Source Profiles of Fugitive Dust in Zhengzhou , China

As a result of rapid urbanization, the road lengths and built-up areas in Zhengzhou are steadily expanding along with increasing fugitive dust emissions. Identifying the physical and chemical characteristics and the chemical profiles of fugitive dust is important in achieving effective atmospheric pollution management. In this study, soil dust, road dust (RD), building demolition dust, and cement (CE) were chosen as the research objects. A total of 94 dust samples were collected from 20 sites. PM2.5 (particulate matter with diameter ≤ 2.5 μm) and PM10 (particulate matter with diameter ≤ 10 μm) samples were obtained by using a re-suspension device and their physical and chemical properties were analyzed. The scanning electron microscopy of four types of dust particles showed that most of the dust particles presented an irregular shape. In terms of particle size distribution, the mass concentration of PM2.5 accounted for less than 10% of the total PM10, whereas the number concentration of PM2.5 accounted for more than 96% of the total number. Chemical component analysis revealed that crustal elements such as Al, Mg, Fe, and K were abundant in all samples, and they were the most abundant species in PM2.5 and PM10 in the reconstruction results. The percentage of NO3 in the RD sample was higher than that in the other three fugitive dust samples because of the influence of vehicles. Furthermore, the CE sample had higher SO4 and Ca percentages than the other three types of fugitive dust samples. Enrichment factor analysis showed that the significant enrichment of Cd and Ag was mainly caused by anthropogenic sources. The coefficients of divergence values between the profiles for different sites of dust ranged from 0.21 to 0.68, indicating the fugitive dust profiles from various sites mostly different. The chemical profiles of four dust sources obtained from this study is limited in Zhengzhou.


INTRODUCTION
As the largest developing country, China has maintained fast economic development, urbanization, and industrialization in the past three decades, and consequently, the country faces serious atmospheric pollution and public health problems (Lin et al., 2010;Hu et al., 2015).Particulate matter (PM) has become the primary pollutant of the atmosphere in recent years (Ji et al., 2014), with a serious impact on global and regional climate changes (Bytnerowicz et al., 2007), reduced visibility (Moosmüller et al., 2009), and human health effects, especially associated with cardiovascular disease and wheezing (Pope and Dockery, 2006;Dunea et al., 2016).The atmospheric particle shape plays a key role in the aspects of radiation effects, climate effects, multiphase chemical reaction mechanism, and clear mechanism (Colarco et al., 2002), and scanning electron microscope (SEM) is one of popular techniques to analyze the single atmospheric particle size, geometric characteristics, and optical characteristics (Germani and Buseck, 1991).Particle size plays an important role on deposition rate of the inhaled aerosol in different regions of the respiratory system (Arhami et al., 2010).Coarse particles (PM 10 , aerodynamic diameter ≤ 10 µm) and fine particles (PM 2.5 , aerodynamic diameter ≤ 2.5 µm) have been associated with hospital admissions for respiratory (Brunekreef and Forsberg, 2005) and cardiovascular disease (Bell, 2012).Ultrafine particles (PM 0.1 , aerodynamic diameter ≤ 0.1 µm), carrying considerable amounts of toxic substance, can penetrate deep into pulmonary alveoli (Sioutas et al., 2005).
The source apportionment model is widely used to analyze the contribution of emission sources to PM in atmosphere and develop pollution control strategies (Khan et al., 2010), and fugitive dust is identified as one of the staple sources of PM (Chen et al., 2010a;Kumar et al., 2012;Kim et al., 2013;Lv et al., 2016).According to previous studies, the contribution of dust to PM 10 and PM 2.5 is 19%-65% and 3-46%, respectively (Samara et al., 2003;Almeida et al., 2006;Bi et al., 2007;Huang et al., 2014;Cheng et al., 2015).Chemical profiles are entered into chemical mass balance (CMB), one of popular receptor models from US. Environmental Protection Agency (Chen et al., 2010b;Rutter et al., 2011), as vital parameters.In previous studies, several results of source apportionment were conducted by CMB without local profiles available (Tony et al., 2005;Rizzo and Scheff, 2007).However, chemical profiles vary with time and location (Watson et al., 2001).Hence, source profiles from literatures result in substantially bias on source contribution (Samara, 2005), and accurate local source profiles are urgently needed.In order to get chemical profiles of dust, the resuspension method has been used to collect PM 10 and PM 2.5 samples from different fugitive dust sources (Chow et al., 1994;Vega et al., 2001).However, most studies focused on road dust (RD) (Wang et al., 2005;Chen et al., 2006;Han et al., 2011;Shen et al., 2016), soil dust (SD) (Ho et al., 2003;Kong et al., 2011;Liu et al., 2016) and cement (CE) (Vega et al., 2001;Ho et al., 2003).Unfortunately, few chemical profiles on demolition dust (DD) are existent, although heavy metals in ambient air contaminated from DD have been evaluated for health risk assessment (Farfel et al., 2003;Brown et al., 2015a;Azarmi and Kumar, 2016).
Henan province has a population of 94.36 million.Zhengzhou, which is the capital of the province, recorded a population of 9.38 million by the end of 2014 (http://www.ha.stats.gov.cn).As the country's major transport hub, Zhengzhou has undergone an increase in its paved roads, the length of which has increased from 490 km to 1,809 km and the area of which has increased from 5.8 × 10 6 m 2 to 4.7 × 10 7 m 2 in the past 23 years (Fig. 1(a), http://www.zzstjj.gov.cn).The motor vehicle population has an annual average growth rate of 16.7% and increased from 1.05 million in 2005 to 2.55 million in 2013 (Zhengzhou Statistical Yearbook, 2006-2014).The population density of Zhengzhou is increasing yearly, especially in the last 10 years (Fig. S1 in Supplemental Materials).With rapid urbanization, the area dedicated to construction and demolition in Zhengzhou occupies a high proportion in urban areas.From 1984 to 2014, the urban built-up area increased from 69.3 km 2 to 412.7 km 2 , and in 2014, the area of construction and demolition accounted for 25% and 17% of the urban builtup area, respectively (Fig. 1(b)).
As a result of these developments, PM has become one of the most important air contaminants in Zhengzhou.According to satellite remote sensing data, China has the most serious PM pollution, and Zhengzhou is one of the most polluted cities in China (Tao et al., 2014).Among all the 74 cities monitored by the Chinese Ministry of Environmental Protection, Zhengzhou ranked tenth and ninth from the bottom in 2013 and 2014, respectively, in terms of air quality.In 2015, the air pollution in the city became particularly serious (fifth from the bottom), with PM as the primary pollutant (http://jcs.mep.gov.cn;http://www.hnep.gov.cn).In fact, Geng et al. (2013) employed positive matrix factorization and determined that in 2010, fugitive dust was one of the major sources of PM 2.5 , given its contribution of approximately 26% to the total PM 2.5 .Hence, the study on source profiles of fugitive dust in this study can provide essential input for source apportionment of PM, and thus eventually play an important role for the government to formulate reasonable and effective measures of prevention and control of dust to decrease PM level and improve atmospheric environment quality in Zhengzhou.
In the present study, we aimed to obtain the geological dust source profiles in an urban area of east-central China, examine the similarities and differences among the samples from different dust sources and different sites for the same dust source, and prepare source profiles for receptor modeling.Four dust sources, namely, paved road, demolition, surface soil, and CE production, were chosen as fugitive dust samples, and PM 10 and PM 2.5 samples were obtained via resuspension experiments.The physical properties, including shape characteristics and size distribution, and chemical constituents of the PM samples were analyzed.Finally, four types of source profiles and their physical and chemical differences were determined, and the coefficients of divergence values between specific source profiles from the various sites for the same type of dust were evaluated.This study is the complete and systematic research, including RD, DD, SD and CE together for both PM 10 and PM 2.5 , and the results provide essential input for source apportionment of PM in atmosphere in this region.

Site Description and Dust Sample Collection
Dust samples were collected in Zhengzhou from September 2015 to January 2016, belonging to autumn and winter.The longitude of Zhengzhou ranges from 112°42′ to 114°13′E, and the latitude ranges from 34°16′ to 34°58′N.Ninety-four dust samples (i.e., 21, 49, 21 and 3 samples for RD, DD, SD and CE, respectively) were collected from 20 sites (i.e., 7, 7, 3 and 3 sites for RD, DD, SD and CE, respectively).The dust samples originated from four sources, namely, RD, DD, SD, and CE.The collection mass of each dust sample was approximately 400 g.The detailed characteristics of fugitive dust and sampling sites are presented in Table S1 in Supplemental Materials.All the samples were collected in the urban district, except for RD-4 and RD-5, which were collected from a provincial road to determine any differences in the source profiles of roads in urban and non-urban areas.Fig. 2 and Fig. S2 in Supplemental Materials show the geographical distribution of all the dust samples, except for those sourced from CE, and the photos of the sampling sites, respectively.

Sample Pre-Treatment and PM Sample Collection
All the dust samples were dried in a clean, closed laboratory at room temperature (10°C-22°C) for seven days.The dried samples were screened with a 20 mesh stainless steel sieve (Tyler standard) to remove large particles and then screened with a 150 mesh sieve to collect total suspended particulate (particle size of less than 100 µm).Each sieve was used only once before a cleaning procedure.
A self-made resuspension installation (Fig. 3) was used to collect PM 10 and PM 2.5 from fugitive dust after pretreatment.For each process, approximately 100 g of dust was kept in a dust storage hopper in the resuspension chamber, and each dust sample was suspended twice to obtain eight PM samples (four filters for PM 10 and four filers for PM 2.5 ; half for duplicate samples).The resuspension chamber and related pipeline were cleaned before each resuspension process to avoid mutual interferences among different dust samples.
PM was collected from the dust samples by using a four-   channel particulate sampler (two channels for PM 10 and two channels for PM 2.5 ) at a flow rate of 16.7 L min -1 (TH-16A, Tianhong, China) and the sampling duration was 30 min.Quartz microfiber filters (47 mm in diameter, Pall, USA) were wrapped in aluminum foil and calcined at 450°C for 5 h to remove organic compounds; these filters were used to analyze the PM 2.5 mass concentrations, WSIIs, EC, and OC.Teflon filters (47 mm in diameter) were used to analyze elements.Before and after the sample collection, all filters were placed in a clean room (temperature: 20 ± 5°C, relative humidity: 50% ± 5%) for 48 h and then weighted in a microbalance (Mettler Toledo XS 205, Switzerland).Each filter was weighted at least twice, and the difference of the values was less than or equal to 0.03 mg.The filters were placed in cold storage at -18°C before analysis to prevent the loss of volatile components.

Analysis of Physical Properties
Fugitive dust was analyzed with a SEM (TM-1000, Hitachi).SEM is effective in not only the study of morphology, size, granulometric distribution, and other physical characteristics but also the studies on the sources of atmospheric particles and aerosol chemistry.

Chemical Composition Analysis
WSIIs (Na + , Ca 2+ , NH 4 + , K + , Mg 2+ , F -, Cl -, NO 3 -, and SO 4 2-) were analyzed with an ion chromatograph (ICS-90, ICS-900, Dionex).Quartz microfiber filters were placed in an individual beaker with 10 mL of pure water (18.2MΩ) and then extracted with an ultrasonic cleaner for 30 min by adding ice to keep the water temperature below 30°C.The extract was filtered with a 0.22 µm membrane filter (Menbrana, German) and then analyzed.The separation column and guard column were IonPacASll-HC4 mm and IonPacAGll-HC4 mm, respectively, for the anion determination.The eluent was composed of 8 mM Na 2 CO 3 and 1 mM NaHCO 3 mixed solution at a flow rate of 0.8 mL min -1 .IonPacCS12A and IonPacCG12A were used as the separation column and guard column, respectively, for the cation determination.The eluent was 20 mM methane sulfonate acid solution with 1.0 mL min -1 flow rate.
OC and EC of each sample were analyzed with a thermooptical method, and two 2.0 cm 2 membranes were punched from each quartz filter and used for analysis using a carbon analyzer (Sunset Laboratory, USA).OC and EC were tested consecutively through two temperature programs.For the first phase, the temperature was gradually increased to 840°C in a non-oxidizing atmosphere of helium.For the second phase, the temperature was gradually increased to 870°C in an oxidizing atmosphere of 98% helium with 2% oxygen.Pyrolysis products into a manganese dioxide (MnO 2 ) oxidizing oven, then quantitatively converted to CO 2 gas.CO 2 was directly measured by a self-contained nondispersive infrared detector system.The analyzer was calibrated via an installed tuned diode laser (red 660 nm).
Elements of the samples were measured by using an inductively coupled plasma mass spectrometer (ICP-MS), a powerful tool for high-precision trace element analysis (Jenner et al., 1990).Teflon filters of PM 10 and PM 2.5 were used for individual analysis of 24 elements (Al, V, Mn, Cu, Zn, Pb, etc.) via ICP-MS (Agilent 7500cx, Santa Clara, CA, USA).Teflon filters were cut into pieces, placed into individual nickel crucibles, and baked in a muffle furnace at 300°C for approximately 40 min.The temperature was gradually increased to 550°C and remained constant for 60 min.Then, the ashed sample was cooled to room temperature.A few drops of anhydrous ethanol and 0.2 g NaOH were added into the nickel crucibles, which were then baked in a muffle furnace at 500°C for 10 min.After cooling, 5 mL of pure water with a temperature of 90°C was added into each nickel crucible, which was then placed on a hot plate for extraction at 100°C.The extract was transferred to a plastic tube containing 2 mL hydrochloric acid solution (V HCl /V H2O = 1).The crucible was rinsed several times with 0.1 M hydrochloric acid to avoid sample loss, and the extracted solution was diluted to 50 mL.

Quality Assurance and Quality Control (QA/QC)
Field blank filters were analyzed to measure blank concentrations.Lab blank filters were measured 7 times by using the same analysis process to calculate method detection limit (MDL) (Long andMartin, 1990, Pekney andDavidson, 2005).For the sample concentration value below MDL, we replaced it with 0.5 * MDL, and this method was usually used in the previous studies (Croghan and Egeghy, 2003;Brown et al., 2015b).Each sample was measured twice, and the error was considered acceptable within 5%.
For WSIIs, standard curve was used to quantitatively analyze the samples, and the R 2 values of the standard curve for all the ions were higher than 0.9996 (except for NH 4 + , 0.9988).The MDLs ranged from 0.06 µg m -3 for NH 4 + to 0.51 µg m -3 for SO 4 2- . For the recovery test, the standard addition recovery of the nine ions ranged between 89% and 110%.
For OC and EC, in order to ensure the accuracy of the instrument, the 10 µL sucrose standard solution (10.0000 g L -1 ) was added to the blank filter, and was measured for every 10 samples.The instrument was accurate with the deviation between the measured value and the actual value (4.2 µg C µL -1 ) less than 5%.The method detection limits for both OC and EC was 0.2 µg cm -2 .
For elements, elements were added into the blank membranes to determine the standard recovery efficiency, and the values ranged from 65% (Ag) to 117% (V).The MDLs ranged from 0.001 µg L -1 (Ag) to 1.547 µg L -1 (Sb).The R 2 values of standard curve of the 24 trace elements were all better than 0.99.

Data Processing Methods
To study the deviations of chemical component between the profiles for different sites of dust, the coefficient of divergence (CD) was calculated as follow (Feng et al., 2007;Kong et al., 2011;Tan et al., 2014): where j and k are the two sampling sites, x ij and x ik represents the average concentration of chemical component i for sites j and k, and p is the total number of chemical components.If CD jk approaches zero, sources j and k are similar; if it approaches one, the sources are significantly different.In the previous study, Tan et al. (2014) showed that the CD values of 0.31, 0.54 and 0.4 indicated significant differences among gas pollutants and size segregated number from those three sites.Hence, the difference between two profiles was considerable when CD > 0.31.Particulate components are often reconstructed (Kong et al., 2011;Shen et al., 2014) to account for emission features, crustal minerals (CM, e.g., Al, Si, Ca, Fe, and Ti), organic mass (OM), secondary inorganic aerosols, and EC, which served as the major constituents of PM 2.5 and PM 10 .In the present study, two summary variables were calculated on the basis of chemical species.The first variable was the dust mass calculated using the crustal species (Malm et al., 1994): where Si is estimated on the basis of the Si-to-Al ratio The second variable is generally OM.OM is calculated as OC multiplied by 1.2 to 2.6 depending on the extent of oxidation and secondary organic aerosol formation (Chow et al., 2015).In this study, the value of 1.8 was applied for the sites.
Enrichment factors (EFs) are commonly used to evaluate the contribution of anthropogenic sources.Given EF > 10, the elements are mostly generated by anthropogenic sources (e.g., fossil combustion and transportation emission).When EF is near to 1, it points to a crustal origin (Nolting et al., 1999).In the present study, EFs were calculated by dividing the relative abundance of the selected elements in the PM samples by their average abundance in the upper continental crust (Zonta et al., 1994).The average crustal abundances were selected from a study on Chinese soil background (Wei, 1990), with Al serving as the reference element (Hsu et al., 2016).The EF was calculated as follows: where C is the element concentration.

Particle SEM Characterization of Dust Samples
The SEM images of four types of dust particles after pretreatment are shown in Fig. 4. The morphology of particle in dust sources showed different size and irregular shape.For SD large particle were dominant because of less anthropogenic interference.The RD was usually mixed with SD (Gunawardana et al., 2012).However, in this study RD size was smaller than SD, indicating that vehicle rolling and tire wearing.DD was generally affected by Fig. 4. SEM images of four dust samples after pretreatment.anthropogenic activities and mechanical operations, and it was smaller than SD as well.The dust from CE production differed from the other dust types.Parts of CE particles were shaped like a regular cube or ball and showed a certain viscosity in the aggregate form.Therefore, the particle size of CE was generally smaller than that of the other types of dust, and this dust type showed adhesion properties.Mineral dust particles with an irregular shape accounted for a large proportion of the fugitive dust samples, mainly from natural soil.Particles with spherical shapes were associated with anthropogenic sources such as construction activities, farming activities, and industrial emission.The shape of dust associated with either natural or anthropogenic source has been descripted by others (Li et al., 2016;Radulescu et al., 2017).

Particle Size Distribution of PM 10
The particle size-resolved number and mass concentration distribution of PM 10 in the four dust types were measured by ELPI+, and the results are shown in Fig. 5.For mass concentration, PM 2.5 accounted for less than 10% of the total PM 10 mass, and PM 1 contributed negligibly to the PM 10 mass concentrations in all the four dust types.The ratio PM 2.5 /PM 10 was highest in RD.For number concentration, ultrafine particle accounted for 64%, 58%, 51% and 34% of the total number of PM 10 in CE, SD, DD and RD, respectively.For PM 1 (aerodynamic diameter ≤ 1 µm) this ratio value was above 90% in all four dust sample, indicating that the number concentrations of PM 1 actually dominated the total particle number.The proportion of PM 0.1 number concentration in the PM 10 of CE was highest (64%), suggesting number concentration in CE relatively mainly distributed in the range of ultrafine particle.The differences in the proportions of number and mass concentration in different dust sources depend on dust type and their physical and chemical formation processes.

Chemical Characteristics of PM 10 and PM 2.5 for Different Fugitive Dust Types
The comparisons of the chemical compositions of PM 2.5 and PM 10 in the four types of fugitive dust are presented in Fig. 6.Chemical composition was classified into crustal elements, trace elements, WSIIs, OC, and EC.Crustal elements included Al, Mg, Fe, K, and Ti; and trace elements included Mn, B, Cr, Cu, Zn, Sr, Ba, Pb.OC contributed to approximately 11% for each fugitive dust type, except for CE dust (7%), in the PM 10 profiles.OC contributed to approximately 10% for both RD and DD, 12% for SD, and 3.9% for CE dust in PM 2.5 composition profiles.Crustal elements contributed to 15%, 14%, 17%, and 12% of the RD, DD, SD, and CE dust of the PM 10 composition profiles, respectively, and accounted for 8%-14% of the four types of fugitive dust in the PM 2.5 composition profiles.WSIIs contributed to 3%-8% for the four types of fugitive dust in PM 10 , and contributed 2%-5% in PM 2.5 , respectively.The elements of the four types of fugitive dust accounted for less than 1% of the PM composition profiles.
Crustal elements such as Al, Mg, Fe, and K were abundant in all samples.The element Al was the highest among the detected elements, in addition, K, Mg, and Fe were highly enriched; this result is similar to those of previous studies on samples of dust from CE production, RD and construction dust (Ho et al., 2003;Kong et al., 2011).Each crustal element in the four fugitive dust samples, except for Al in soil, had a higher percentage in the PM 2.5 samples than in the PM 10 samples; this outcome is similar to the research on resuspended RD by Chen et al. (2012).For the trace elements, the percentages of Zn, Mn, B, Cr, Sr, Ba, and Pb were high compared to other elements.The percentages of Mn, Zn, Ba, and Pb in the RD and SD samples were higher than those in the other two fugitive dust samples.Furthermore, the percentage of trace elements in PM  higher than that in PM 10 .The discharge from hightemperature processes is mostly in the form of fine particles emitted into the atmosphere and finally imported into fugitive dust via dry and wet deposition.Hence the result mainly indicates the influence of anthropogenic activities, including high temperature combustion of fossil fuels, and other high-temperature calcination processes, such as those in smelting plants, steel plants, and so on (Fan et al., 2016).
The ratios of WSIIs/PM 2.5 were higher than WSIIs/PM 10 , indicating WSIIs in fine particles more abundant than in the coarse fraction for the same sample.Similarly, the study researched by Wang et al. (2003) showed that K + was mostly concentrated in fine particles ranges.The SO 4 2concentration was higher than the other ion concentrations in all fugitive dust samples, as similarly determined by other studies (Gupta et al., 2007;Kong et al., 2011).Compared with the other three types of fugitive dust samples, the CE dust sample showed higher SO 4 2-and Ca 2+ percentages.The percentage of NO 3 -in the RD sample was higher than that in the other three fugitive dust samples.This result was likely related to vehicle emissions.Nitrogen oxides emitted by vehicles, and then transformed into NO 3 -by photochemical reaction in atmospheric environment.
The percentage of OC was the highest in all fugitive dust samples except in CE samples, and the value in DD and CE might be overestimated without acid modification (Fig. 6).For SD, the percentage of OC in PM 2.5 was higher than that in PM 10 , as expected.For RD, the percentages of OC in PM 10 and PM 2.5 were both more than 10%.The same result of relative high content of OC in the resuspended dust was reported by Han et al. (2011) and the reason was the influence by the deposition of PM in the atmosphere and anthropogenic source of PM near the sampling site (for example, vehicle emissions).In addition, the percentage of EC in the CE dust sample was higher than those in the other three fugitive dust samples.
The PM 2.5 and PM 10 composition profiles for four kinds of fugitive dust samples from Zhengzhou and other cities are listed in Tables 1 and 2. NO 3 -content in the PM composition profiles for Zhengzhou was higher than those in the PM composition profiles for Hong Kong (Ho et al., 2003), Mexico (Vega et al., 2001), and Fushun (Kong et al., 2011).These results may be related to the increase in vehicle emissions in recent years.SO 4 2-and NH 4 + content in the PM composition profiles of the fugitive dust samples in this study, except in the profile of CE, were higher than those in the PM composition

Species
This study Fushun (Kong et al., 2011) Hong Kong (Ho et al., 2003) RD SDSS is for surface layer soils around the city; RDCS is for dust from the city's main streets; CDBS is for dust collected at building sites.
profiles of Hong Kong (Ho et al., 2003), Mexico (Vega et al., 2001), and Fushun (Kong et al., 2011).Other WSIIs in the PM composition profiles of Zhengzhou were almost similar to those reported in the previous studies (Vega et al., 2001;Ho et al., 2003;Kong et al., 2011).In this study, content of several elements in PM composition profiles was different from those for other cities.Mg content in PM 2.5 composition profiles for Zhengzhou was higher than those reported for Mexico (Vega et al., 2001) and was lower in PM 10 composition profiles than those in Fushun (Kong et al., 2011).Al content in SD (4.4 ± 0.1%; 5.2 ± 0.8%) was nearly one third of that in Hong Kong (11.5 ± 1.0%; 15.1 ± 4.3%) (Ho et al., 2003) both for PM 2.5 and PM 10 , and that in RD (2.9 ± 1.4%) was less than one half of that in both Fushun (6.5 ± 3.3%) (Kong et al., 2011) and Hong Kong (7.4 ± 2.1%) (Ho et al., 2003) for PM 10 .Fe content in this study was lower than that in Hong Kong (Ho et al., 2003) for PM 2.5 and PM 10 composition profiles but was higher than that in Fushun (Kong et al., 2011) for PM 10 composition profiles.Ti content in this study was lower than those in Hong Kong (Ho et al., 2003), Mexico (Vega et al., 2001), and Fushun (Kong et al., 2011) for PM 2.5 and PM 10 composition profiles.These results of contrast indicated that crustal elements content in PM dust profiles of these cities differed with each other.Zn and Cu content were lower in the PM profiles of RD for Zhengzhou compared with Hong Kong (Ho et al., 2003).There were almost no obvious differences for other elements in the PM 2.5 and PM 10 profiles in this study and in other studies.
OC content in the PM 2.5 and PM 10 profiles of CE from Zhengzhou was higher than that in CE from Fushun (Kong et al., 2011) and Hong Kong (Ho et al., 2003), but was considerably lower than those in the PM 2.5 profiles of CE from Mexico (Vega et al., 2001).CE and DD from Zhengzhou had relatively high OC content, which may be related to the decomposition of CaCO 3 in CE and DD.Under OC test conditions, the temperature was gradually increased to 870°C.At this temperature, CaCO 3 will decompose into CaO and CO 2 (Basu et al., 2011).Therefore, CO 2 from CaCO 3 decomposition resulted in OC contents in CE and DD overestimated.The overestimation phenomenon of OC by using the same test method in CE also existed in the previous study (Vega et al., 2001) without enough attention.The OC content of road dust in PM 10 composition profiles was higher than that in Fushun (Kong et al., 2011) and Hong Kong (Ho et al., 2003), which may be related to the increasing motor vehicle population influence in Zhengzhou.The OC content of road dust in PM 2.5 composition profiles was lower than that in Hong Kong (Ho et al., 2003) but higher than that in Mexico (Vega et al., 2001).

Comparability among Fugitive Dust Samples
The CD values for PM 2.5 and PM 10 of the four fugitive dust samples were calculated using chemical profiles tested (Table S2 in Supplemental Materials).CD values of RD profiles ranged from 0.29 to 0.60 for PM 2.5 and ranged from 0.21 to 0.68 for PM 10 .A minimum CD value (0.21) was presented between RD-4 and RD-5, both provincial roads with low traffic (Table S1 in Supplemental Materials), which suggested the two sites with less difference in the chemical profiles of dust.However, other profiles of RD sites showed obvious differences.CD values of DD ranged from 0.26 to 0.56 for PM 2.5 and ranged from 0.28 to 0.52 for PM 10 .Three profiles of SD were almost dissimilar to each other for both PM 2.5 and PM 10 , except SD-1 and SD-2 for PM 10 with CD value (0.21) less than 0.31.Three profiles of CE were also dissimilar, with CD values higher than 0.31.Besides, the CD values among chemical composition profiles of four kinds of fugitive dust were listed in Table S2(e) in Supplemental Materials, with obvious differences.Thus, the average profile was usable and significant differences existed widely among chemical profiles at different sites.As a result, special source profiles should be set up in special regions.

Reconstructed Chemical Composition
The reconstructed chemical composition based on the average concentrations of fugitive dust in PM 2.5 and PM 10 is shown in Fig. 7.The total reconstructed masses of RD, DD, SD, and CE for PM 2.5 were 78%, 77%, 89%, and 69% respectively.CM was the most abundant species, accounting for 47%-56%, and OM was the second major component in PM 2.5 .The ratios in PM 2.5 are as follows: SD (22%) ˃ RD and DD (both 18%) ˃ CE (7%).The total reconstructed masses of RD, DD, SD, and CE for PM 10 were 69%, 81%, 95%, and 79%, respectively.CM was the most abundant species, accounting for 40%-69% in PM; its average value (55%) in PM 10 was slightly higher than that in PM 2.5 (51%) and close to the values four cities of southwestern China (56%-68%) (Liu et al., 2016).The ratios (OM/PM 10 ) of RD and SD were both approximately 20%.The ratios of the other components, including SO 4 2-, NH 4 + , NO 3 -, EC, etc., in PM 2.5 and PM 10 were all less than 5%; the small fraction, however, should not be neglectful because of the environmental and health effects of these components, especially EC and heavy metals (Jacobson, 2001;Hu et al., 2012).

Enrichment Factors of Trace Elements
The calculated EFs of the elements of fugitive dust in this work are displayed in Fig. 8. Cd, Ag, and Zn were significantly enriched with high EFs (between 11 and 380) for all the four types of dust (excluding PM 10 of CE), thereby indicating the three elements from anthropogenic sources.For example, the EFs of Cd were highest in RD, with the value of 259 and 202 for PM 2.5 and PM 10 , respectively, suggesting the effect of vehicle, in accordance with the previous study, as Cd mainly derived from wear (Sternbeck et al., 2002).The EFs of Si, K, Ti, V, and Fe were close to unity, thus suggesting a crustal origin.The EFs of Ca, Mg, Mn, Cr, Ba, As, Co (excluding SD of PM 2.5 ), Sr, and Pb (including DD and CE for PM 2.5 and PM 10 , and SD for PM 10 ) were lower than 10 and higher than 1, thereby indicating that these elements were emitted from both natural and anthropogenic sources.However, the Pb content of RD and SD in PM 2.5 with EFs values of 30 and 12, respectively, was generated by anthropogenic sources, suggesting the effect of smelting and coal combustion processes (Zhang et al., 2009).Cd, Mo, Cu, and Ba were most enriched in RD,  thereby suggesting the effect of vehicle emissions, i.e., tailpipe emissions of Cd and Cu (Cadle et al., 1999;Sternbeck et al., 2002), and abrasion emissions of Cu, Ba, and Mo from brake linings and tire tread wear (Garg et al., 2000;Bozlaker et al., 2013).

CONCLUSIONS
This study obtained the geological dust source profiles of RD, DD, SD, and CE for PM 2.5 and PM 10 in Zhengzhou and analyzed the physical and chemical characteristics of different dust sources.Based on SEM results of the original dust, CE particles with a regular cube or ball were generally smaller than that of the other types of dust.Particles of RD and DD were smaller than SD, indicating more influence by anthropogenic activities.For mass concentration, PM 2.5 accounted for less than 10% of the total PM 10 mass, and PM 1 contributed negligibly to the PM 10 mass concentrations in all types of dust.For number concentration, PM 1 actually dominated the total particle number, accounting for more than 90% in the four types of dust, and the proportion of PM 0.1 number concentration in the PM 10 of CE was highest, suggesting number concentration in CE relatively mainly distributed in the range of ultrafine particle.
Crustal element was one of the most dominant for four fugitive dust samples, accounting for 8%-14% of PM 2.5 and 12%-17% of PM 10 with the ratio different from other cities.Moreover, CM was the most abundant species in the result of reconstructed chemical composition, accounting for 40%-69% in PM, with average value (55%) in PM 10 slightly higher than that in PM 2.5 (51%).The percentages of OC and NO 3 -in PM 10 and PM 2.5 for RD were relative high, indicating the influence by the deposition of PM in the atmosphere and vehicle emissions.However, OC content in CE and DD was also high because of overestimate.The content of EC and heavy metals was less than 5%, which still should not be neglectful because of the environmental and health effects.The PM 2.5 and PM 10 profiles from four types of fugitive dust were different from each other, with all CD values more than 0.31.Moreover, the PM profiles from the various sites for the same type dust are mostly different, with few CD values relative low (e.g., RD-4 and RD-5 with CD of 0.21).Hence, the specific source profiles, and not average value for each type dust should be used for source apportionment according to the site location, instead of average profile.Cd, Ag, and Zn were significantly enriched from anthropogenic sources, with EFs ranging from 11 to 380.The Pb content of RD (EF of 30) and SD (EF of 12) in PM 2.5 was generated by smelting and coal combustion processes.Cd, Mo, Cu, and Ba were most enriched in RD, thereby suggesting the effect of vehicle emissions.This study can provide essential input profiles for source apportionment of PM in Zhengzhou, eventually playing an important role for the government to improve atmospheric environment quality.

Fig. 3 .
Fig. 3. Sketch map of dust resuspension installation and sampling equipment.

Fig. 5 .
Fig. 5. Size distribution of PM 10 of the four types of dust sources (RD: road dust; DD: demolition dust; SD: soil dust; CE: cement).

Fig. 6 .
Fig. 6.Chemical species concentration of PM 2.5 and PM 10 of fugitive dust.

Fig. 7 .
Fig. 7. Reconstructed chemical composition of PM 2.5 and PM 10 for the four types of fugitive dust (RD: road dust; DD: demolition dust; SD: soil dust; CE: cement).

Table 1 .
Comparison of PM 2.5 composition profiles for fugitive dust samples with other studies (percentage, %).

Table 2 .
Comparison of PM 10 composition profiles for fugitive dust samples with other studies (percentage, %).