Concentration and speciation of mercury in atmospheric particulates in the Wuda coal fire area, Inner Mongolia, China

Coal-seam fire is a source of atmospheric mercury that is difficult to control. The Wuda Coalfield in Inner Mongolia, China, is one of the most severe coal fire disaster areas worldwide and has been burning for more than 50 years. To investigate atmospheric mercury pollution from the Wuda coal fire, gaseous elemental mercury (GEM) concentrations and atmospheric particulate mercury (PHg) speciation were measured using a RA-915+ mercury analyzer and the temperature-programmed desorption method. Near-surface GEM concentrations in the Wuda Coalfield and adjacent urban area were 80 ng m−3 (65–90 ng m−3) and 52 ng m−3 (25–95 ng m−3), respectively, which are far higher than the local background value (22 ng m−3). PHg concentrations in the coalfield and urban area also reached significantly high levels, 33 ng m−3 (25–45 ng m−3) and 22 ng m−3 (14–29 ng m−3), respectively (p < 0.05). There is no clear evidence that PHg combines with organic carbon or elemental carbon, but PHg concentration appears to be controlled by air acidity. PHg mainly exists in inorganic forms, such as HgCl2, HgS, HgO, and Hg(NO3)2·H2O. This work can provide references for the speciation analysis of atmospheric PHg and the safety assessment of environmental mercury.


Introduction
Coal fires are a global catastrophe and have been reported in many countries and regions, such as the USA, Australia, South Africa, India, and China (Heffern and Coates 2004; Kuenzer and Stracher 2012;Mishra et al. 2011;Pone et al. 2007;Sharygin et al. 2009;Zhang et al. 2004). Heat generated by the oxidation of sulfides accumulates in the coal-seam, and fires occur as the heat intensifies (Heffern and Coates 2004;Walker 1999). Coal fires consume large amounts of coal resources and cause ground collapse (Kuenzer and Stracher 2012); they also release large concentrations of toxic and harmful gases to the atmosphere, including Hg, CO, SOx, NOx, and volatile organic compounds (Finkelman 2004;Hower et al. 2011;O'Keefe et al. 2011). The Wuda Coalfield, Inner Mongolia, is one of 14 large-scale coal bases in China, and it is also one of the most severe coal fire disaster areas worldwide (Liang et al. 2014). Underground coal fires in this area have been burning for more than 50 years (Zhang et al. 2008), and the annual coal loss is approximately 20 × 10 7 t, which exceeds the annual coal output of Germany (Cao et al. 2018). Local soil, dust fall, and air mercury pollution caused by coal fires have attracted significant attention from the research community (Liang et al. 2014;Liang et al. 2016;Liang et al. 2018;Li et al. 2018). Unfortunately, no effective measures are in place to curb coal fire pollution.
Mercury is a heavy metal pollutant with high neurotoxicity and a strong affinity for bioaccumulation (Fu et al. 2011). It is prone to long-range transportation and can be found in remote areas away from emission sources (Chance et al. 2015;Soriano et al. 2012). Atmospheric mercury can enter surface soil and water via dry-wet deposition (Ariya et al. 2004;Harada 2005). It can also return to the atmosphere via evasion fluxes from the water-air and soil-air interfaces, thus forming the global mercury cycle (Kalinchuk et al. 2021;Bowman et al. 2020;Ci et al. 2011;Windham-Myers et al. 2014;Yang et al. 2020). Generally, gaseous elemental mercury (GEM) is the primary form of atmospheric mercury, and atmospheric particulate mercury (PHg) often contributes less than 10% to the total atmospheric mercury Fu et al. 2010;Miller et al. 2021). However, GEM is known to convert to PHg under certain oxidative conditions (Ebinghaus et al. 2002;Obrist et al. 2011;Weiss-Penzias et al. 2015). The regional sedimentation and water solubility of PHg are stronger than those of GEM (Sun et al. 2020), and PHg can enter and permanently damage the human body through inhalation, dietary consumption, and skin exposure (Trasande et al. 2005).
Regarding atmospheric mercury concentration measurement, previous studies have recorded varying levels of success globally (Landis et al. 2002;Cheng and Zhang 2017;Zhang et al. 2017;Fu et al. 2010). However, reports that reveal the specific chemical form of atmospheric PHg are rare mainly because PHg concentration in ordinary environments is very low, often < 2 ng/m 3 , making it difficult to enrich; there is also a lack of effective analysis methods. The Wuda Coalfield, a typical PHg enrichment area, is suitable for analyzing the speciation of PHg. In this study, PHg concentration in the Wuda Coalfield fire area was tested, and specific PHg speciation was further revealed in combination with various experimental data. The results can serve as a reference for the speciation analysis of atmospheric PHg and the safety assessment of environmental mercury.

Study area
The Wuda District of Wuhai City (39°20′-39°40′ N, 106°3 0′-106°50′ E) is located in central Inner Mongolia in northern China; it is situated at the southern edge of the Ulan Buh Desert, the northern end of Helan Mountain, and the eastern border of the Yellow River. The district is in a temperate zone, with an annual average wind speed of 4.8 m s −1 (northwesterly winds prevail in the local area), average annual rainfall of 176.9 mm, and average annual evaporation of 3,641 mm. The Wuda Coalfield is located in the northwest of the Wuda District and has an area of 35 km 2 . It is rich in Carboniferous-Permian coal, with a reserve of 660 million tons, and mainly produces bituminous coal. The Wuda urban area is located approximately 6 km southeast of the coalfield and is one of the economic development zones approved by Inner Mongolia in 1998. The coal-seam fire in the Wuda Coalfield began in 1961 and was caused by the non-standard mining of small coal kilns (Liang et al. 2014). Six fire zones had formed in the surface mining area by 1978, which increased to 26 in 2004, covering a total area of~4 km 2 . Numerous attempts have been made to control the localized coal fire since 2009, but the coal fire trends remain volatile and uncontrollable.

Sample collection
Atmospheric PHg samples were collected in coal fire areas (located in the northwest) and neighboring urban areas (located in the southeast) in August and September 2015. In the coal fire area, on-site collection and testing activities avoided the underground flue gas discharge channel to ensure this investigation was objective and representative. An MLY-60 atmospheric particulate sampler was used to collect the total suspended particles (TSP) and PM 2.5 samples. The collection flow of the sampler was 20 L min −1 , and the collection height was 1.5 m above the ground. The atmospheric particle sampling membrane was a quartz filter membrane with a diameter of 47 mm. The quartz filter membrane was baked in a muffle furnace at 550°C for 6 h before use. After natural cooling, it was placed in an equilibrium chamber at a relative humidity of 35% and a temperature of 22°C for 24 h. The filter membrane was then weighed with a precision of 0.0001 g using an electronic balance and loaded into the membrane box. Mercury was not detected in the treated quartz filtration membrane. The collected samples were rebalanced and weighed in a balance room under constant temperature and humidity before being placed in cold storage. A total of 12 samples from the coalfield fire areas and 23 samples from the urban area (12 TSP samples and 11 PM 2.5 samples) were collected.
The filter membrane was divided into multiple parts for various experiments and tests. First, the filter membrane area was evenly divided into four parts using a quarter divider. Then, a pair of scissors was used to cut and separate the membrane along the dividing lines.

A DRI Model 2001 carbon analyzer (American Desert
Research Institute) was used to determine the OC and EC content of the atmospheric particulate samples. At present, thermal decomposition optical analysis is a recognized and mature method for determining OC and EC contents. Depending on the optical correction methods, the process can be divided into thermal-optical reflection (TOR) and thermal-optical transmission (TOT). The measurement principles of the two methods are similar, but the obtained results are different due to the time difference of the laser segmentation point (Shen et al. 2011). The DRI Model 2001 can provide both TOR and TOT data based on the thermal-optical analysis principle, resulting in more comprehensive data analysis. The analysis area was a quarter of the sample membrane.
The test results were converted to air concentrations according to the sampling volume.

Water-soluble ion and pH testing
For the atmospheric particulate samples, 10 water-soluble in- , Na + , NH 4 + , K + , Ca 2+ , and Mg 2+ ) were determined using an IC-8610 ion chromatograph (Qingdao Luhai Optoelectronics Technology Co., Ltd.). A quarter of the sample membrane was placed in a beaker, 10 mL of deionized water was added, and the mixture was then sonicated for 30 min. After standing for 60 min, the supernatant was collected for the experimental test. The anion test was performed using an IonPac AS19 separation column. The injection volume was 25 μL, and the eluent was a 25 mmol L −1 NaOH solution with a flow rate of 1.2 mL min −1 . The cations were separated using an IonPac CS12 column. The eluent was 30 mmol L −1 MSA solution with a flow rate of 1 mL min −1 and an injection volume of 25 μL. The linear fitting degree of the anion and cation standard curves was better than 0.999, and the detection limit (S/N = 3) was less than 0.02 mg L −1 . Three blank experiments were also performed using the blank quartz filter membrane, and their experimental mean value was used as the background value. The equivalent concentration obtained by the test was converted to air concentration. In addition, a pH meter (Mettler Toledo) was used to measure the acidity of the aqueous solution after sonication. Before measurement, the instrument was calibrated with a standard pH solution (pH = 4.01, 7.00, 9.02).

Mercury concentration testing
The GEM concentration in the air was measured using an RA-915+ mercury analyzer (Lumex Ltd., Russia), which employs Zeeman atomic absorption spectrometry with a highfrequency modulation of light polarization. The gas detection limit of the instrument was 2 ng m −3 , and the signal response time-frequency was 1 s −1 , making it suitable for real-time measurements of GEM (Hong et al. 2017;Li et al. 2018;Liang et al. 2014). The mercury content of the particulate matter samples was measured using the RA-915+ combined with a Pyro-915+ pyrolysis device, and 1/4 of the sample area was used. The standard curve was calibrated using GBW07447 (GSS-18) standard Hg soil (15 ng g −1 ), with a linear correlation coefficient (R) of 0.999.

Temperature-programmed desorption (TPD)
TPD combined with atomic absorption spectrometry (AAS) technology is suitable for the comprehensive analysis of Hg species in solid samples (Cao et al. 2020a;Cao et al. 2020b). The identification of mercury species needs to be based on the pyrolysis characteristics of standard mercury compounds. The characteristic pyrolysis peaks of some standard mercury compounds are shown in Table 1. The TPD experimental system was predominately composed of an argon gas supply device, an SK-G04123K tube pyrolysis furnace, an RA-915+ mercury analyzer, and an exhaust gas adsorption device. The heating range was 25-650°C at a heating rate of 10°C min −1 , and the argon purge rate was 1.0 L min −1 . The overview diagram of the TPD experimental system can be found in the report by Cao et al. (2020b).

Atmospheric Hg concentration
The wind direction was northwesterly during the entire sampling period, consistent with the perennial wind direction of the region, with a wind speed range of 0.6-5.2 m s −1 . To evaluate the level of atmospheric mercury pollution in the Wuda region, the GEM content at the coalfield boundary northwest of the Wuda District was measured for use as the background value. The data were recorded every 5 s, and the real-time measurement lasted 3 min. According to the on-site monitoring data, the background GEM concentration ranged from 13 to 29 ng m  (Fang et al. 2001), and Taiwan (6.3-9.4 ng m −3 ) (Kuo et al. 2006).
Using a scanning electron microscope (SEM-EDX) to analyze the morphology and chemical composition of 1,600 single atmospheric particles in the Wuda urban area, Wang et al. (2018) found that mineral particles, combustion particles, and sulfur-containing particles were the main particle forms in the region, which contained abundant weathered coal gangue and raw coal dust particles; this suggests that the atmospheric pollutants from the coalfield fires were transmitted to the downwind urban area. It should be noted that this study omitted the contribution of other potential mercury pollution sources (such as transportation, biomass combustion, industrial waste gas, and other long-distance transportation of particulate matter). The Wuda area has significant atmospheric mercury concentrations, despite its much lower population size and total gross domestic product (GDP) than those of the other cities listed in this manuscript. In addition, combined with previous investigations, the GEM content at the surface cracks in the Wuda coalfield fire area reached 464 ng m −3 (Liang et al. 2014), and the exhaust gas vents on the gangue hill slope reached 5908 ng m −3 (Liang et al. 2016), far exceeding the local GEM background value (22 ng m −3 ). The total mercury content in the coal fire sponge (CFS), a type of sponge-like contaminated soil found at surface vents of coal fires commonly seen in global coal fire regions, reached 13967 ng g −1 (2653-38470 ng g −1 ) ), approximately 1400 times the local soil background level (10 ng g −1 ) ). This suggests that coal fires are the highest mercury contributors in the area, with other source factors contributing significantly less in comparison.

Correlation between PHg content and OC/EC content in TSP
OC and EC are mainly derived from coal combustion, motor vehicle emissions, and biomass combustion. Moreover, EC is more stable and is principally sourced from the primary combustion emissions of coal (Gray et al. 1986;Salma et al. 2004;Viana et al. 2006  We observed no correlation between PHg content and OC/ EC content in TSP. The correlation coefficients (r) did not exceed 0.1 in the TOR and TOT modes (Fig. 1). This suggests that mercury in atmospheric particles is less likely to exist in its organic state and is also unlikely to be adsorbed on the surface of carbon particles in its elemental state. Therefore, mercury in atmospheric particles is more likely to occur in the form of inorganic compounds. , Cl − , and Na + are largely derived from sea salt particles, whereas, in inland cities, they are largely derived from fossil fuel combustion, the chlor-alkali industry, and other production activities. Moreover, K + is mostly derived from biomass combustion Behrooz et al. 2017b;Bisht et al. 2015;Zhang et al. 2013). The content ranges of F − , Cl − , NO 2 − , NO 3 − , and SO 4 2− in TSP in the Wuda urban area were 0.08-1. 41, 15.48-32.94, 0.09-0.38, 6.73-12.38, and 17.85-40.7 μg m −3 , respectively, and their average values were 0. 83, 22.75, 0.24, 9.50, and 25.15  Based on Pearson's correlation analysis, the PHg content was significantly correlated with Cl − and NO 3 − in PM 2.5 , and the corresponding r values were 0.854 and 0.745, respectively (Table 2). This indicates that Hg tends to combine with Cl − and NO 3 − to form mercury particles, likely as HgCl 2 and Hg(NO 3 ) 2 . We also observed a high correlation between PHg and NO 3 − in TSP (r = 0.643). The Cl − in TSP showed no clear correlation with PHg but was closely related to Na + (r = 0.890) ( Table 3), suggesting that the association between Cl − and PHg in TSP may be masked by the interference of NaCl particles. In addition, PHg was correlated with Ca + , Na + , and Mg 2+ contents in PM 2.5 (Table 2). However, this may be an indirect effect caused by the correlation between Ca + , Na + , Mg 2+ , and NO 3 − /Cl − ( Table 2). The contents of Cl − and NO 3 − in PM 2.5 accounted for 47 and 92% of the corresponding ion concentrations in TSP, respectively, indicating that PHg predominantly occurs as fine particulate matter. Some studies have shown that burning coal produces SO 2 and NO 2 , which subsequently forms H 2 SO 4 and HNO 3 through oxidation and moisture absorption (Fang and Liu 2010;Pathak et al. 2009;Anttila et al. 2006). As shown in Fig. 2, the TSP sampling volume from the coalfield fire area ranged from 11,476 to 25,490 L (average 17,849 L), with corresponding solution pH of 3.28-4.81; the TSP sampling volume from the urban area was 43,859-45,533 L (average 44,515 L), with corresponding solution pH of 5.43-6.41. The sampling volumes from the coalfield fire area were smaller but produced more acid; this suggests that the acidic gas produced by the combustion of underground coal fires was the main source of acid in this area, and it may also be transported to other areas through airflow migration. We observed a significant negative correlation between PHg content and acidity in TSP (r = −0.906), as PHg more easily accumulates in acidic environments. Fig. 3 shows the thermal decomposition curves of Hg in the TSP samples collected from the Wuda mining and urban areas under an argon atmosphere. Mercury decomposition predominantly occurs in the range of 150-450°C. Owing to the coexistence of various mercury species with different thermal stability characteristics, multiple Hg signal peaks appear in this temperature range (Cao et al. 2019a;Cao et al. 2019b).  To clarify the individual mercury peak ranges, Origin 6.0 software was used to deconvolve the mercury decomposition curves and obtain four peaks (A, B, C, and D) (Fig. 3). The starting temperature ranges of the four TSP peaks from the coalfield area were 173-253, 276-335, 320-380, and 367-379°C, with (Liu et al. 2013;Lopez-Anton et al. 2011;Lv 2017;Meng and Wang 2012;Wang 2016). The starting ranges of the characteristic peaks are highly consistent with peaks A, B, C, and D. The mercury in atmospheric particles in the study area therefore likely exists mainly as HgCl 2 , HgS, HgO, and Hg(NO 3 ) 2 ·H 2 O. This is consistent with the high correlation between PHg, Cl − , and NO 3 − obtained by water-soluble ion analysis.

Analysis of PHg speciation based on AAS-TPD
Based on the deconvolution results, the contents of HgCl 2 , HgS, HgO, and Hg(NO 3 ) 2 ·H 2 O in atmospheric particles from the coalfield area accounted for 9%, 23%, 58%, and 10%, respectively. The corresponding proportions in the urban area were 15%, 30%, 41%, and 14%, respectively. Thus, overall, the proportions of various mercury species in the coalfield area and urban area are similar. The difference in the proportion of mercury species may be due to related factors, such as industrial activities, the uniformity of the spatial distribution of particulate matter, and the accuracy of signal curve fitting.

Conclusion
The Due to the small number of samples, these results were obtained through a comprehensive analysis of multiple methods. The PHg content was related to NO 3 − and Cl − in atmospheric particles, and this was more apparent in PM 2.5 . The PHg concentration in atmospheric particulates is closely related to acidity. PHg in the atmosphere may predominantly exist as inorganic forms, such as HgCl 2 , HgS, HgO, and Hg(NO 3 ) 2 ·H 2 O, and also tends to occur as fine particles (≤ 2.5 μm). This work can, thus, act as a reference for analyzing the speciation of atmospheric PHg and the safety assessment of environmental mercury.
Author contribution Yahui Qian, sample analyses, data interpretation and presentation, and writing original draft; Qingyi Cao, conceptualization, data analysis, interpretation, and editing; Yanci Liang, review and editing; Zhe Wang and Yunyun Shi, sample collection and experiments; Handong Liang, supervision, validation, review, and editing Funding This study was financially supported by a project supported by the National Natural Science Foundation of China (41371449) and the State Key Laboratory of Coal Resources and Safe Mining (SKLCRSM19ZZ03).
Availability of data and materials The datasets used or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations
Ethics approval and consent to participate Not applicable.
Consent for publication Written informed consent for publication was obtained from all participants.

Competing interests
The authors declare no competing interests.