Seasonal Variations and Sources of 17 Aerosol Metal Elements in Suburban Nanjing, China

Abstract

In this work, the seasonal variations and sources of trace metal elements in atmospheric fine aerosols (PM_2.5) were investigated for a year-long field campaign from July 2012 to June 2013, conducted in suburban Nanjing, eastern China, at a site adjacent to an industry zone. The PM_2.5 samples collected across four seasons were analyzed for 17 metal elements, namely, Sodium (Na), Magnesium (Mg), Aluminum (Al), Vanadium (V), Chromium (Cr), Manganese (Mn), Nickel (Ni), Copper (Cu), Zinc (Zn), Arsenic (As), Selenium (Se), Strontium (Sr), Cadmium (Cd), Barium (Ba), Lead (Pb), Molybdenum (Mo), and Antimony (Sb) using an inductively coupled plasma mass spectrometry (ICP-MS). We found that the total concentration of all 17 metal elements was 1.23 μg/m³, on average accounting for 1.0% of the total PM_2.5 mass. For our data, mass concentrations of Al, Cd, Ba were highest in summer, Mg, Cu, Zn, Se, Pb peaked in autumn, Cr, Mn, Ni, As, Sr, Sb increased significantly in winter, while the concentrations of Na, V, Mo were at their highest levels in spring. Air mass back trajectory analysis suggested that air parcels that arrived at the site originated from four dominant regions (Japan, yellow sea and bohai; Southeast of China, the Pacific Ocean; Southwest of Jiangsu and Anhui province; Northern Asia inland and Mongolia region), in particular, the one from Northern Asia inland and Mongolia contained the highest concentrations of As, Sb, Sr, and was predominant in winter. Positive matrix factorization (PMF) analyses revealed that the industrial emission is the largest contributor (34%) of the observed metal elements, followed by traffic (25%), soil dust (19%), coal combustion (10%), incineration of electronic waste (9%), and a minor unknown source (3%). In addition, we have also investigated the morphology and composition of particles by using the scanning electron microscopy (SEM)/energy-dispersive spectrometry (EDS) techniques, and identified particles from coal burning sources, etc., similar to the PMF results.

1. Introduction

Ambient particulate matter (PM) pollution, is emerging as one of the leading risk factors of mortality, ranking in the sixth place in 2010 [1,2] from the 13th place in 2000 among all risk factors [3]. Atmospheric PM came not only from natural sources, including volcanic eruptions, crustal materials, surface dust, sea spray, wildfires, etc., but also from a variety of anthropogenic sources such as coal combustion, vehicular emissions, biomass burning and industrial emissions, and so on [4,5,6,7,8]. Various metal elements are ubiquitous constituents of ambient PM, and most of them are chemically stable, thus the transport and dispersion of atmospheric aerosol particles can largely expand their influences [6,9,10]. For instance, as atmospheric fine particles can travel over a long distance downwind from the sources [11,12,13], the airborne metallic species can affect the terrestrial and aquatic ecosystems, at both regional and global scales via dry and wet depositions [14,15,16,17,18,19]. Inhalation of the airborne metallic species is also detrimental to human health. Hence, a detailed elucidation of the temporal variations and sources of trace metals is of central importance to evaluate their environmental and health consequences [20,21].

Monitoring networks for PM-metal elements have already been established at urban, suburban, rural, mountain, forest, industrial sites, or near heavy traffic roads in many countries including America [22,23], India, United Kingdom [24] and other European countries [25,26,27]. Previous studies that focused on the distribution of metal elements in different sizes of aerosol particles [9,28,29,30] found that elements such as aluminum, iron, strontium, zinc mainly existed in coarse particles, whereas vanadium, nickel, lead, cadmium mostly resided in fine particles [31,32]. Sources of metal elements have been investigated by methods including Enrichment Factors (EF), Principal Component Analysis (PCA) and Positive Matrix Factorization (PMF), etc., and typical sources identified include coal combustion, vehicle emission, mineral dust and so on [13,33,34,35]. The connections between atmospheric deposition of metal elements and contamination of soil and bryophyte have also been investigated [12,36]. Furthermore, the bioavailability, health risk evaluation, and epidemiological inter-disciplinary studies of metal elements were also reported previously [2,37,38].

Recently, air pollution has become a serious and pressing issue along with the fast economic development in China. However, the airborne heavy metal pollution still remains not fully understood, in particular, the detailed emission inventories due to the fact that the corresponding monitoring system has not been set up [39,37]. Moreover, many studies that focused on analysis of the concentrations and size distributions of metal elements were conducted in a few major heavily polluted megacities such as Beijing, Shanghai and Guangzhou, and covered relatively short time periods [40,41,42,43,44,45]. Nanjing, as the second largest city in the Yangtze River Delta region in Eastern China, with a river shoreline of about 2 hundred kilometers, an area of 752.83 square kilometers and a population of 8.19 million, also suffered serious air pollution in the past decades due to increasing emissions from traffic (gasoline and diesel combustions, tire and brake lining wear, etc.), industry (petrochemical activities), coal burning, and so on [46,29]. However, studies regarding the temporal, seasonal variations and source contributions of atmospheric metal elements in Nanjing are very scarce.

Previously, we have investigated the mass loadings and sources of 15 trace metals in the PM_2.5 samples during the second Asian Youth Games period in urban Nanjing [47]. Here, we further performed analyses of 17 metal elements in PM_2.5 samples collected near one-year period from 28 July 2012 to 31 June 2013 in suburban Nanjing. The metal elements analyzed include sodium (Na), magnesium (Mg), aluminum (Al), vanadium (V), chromium (Cr), manganese (Mn), nickel (Ni), copper (Cu), zinc (Zn), arsenic (As), selenium (Se), strontium (Sr), cadmium (Cd), barium (Ba), lead (Pb), molybdenum (Mo) and antimony (Sb). In this work, we aim to: (1) determine the mass concentrations of 17 metal elements, and their contributions to the total PM_2.5 mass loading; (2) investigate the seasonal variations by combining with the meteorological parameters; and (3) identify the major sources of aerosol metal elements in this region using PMF model. Our results are valuable for the implementation of appropriate measures to reduce the atmospheric metal and PM_2.5 pollution in this region.

2. Methodology

2.1. Site Description

The sampling site is located at the campus of Nanjing University of Information Science and Technology in suburban Nanjing (32.21°N, 118.72°E), in the provincial capital of Jiangsu Province, as shown in Figure 1. The site is east of a rural field, and east/northeast of an industry zone with a large number of factories including iron and steel making plants, petrochemical and chemical engineering factories, and electronic plants, etc. The site is also influenced by other human activities, such as traffic, cooking, and biomass (straw and crop) burning, etc.

During the sampling period, meteorological parameters (air temperature, relative humidity, wind speed, wind direction) were recorded (Figure 2) at the meteorological station located at the same site.

2.2. Sample Collection and Preparation

The PM_2.5 samples were collected in four seasons, including Summer (23 August–8 September 2012), Autumn (17–28 November 2012), Winter (25–30 January 2013), and Spring (20 May–6 June 2013). Note that although we only collected samples for a limited number of days in each season, the sampling days and results should be representative of the corresponding seasons—as shown clearly in Figures S1–S5 in the supplement, the meteorological parameters, including wind speed, wind direction, temperature and RH, averaged over the sampling days are very similar to the corresponding seasonal average values. In total, 64 PM_2.5 samples (details in Table S1) were collected on the Whatman mixed cellulose ester filters (GE healthcare UK limited, Amersham place, Little Chalfont, England) in summer, autumn, spring and on the Whatman quartz fiber filters in winter (9 cm diameter) using a median volume atmospheric particle sampler (HY-100, Qingdao Henyuan Instruments Co. Ltd., Qingdao, China) equipped with a PM_2.5 cyclone. The flow rate was 100 L·min⁻¹, and was calibrated every day during sampling.

All filters before sampling were firstly baked in a muffle furnace (450 °C) for four hours, wrapped in tinfoil and put into a constant temperature (at 25 °C) and relative humidity (50%) chamber for 48 h, and then were weighed. The sampling time for each sample was 12 h (we did not observe significant differences between daytime and nighttime samples as the sampling time resolution is low and metals are relatively stable, we reported results below without daytime-nighttime comparisons). Sampling was not conducted at rainy and snowy days as wet deposition can efficiently scavenge the atmospheric pollutants and high air relative humility may damage the filters. After sampling, the filters were equilibrated in the constant temperature and relative humidity chamber for 48 h, weighed, placed individually inside zip-sealed bags, and were then stored in the refrigerator at −4 °C in laboratory prior to analysis. Note Illuminati et al. [48] provided a very strict, precise aerosol mass measurement procedure with careful decontamination step, and filter blanks were also treated with more accuracy test—to ensure the accuracy of the measured aerosol mass concentrations. These procedures will be applied to our future studies if possible.

2.3. Sample Analyses

For the measurement of metal elements in environmental samples such as aerosol particles, soil, and municipal water, microwave digestion has recently been widely employed for sample pretreatment [47,49,50,51]. In this study, a MARS6 Xpress CEM corporation Microwave Digestion System was used to digest the PM_2.5 constituents. About 1/8 of a filter was weighed, cut into pieces and put into the PTFE pressure digestion tank, and then dissolved in 5 mL 69% nitric acid (HNO₃) and 1 mL 49% hydrofluoric acid (HF). The digestion tanks (16 samples in a batch) were sealed and processed according to the programmed procedure consisted of three steps: (1) raise the temperature to 190 °C in 20 min; (2) maintain the temperature at 190 °C for 25 min to achieve a complete digestion; (3) cool down to room temperature in about 30 min. The digested solutions were then transferred into a vacuum filtration system, to remove the excess HF and HNO₃ solutions, and then diluted to 45 mL with purified water in a 45 mL centrifuge tube kept at 4 °C. In addition, blank filters were treated in the same way. All sample containers are made of polypropylene, high-density polyethylene Teflon to minimize possible contaminations.

The concentrations of 17 metal elements in the solutions were analyzed using an Thermo Fisher X2 Series Inductively Coupled Plasma Mass Spectrometry (ICP-MS), which has an ICP source, Optimization system (extracting and focusing of the ion beam), a quadrupole mass spectrometer and a detector. The instrument has high mass sensitivity and selectivity, and was calibrated daily according to the standard protocol provided by the manufacturer. External multi-element standard solution (XCCC-13A, XCCC-14B, SPEX CertiPrep Group, Metuchen, NJ, USA) was used to establish a four-points (2 ppb, 5 ppb, 10 ppb, 18 ppb) standard curve, and internal standard, which was 2 ppb ⁴⁵Sc, ¹¹⁵In mixed solution (60201SS, Inorganic Ventures Corporation, Christiansburg, VA, USA) was used to eliminate the interferences of other elements to the targeted elements.

2.4. Quality Assurance and Quality Control

Quality assurance and quality control procedures were applied during sampling, handling and measurement steps, in order to minimize possible artifacts. Two certified materials of soil (GBW07403) and fly ash (GBW08401) were measured in parallel to ensure high quality. NIST 1648 and 1648a as pointed by Illuminati et al. [49] can also be used as urban particulate reference material to check the accuracy of the measurements, which should be investigated further in the future. Nevertheless, under the current digestion and the analysis procedures mentioned above, all elements are within the target recovery of 100% ± 15%. Four levels of calibration standards were used, and the correlation coefficients of standard curves were all greater than 0.999.

In this work, the limit of detection (LOD) of each element was calculated according to ICH Q2B (ICH 2005) using the following equations [52]:

LOD = 3.3 Sa/b

where Sa = standard deviation of the intercept of regression line and b = slope of the calibration curve [53,54]. The LOD in air was calculated by the formula: LOD in air (ng·m⁻³) = concentrations in solution (μg·L⁻¹) × 0.045 L/72 m⁻³. By using this approach, the LODs of Na, Mg, Al, V, Cr, Mn, Ni, Cu, Zn, As, Se, Sr, Cd, Ba, Pb, Mo, and Sb were determined to be 2.36, 0.98, 1.20, 0.17, 1.09, 0.88, 1.19, 1.08, 1.15, 0.92, 0.54, 0.98, 1.14, 0.27, 0.95, 0.19 and 0.25 ng·m⁻³, respectively. The average concentrations of all 17 elements in the PM_2.5 samples were higher than the LODs. The blank filters were treated in the same way as for the samples, and the mean values of Na, Mg, Al, V, Cr, Mn, Ni, Cu, Zn, As, Se, Sr, Cd, Ba, Pb, Mo, Sb in blank filter of mixed cellulose ester were 34, 11.4, 16.8, 0.15, 1.2, 0.55, 0.3, 0.12, 1.57, 0.85, 0.11, 0.03, 0.08, 0.25, 0.06, 0.19, 9.2 ng·m⁻³, respectively; the background data of quartz fiber filter were 22.31, 8.55, 3.75, 0.08, 0.33, 0.14, 0.78, 0.85, 4.88, 0.19, 0.12, 0.27, 0.01, 0.3, 0.07, 0.05, 5.6, 9.2 ng·m⁻³, respectively. These background values were well below the corresponding values in the filter samples, indicating that filter itself did not interfere the observed concentrations significantly. All results reported in this study were subjected to blank filter correction.

2.5. Scanning Electron Microscopy

Particle morphology of aerosol samples was investigated by using the scanning electron microscopy (SEM) (S-3400N II, Hitachi High. Technologies Corporation, Tokyo, Japan). The SEM technique can show the detailed morphology of a single aerosol particle. The samples were analyzed by cutting 1 cm² of the filter, and mounted onto an Al stub, coated by gold nanoparticles, and then placed into the instrument operated with a resolution at 20 kv, 1280 × 960 pixel. Compositions of the SEM-imaged particles were also analyzed by the Energy Dispersive Spectrometer (EDS) (Hitachi, Tokyo, Japan).

2.6. PMF Analysis

Positive matrix factorization (PMF) [55,56] is a multivariate factor analysis model recommended by the EPA. A unique advantage of the PMF algorithm is that the model considers the measurement uncertainties corresponding to the observed data to weigh the data points, and thus output the source profiles that can well represent the real characteristics. Of course, Principal Component Analysis (PCA) related methods (e.g., Padoan et al. [57]) are also useful for the source apportionment analyses. In this study, we used EPA (Environmental Protection Agency) PMF 3.0 to identify and apportion possible sources to the observed metal elements. The data matrix X of the samples can be represented as the product of two matrices (the concentration series of G and the factor profiles F) plus a residual matrix E that accounts for the unexplained portion of the measured data [46].

$${\mathrm{x}}_{ij}={\displaystyle \sum _{k=1}^p{g}_{ik}}{f}_{kj}+e_{ij}\left(i=1,\dots ,n;j=1,\dots ,m;k=1,\dots ,p\right)$$

$$\mathrm{Q}={{\displaystyle \sum _{i=1}^n{\displaystyle \sum _{j=1}^m\left[\frac{e_{ij}}{u_{ij}}\right]}}}^2$$

Supposing matrix X is n × m (n rows are concentration series of the samples; m columns are the elements determined), G is the n × p matrix (p, numbers of possible sources) and F is a p × m matrix. u_ij is the estimated error (standard deviation) of the corresponding data points in the matrix X. One constraint of PMF model is that the data of metal elements are all non-negative. Both the data and error matrices were pre-processed following the protocols detailed in the EPA PMF user manual. The PMF algorithm was run in robust mode, and a few outliers were excluded during calculations, in order to eliminate unreasonable influences of these data on the PMF results.

3. Results and Discussion

3.1. Characteristics of Metal Elements

3.1.1. Annual Characteristic of Metal Elements

The measured mass concentrations (minimum, maximum, mean, median, standard deviation, skewness and kurtosis) of the 17 metal elements and PM_2.5 were summarized in

Table 1. Concentrations of PM_2.5 metal elements in suburban Nanjing during 2012–2013 (ng/m³).

	Min	Max	Mean	Medium	SD	Skewness	Kurtosis
PM2.5/104	4.56	40.30	12.50	17.50	6.42	1.59	-
Na	170.53	777.89	369.10	355.71	122.17	1.02	1.23
Mg	137.89	453.72	246.40	243.28	67.85	0.59	0.42
Al	2.96	607.16	114.95	127.71	95.92	2.22	10.31
V	0.84	8.82	3.95	3.81	1.81	0.55	−0.07
Cr	5.61	54.63	24.76	19.60	15.48	0.28	−1.51
Mn	17.00	153.54	58.97	56.02	30.05	0.62	0.32
Ni	3.53	18.89	8.10	7.01	3.53	1.32	1.49
Cu	7.80	206.20	29.44	22.27	29.30	4.17	22.20
Zn	64.87	850.00	254.80	237.98	167.23	2.89	11.80
As	1.47	21.06	6.30	4.06	4.95	1.26	0.62
Se	1.23	8.41	4.51	4.07	1.91	0.50	−0.74
Sr	2.56	15.14	5.99	4.92	2.87	1.10	0.74
Cd	0.46	6.62	2.09	1.72	1.38	1.64	3.19
Ba	7.52	65.57	20.00	16.25	10.44	1.88	4.93
Pb	15.26	145.50	67.37	66.18	25.53	0.36	0.38
Mo	0.96	51.19	8.48	6.74	8.42	3.02	11.55
Sb	0.31	10.14	2.37	1.50	2.30	1.82	3.08
Total metal	571.04	3324.24	1227.58	1178.79	-	-	-

SD: Standard deviation.

. Concentrations of four elements, e.g., Al, Cu, Mo, Sb, varied dramatically in 2 orders of magnitude in different days, while variations of other elements were on one order of magnitude. The skewness coefficients of 13 elements were within 0~2, indicating that there is little right-skewed distribution in the samples; hence the median values were typically lower than the mean values except for Al. Most of the kurtosis coefficients were larger than zero. This result demonstrates that concentration distribution of samples was narrower and steeper than the normal distribution, especially for the elements featured with relatively high concentrations, such as Al (10.31), Cu (22.2), Zn (11.8) and Mo (11.55), likely indicating emission sources of soil dust, traffic and Mo source (detailed in Section 3.2) in suburban Nanjing.

The PM_2.5 mass concentrations ranged from 45 to 403 μg/m³ with an average mass loading of 125 μg/m³, and the total concentration of 17 metal elements was 1.227 μg/m³, accounting for 1.0% of the total PM_2.5 mass. These findings are similar to a study conducted at Ji’nan industrial and urban sites (130 μg/m³ of PM_2.5 mass concentrations and 1.62 μg/m³ of metals), where ferrous metal smelting, traffic equipment production, and petrol processing are large emitters of metals, in particular Sb, Cd, As, Se [58]. In the samples (Table 1), Na, Mg, Al were the crustal elements, with higher concentrations than other elements except for Zn. Zn was found to be the most abundant heavy metal element, with concentrations ranging from 64.87 to 850 ng/m³ and an average of 254.80 ng/m³. Pb and Mn also had relatively high concentrations, on average, of 67.37 ng/m³ and 58.97 ng/m³, respectively, following by Cu, Cr, and Ba, all of which varied in a relatively wide range. The concentration of Pb in Nanjing was far below the limits of current National Ambient Air Quality Standard (NAAQS) of China (GB3095-2012) and that of World Health Organization (WHO) of 500 ng/m³. The average concentrations of Mo, Ni, As, Sr, Se, V, Sb, Cd were all less than 10 ng/m³. However, the average concentration of As (6.3 ng/m³) was higher than the NAAQS limit of 6 ng/m³. Details regarding the sources of metal elements are presented in Section 3.2.

Table 2. Concentrations of metals in Nanjing PM_2.5 and aerosol metal concentrations in a selection of previous studies of China and other countries.

	Na	Mg	Al	V	Cr	Mn	Ni	Cu	Zn	As	Se	Sr	Cd	Ba	Pb	Mo	Sb
Nanjing	390	255.8	121.3	4.1	28.7	65.5	8.1	32.2	281.9	7.2	4.7	6.2	1.9	18.3	68.2	9.5	2.8
Beijing, China (2011), suburban [35]	1449	637	2182	25	21.8	35.6	20.2	10	27.8	2.9	1.1	9.8	1.1	18.4	52.9	6.5	0.6
Switzerland (2006), rural [46]	298	48	91	0.7	-	2.8	1.2	6	-	0.53	0.16	-	0.32	-	-	-	-
Cordoba, Argentina (2010), rural [23]	1928	-	-	-	-	95.78	5.71	5.55	17.66	1.26	0.36	-	0.8	45.8	4.63	-	-
Los angles, USA (2009), rural [23]	66.5	30.8	107	2.36	1.48	3.76	1.23	8.93	10	0.27	-	1.16	0.09	5.44	2.52	0.45	1.08
Xi’an, China (2012), urban campus [60]	-	-	-	68	145.4	538.9	31.2	57.4	319.5	10.2	-	-	-	914.6	131.8	-	-
Ji’nan, China (2010), industrial [58]	-	-	670	-	10	110	10	40	440	60	-	10	-	20	200	-	-
Zhuzhou, China (2012), industrial [33]	-	-	-	-	115	-	35	98	1140	42	-	-	10.3	-	254	-	9.8
Oxford, UK (2007), urban [31]	-	-	-	3.9	1.2	3.5	67.3	39.5	30	-	-	25.4	1.06	-	186	0.4	-
Shanghai, China (2012), urban [39]	-	-	-	10.3	27.3	60.3	10	35.5	418.5	30.8	-	-	2.9	-	108.5	-	-
Mountain XL, China (2012) [29]	888	540	323	18.9	78.6	29.4	12.9	43.7	140	-	5.09	-	11.26	33.3	57.2	1.07	21.3
S-E Finland (2006), forest [28]	-	100	-	2.7	-	7.3	1.3	2.6	20	0.85	-	-	-	-	13	-	-
Venice Lagoon (2002) [59]	370	-	400	12	-	16	14	10	60	3	-	2	2	-	19	-	-
Los angles, USA (2009), urban [23]	41.4	11.8	46.5	3.72	1.22	1.78	1.38	9.02	8.05	0.22	-	0.83	0.06	6.22	1.79	0.44	1.17

compares our results with previous measurements of the metals in PM_2.5 samples collected from various sites, including rural, urban, industrial, forest, suburban, mountain, and island sites cities in China, Europe and United States [28,31,33,59,60]. Overall, the concentrations of dust-related metal elements in suburban Nanjing were at similar levels compared with those from some rural sites in foreign countries like Switzerland, Los Angles, USA, but with higher concentrations of some heavy metals derived from vehicle and industrial emissions, although levels of these metals appear to be lower than those in a few cities, such as Xi’an, Ji’nan and Zhuzhou [33]. In details, concentrations of Na, Mg, Al measured in this study were much lower than those found at the rural sites of China, while generally higher than the reported concentrations in other countries such as Switzerland [46] and Los Angles, USA [23]. As these three elements are relevant to the soil or dust-derived sources, it is obvious that their concentrations were higher in rural area (Table 2). Concentrations of V, Mn, and Ni were closer to those measured in rural site, lower than in urban locations. The average concentrations of Cu, Se, Cd, As and Sb were higher at urban sites than those at rural sites presumably indicating the significant contributions from local emission sources to these elements in Nanjing. Zn and Pb behaved differently from other elements, which is discussed in detail later.

Overall, our results together with previous observations in Table 2 showed that particle-phase metal concentrations in urban/industrial areas were much higher than those in rural/mountain areas, indicating that anthropogenic emissions such as industrial production contribute significantly to urban atmospheric metals. For example, in industrial areas of Ji’nan [58] or Zhuzhou, China, the metal elements like Mn, Cu, Zn, As, Pb were several orders of magnitude higher than those at mountain or rural sites, such as Mountain XL and rural areas near Beijing [29]. As mentioned earlier, high concentrations of As exceeding the limits of 6 ng/m³ of the national standard were observed in many cities in China, in particular Hangzhou (120 ng/m³), Zhengzhou (185.2 ng/m³) [39], Ji’nan (60 ng/m³), Shanghai (30.8 ng/m³), and Zhuzhou (42 ng/m³).

3.1.2. Seasonal Variations of Atmospheric Metals

The average seasonal mass concentrations (±SD) of PM_2.5 and the 17 metal elements are shown in

Table 3. Seasonal variations of atmospheric PM_2.5 metals (mean ± SD, ng/m³) during the study period. (SD: Standard deviation).

	Summer	Autumn	Winter	Spring
PM2.5/104	8.2	16.9	16.8	11.3
Na	312.73 ± 134.34	307.24 ± 44.63	395.39 ± 77.93	444.79 ± 114.94
Mg	224.21 ± 85.21	271.05 ± 66.55	262.36 ± 55.66	252.37 ± 50.29
Al	140.58 ± 129.15	97.74 ± 71.56	113.2 ± 51.34	87.84 ± 87.42
V	3.48 ± 2.21	3.08 ± 1.27	4.19 ± 0.99	4.77 ± 1.74
Cr	9.84 ± 4.2	39.21 ± 2.22	39.21 ± 6.97	37.81 ± 10.39
Mn	37.97 ± 22.1	66.92 ± 23.46	70.31 ± 23.23	68 ± 33.49
Ni	7.69 ± 3.63	7.82 ± 2.61	8.76 ± 3.26	8.25 ± 4.06
Cu	17.84 ± 9.37	52.65 ± 63.88	34.58 ± 12.08	24.17 ± 11.49
Zn	166.60 ± 68.24	363.48 ± 203.79	309.99 ± 111.10	257.15 ± 212.85
As	2.90 ± 0.92	8.95 ± 6.01	12.1 ± 4.78	4.77 ± 3.1
Se	3.41 ± 1.35	5.62 ± 1.85	5.48 ± 1.73	4.21 ± 1.74
Sr	4.85 ± 2.74	6.51 ± 2.92	7.55 ± 2.41	5.44 ± 2.05
Cd	2.68 ± 1.9	2.26 ± 0.72	2.52 ± 1.06	1.15 ± 0.56
Ba	23.14 ± 14.09	18.26 ± 9.11	23.06 ± 7.12	15.17 ± 6.14
Pb	62.95 ± 31.66	76.61 ± 20.66	72.19 ± 14.46	62.46 ± 24.68
Mo	4.37 ± 3.01	7.31 ± 3.85	10.06 ± 4.1	11.48 ± 12.38
Sb	0.82 ± 0.45	2.61 ± 1.45	5.71 ± 2.64	1.84 ± 1.34
Total	1066.74	1356.79	1421.77	1353.68

. Figure 3 shows the time-dependent concentrations of the 17 metal elements during the four seasons. The highest PM_2.5 levels were found in winter (162 μg/m³), followed by autumn (139 μg/m³), spring (119 μg/m³) and summer (106 μg/m³). These values are a few times higher than the PM_2.5 standards of the Air Quality Guideline of World Health Organization (WHO) (25 μg/m³), and also much higher than the Ambient Air Quality Standard set by the Chinese Environmental Protection Bureau (75 μg/m³), reflecting the serious air pollution in suburban Nanjing.

For the 17 metal elements, the highest levels of Cr, Mn, Ni, As, Sr and Sb occurred in winter, average concentrations of Mg, Cu, Zn, Se and Pb peaked during autumn, Al reached its maxima in summer, while Na, V, Cd, Ba and Mo were found at high levels during both winter and spring. The variations of concentrations in different seasons can be represented by the standard deviations (SD), as shown in Table 3. Large SD values were mostly found during summer and spring, except for Cu, As, Se, and Sr in autumn, and Mo in winter. The Cr concentrations were elevated significantly during winter and spring, similar to Na; the concentrations of Zn, as well as As, Se, and Sb, also varied dramatically in autumn and winter, (Figure 3). On the contrary, V and Ni showed a similar trend, with low levels observed in autumn and winter and high levels in summer and spring. So the opposite trends from the two groups of metal elements indicate that they may originate from different sources. Both Cd and Ba had maximum values in summer and showed a higher degree of variation in summer. It is worth mentioning that Cu had a unique distribution during four seasons (Figure 3) with the highest concentrations occurring in autumn, but appeared to be relatively flat during other seasons. Source contributions of metal elements varied during different seasons. For example, coal burning is remarkably high in cold seasons (such as winter), so it may emit high amounts of As and Sr, and industrial emissions may contain more V and Ni. The relationship between the sources and seasonal variations of the metals is further discussed in Section 3.2. In addition, it also should be noted that, besides the different emission sources, the average boundary layer heights in different seasons may influence the measurement results as well.

3.1.3. Meteorological Influences

The temporal variations of meteorological parameters throughout the campaign are presented in Figure 2. The average temperatures in summer, autumn, winter and spring are 25.4 °C, 8.3 °C, 2.8 °C and 21.5 °C, and average relative humidity are 72.9%, 70.2%, 70.3%, and 64.7%, respectively. In summer, both the air temperature and relative humidity are higher than those in other seasons, which may enhance the secondary formation of other aerosol components, such as sulfate or secondary organic aerosols, altering the PM_2.5 compositions and thus the fractional contributions of metals. However, on the other hand, high temperature may also lead to the evaporation of semi-volatile species and decrease the PM_2.5 mass loadings. Overall, the influences of temperature and humidity of the variations and contributions of metals in PM_2.5 are complex, and should be investigated in the future with other supporting data.

The air mass back trajectories in the four seasons were calculated using the Hybrid Single-Particle Lagrangian Intgrated Trajectory (HYSPLIT) model (Figure 4), based on the meteorological parameters [61,62]. In this study, the air parcels were traced 72 h back in time to estimate trajectories at three different heights (500 m, 1000 m, 1500 m) every 8 h to provide information on the air parcel origins (Figure 4a). The air mass trajectories could be classified into four sectors: Sector A: Japan, yellow sea and bohai; Sector B: Southeast of China, the Pacific Ocean; Sector C: Southwest of Jiangsu and Anhui province; Sector D: Northern Asia inland and Mongolia region.

In summer, the air parcels mainly came from the ocean, including Sectors A and B, except for 28 August from Anhui province (Sector C), and 29 August, 30 August, 5 September and 8 September from the north of China (Sector D). The air masses during these special days (Figure 2) can be traced back to NW of Mongolia. Since this trajectory intercepts with numerous industrial and suburban areas of northern China, which delivered polluted air to the receptor site, the concentrations of metal elements in Figure 3, like Mn, Pb, Ba, Cd, Ni, Sr, V, showed peak values in summer. The back trajectories in autumn and winter indicate three air mass pathways: Sector B, Sector C and more predominantly Sector D. It has been mentioned earlier that some elements had higher values in these two seasons, probably because of the air parcels from North of China, which not only transport the cold air but also the pollutants, such as heating-related or coal burning emissions during these cold seasons. This influence can be observed in Figure 3 by the characteristics of elements like As, Sb and Sr. Cu had a peak concentration on 18 and 19 November in autumn, similar to Mg, Mn, Pb, which is likely due to two reasons: (1) the wind mostly blew form the North in these two days, thus leading to high loadings of aerosol metals; (2) in the next two days, the wind direction changed to the NE, and the air parcel came from the ocean (Figure 4a), which carried clean air and led to the sudden decrease of the concentrations of these metals. In spring, Sector B was the main origin of air parcels, with the air coming only from the NW in May 28. On the other hand, as our sampling location resided in the suburban area and was adjacent to the farmland, the aerosol characteristics were inevitably influenced by local biomass burning emissions which released metal elements such as As, Cd, etc., into the atmosphere.

3.2. Source Apportionment of Selected Atmospheric Metal Elements

In this study, 64 samples with 17 metal species were inputted into the PMF 3.0 model. The PMF algorithm is able to segregate the total airborne metals into the sum of a few dominant sources, and quantify the temporal variations and contributions of each source [63,46]. In general, inclusion of more samples (>100 samples) can provide better PMF results and more scientifically sound interpretation of the sources. Nevertheless, applications of PMF model on a limited number of samples were also reported previously, and can also provide very valuable insights into the sources and processes of the aerosols. For example, [64] analyzed about in 57 samples from four cities, and identified different sources during heavy haze formation in China; [65] analyzed in total 24 samples from four sites to elucidate the sources and transformation processes of the water-soluble organic aerosols. In this study, the PMF results were carefully evaluated by investigating the characteristics of different solution spaces, for examples, five-, six- and seven-factor solutions. However, the five- and seven-factor solutions appeared to be physically unreasonable, and six sources were determined as the optimal solution, with the fpeak of −0.1, the Qrobust and Qtrue values being almost equal to the Qtheory. The six factors were physically reasonable and also consistent with the potential aerosol sources in Nanjing (United States Environmental Protection Agency US EPA 2008). The profiles and corresponding time series of the six sources are shown in Figure 5 and Figure 6, respectively.

Factor 1 contains high contributions of Pb (45.62%), Mn (43.33%), V (41.44%), Cu (38.38%), Se (38.88%), Mg (34.08%), Zn (33.40%), Cd (30.39%), and Ni (29.56%), indicating a predominant traffic source. Tire abrasion and brake wearing can generate debris particles that contain significant amounts of Zn, Cd, Pb, and Cu [34,66,67]. Vehicular exhaust emissions may contain V, Ni, Zn, Pb, and Se from gasoline, oil or fuel additives [68]. Seasonal variations shown in Figure 6 indicate that the traffic-related metal concentrations were higher in summer and spring, and lower in colder seasons (i.e., winter). This trend could be related to increased advection from upwind urban sites [23].

Factor 2 corresponds to soil/dust source as it contains the highest concentrations of a key crustal element—Al (82.88%) mixed with Na (17.61%) and Mg (16.47%). The soil dust could be contributed by fly ash from industrial activities nearby or other regions, or re-suspended dust due to traffic or construction activities. According to the time series of soil dust contribution (Figure 6), there was a sharp rise in summer; but the air parcels came from the Sector A (Figure 4b), the wind direction was NE and the wind speed was low enough, so this bump may have resulted from the behaviors of anthropogenic emissions around the sampling site.

Factor 3 was identified as coal burning, as it contains high fractions of the heavy metals, including Sb (56.21%), As (51.63%), Zn (24.65%), Sr (24.06%), Se (23.15%), and Cd (23.10%). Sb, Se and Sr are commonly used as markers of coal burning emissions [29]. In addition to coal burning, As can be also emitted by iron making and steelmaking processes [69]. As shown in Figure 6, the contributions of coal burning factor were significantly high in autumn (31.49%) and winter (35.07%), in accordance with the enhanced coal consumption for energy supply during the cold seasons. Furthermore, during autumn and winter, the wind and air parcels (Figure 2 and Figure 5a) were mainly derived from north China where coal burning is very active. It also should be mentioned that a large iron and steel plant is only a few kilometers away from our sampling site, and coal burning is a primary energy source for iron and steel production. A previous report (Yang et al. [63]) showed that in Beijing, coal burning from coal power plants, iron and steel industries contributed to 25% of the emitted PM_2.5 mass.

Factor 4 with markedly high contributions from elements like Cd (41.48%), Ba (32.06%), and relatively high contributions from Sr (17%), Pb (16.17%), and Ni (15.06%), likely represents the source of incineration of electronic waste. Earlier studies identified a variety of electronic materials, including batteries (Ni-Cd, Pb batteries), electric switches, electronic components of TV which contain Cadmium compounds [70,71], so incineration of these different types of electronic wastes can emitted metal-containing particles. Furthermore, this incineration is also associated with elements of Ba, Ni, Pb, Sr, which were usually used to produce the electric products like light bulb, TV tube or computer and so on [72]. Figure 6 showed the seasonal trend of this factor—the peak occurred in summer, likely due to enhanced applications of various electronic devices during summer.

Factor 5 was characterized by high percentages of Cr (56.38%), Na (32.35%), V (27.84%), Mn (20.41%), and Ni (15.87%). These elements were strongly associated with the industrial emissions in the vicinity of our site, especially for Cr and V [39,73]. Combustion of fossil fuels or oil burning can emit particles containing Cr, V, and Ni, and those activities often occur in the petrochemical and chemical plants [74]. Also, iron and steel production is likely an important source of Mn and Ni. Note that our sampling site is located at a region close to many industrial plants and thus may experience serious industrial pollutants. The seasonal variations were (Figure 6) showed that industry contributed more significantly in winter and spring than in other seasons, which might be attributed to the low mixing layer height and stagnant wind conditions, resulting in the accumulation of the industry-related metal-containing particles near our sampling site.

Factor 6 (Unknown) was a main contributor to Mo (61.93%), Ni (21.43%), Cu (17.08%) in ambient PM_2.5. In previous studies, Mo was identified as a tracer of a hard alloy plant which produces molybdenum alloy and Ni [75]. Mo was also observed in particles from diesel emissions (Duan and Tan [39]); however, this source needs further investigation. Factor 6 had a high contribution only in spring.

Overall, the contributions of different sources to the total aerosol metals were factor 5 (Industry, 34%), factor 1 (Traffic, 25%), factor 2 (Soil dust, 19%), factor 3 (coal burning, 10%), factor 4 (Incineration of electronic waste, 9%), and factor 6 (Unknown, 3%).

3.3. Particle Morphology and Composition

SEM technique was used to analyze the particle morphology [76,77], and EDS analysis was applied to provide the elemental composition of aerosol particles quantitatively (Figure 7), so as to aid the source apportionment analyses.

Our SEM measurements identified three major types of morphology of the sampled particles. The first type appears to be mineral grains with the typical shape of single or aggregated regular columnar crystal (Figure 7a). EDS showed that these particles contain mainly silicon aluminate, quartz, calcium, iron and a small amount of other metal elements. For example, EDS analyses that identified the elemental compositions of a typical mineral grain were Ca, S and O, indicating the presence of gypsum (CaSO₄·H₂O) [9,78] The gypsum particles could be traced to several sources, such as secondary reactions involving SO₂, airborne limestone, photocatalytical conversion by Fe₂O₃ or TiO₂ of dust-related species, or directly from coal combustion [79] and desulfurization. Fe mainly came from industries; and Ti or K may be associated with crustal particles. Although previous studies indicated that mineral particles may generate from construction activities [31,79], it has been demonstrated that the morphology of such particles is typically porous, therefore the gypsum particles observed in this study are unlikely from construction source. The second type of particles are spherical with a smooth surface by the SEM (Figure 7b), and EDS further showed that they were predominately alumino-silicates classified as fly ash, supporting that these particles were mainly derived from coal combustion and metallurgical processes. Another type of spherical particles (Figure 7c) was mainly composed of C and O and thus likely consisted of organics. Overall, the qualitative analyses by the SEM/EDS results indicated the particle sources of coal burning, industrial emissions and soil dust, which is consistent with the PMF results.

4. Conclusions

This study presents measurement results of 17 metal elements in PM_2.5 samples collected over a one-year period. Chemical characteristics including concentrations, temporal variations, sources, morphology, and air mass trajectories were thoroughly discussed. We found that the airborne metal mass loadings in suburban Nanjing was on average, 1.23 μg/m³, accounting for 1.0% of the total PM_2.5 mass, which is comparable with those observed in other polluted cities in China. In particular, we found that the element of As concentrations was above the reference values of NAQQS (GB3095-2012) and WHO guidelines. Temporal changes of these metals behaved differently due to variations of emission sources, air mass origins, and meteorological conditions. Across the sampling period, we identified and quantified six major sources for the observed metal elements, including industry (34%), traffic (25%), soil dust (19%), coal burning (10%), incineration of electronic waste (9%), and a minor unknown source (3%). Further SEM/EDS analyses of the particles verified the source apportionment results. As the heavy metals have multiple adverse effects on the environment and public health, our findings are thus valuable for the air quality remediation in the densely populated YRD region, and may provide insights for evaluating the environmental influences of metals on soil, water and the ecosystem.