Characterization and source apportionment of oxidative potential of ambient PM_2.5 in Nanjing, a megacity of Eastern China

ABSTRACT

Oxidative potential (OP) reflects integratedtoxicity of PM_2.5. Therefore, OP of PM_2.5 samples collected from traditional central and suburban areas (Gulou and Xianlin) in Nanjing, a megacity of China, from Jan. 2020 to Jun. 2021 was analyzed by using two abiotic assays, DTT and AA assays (OP^DTT and OP^AA). Values of OP^DTTv and OP^AAv were  0.82 ~ 3.82  and    0.15 ~ 2.22 nmol/min/m³, respectively. The OP values of PM_2.5 samples collected from the Xianlin were higher than those from the Gulou by using both the DTT assay and the AA assay in this study, suggesting the seasonal and spatial variation. Moreover, OP values analyzed by using DTT assay were higher than those by using AA assay. Based on Positive Matrix Factorization (PMF), the major pollution source of the OP of PM_2.5 in Nanjing was combustion source. This study may be helpful for evaluating adverse health risks of PM_2.5 in a megacity.

KEYWORDS:

• Oxidative potential
• PM_2.5
• seasonal variation
• spatial variation
• positive matrix factorization

1. Introduction

Epidemiological studies show that long-term exposure to the atmospheric fine particulate matters (PM_2.5) can result in various adverse health risks such as asthma, airway rhinitis, and congestive heart failure [1–3]. Although the specific mechanisms of those toxic effects of the inhaled PM_2.5 is currently unclear, it is widely accepted that the adverse health effects of PM_2.5 are related to oxidative stress [4–7]. Inhaled PM_2.5 produces excess reactive oxygen species (ROS) in the human body, leading to an imbalance in the body’s antioxidant defenses, which will result in damage to DNA and lipids and inflammation in the body, and then ultimately cause various diseases [8,9]. Now, the ambient air quality standards established by governments around the world and World Health Organization (WHO) focus on the mass concentrations of airborne particles/gaseous pollutants (e.g. PM₁₀, PM_2,5, SO₂, NOx, O₃) and particle-bound contaminants (e.g. inorganic toxic elements such as Pb, F, As, Cd, and Hg and organic toxic compounds such as BaP) [9–11]. These single indicators are hard to reflect the integrated toxicity of PM_2.5. For example, oxidative stress posed by inhaled airborne particles may result from the integrated effects of the mass concentrations of airborne particles and particle-bound contaminants [4]. Recently, oxidative potential (OP) has been advocated to assess the oxidation activity of airborne particles to reflect the integrated toxicity of airborne particles [12,13].

Oxidative potential of PM_2.5 may be posed by PM_2.5-bound persistent free radicals (e.g. semiquinone radical and aromatic radical), PM_2.5-bound hazardous trace elements (e.g. As, Pb, Cd, and Hg), PM_2.5-bound toxic organic compounds (e.g. PAHs and PCBs), and airborne microorganisms [14–18]. For example, the previous studies show that persistent-free radicals, transition metals (such as copper), quinones, and polycyclic aromatic hydrocarbons have relatively higher oxidative potential than other compounds [19,20]. It is reported that bacterial cells significantly reduced the oxidative potential of particulate matter, while fungal spores increased the oxidative potential of particulate matter [21]. Moreover, factors such as particle size, particle aging time, and particle-derived products may also alter particulate oxidative potential [22]. As an index of the integrated toxicity of oxidation activity, OP of PM_2.5 is a better indicator for potential inhalational health risk assessment of PM_2.5 than mass concentrations of PM_2.5 and/or PM_2.5-bound toxic chemical components [3]. Therefore, more investigations on OP of PM_2.5 should be carried out in different cities, as extensive investigations of PM_2.5-bound hazardous trace elements and organic compounds.

Recently, abiotic and biotic assays have been developed to assess the oxidation activity of atmospheric particulate matters [13]. The most widely used abiotic assays are dithiothreitol assay (DTT) and ascorbic acid (AA) and they were labeled as OP^DTT and OP^AA [23,24]. It is reported that DTT assay was sensitive to transition metals (e.g. Cu and Mn) and organic compounds (e.g. water-soluble organic carbon and quinone compounds) [25,26], while the AA assay was more sensitive to transition metals (e.g. copper) [27]. Therefore, abiotic DTT and AA assays may reflect different response of PM_2.5-bound components to potential health risk. For example, there was a correlation between OP^DTT of particulate matter and multiple adverse health outcomes, such as airway inflammation, asthma, and congestive heart failure [1,3]; however, no good correlation was reported between OP^AA and asthma, cardiopulmonary emergency departments [28]. So, the choice of assays may impact the performance of the oxidative potential of the particulate matter during the health risk assessment. Therefore, determination of OP using DTT and AA assay synchronously may offer more information on evaluating spatiotemporal variation of OP, which reflects the response of different PM_2.5-bound components on the potential health risk assessment of PM_2.5.

Yangtze River Delta is one of the economically developed areas in China. Nanjing, one of the metropolitan cities in the Yangtze River Delta, is the capital city of Jiangsu Province locating in 118°22’ and 119°14ʹE and 31°14’ and 32°37ʹN. PM_2.5 is an important atmospheric pollutant in Nanjing and may pose potential health risks to local residents [29,30]. Therefore, PM_2.5 samples were collected from central and suburban areas (Gulou and Xianlin) in Nanjing, a Chinese megacity, from January 2020 to June 2021. The oxidative potentials of PM_2.5 samples were analyzed and compared by using two abiotic assays, DTT and AA assays. The aim of this study is to evaluate the seasonal and spatial variations of OP determined by using two abiotic assays and to discuss the potential resources of OP based on Positive Matrix Factorization (PMF). The novelty of this study is to evaluate redox activity of PM_2.5 in a typical metropolitan city in China and it may be helpful for assessing the potential impact of PM_2.5 on adverse health risks, providing a scientific basis for the management and control of polluted atmospheric particulate matters in a megacity.

2. Materials and methods

2.1 Sampling of PM_2.5

Therefore, PM_2.5 samples were collected from central areas (Gulou campus of Nanjing University) and suburban areas (Xianlin campus of Nanjing University) shown in Figure 1 from January 2020 to June 2021. Concentrations of PM_2.5, PM₁₀, SO₂, NO₂, and O₃ were recorded from Nanjing Environmental Monitoring Center Station. A total of 198 samples of PM_2.5 were collected using a high-flow sampler (HY-1000, Qingdao Henyuan Instruments Co. Ltd., Qingdao, China) in Gulou sampling site and a medium-flow particle sampler (LB-120 F, Qingdao Henyuan Instruments Co. Ltd., Qingdao, China) in Xianlin sampling site, loaded with a quartz filter membrane with a daily air volume of 1000 and 120 m³, respectively. All Whatman quartz microfibre filters (Whatman International Ltd., Maidstone England) wer precombusted in a muffle furnace at 500°C for 2 h. The filter membrane is wrapped in tin foil and stored in a refrigerator. The mass of PM_2.5 collected on quartz microfibre filters was obtained via the mass difference of the filter membrane before and after sampling after equilibrated in a desiccator for 48 h.

Figure 1. Sampling station of PM_2.5 in Xianlin and Gulou of Nanjing, China.

2.2 Measurement of oxidative potential

A sampling membrane was immersed in deionized water and sonicated for 30 min using an ultrasonic cleaner (SB-5200DTD, YMNL, China). The extraction solution was filtered with a 0.22 μm polyethersulfone membrane (Polyethersulfone Membrane, PES NAVIGATOR, Germany) to remove insoluble substances, and the filtrate was used for the measurement of oxidative potential. In this study, two abiotic assays (DTT and AA assays) were stored to measure the oxidative potential of water-soluble components of PM_2.5. One method (reagent) blank was carried throughout the entire sample preparation and analytical process to identify possible contamination resulting from either the equipment used during sample processing or the reagents for each batch of samples processed. Each sample in triplicates was processed and if the relative percent difference of OP values in parallel samples was higher than 10%, the batch treatment would be redone.

DTT assay: DTT assay was performed following the protocol in previous reports [31,32]. Briefly, PM extraction solution, potassium phosphate buffer (0.2 M, pH 7.4), and DTT solution were mixed and incubated at 37°C, and a portion of the mixture was removed every 5 minutes into a centrifuge tube pre-filled with trichloroacetic acid (10%). Then DTNB solution and Tris-HCl (pH 8.8) buffer were added. The absorbance of 5-thio-2-nitrobenzoic acid (TNB) formed by the reaction of the remaining DTT and DTNB was measured at 412 nm using a UV-Vis spectrophotometer (UH5300, China). The OP values were calculated by the consumption rate of DTT, and then normalized by using sampling air volume.

AA assay: AA assay was based on the protocol of Pietrogrande’ report [33]. Briefly, AA solution, PM extraction solution, and potassium phosphate buffer (0.2 M, pH 7.4) were mixed in a 96-well plate and incubated at 37°C. The absorbance of the mixture at 265 nm was measured every 5 minutes using a full-wavelength scanning multi-function reader (Thermo Fisher Scientific, Varioskan Flash, USA). The OP values were calculated by the consumption rate of AA, and then normalized by using sampling air volume.

2.3 Element determination

The quartz filters were immersed into 1 M dilute hydrochloric acid solution in centrifuge tubes. After sonication for 30 minutes, the tubes were transferred to a thermostatic graphite-digestion device and digested at 130°C for 1 h. After cooling to room temperature, those tubes were sonicated for 15 minutes again and the supernatant was separated via centrifugation and stored for further analyses. Soluble elements in the supernatant were determined by using an ICP-OES (Avio 500, PerkinElmer, USA). The measured elements include Al, As, Ca, Cr, Cu, Fe, K, Mg, Mn, Mo, Na, P, Pb, S, Sb, Sr, Ti and Zn. Each sample shall be tested 3 times in parallel. The analytical curve for each batch of sample analyses was set up via the dilution of Multi-Element Standard Solution (1000 mg/L) bought from SPEX CertiPrep, USA. One method (reagent) blank was set up and a continuing calibration verification standard (CCVS) prepared via the dilution of the Multi-Element Standard was determined every 30 of sample solution. If the relative percent difference of parallel samples or CCVS was higher than 10%, the batch treatment would be redone.

2.4 Statistical analysis

Descriptive statistics and correlation analysis of experimental data were processed by using IBM SPSS Statistics 26 software. The potential resources of OP were identified by using a model of EPA PMF5.0. Positive Matrix Factorization (PMF) model, a mathematical receptor model based on factor analysis method, is advocated and developed by United States Environmental Protection Agency (US EPA), which has been widely used in environmental samples including ambient and indoor airborne particles, soil, sediments, and surface water to reveal potential source types and source contributions. PMF model does not need to the source profiles or fingerprints and only files of sample species concentrations and uncertainties of receptor samples and the number of sources need to be provided by PMF users. The fundamentals and technical support of PMF model have been interpreted detailly in the US EPA website (https://www.epa.gov/air-research/positive-matrix-factorization-model-environmental-data-analyses). OP values and inorganic elements of PM_2.5 were run on EPA PMF 5.0 for 100 times to discuss the potential sources in Nanjing. The basic principle of the model is that the environmental sample X is assumed to be an n × m matrix, n is the number of samples, and m is the chemical composition, so X can be decomposed into the pollution source contribution matrix G (n × p. P is the number of pollution sources) and the pollution source component spectrum matrix F (p × m), the measured sample concentration can be expressed as [34]:

$x_{ij}={\sum }_{k=1}^p{g}_{ik}{f}_{kj}+e_{ij}$

Where, x_ij is the sample concentration matrix X, that is, the concentration of the jth species in the ith sample, p is the number of pollution sources, g_ik is the contribution of the kth pollution source to the ith sample, f_kj is the concentration of the jth species in the kth pollution source, and e_ij is the residual.The model imposes non negative constraints on G and F matrices (g_ik≥0 and f_kj≥0). When the weighted value Q of the square of the ratio of residual error to uncertainty reaches the minimum, it can be considered that the factorization of the model achieves the optimal result. The PMF algorithm determines matrices G and F by continuously minimizing Q:

$Q={\sum }_{i=1}^n{\sum }_{j=1}^m{\left(\frac{e_{ij}}{u_{ij}}\right)}^2$

with $e_{ij}=x_{ij}-{\sum }_{k=1}^p{g}_{ik}{f}_{kj}$

with u_ij corresponding to the uncertainty of the concentration x_ij.

3. Results and discussion

3.1 Major ambient air contaminants in Nanjing

Concentration variation of major ambient air contaminants (24-hour average for PM_2.5, PM₁₀, SO₂, and NO₂ and 8-hour average for O₃) from 1 January 2020 to 1 July 2021 in Nanjing is shown in Figure 2. Figure 2a shows that concentrations of PM_2.5 and PM₁₀ varied greatly among different date, even in the same season. For example, concentrations of PM_2.5 and PM₁₀ ranged from 6 to 133 μg m³ with a medium of 27 μg m³ and from 13 to 168 μg m³ with a medium of 51 μg m³, respectively, in 2020 (from 1 January 2020 to 31 December 2020). Figure 2a shows that the lowest 24-hour average concentrations of PM_2.5 and PM₁₀ were observed frequently in summer while PM_2.5 and PM₁₀ had higher 24-hour average concentrations in winter, which was due to urban central heating in winter. 24-hour average threshold values of PM_2.5 and PM₁₀ are 35/75 and 50/150 μg m³ (Level 1/Level 2 standard), respectively, in ‘Ambient air quality standards of China (GB-3095-2012)’. So, PM_2.5 and PM₁₀ are beyond the threshold value of Level 1 standard) frequently, while lower than the threshold value of Level 2 standard in most days of a year (Figure 2a). PM_2.5/PM₁₀ values in winter was obvious higher than those in summer (Figure 2a). Great variation on the average concentrations of SO₂, NO₂ and O₃) was observed (Figure 2b). However, regular temporal variation on 24-hour average concentrations of SO₂ and NO₂ was not observed (Figure 2b); while 8-hour average concentrations of O₃ has higher values in summer and lower values in winter (Figure 2b), which was opposite to PM_2.5 and PM₁₀. The average threshold values of SO₂, NO₂, and O₃ are 20/50, 80/80, and 100/160 μg m³, respectively, in ‘Ambient air quality standards of China (GB-3095-2012)’. Average concentrations of NO₂ fall into its threshold value and those of SO₂ are beyond its threshold value in some days (Figure 2b). Average concentrations of O₃ are beyond its threshold value frequently in summer. So O₃ is the main ambient atmospheric contaminants in summer while PM_2.5 and PM₁₀ are the main ambient atmospheric contaminants in winter, spring, autumn in Nanjing. So PM_2.5 is an important ambient atmospheric contaminant in Nanjing and the potential health risks posed by airborne particles cannot be ignored, especially PM_2.5 which can go into the alveoli.

Figure 2. Variation of major ambient air contaminants (24-hour average for PM_2.5, PM₁₀, SO₂, and NO₂ and 8-hour average for O₃) from 1 January 2020 to 1 July 2021 in Nanjing.

3.2 Oxidative potential of PM_2.5

Figure 3 (a & b) shows volume-normalized DTT activity (OP^DTTv, nmol/min/m³). The seasonal variation of OP^DTTv in Xianlin sampling site was in the order of summer> autumn> spring> winter. The average values of OP^DTTv were 2.67 nmol/min/m³ (2.00 ~ 3.82 nmol/min/m³) for summer, 2.48 nmol/min/m³ (1.58 ~ 3.67 nmol/min/m³) for autumn, 2.32 nmol/min/m³ (1.55 ~ 3.02 nmol/min/m³) for spring, and 2.31 nmol/min/m³ (1.50 ~ 3.42 nmol/min/m³) for winter. The average values of OP^DTTv were 1.56 nmol/min/m³ (1.25 ~ 2.08 nmol/min/m³) for winter, 1.23 nmol/min/m³ (0.82 ~ 1.75 nmol/min/m³) for summer, and 1.32 nmol/min/m³ (0.88 ~ 1.82 nmol/min/m³) for spring. The order of OP^DTTv of the samples from Gulou was winter> summer> spring. Therefore, OP^DTTv in PM_2.5 of Xianlin and Gulou sampling sites indicated the seasonal difference of different sampling seasons and spatial difference between Xianlin and Gulou. Our previous study reported seasonal and spatial variation of mass concentrations of TSP and PM_2.5 and TSP/PM_2.5 bound toxic chemical components (e.g. As, Pb, Cd, and Hg) in the traditional center district (Gulou) and the suburb (Pukou) in Nanjing [35]. These may be the important reason for seasonal and spatial variation of OP^DTTv of PM_2.5 in the traditional center district and the suburb (Pukou) in Nanjing.

Figure 3. Oxidative potential of PM_2.5 (OP^DTTv and OP^AAv) in Xianlin and Gulou.

Figure 3 (c & d) shows volume-normalized AA activity (OP^AAv, nmol/min/m³). The seasonal variation of OP^AAv in Xianlin sampling site was in the order of Autumn> Summer> Spring> Winter. The average values of OP^AAv were 1.41 nmol/min/m³ (0.75 ~ 1.90 nmol/min/m³) for Autumn; 1.37 nmol/min/m³ (0.15 ~ 1.85 nmol/min/m³) for summer, 1.24 nmol/min/m³ (0.47 ~ 2.22 nmol/min/m³) for spring, and 0.93 nmol/min/m³ (0.44 ~ 1.93 nmol/min/m³) for winter. The average values of OP^AAv were 0.70 nmol/min/m³ (0.45 ~ 1.16 nmol/min/m³) for winter, 1.22 nmol/min/m³ (0.47 ~ 1.80 nmol/min/m³) for summer, and 0.78 nmol/min/m³ (0.45 ~ 1.03 nmol/min/m³) for spring. The order of OP^AAv of the samples from Gulou was summer> spring> winter. Therefore, these indicated the seasonal and spatial variation of OP^AAv in PM_2.5. The potential reasons may be consistent with these for OP^DTTv discussed above.

To DTT assay, the average OP^DTTv value was 2.42 nmol/min/m³ (1.50 ~ 3.82 nmol/min/m³ with a median value of 2.39 nmol /min/m³) for PM_2.5 in Xianlin, while the average OP^DTTv of was 1.33 nmol/min/m³ (0.82 ~ 2.08 nmol/min/m³ with a median value of 1.34 nmol/min/m³) for PM_2.5 in Gulou (Figure 3). The average OP^DTTv value of PM_2.5 in Xianlin was 1.8 times higher than that in Gulou. To AA assay, the average OP^AAv value was 1.17 nmol/min/m³ (0.15 ~ 2.2 nmol/min/m³ with a median value of 1.17 nmol/min/m³) for PM_2.5 in Xianlin, while the average OP^AAv of was 0.93 nmol /min/m³ (0.45 ~ 1.80 nmol/min/m³ with a median value of 0.86 nmol/min/m³) for PM_2.5 in Gulou. The average OP^AAv value of PM_2.5 in Xianlin was 1.26 time higher than that in Gulou. These indicated the spatial variation of OP^DTTv and OP^AAv in of PM_2.5 between sampling sites. PM_2.5-bound Zn (average: 507 ng m⁻³) in Xianlin were significantly higher that (126 ng m⁻³) in Gulou and there were no significant differences on concentrations of As, Pb, Cd, Mn, Cu and Fe between Gulou and Xianlin in Nanjing. PM_2.5-bound As, Pb, Cd, Cr, and Zn in the suburb (Pukou) were generally higher these in the traditional center district (Gulou) in Nanjing [35]. Moreover, Xianlin sampling site is located in the downwind of Jinling Petrochemical Group and about 200 meters away from the highway around the city; while Gulou sampling site is situated at the traditional commercial and living areas and low speed traffic exhaust and atmospheric diffusion within the city are the main local pollution source. These may result in the differences on OP between Gulou and Xianlin sampling sites. The average OP^DTTv value were higher than the average OP^AAv values in PM_2.5 of both sampling sites, suggesting the sensitivity of the selected method (Figure 3). Therefore, OP values analyzed by using DTT assay were higher than those by using AA assay. The previous investigations reported that DTT assay was sensitive to transition metals and organic compounds while the AA assay was more sensitive to transition metals [25–27]. A sampling membrane was extracted with deionized water and the extraction solution was use for the determination of OP using DTT and AA assay. So water-soluble organic carbon and quinone compounds may involve into the formation extract OP^DTTv. Moreover, this study confirms the seasonal and spatial variation of OP^DTTv and OP^AAv of PM_2.5 in the traditional center district and the suburb in Nanjing.

The average OP of atmospheric particulate matter in this study was compared with those in other regions in the related literatures (Table 1). The average DTT activity in Nanjing was 2.03 nmol/min/m³, higher than those for Beijing (0.19 nmol/min/m³), Delhi (1.57 nmol/min/m³), Xi’an (0.51 nmol/min/m³), Toronto (0.03 nmol/min/m³), Athens (0.33 nmol/min/m³), Ala (0.61 nmol/min/m³), Trento (0.58 nmol/min/m³) and Lecce (0.40 nmol/min/m³), respectively, but lower than that for Sichuan (9.6 nmol/min/m³), Patiala (3.8 nmol/min/m³) (Table 1). Therefore, OP values in this study differed from those in the related literatures, revealing the regional characteristics of OP. So, more investigations should be carried out in different regions to accumulate abundant data to reveal how OP works.

Table 1. OP values in the related literatures.

OP	Mean	Range	Site	Site type	Source
DTTv	9.6	8.50 ~ 10.90	Sichuan	Rural Individual	[36]
	9.7	8.60 ~ 11.00		Rural Family
DTTv	0.19	0.11 ~ 0.49	Beijing	Suburb	[37]
	0.24	0.09 ~ 0.32		Downtown
DTTv	1.57	0.49 ~ 3.60	Delhi	Suburb	[38]
DTTv	0.51	0.24 ~ 1.10	Xi’an		[39]
DTTv	0.03	0.16 ~ 0.70	Toronto		[40]
AAv	0.016	0.012 ~ 0.033
DTTv	0.61 ± 0.23		Ala	Industry	[41]
AAv	1.4 ± 1.1
DTTv	0.58 ± 0.22		Trento	Urban Background
AAv	0.7 ± 0.4
DTTv	0.40 ± 0.26		Lecce	Mountain	[42]
DTTv	0.33	0.02 ~ 1.16	Athens	Urban Backgroud	[43]
DTTv	3.8	1.3 ~ 7.2	Patiala	Semi-Urban	[44]
DTTv	2.42	1.50 ~ 3.82	Nanjing	Suburb	Current study
AAv	1.17	0.15 ~ 2.22
DTTv	1.33	0.82 ~ 2.08		Downtown
AAv	0.93	0.45 ~ 1.80

3.3 Correlation between OP and element concentration

Table 2 listed the correlation coefficients between OP values and element concentrations in PM_2.5 collected from Xianlin and Gulou. To Xianlin sampling site, OP^DTTv was moderately correlated with S (0.461) and weakly correlated with Cr (0.275), Cu (0.280), Pb (0.205), Sb (0.209), and O₃ (0.317). This indicates that secondary sulfate may play an important role in OP, which is consistent with the previous research result that sulfate driven metal dissolution plays a key role in the OP of particulate matter [45]. There is no significant positive correlation between OP^AA and the measured elements. OP^DTTv was correlated with various elements, including S, Ca, Mn, Fe, Cu, and Zn; while OP^AAv was nearly exclusively correlated with Cu [28]. Romano et al. reported that OP^DTTv was strongly correlated with metals from anthropogenic sources (e.g. Cr, Ni, Cu, and Zn) and with mineral elements related to natural sources (e.g. K and P) in winter while less element mass concentrations (Pb, P, Ca, Fe, and Zn) were correlated with OP^DTTv in spring; no correlation between OP^AAv and element mass concentrations was observed [46]. These were consistent with our research.

Table 2. Correlation coefficients between OP values and concentrations of PM_2.5-bound elements and major ambient air contaminants.

	OP^DTTv		OP^AAv			OP^DTTv		OP^AAv
	Xianlin	Gulou	Xianlin	Gulou		Xianlin	Gulou	Xianlin	Gulou
OP^DTTv	1	1	0.209	0.074	Na	−0.301*	0.204	−0.178	−0.638**
OP^AAv	0.209	0.074	1	1	P	−0.127	0.090	0.078	−0.625**
Al	−0.143	0.470**	−0.267*	−0.294*	Pb	0.205	0.495**	−0.047	0.024
As	0.160	0.346*	−0.039	0.088	S	0.461**	0.401**	0.088	−0.179
Ca	−0.179	0.489**	−0.266*	−0.275*	Sb	0.209	0.310*	0.053	0.005
Cr	0.275*	0.113	0.066	0.272*	Sr	−0.124	0.454**	−0.256*	−0.262*
Cu	0.280*	0.555**	0.026	0.169	Ti	−0.085	0.488**	−0.230*	−0.281*
Fe	−0.035	0.484**	−0.196	−0.239	Zn	0.082	0.601**	0.024	−0.266*
K	0.218*	0.611**	−0.154	−0.362**	SO2	0.168	0.481**	−0.078	0.085
Mg	−0.145	0.459**	−0.265*	−0.307*	NO2	0.137	0.448**	−0.068	−0.386**
Mn	0.089	0.566**	−0.070	−0.238	O3	0.317*	−0.011	0.195	0.490**
Mo	0.175	0.230	−0.082	−0.052

* Correlation is significant at the 0.05 level (2-tailed).

** Correlation is significant at the 0.01 level (2-tailed).

To Gulou sampling site, OP^DTTv had a significant correlation with Al (0.470), Ca (0.489), Cu (0.555), K (0.611), Mg (0.459), Mn (0.566), Pb (0.495), S (0.401), Sb (0.310), Sr (0.454), Ti (0.488), Zn (0.601), SO₂ (0.481), and NO₂ (0.488). K is a marker of biomass combustion [47–49] and particles emitted from combustion sources may contain abundant persistent free radicals which have high reactivity. So combustion sources may have a great impact on OP in the region. The good correlation between transition metals (Cu, Pb, Mn and Zn) and OP^DTT shows that transition metals may have an important contribution to the OP of particulate matter. These correlation results are consistent with previous studies, for example, Yang reported that the correlation coefficients between OP^DTT and Cu, Fe, Mn and Zn ranged from 0.61 to 0.68 [50] and Gao et al. reported the similar results [32]. The correlation between OP^AA values and element concentrations showed that OP was highly negatively correlated with Na (−0.638) and P (−0.625), and no good correlation between OP^AA and the measured elements. In a word, the significant correlation between OP^DTTv and the elements measured was observed while weak and/or no good correlation between OP^AA and the elements measured was found for PM_2.5 from both Xianlin and Gulou sampling sites.

3.4 Source apportionment of OP via PMF

153 observation samples and 17 variable species were inputted into the PMF model as the input basis to infer their possible sources. The uncertainty is jointly determined by the error coefficient and the method detection limit. The error coefficient is set to 0.1, and the method detection limit is determined by three times of the standard deviation of seven blank tests. The solution of 4 factors is determined as the optimal solution, and the number of factors is determined by Q (goodness of fit parameter) and the expected knowledge of sampling points. Q (Robustness) is the goodness of fit parameter for excluding outliers, and Q (True) includes all data points. The consistency of the two parameters shows that the model fits the input data well.

Announcement of source apportionment of PM_2.5 in Nanjing announced by Environmental Protection Bureau of Nanjing shows that the main pollution sources of PM_2.5 in Nanjing in 2015 were coal-burning emissions (27.4%), automobile exhaust (24.60%), industrial emissions (19.0%), dust (14.1%), and other sources. So 4 factors were set up when using PMF model in this study (Figure 4). As shown in Figure 4, potential sources in Nanjing were identified based on particle-bound elements. Factor 1 (F1) contains more Al, Ca, Fe, Mg, P, Sr, and Ti. Al, Fe, Mg, Sr, and Ti are crustal elements, which are generally related to site dust [51]. Therefore, F1 is defined as soil dust. Factor 2 (F2) was dominated by S, Sb, Zn and Pb. It has been reported that Zn and Sb were associated with lubricating oil, fuel additives, tire and brake wear of traffic sources [52,53]. S is included in the source profiles of coal and oil combustion in PM_2.5 is general secondary sulfate and sulfate particles mainly comes from SO₂ which is oxidized to sulfate in the atmosphere in clouds. The combustion of oil and coal is the main source SO₂, which was oxidized to sulfate in air via through gas-phase homogeneous and heterogeneous oxidation. Fine particles emitted from oil and coal combustion also contain abundant Zn and Pb [54–57]. Hence, F2 was assigned as a combustion source of fossil fuel. Factor 3 (F3) had the higher contribution of As, Cu, Mn, and Zn. According to relevant literature reports, Chinese coal contains high contents of As [58], which is considered as a tracer of the coal combustion source (power plant boilers, industrial boilers and residential stoves) [59], So F3 was classified as coal combustion source. Factor 4 (F4) mainly contained Cr, Cu, and P. Cr and Cu many come from a wide range of sources, including metallurgy, vehicle exhaust, coal burning and rubber tire wear [60]. So F4 was the mixed source of metal smelting and traffic-related source.

Figure 4. Source apportionment of OP by using PMF.

According to the results of the PMF model (Figure 4), the contribution of sources to OP^DTTv was 5.6% for F1, 75.7% for F2, 0.5% for F3, 18.2% for F4, respectively; the contribution of sources to OP^AA was 66.4% for F2 and 33.6% for F4, respectively. So more than 50% of OP of ambient PM_2.5 in Nanjing was resulted from F2, indicating that combustion source of fossil fuel was the main contributor of OP of ambient PM_2.5 in Nanjing. Strategies of control reducing toxicity of PM_2.5 should pay more attention to the combustion sources of fossil fuel.

4. Conclusions

OP of PM_2.5 collected from the unban traditional center district and the suburb in Nanjing from Jan. 2020 to Jun. 2021 was determined by using DTT and AA assays. The seasonal and spatial variation of both OP^DTTv and OP^AAv of PM_2.5 in the unban traditional center district and the suburb in Nanjing were observed. OP values analyzed by using DTT assay were higher than those by using AA assay. Moreover, different correlation for OP^DTT and OP^AA with PM_2.5-bound components was found, which may be the reason for higher OP^DTT values than OP^AA. The PMF analysis shows that the main potential source for OP in Nanjing was the combustion sources of fossil fuel. This study shows that OP provide a good index for evaluating the integrated toxicity of PM_2.5 in different district (e.g. the unban traditional center district and the suburb) and OP using AA assay may underestimate the potential health risks posed by PM_2.5.

Data availability

All data reported here can be made available on request

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by the National Key Research and Development Program of China [2018YFC1800603]; National Natural Science Foundation of China [21874065 and 22176085].