of χlf versus χfd (b, d), and average values and
standard deviations of χlf and χfd (e, f) for NSS in
each source and in urban dust aerosols, including AD, STD, and AP. Frequency
distribution statistics of χlf for NSS, AD and STD, and AP are
generated using intervals of 10×10-8 m3 kg-1,
100×10-8 m3 kg-1, and 200×10-8 m3 kg-1, respectively.
χlf values from the Taklimakan Desert and northern Chinese desert
samples exhibit a unimodal distribution (Fig. 5a), and those from the Mongolian
Gobi exhibit a bimodal distribution. However, χlf values have a
multimodal distribution in the Tibetan Plateau (Fig. 5a). Different
distribution patterns and peak values of χlf (Fig. 5a) indicate
that the assemblage of magnetic minerals may differ in these four natural
sources. Average χlf in individual sources shows a decreasing
trend from the Mongolian Gobi, to northern Chinese deserts and the Tibetan
Plateau, and then to the Taklimakan Desert (Fig. 5e). The mean values of χfd in different natural sources show a decreasing trend of
superparamagnetic components from the Taklimakan Desert to Mongolian Gobi and
Tibetan Plateau, and then to northern Chinese deserts (Fig. 5f).
Time series of magnetic susceptibility and dust flux of
atmospheric dustfall at XA1 and XA2 from 2009 to 2014.
The frequency distributions of χlf for the street dust and
atmospheric dustfall are both unimodal (Fig. 5c). The average χlf
and χfd values of the street dust are higher than those of the
atmospheric dustfall and natural surface sediments (Fig. 5e). Low χlf (<500×10-8 m3 kg-1) occurs in
the Ecological District, Han Chang'an city ruins park, and Cultural
District, while samples with intermediate χlf values (500–800×10-8 m3 kg-1)
are from the moderately developed
Industrial District and the periphery of the Commercial District. In
contrast, the central areas of the Industrial District and the Commercial
District (particularly the area of high traffic density at the Bell Tower)
are characterized by relatively high χlf values (>800×10-8 m3 kg-1). χlf of atmospheric
dustfall from XA1 and XA2 exhibits significant and consistent seasonal
variations (Fig. 6). The lowest (highest) χlf values correspond
to the highest (lowest) dust flux in spring (autumn).
The representative anthropogenic pollutants, i.e., vehicle exhaust, fly
ash, and nearby street dust at the Bell Tower and thermal power plant, have
high χlf and χfd (Fig. 5c, d). The χlf
and χfd of vehicle exhaust and fly ash are higher than the mean
values of other sources of dust (Fig. 5e, f).
Morphology and mineralogy of representative samples of the natural
surface sediments (a), street dust (b), and atmospheric dustfall with low χlf (c) and high χlf (d). Qtz – quartz, Fsp – feldspar, Cal
– calcite, Dol – dolomite, Cm – clay minerals, Dmm – detrital magnetic
mineral, Irs – iron-rich sphere, As – aluminosilicate sphere, An – anomalous
particles with a porous and loose structure.
Morphology and mineralogy of the dust samples
SEM provides morphology information based on gray-scale intensity. The
elemental composition is determined by the EDS detector. In order to compare
the morphology and mineralogy characteristics of different dust aerosols, more
than 40 fields of views of the representative bulk samples were randomly
obtained for various types of particles. The morphologies and mineral
compositions of the natural surface sediments, street dust, and atmospheric
dustfall with low and high χlf are illustrated in Fig. 7. The
particles are typically angular and irregularly shaped in the surface
sediments, with a broad size range (around 1–100 µm). Based on the
EDS analysis for each particle in the selected field, clay minerals,
quartz, calcite, dolomite, and magnetic grains (Fig. 7a) were clearly
identified (Welton, 1984).
The SEM–DES analysis shows that the morphology and constituents of the
particles in the street dust are complex and heterogeneous. Three categories
of particles can be morphologically differentiated, including irregular and
aggregate mineral particles, spherical particles, and anomalous particles
with porous and loose structures (Fig. 7b). Particles with irregular shapes
are mainly minerals and commonly present in street dust samples. Compared to
the natural surface sediments, the grain size of mineral particles in the
street dust is finer and mostly ranges from 1–50 µm, with some up to
80 µm. Spherical particles are mainly amorphous silicon–aluminum and
iron-rich spheres, whose grain size varies mostly from 1–20 µm, with
some up to 50 µm. There are a small number of anomalous particles with
diameters of 10–100 µm.
The morphology and mineral composition of atmospheric dustfall are similar
to those of the street dust, except that atmospheric dustfall with low χlf has a higher content of irregularly shaped detrital minerals
(Fig. 7c), while that with high χlf contains more spherical and
anomalous particles (Fig. 7d).
SEM photograph and elemental spectra for a typical sample of
street dust. In the subplots, the green plus symbols denote the locations of
the beam used in the EDS analysis.
Elemental compositions of mineral particles
Since the elemental compositions of mineral particles can be clearly
distinguished using SEM–EDS analysis (Blanco et al., 2003; Kutchko and Kim,
2006), a street dust sample
dominated by anthropogenic inputs with the
highest χlf was selected for
EDS analysis. The results indicate that various mineral particles
exhibit distinct chemical compositions (Fig. 8). The plate-like aggregates (labeled
a) with high levels of Si and Al and low levels of K, Ca, Mg, and Fe are
clay minerals composed of crystalline sheet-structure silicates with a small
particle size (Fig. 8a). The angular and sharp-edged particle (labeled b)
with high Si and O is quartz (Fig. 8b). The angular particle consisting of
Si, Al, and K is potassium feldspar (Fig. 8c). Particles with high
levels of Ca and Mg are calcite (Fig. 8d) and dolomite (Fig. 8e).
The irregular particles (labeled f) that are abundant in Fe are identified
as magnetic grains (Fig. 8f), although some of the particles show low levels
of crustal elements, including Si, Al, Ca, and K. Two types of spheres were
observed. One (labeled g) is an amorphous aluminosilicate particle
(Fig. 8g) with predominant Si and Al and lesser amounts of K, Mg, Na, and Ti. The
other (labeled h) is an iron-rich sphere (Fig. 8h), which is mainly composed
of Fe. These particles exhibit various surface textures. In addition, almost
all particles contain O and C.
Frequency distributions of χlf (a) and bivariate plots
of χlf versus χfd (b) of NSS, STD, AD, and AP.
Discussion
Contributions of local anthropogenic sources estimated by dust
flux and χlf
On the bivariate plot of χlf vs. χfd, atmospheric
dustfall is intermediate between the surface sediments and street dust
(Fig. 9b), implying that atmospheric dustfall is a mixture of distal natural
dust and local anthropogenic dust, but much closer to the latter. The local
anthropogenic contribution (LC) is mainly derived from local stable and
sustained pollutant sources, including vehicle emissions and fly ash.
Considering that natural dust comes primarily from natural dust sources with
a minor local soil contribution (Wang et al., 2004; Ginoux et al., 2012), we
attribute the natural contribution entirely to the distal natural dust.
The dust flux background can be taken as the average input from the end-member of LC.
The time-dependent background estimation was calculated using
xibg=MEDj=i-kj=i+kxj,
where i=k+1,…,n-k, x(i)bg is the background of x(i)
at time t(i). MEDj=i-kj=i+kxj is the running median with window points
of 2k+1 (k≤(n-1)/2) (Härdle and Steiger, 1995); cross-validation
can be used to choose k. We used two such criteria: the median criterion (Zheng
and Yan, 1988) and L1 norm (Marron, 1986; Dodge, 2012).
CVmk=medianxi-MEDj=i-k,j≠ij=i+kxjCV1k=∑i=1nxi-MEDj=i-k,j≠ij=i+kxj/n
MEDj=i-k,j≠ij=i+kxj is the delete-one background estimate. The cross-validation
functions are to measure the average performance of the delete-one estimate
to predict the observation x(i). Optimal k values should minimize CVm(k) or
CV1(k) (Mudelsee, 2006).
Through the cross-validation calculation on the dust flux series of
atmospheric dustfall, we find that the cross-validated number of window width
(Eq. ) is k=19. On this basis, we calculate the monthly LC using the
ratio of monthly background and total dust flux as
LCflux=x(j)bg/DF×100%,
where LCflux is the percentage of the monthly local anthropogenic
contribution estimated by dust flux (Fig. 10a). Note that when the
background is larger than the dust flux, LC is taken to be 100 %.
The estimated local contributions by dust flux (a) and χlf (b). From bottom to top: (a) dust flux (pink) and background
estimate by the running median with a cross-validated number of window
points (k=19) (black), monthly averaged dust flux (blue) and background
(brown), average monthly local contribution (red) estimated by dust flux at
XA1, and the uncertainty bounds calculated by standard deviation (gray
area). (b) χlf values (light blue), averaged χlf
values of natural distant dust (green dotted lines), monthly averaged
χlf values (dark red), averaged χlf values of local
street dust (orange dotted lines), monthly local contribution (violet)
estimate by χlf at XA1, and the uncertainty boundaries calculated
by standard deviation (gray area).
Ms values of representative samples (Fig. 3h–m) are measured to identify
the concentration of ferrimagnetic minerals. We find that the averaged values of
Ms in different sources show a rising trend from the natural surface
sediments (0.04 Am2 kg-1) to atmospheric dustfall (0.81 Am2 kg-1) and
street dust (1.03 Am2 kg-1), and then to anthropogenic pollutants
(1.58 Am2 kg-1), which correspond to the characteristics of averaged χlf in different sources. This indicates that the high χlf
of urban dust is caused by ferrimagnetic minerals from a local
anthropogenic source. In consequence, the LC contribution could also be
estimated by the peak values of χlf frequency distribution,
with 20–30×10-8 m3 kg-1 in the distant natural
surface sediments and 500–600×10-8 m3 kg-1 in
local street dust (Fig. 9a). On this basis, we calculate the average LC
using the following equation:
LCχ=χm-25/550-25×100%,
where LCχ is the percentage of the monthly local contribution
estimated by χlf (Fig. 10b), and χm is the monthly
average χlf value in 10-8 m3 kg-1. 25×10-8 and 550×10-8 m3 kg-1
are the average χlf values of surface sediments from source
regions and local street dust, respectively. Note that when χm is
larger than the average χlf of the street dust, LC is taken to be 100 %.
The LCflux and LCχ values have the same trend and show a
distinctive seasonal pattern (Fig. 10a, b), with a maximum in autumn
(92.4 %, 92.3 %), followed by winter (90.8 %, 74.7 %), summer
(83.5 %, 71 %), and spring (73.0 %, 53.1 %). Both the LCflux and
LCχ are the lowest in spring, implying that distant natural dust
input makes a great contribution to atmospheric dustfall during this period.
The LC variation exhibits a similar seasonal pattern to χlf,
but an opposite trend to that of dust flux (Fig. 10a, b). This suggests that
the major sources of atmospheric dustfall varied seasonally between the
distant natural sources in spring and local anthropogenic sources in other
seasons. In spring, dust is emitted from the natural sources by strong
winds, and after long-range transport it contributes to the elevated dust
flux in Xi'an and decreases the LC in atmospheric dustfall. However, from
summer to winter, dust input from local anthropogenic sources is low and
stable as indicated by the high LC.
SEM images and typical elemental spectra (d) of magnetic extracts
from street dust and (a) atmospheric dustfall with high χlf (b) and
low χlf (c). From left to right, the particle morphologies
represent detrital particles with relatively smooth surfaces from natural
source regions, anthropogenic particles with angular shapes and coarse
surface textures, aggregates, spherules, and porous features.
Magnetic characteristics of anthropogenic particles
SEM–EDS analysis shows that the extracted magnetic particles from the street
dust and atmospheric dustfall can be divided into detrital and anthropogenic
types (Fig. 11a–c). Detrital particles are angular and characterized by
relatively smooth surfaces, with Fe and O as the major elements and minor
amounts of Ti (Fig. 11d), indicating the presence of magnetite, hematite, and
titanomagnetite (Maher and Thompson, 1991; Liu et al., 2015). Anthropogenic
particles include angular particles with coarse surface textures, spherules,
aggregates, and porous particles with complex internal structures. The major
elements identified in these particles are Fe and O, which indicate the
occurrence of magnetite or hematite, consistent with previously identified
anthropogenic magnetic particles (Kim et al., 2007; Koukouzas et al., 2007;
Maher, 2009). Minor concentrations of S, Zn, Cu, and Cr were also
observed in this type of particle, which is typically attributed to
anthropogenic activities (Fig. 11d). The relatively weaker signal intensity
of Fe in the EDS spectra of porous particles indicates a much lower Fe
concentration (maximum less than 10 %), while their concentrations of Si,
Al, Ca, Ti, and Mn are higher.
The morphology and concentration of magnetic
materials in urban dust aerosols varied with sampling sites and over time.
Among more than 20 images of analyzed magnetic extracts from urban dust
samples, angular particles with coarse surface textures were the most
frequently observed (>50 %, some up to 80 %), with a wide
range of grain size (1–100 µm). Spherules were also commonly observed
in all samples, ranging from 10–40 %, mainly with diameters from
10–30 µm. Aggregates with diameters of 5–30 µm account for less
than 10 %. Detrital particles characterized by smooth surfaces range
from 1–5 % and have small diameters (1–20 µm). Porous particles
are the least observed magnetic particles (<1 %) with diameters
of 30–120 µm. The SEM–EDS data show that the morphology and
concentration of magnetic particulates in atmospheric dustfall with high
χlf values are similar to those of the street dust, whereas
atmospheric dustfall with low χlf contains more
angular–subangular magnetic particles of detrital origin.
SEM images and elemental spectra of magnetic extracts from
atmospheric dustfall (a–d), vehicle exhaust (e–g), and fly ash (h–k). Black
lines are elemental spectra of atmospheric dustfall. Blue and red lines are
elemental spectra for vehicle exhaust and fly ash.
Potential sources of anthropogenic magnetic particles
Anthropogenic magnetic particles in the urban environment are mainly derived
from the combustion of fossil fuels (Flanders, 1994; Matzka and Maher, 1999;
Muxworthy et al., 2001), vehicle emissions (Harrison et al., 1997b; Moreno
et al., 2003; Diapouli et al., 2008; Pant and Harrison, 2013; Maher et al., 2013),
and industrial activities (Hanesch et al., 2003; Desenfant et al., 2004). To
clarify potential sources, microscopic and elemental investigations of
magnetic extracts from anthropogenic pollutants were performed using
SEM–EDS. Compared with the magnetic particles in atmospheric dustfall (Fig. 12a–d),
those from vehicle exhaust consist of only three types of particles,
including angular particles with coarse surface textures, spherules, and
aggregates (Fig. 12e–g), while all magnetic particle types in dustfall
samples were identified in fly ash (Fig. 12h–k). The EDS analysis showed
that the major elements of the same three types of magnetic particles in
vehicle exhaust and fly ash are Fe and O, consistent with elemental features
of those in atmospheric dustfall (Fig. 12l–n). This suggests that vehicle
exhaust and fly ash are the main pollutant sources in dustfall. However,
there are some differences in the compositions of the minor elements in the
three types of particles between vehicle exhaust and fly ash. Angular
particles with coarse surface textures from vehicle exhaust contain more S,
Cr, Cu, Zn, and Mn, while those from fly ash have more Ca and Mn. Aggregates
consist of more Cr, Zn, and S in vehicle exhaust, whereas Ca and S are
enriched in fly ash. Spherules from vehicle exhaust contain higher amounts
of heavy metals (Cr, Ni, Mn, and Zn), while those from fly ash have higher Ca
and Mn. Coarse-grained porous magnetic particles were only observed in fly
ash, which are relatively low in Fe and high in crustal elements (e.g., Si,
Al, K, Ca, Mg, and Ti).
The EDS elemental data clearly indicate that the magnetic particles from
vehicle exhaust contain higher concentrations of a greater range of elements
from anthropogenic activities (S, Cr, Cu, Zn, Ni, and Mn) than those from fly
ash, whose EDS spectra show a substantial peak of Ca. The χlf
(925.7×10-8 m3 kg-1) and Ms (2.5 Am2 Kg-1)
values of vehicle exhaust are significantly higher than those
of fly ash (769.9×10-8 m3 kg-1 and
0.66 Am2 Kg-1), indicating a higher content of ferrimagnetic contaminants. In
summary, the magnetic particles emitted by vehicle exhaust and thermal power
plants can be distinguished by a combination of morphological and elemental
characteristics, which indicates that SEM–EDS can be used to trace the
sources of anthropogenic pollutants in Xi'an.
Conclusions
By comparing the magnetic properties of surface sediments in natural dust
sources in East Asia and various urban dust samples in Xi'an, we found that
distal natural dust and local anthropogenic dust have different magnetic,
morphological, and elemental characteristics. We take natural surface
sediments as representative of distal natural dust, with background
atmospheric dustfall and polluted street dust as representative of local
anthropogenic dust. Based on this end-member configuration, the relative
contributions of local anthropogenic sources to urban atmospheric dustfall
can be quantitatively estimated.
The results show that local anthropogenic contribution decreases in spring
and increases in other seasons. Local anthropogenic contribution variation
exhibits a similar seasonal pattern to χlf, but an opposite trend
to that of dust flux with a maximum in spring. This means that a great
amount of distant natural dust input with less magnetic content makes a
great contribution to atmospheric dustfall in spring, which results in minimum
χlf and anthropogenic contributions during this period. Hence,
the local contribution is reduced as a result of increasing natural dust
flux.
SEM–EDS analysis of urban dust indicates that magnetic particles produced by
anthropogenic activities have distinct morphological and elemental
characteristics. The anthropogenic particles exhibit angular, spherical,
aggregate, and porous shapes and contain distinctive marker elements such
as S, Cr, Cu, Zn, Ni, Mn, and Ca. The porous particles are likely derived
from the thermal power plant, while others may be attributed to both vehicle
exhaust and the thermal power plant. Our results suggest that magnetic
signatures combined with morphological and elemental compositions can be
used to quantitatively estimate local and anthropogenic contributions to
urban dust aerosols.