Seasonal concentration distribution of PM1.0 and PM2.5 and a risk assessment of bound trace metals in Harbin, China: Effect of the species distribution of heavy metals and heat supply

To clarify the potential carcinogenic/noncarcinogenic risk posed by particulate matter (PM) in Harbin, a city in China with the typical heat supply, the concentrations of PM1.0 and PM2.5 were analyzed from Nov. 2014 to Nov. 2015, and the compositions of heavy metals and water-soluble ions (WSIs) were determined. The continuous heat supply from October to April led to serious air pollution in Harbin, thus leading to a significant increase in particle numbers (especially for PM1.0). Specifically, coal combustion under heat supply conditions led to significant emissions of PM1.0 and PM2.5, especially heavy metals and secondary atmospheric pollutants, including SO42−, NO3−, and NH4+. Natural occurrences such as dust storms in April and May, as well as straw combustion in October, also contributed to the increase in WSIs and heavy metals. The exposure risk assessment results demonstrated that Zn was the main contributor to the average daily dose through ingestion and inhalation, ADDIng and ADDinh, respectively, among the 8 heavy metals, accounting for 51.7–52.5% of the ADDIng values and 52.5% of the ADDinh values. The contribution of Zn was followed by those of Pb, Cr, Cu and Mn, while those of Ni, Cd, and Co were quite low (<2.2%).

concentration distribution of inorganic/metal elements and WSis in pM 1.0 and pM 2.5 fine particles. To detect the elements trapped on the fine particles, the concentrations of inorganic elements, as well as WSIs within PM 2.5 and PM 1.0 obtained from different months, were analyzed, and the distribution of the typical trace elements of PM 1.0 was selected as an example and discussed (PM 1.0 in Table 1 and PM 2.5 in Table S2). Generally, the monthly concentration variation in different trace elements could reveal the origin of those pollutants and the possible sources of those elements 27,28 .
As shown in Table 1, the maximum concentration of the majority of inorganic/metal elements in Harbin was found in the heat supply period, except for Fe, Ti, Mn and Cr. In brief, the concentrations of Cu, K, and Zn in the fine particles were highest in Oct. and Nov., and As, S, Al, Ba, Sr and Si occurred in Dec., Jan. and Feb., while Ca, Mg, Pb and Mn were found in Mar. and Apr. Ca exhibited the highest concentration among all 18 non-WSIs, which is related to its wide application in building materials. Similarly, elemental Na ranked as the second-highest emission in particulates during the whole year, mainly originating from natural emissions and human activities. By contrast, the existence of Pb and Zn, specifically in summer, was partially ascribed to automobile exhaust (Pb in the heat supply period was also related to coal combustion). In addition, the continuous heat supply in Dec., Jan. and Feb. contributed to the higher emissions of S, As, Si to the atmosphere, which was mainly ascribed to coal combustion.
Among all 9 selected WSIs, the secondary atmospheric ions of SO 4 2− , NO 3 − , NH 4 + were predominant in the fine particles, which originated mainly from physical/chemical reactions 29 . Generally, WSIs were higher in the heat supply period than in the seasons with no heat supply, both for PM 1.0 and PM 2.5 (Fig. 2). For example, the mass concentration of SO 4 2− reached its maximum value in Jan., with an average value of 28.21 μg/m 3 for PM 1.0 and 33.88 μg/m 3 for PM 2.5 . Similarly, the highest values of NO 3 − and NH 4 + were observed in Feb., exhibiting maximum concentrations (in μg/m 3 ) of 15.86 and 14.25 for PM 1.0 , respectively. Since the atmospheric condition of the inland city Harbin was insignificantly affected by sea salt, the WSIs of Na + , Mg 2+ , and Cl − in the collected particles originated from industrial emissions and natural sources 30,31 . Na + ions, mainly originating from crustal  Fig. S2). Since the concentrations of Zn, Ti, Pb, As, Cu, Cr and Ni in the chemical composition of PM 1.0 /PM 2.5 are quite low, the sum of those proportions are defined as other concentrations. K mainly originated from the burning of biomass; thus, K was higher (>14%) in autumn for both PM 1.0 and PM 2.5 than in the other three seasons. S mainly originated from coal combustion and exhibited the highest value in winter. For comparison, Ca was highest in spring, which is related to the windy weather and dust emissions from building materials. Although Si basically originates from natural sources, coal combustion in winter contributed a significant amount of emissions into the atmosphere; thus, a higher concentration in winter than in the other seasons was finally observed. As shown in Table S3, the majority of the elements in PM 1.0 and PM 2.5 existed as inorganic elements regardless of the seasonal variation (≥58.0%), except for S and Si in spring. It should be noted that most of the inorganic elements in PM 1.0 are highly enriched, demonstrating that the majority of those inorganic particles were attached to PM 1.0 , and the seasonal variation insignificantly affected their distribution. ). Factor 5 was characterized by high loadings of Al, Na, K and Si, which points to crust/soil emissions. Factor 6, characterized by a high F − loading, is related to fossil fuel (especially coal) combustion emissions.
The contributions of each source/group of sources were predicted according to Bhuyan et al. 34 and are listed in Fig. 3. Overall, the percentage contribution of Factor 1, road dust, changed insignificantly with seasonal variations and was the highest in spring and lowest in autumn. Similarly, Factor 5 of crust/soil emissions accounted for as much as 16.1% of the PM 2.5 pollution in spring and was the lowest in winter (11.2%). For comparison, both the secondary formation (Factor 2) and fossil fuel combustion (Factor 6) exhibited the highest percentage contribution in winter (24.2% and 22.8%, respectively), followed by spring, whereas it was lowest in summer, ascribed to the heat supply from Oct. to Apr. for Harbin. In addition, the highest percentage contribution of biomass burning and vehicle emissions in autumn (29.9%) was related to the burning of straw, as well as the presence of 1.8 million automotive vehicles in Harbin City. The above results were similar to the previous observations of Shi et al. 36 , who found that the top five significant contributors to PM 10 in Chengdu were vehicle exhaust (28.71%), coal combustion (24.45%), resuspended dust (19.24%), secondary sulfate+nitrate (18.20%), and soil dust (16.53%). Similarly, the observations of Bhuyan et al. 34 revealed that the most dominant sources of the mid-Brahmaputra Valley of India were combustion sources, followed by road dust, construction dust and soil/crust input. The seasonal percentage contributions of the six sources of PM 1.0 were similar to those of PM 2.5 , in which the contributions of secondary formation and fossil fuel combustion were higher for PM 2.5 than for PM 1.0 , whereas soil/crust, combustion (biomass burning + vehicles), road dust and construction contributions were lower. www.nature.com/scientificreports www.nature.com/scientificreports/ Overall, the percentage contributions of the six main PM 2.5 sources of Harbin in the heat supply period were distributed as secondary formation (22.0%) > fossil fuel combustion (21.2%) > combustion (biomass burning + vehicles) (17.6%) > road dust (14.6%) > soil/crust (12.5%) > construction contribution (12.0%). By contrast, the contributions of secondary formation and fossil fuel combustion (coal) decreased to 16.1% and 15.1%, respectively, in the periods with no heat supply. Based on the above analysis, we conclude that coal combustion and secondary aerosols were the main sources of PM pollution in Harbin, especially in seasons with heat supply. Moreover, PM 1.0 is the most important fine PM emitted for most pollution sources. Thus, prevention and control measures should be urgently developed. www.nature.com/scientificreports www.nature.com/scientificreports/ Speciation characteristics of heavy metals in PM 2.5 . Excess heavy metals in PM 2.5 , especially the weakly bonded exchangeable fractions, endanger natural systems and human health 37 . Speciation distributions of heavy metals in PM 2.5 were characterized, and the corresponding speciation information is given in Fig. 4. Overall, the majority of the eight selected metals preferentially existed as a residual fraction (mineral crystal lattice), followed by water-and acid-exchangeable phases, while the reducible, oxidizable fractions had substantially lower contents. Briefly, the residual fraction of Cr, accounting for as much as 71.5% of the bulk Cr, was the highest among all eight metals, followed by Cu (41.1%), Zn (33.1%), Cd (31.1%), Mn (29.5%), Co (24.0%), Pb (17.0%), and elemental Ni (9.6%). For comparison, the water-exchangeable fraction of eight elements decreased in the order of Pb(40.3%) > Zn(33.6%) > Cd(29.7%) > Co(29.4%) > Mn(22.3%) > Cu (11.3%) > Ni(8.0%) > Cr(3.2%). Overall, Cd, Co, Pb, Mn and Zn exhibited a higher percentage distribution in the water/acid-exchangeable fractions (>46.4%) than in the other fractions.
According to Sabiene et al. 38 , residual, oxidizable and reducible heavy metals are quite refractory and are generally recognized as a stable fraction due to their weak bioavailability and low solubility, while water/ acid-exchangeable heavy metals are defined as a labile fraction 39 . As shown in Fig. 4, Pb was the most labile element among all eight heavy metals in PM 2.5 , accounting for 40.3% as water-exchangeable forms and 27.7% as acid-exchangeable forms. Zn was the second-most labile heavy metal in PM, exhibiting a relatively high percentage distribution of water-and acid-exchangeable fractions (33.6% and 22.1%, respectively). By contrast, Ni, Cu and Cr showed a lower percentage of labile fractions (≤34.1%). Overall, Cr was the most refractory metal in PM 2.5 (3.2% for water-and 9.2% for acid-exchangeable Cr). Based on the binding strength and solubility in different geochemical fractions of heavy metals, the potential toxicity of different heavy metal fractions is expected to decrease in the following order: exchangeable (F 1 ) > carbonates (F 2 ) > reducible (F 3 ) > oxidizable (F 4 ) > residual metal phase (F 5 ) 40 . Thus, we concluded that the elements Ni, Cu and Cr in PM 2.5 had a low hazard potential. For comparison, elemental Pb exhibited a converse trend, with the highest percentage ratio for F1 and F2 (40.3% and 27.7%, respectively).
Metal fractionation occurring in PM 2.5 and PM 1.0 is, in turn, likely to influence metal toxicity 41 . Toxicities of heavy metals in PM can be calculated according to the distribution of the fractional species. The equations for the risk assessment code (RAC) calculation can be summarized as follows: where RAC is the risk assessment code (%) of heavy metal pollution. A value of RAC < 1% usually demonstrates no risk; values of 1%≤RAC < 10% and 10%<RAC < 30% refer to low and medium risk, respectively; and an RAC ≥ 30% indicates a high risk, especially when RAC > 50% (very high risk). F 1 , F 2 , F 3 , F 4 and F 5 represent the percentage distributions of the water-, acid-, reducible-, oxidizable-and residual heavy metal fractions in PM 2.5 , respectively . C f refers to the pollution coefficient. As shown in Fig. 5, the results obtained from the RAC calculation demonstrated that Pb, Zn, Co and Cd in PM 2.5 were at a very high-risk level, which decreased in the order of Pb (68.0%)>Zn (55.7%)>Co (54.7%)>Cd (50.2%). For comparison, the elements Mn (46.4%), Ni (34.1%) and Cu (31.9%) were at a high level, whereas the RAC of elemental Cr was at a medium risk level (RAC = 12.4%). The distribution trend of C f was quite different from that of RAC and decreased in the order of Ni(9.39)>Pb(4.88)>Co(3.18)>Mn(2.39)>Cd(2.23)>Zn(2.02)>Cu (1.43)>Cr(0.40). Although Ni was at a low concentration level in PM 2.5 , its high RAC pollution potential implied that we should be highly concerned. In addition, the higher toxicities of Cd, Co, Pb, and Zn demonstrated that the emissions from coal combustion, the electroplating industry, metallurgy, the chemical industry and private cars in Harbin City should definitely be controlled, especially in the winter season. www.nature.com/scientificreports www.nature.com/scientificreports/ The distribution of the fractional species of heavy metals obtained from the periods with no heat supply was similar to the distribution in the heat supply periods (seen in Fig. 5). Specifically, the RAC of Cu, Co, Ni and Zn exhibited a slight increase during the periods with no heat supply compared to that during the heat supply period (<12%), and that of Cd, Mn and Pb increased significantly (41.5%-64.6%). The higher concentration of Cr observed in the heat supply period (listed in Table 1), as well as the corresponding higher RAC value, implied that the hazardous element Cr within the particles should be given more attention. All the tested heavy metals except Pb exhibited an increased C f value (especially for Cd, Mn and Co) during the period with no heat supply, implying that coal combustion during heat supply periods contributed to the emission/production of the residual fraction of heavy metals.

Risk assessment.
A carcinogenic and noncarcinogenic risk assessment for Cr, Ni, Cd, Co, Pb, Cu, Zn, and Mn in PM 2.5 samples was performed using the health risk assessment model of the US EPA, and three exposure pathways, namely, ingestion, inhalation, and dermal contact, were separately studied ( Table 2). In general, the hazard quotient results demonstrated that the three exposure pathways had the same trends of ingestion> dermal contact> inhalation, implying that ingestion was the most health-threatening exposure route for heavy metals in PM 2.5 in Harbin. Moreover, children suffered from a higher risk from ingestion and dermal contact from PM 2.5 than adults. Specifically, the ADD Ing values for children were 7.23 and 8.21 times higher than the values for adult males and females, respectively, indicating that children not only experienced a higher carcinogenic and noncarcinogenic risk but were also more vulnerable to that risk than adults. Similarly, the ADD Derm for children was 5.63 and 5.82 times higher than the ADD Derm for adult males and females, respectively. Conversely, the ADD Inh values obtained for children were 48.4% and 41.9% lower than the values for adult males and females, respectively.
As shown in Table 2, Zn was the main contributor to the ADD Ing and ADD inh among the 8 heavy metals, accounting for 51.7-52.5% of the bulk ADD Ing values for children, males and females; the next most significant contributors were Pb, Cr, Cu and Mn, with values ranging from 16.4-16.7%, 10.3-11.4%, 8.7-8.8%, and 8.1-8.2%,   www.nature.com/scientificreports www.nature.com/scientificreports/ respectively, while the ADD Ing values of Ni, Cd, Co were quite low (<2.2%). Similarly, Zn was also the predominant source of the ADD Inh risk and accounted for 52.5% of the bulk ADD Inh risk. Pb was the second highest risk source, and Cr ranked third, accounting for 16.7% and 10.3%, respectively. Cu and Mn were also important sources of the ADD Inh risk and contributed 8.8% and 8.2% of the bulk risk, respectively. For comparison, Pb played the most important role in ADD Derm risk and accounted for 66.5-69.7% of the bulk ADD Derm risk, followed by Zn (20.9-21.9%), while the ADD Derm risk of Cr, Ni, Cd, Co, Cu and Mn was quite low. In addition, Pb, Cu, Zn and Mn exhibited a higher ADD Derm risk for adults than for children, while the ADD Derm of Cr, Ni, Cd, and Co was significantly lower (especially for Cr and Ni). From the above, we can conclude that Zn and Pb in the PM 2.5 pollutants of Harbin City should be preferentially considered for their high environmental risk and RAC value; Cr, Cu, and Mn should be controlled first, while the environmental risk of Ni, Cd and Co was quite low.
Because the toxicity of the heavy metals was significantly affected by the fractional species distribution of the heavy metals, the risk assessments were further combined and analyzed with the fractional results. Compared to the periods with no heat supply, the RAC value of Cr, as well as the C f value of Pb, both exhibited a noticeable increase during the heat supply periods; thus, the pollution of Cr and Pb from coal combustion in heat supply periods should be of great concern. In contrast, Zn, Cu, Co and Mn from street dust should also be at a high level of concern in periods with no heat supply for controlling health risks 42 due to their high C f value and high ecological risk.
conclusions (1) The continuous heat supply from Oct. to Apr. led to serious air pollution in Harbin. The tremendous amount of emissions of the fine particle PM 1.0 during the heat supply period led to a higher PM 1.0 /PM 2.5 value of 0.832 than that during the months with no heat supply. The PM 1.0 particle number was highest in Dec. (50714/cm 3 ), which was 4.91 times higher than the minimum value found in August.

Materials and methods
Description of the city of Harbin. Harbin, one of the ten most populated cities in China, is the capital of Heilongjiang Province and is the commercial, industrial, and transportation center of Northeast China. The city is situated on the Songnen Plain, surrounded by mountain chains (the Lesser Khingan Mountains, the Changbai Mountains and the Higher Lesser Khingan Mountains in the north, east and west, respectively, to form a low-lying center), so special geographical and meteorological factors slow the wind speed. The mean annual temperature in Harbin is 3.5 °C; thus, heat supply is necessary during the period of Oct. 20 to Apr. 20 43  Twenty-four-hour PM 1.0 and PM 2.5 sampling was conducted every 7 days during the normal periods and every day during the polluted days from Nov. 2014 to Feb. 2015, using a fine particulate dust sampler (TH-150CIII and TH-150A, Mingxuan Company, Chengdu, China). The fine particulate dust sampler ran at a constant flow rate of 100 L/min. The 47-mm Teflon filters, with low inherent contaminants, were purchased from Pall Company (USA) and were dried/balanced to a constant weight (accuracy of 0.00001 g) at 298 K for 24 h before sampling. Filters were handled with tweezers coated with Teflon tape to reduce the possibility of contamination. In total, more than 150 samples were collected during the whole observation period. Number concentrations of PM 1.0 and PM 2.5 were collected by hand-held number concentration meters (TSI Company, USA), accompanied by the sampling of fine particles. Samples were taken three times a day (8:00 am., 12:00 am and 4:00 pm), each time for 5 min.
Chemical analysis. After weighing, the filters were sectioned, and one-fourth of the filter was cut into small pieces and soaked in a 60 mL polytetrafluoroethylene (PTFE) vial. The WSIs of the PM were obtained by ultrasonicating the filters for 150 min in 50 mL deionized water and then filtering the water extracts through a 0.45-μm cellulose acetate filter before chemical ion analysis. Concentrations of the cations (K + , Na + , Mg 2+ , Ca 2+ ) were analyzed by inductively coupled plasma-atomic emission spectrometry (ICP-AES); the anions (SO www.nature.com/scientificreports www.nature.com/scientificreports/ F − ) were determined using an ion chromatograph (Dionex 4500i, USA); and NH 4 + was determined by Nessler's reagent spectrophotometry.
To detect the species concentration distribution of the heavy metals within the atmospheric PM, one-quarter of the sampling filters was cut into small pieces, and the metals within the PM were extracted and analyzed according to Hu et al. 16 . Briefly, one-quarter of the sampling filters was cut into small pieces and immersed in a 15 mL digestion solution of concentrated HNO 3 and concentrated HClO 4 (3:1) for 10 h in a fume hood. Then, the beaker was baked with the temperature maintained at 100 °C until white smoke appeared, and the temperature was increased to 150 °C to volatilize the acid. When the sample solution had evaporated to near dryness, 2 mL of 2% nitric acid was added three times with continued heating. After being digested and cooled, the remaining solution was transferred to a 50 mL volumetric flask. In the next step, the flask was brought to the final volume by employing 2% HNO 3 . After filtration, the filtrate was stored in a 15-mL centrifugal tube and subjected to ICP-OES (VISTAMPX, US) for Al, Ca, Fe, K, Mg, Na, S and Si detection and ICP-mass spectrometry (MS) (VG PQ ExCell, Thermo Fisher Scientific Inc., USA) for Ti, Cr, Mn, Ni, Cu, Zn, As, Sr, Ba and Pb analysis.
Chemical-mineralogical speciation of heavy metals in the collected PM 1.0 and PM 2.5 samples was operatively measured by a five-step sequential chemical extraction procedure, which followed the extraction methods proposed by Tessier et al. 20 Specifically, sequential chemical extraction was carried out with 50 mL polypropylene centrifuge bottles. The water-exchangeable metal species were extracted with 20 mL MgCl 2 (1 M, pH 7.0) and with oscillation at 25 ± 5 °C for 16 h, after which the extracted solutions were separated from the filters by centrifugation at 4000 rpm at 25 °C for 10 min. Similarly, the acid-exchangeable, reducible, oxidizable and residual heavy metal fractions were extracted using NaOAC (1 M, pH 5.0), NH 2 OH·HCl (0.04 M), H 2 O 2 (8.8 M) and HF-HClO 4 mixtures, respectively. The combined supernatants were heated until 1-2 mL of solution remained, and then they were diluted to a volume of 10 mL with 2% HNO 3 to be stored in a polyethylene bottle at 4 °C before analysis.

Source apportionment analysis.
To understand the probable contributions from local point sources, PCA-MLR and the chemical mass balance (CMB) model were applied for air pollution source apportionment, according to the previous studies of Shi et al. 35 and Bhuyan et al. 34 . The CMB model was freely available for use on the US EPA official website. Seasonal pollution sources of PM 2.5 and PM 1.0 were analyzed and measured (Mar., Apr. and May were regarded as spring; Jun., Jul. and Aug. as summer; Sep., Oct. and Nov. as autumn; and Dec., Jan. and Feb. as winter).
Risk assessment. The risk assessment model developed by the US EPA was applied to evaluate the health risks posed by heavy metals in PM 1.0 and PM 2.5 . Considering the variety of physiological characteristics and lifestyles of Harbin City residents, we divided them into three groups: male (>16 years), female (>16 years) and children (<16 years). Since metal exposure can occur through direct inhalation, ingestion, and dermal contact, the average daily dose (in mg·kg −1 ·d −1 ) through ingestion (ADD ing ), inhalation (ADD inh ), and dermal contact (ADD derm ) can be calculated as follows 27 where C stands for the concentration of the contaminant in PM 2.5 (mg/kg for ADD ing and ADD derm , μg/m 3 for ADD inh ). For ingestion, the intake rate (IngR) was 100, 100 and 200 mg/day for males, females and children, respectively. For inhalation, InhR was 15.2 m 3 /day for males. BW, the average body weight, was 62.7, 54.4 and 15 kg for males, females and children, respectively. SA, the surface area of the skin that contacts the airborne particles (cm 2 ), was 4220, 3820 and 2160 for males, females and children, respectively. AF is the skin adherence factor for airborne particulates (mg/cm 2 ), which was 0.07, 0.07 and 0.2 for males, females and children, respectively. PEF refers to the particle emission factor and was 1.36 × 10 9 m 3 /kg. SL is the skin adherence factor, equal to 0.2 mg/ cm 2 d. EF is the exposure frequency, in days/year, and was 180 days. ED is the exposure duration in years, equal to 24, 24 and 6 years for males, females and children, respectively. ET stands for exposure time, 24 h/day. AT refers to the averaging time in days, equal to ED × 365. ABS refers to the dermal absorption factor, which was 0.001 for Cd, and 0.01 for the other metals. CF, the conversion factor (kg/mg), is 10 -6 .

Data availability
The datasets generated and/or analyzed during the current study are available from the corresponding author upon reasonable request.