Characterization of PM₁₀ Sampled on the Top of a Former Mining Tower by the High-Volume Wind Direction-Dependent Sampler Using INNA

Abstract

The PM₁₀ concentrations in the studied region (Ostravsko-karvinská agglomeration, Czech Republic) exceed air pollution limit values in the long-term and pose a significant problem for human health, quality of life and the environment. In order to characterize the pollution in the region and identify the pollution origin, Instrumental Neutron Activation Analysis (INAA) was employed for determination of 34 elements in PM₁₀ samples collected at a height of 90 m above ground level. From April 2018 to March 2019, 111 PM₁₀ samples from eight basic wind directions and calm and two smog situations were sampled. The elemental composition significantly varied depending on season and sampling conditions. The contribution of three important industrial sources (iron and steelworks, cement works) was identified, and the long-range cross-border transport representing the pollution from the Polish domestic boilers confirmed the most important pollution inflow during the winter season.

1. Introduction

The studied area—Ostravsko-karvinská agglomeration—is situated in the northeast part of the Czech Republic, in the Moravian-Silesian Region. The region, together with the adjacent cross-border Polish region of Silesia, is historically connected with the accumulation of black coal mining and heavy industry, namely energetics, coking plants and ironworks [1,2]. The particular industrial character of the region along with its topography (basin surrounded by fairly high mountain ranges) and local meteorological conditions [3] causes its specific air pollution problems. The strategic industrial development of the region in the 1950s [4,5] initiated intensive population growth connected with substantial emissions from households. This effect has persisted until the present, as coal is still the most widely used fuel in the Polish border area [6,7]. Thus, the region is one of the most polluted in Europe [8]. The air pollution significantly exceeds the limit values of particulate matter (PM₁₀, PM_2.5), benzo[a]pyrene and ozone [4,8,9] according to European legislation [10] and World Health Organization (WHO) guidelines [11,12]. Both daily and long-term exposure to the substances mentioned above has a number of substantiated adverse biological effects [5,13]. Airborne particles increase mortality and morbidity, especially of the respiratory system, even at short-term exposures. The population exposed to PM shows a higher incidence of infectious diseases [14,15,16], and atmospheric pollution and PM pollution are factors classified as proven human carcinogens (category 1) [17]. Being one of the most densely populated regions in the Czech Republic (except its capital city) [18], the air pollution here poses an essential and long-term environmental problem.

Substantial pollution in the region has challenged researchers to identify causes and look for solutions, starting in the 1990s with the US EPA project Silesia [19,20]; later on, the effort continued with projects AIR SILESIA [21,22] and AIR TRITIA [6,23]. Moreover, numerous case studies focused on the pollution origin were performed [24,25,26] together with the recently published study of the Czech Hydrometeorological Institute (CHMI) [27,28]. All the studies emphasized the role of both the industry and the transboundary pollution from Poland (caused mainly by domestic boilers). According to these studies and state air quality monitoring [9,29], the highest PM concentrations occur near the Czech–Polish border (characterized by more prominent growth in the colder half of the year and during smog events) and also close to important industrial sources where the limit values of PM happened to be reached not only during the winter season. The air quality in the region is significantly influenced by the rate and nature of cross-border pollution transmission along the most frequent wind directions (typically SW/NE), together with the inverse character of the weather with steady atmosphere and subsequently worsened dispersion conditions, which significantly contribute to increased air pollution during the winter. According to the available studies [6,22,27,30], the contribution of the cross-border PM pollution to the annual average values can vary from 20–40% depending on the location in the region, emissions and meteorological conditions in the year.

The air quality monitoring presented in this study was performed within the AIR BORDER project focused on the cross-border pollution transport from Poland to the Ostravsko-karvinská agglomeration (the Czech borderland) and vice versa. The aim of the study was to run a special monitoring campaign in order to characterize the transmission of the PM₁₀ particles from different groups of air pollution sources specific for the region, excluding the influence of the local sources [31]. This presumption was possible by locating a monitoring device on the top of a tower that reaches a height of over 85 m above the ground level. The device collects the PM₁₀ particles depending on the wind direction, which enables one to investigate from which directions and from which sources the pollution comes and to more precisely quantify its transfer within the region.

2. Experiments

2.1. Sampling

The monitoring device was situated on the top of a former mining tower (90 m above ground level) located in Horní Suchá village in the centre of the Ostravsko-karvinská agglomeration.

(World Geodetic System 1984 coordinates 49.805166 N, 18.473954 E). The location of the tower in the region can be seen in Figure 1, the tower in Figure 2.

For the particulate matter sampling, the high-volume sampler SAM Hi 30 AUTO WIND (Baghirra Ltd., Prague, Czech Republic) was used. The sampler operates respecting the Guidelines on high-volume sampling methods that can be found in the Compendium of Methods for the Determination of Inorganic Compounds in Ambient Air compiled by the United States Environmental Protection Agency (US EPA) [32]; a summary of the research of high-volume samplers performance can be found elsewhere [33].

The SAM Hi 30 AUTO WIND is a fully automatic, remote-controlled sampler intended for gravimetric and chemical analyses of aerosol particles. The device samples particulate matter with a <10 μm diameter (PM₁₀) using the DIGITEL DPM10/30/00 PM₁₀ pre-separator for 30 m³/h according to EN 12341 [34]. The sampler was designed to work depending on wind conditions. It has a magazine of 15 filters (glass microfiber, Whatman GF/A, Ø 150 mm) stretched in filter holders that are automatically changed to the sampling position according to actually evaluated wind conditions. Thus the sampler was able to collect PM₁₀ particles from eight basic wind directions (N, NE, E, SE, S, SW, W, NW) and CALM (wind speed < 0.2 m·s⁻¹). The sampler can be seen in Figure 3. Both wind speed and wind direction were measured via the WindSonic™ SDI-12 anemometer (Gill Instruments Ltd., Lymington, UK). The wind direction and wind speed for selecting a particular sampling filter from the magazine were determined according to one-hour moving averages calculated from 10-min data in accordance with US EPA guidance [35]. Thus, the actual wind may be different. The wind roses representing the measured wind direction in comparison with the sampled sector can be seen in Figure 4. Moreover, a special filter was designed for the episodes with extreme air pollution, defined as three successive average hourly PM₁₀ concentrations exceeding 100 µg·m⁻³, using 10-min data from continuous ground monitoring.

Employing this method, 111 PM₁₀ samples from eight wind directions and calm (and extreme air pollution if occurring) were collected for each month from April 2018 to March 2019 (total of 12 months).

2.2. Determination of PM₁₀ Mass Concentrations

PM₁₀ mass concentrations were determined following the guidelines of EN 12341 [34]. Filters were weighed using an analytical balance (Sartorius MC 210P) before and after sampling. The filters were conditioned for ≥ 48 h under controlled relative humidity (50% ± 5%) and temperature (20 °C ± 1 °C), then weighed for a first time, followed by a second weighing after additional conditioning for ≥ 12 h for filters prior to the sampling and for 24 to 72 h for filters after the sampling. In accordance with the requirements, the difference of weighing results was ≤40 µg for filters prior to the sampling and ≤60 µg for filters after the sampling. The filter weight was calculated as an average of the two results. Weights for the blank filters were also recorded. The collected PM₁₀ mass was calculated by subtracting pre-weight from the post-weight of the filters.

2.3. Element Content Determination Using Neutron Activation Analysis

One of the premises of this study was to apply NAA at the IBR-2 reactor of the Joint Institute for Nuclear Research (Russia) for the characterisation of sampled PM. Thus, the testing of filters convenient for the analyses preceded the sampling campaign. Six different filters comprising five different materials were tested (glass microfiber, quartz, PTFE membrane, cellulose and paper). Only glass microfiber filters were determined as being fully suitable; a quartz filter was more problematic for repacking and considering the price, the glass microfiber filters were chosen to be used for the sampling.

Before subjecting the samples to NAA, the preparation of the subsamples of exposed and blank filters was done, as the irradiation capsules of the applied pneumatic transport system have limited volume (Ø 18 mm) and the whole filter is not able to fit in it. This also allows putting more subsamples in one capsule so the subsamples from one month and the corresponding blank filter can be irradiated all together under the same conditions. For this purpose, a special automatic punching head was designed and made (used materials: stainless steel, Teflon and surface finish synthetic rubber). Prior to cutting, filters were folded in half to avoid the loss of collected material, and then 4 circles (Ø 16 mm) were cut from the folded filter using the layering of subsamples. This way, one subsample of each filter (counting eight layers of the filter) was prepared reaching a weight of 0.06–0.07 g depending on the exposure. The preparation of the filters took place under a relative humidity of 50% (±5%) and temperature of 20 °C (±1 °C). After the preparation, each subsample was vacuum-packed to be transported for NAA.

The NAA procedure started with unpacking the subsamples and weighing them under controlled relative humidity and temperature. Then the subsamples were immediately packed in polyethylene and aluminium cups for short-term and long-term irradiation, respectively. Once packed, they were put into irradiation capsules and transported to the reactor.

The employed NAA provides the activation with thermal and epithermal neutrons at low temperatures convenient for this type of samples. Complete information about samples irradiation, measurement and quality can be found elsewhere [36,37].

For short-term irradiation, Channel 2 (epithermal neutrons, flux density φ_epi = 2.0 × 1011 cm⁻²·s⁻¹) was used with an irradiation time of about 3 min. Samples were measured with 3–5 min. decay after irradiation for 15 min. Al, Ca, Cl, I, Mg, Mn, Si, Ti and V isotopes were determined in this way. For long-term irradiation, Cd screened Channel 1 (epithermal neutrons, flux density φ_epi = 2.0 × 1011 cm⁻²·s⁻¹) was used with an irradiation time of around 4 days. After 4-days cooling, the samples were repacked and measured twice. The first time, they were measured directly after repacking for 30 min to determine As, Br, K, La, Na, Mo, Sm, U and W, and the second time 20 days after the end of irradiation for 1.5 h to determine Ba, Ce, Co, Cr, Cs, Eu, Fe, Hf, Ni, Rb, Sb, Si, Sc, Se, Sr, Ta, Tb, Th, W, Zn and Zr.

Gamma spectra of activated samples were measured on HPGe detectors (resolution of 1.9 keV for the ⁶⁰Co 1332 keV line, efficiency 40%). The gamma spectra obtained were then processed using GENIE-2K software (CANBERRA) with the verification of the peak fit in an interactive mode. The concentrations of elements were calculated using certified reference materials irradiated simultaneously with samples via “CalcConc” software developed in the Frank Laboratory of Neutron Physics, Joint Institute for Nuclear Research [37]. The element concentrations were calculated by subtracting the corresponding blank filter values from determined values of the element in the subsample and recalculating using the mass concentrations of PM₁₀. In cases of values below the detection limit, half of the detection limit was used. In case of missing values (no data) caused by technical problems during the analysis, the data imputation was employed in order to retain the information in the dataset and allow the proper multivariate assessment [38]. The k-nearest neighbours algorithm (knn) [39] was applied for the missing non-sub-limit data.

Recommendations of US EPA [40] were kept with respect to the workplace standard operating procedures. The quality control of the NAA results was ensured by triplicating standards per batch of unknown samples and carrying out simultaneous analysis. For filter analyses, standard reference materials were used: 2709a–San Joaquin Soil Baseline Trace Element Concentrations from the National Institute of Standards and Technology (NIST), 2710a–Montana I Soil Highly Elevated Trace Element Concentrations (NIST), 2711a–Montana II Soil Moderately Elevated Trace Element Concentrations (NIST), 1632c Trace Elements in Coal (Bituminous) (NIST), 1633c Trace Elements in Coal Fly Ash (NIST), AGV-2 Andesite from the United States Geological Survey and 433 from the Institute for Reference Materials and Measurements (IRMM). Satisfying agreement between the experimental results and certified material was obtained. The accuracy was formulated as the percentage deviation from the certified value amount up to 10%.

2.4. Statistical Analyses and Visualization

Pearson correlations calculation, Principal Component Analysis (PCA) and the visualizations of the results were performed in the R environment [41], package FactoMineR [42,43], Openair [44] and ggplot2 [45]. Prior to the PCA, the data were transformed according to the compositional data analysis (CoDa) principles [46] using the centred log-ratio (clr) transformation [47]. Only the clr-transformed elemental concentration data were used for the construction of the model, and the information on the wind direction, calm and inversion situation were added as supplementary variables.

3. Results

3.1. PM₁₀ Mass Concentrations

The values of PM₁₀ concentrations corresponding to wind directions, calm and inversion situations are shown for year average, warm (from April to September) and cold (from October to March) seasons for the period of observation (04/2018–03/2019) in

Table 1. Average PM₁₀ concentrations for the observed period (µg·m⁻³).

Wind Conditions	Average 1	Warm Season Average	Z-Score 2	Cold Season Average 1	Z-Score 2
CALM	23.3	19.0	−0.35	27.6	1.16
N	22.9	14.6	−1.12	31.2	1.79
NE	21.7	16.5	−0.79	27.0	1.05
E	25.0	17.8	−0.56	32.3	1.97
SE	21.1	17.5	−0.60	24.6	0.64
S	21.9	16.9	−0.71	27.0	1.05
SW	14.8	14.5	−1.14	15.1	−1.04
W	21.3	23.6	0.45	19.0	−0.35
NW	16.8	15.9	−0.89	17.8	−0.56
Inversion	-	-		59.0

¹ Averages calculated for the concerned period (04/2018–03/2019), the inversion values not included. ² Z-score related to the average calculated for the concerned period (04/2018–03/2019).

and Figure 5a. The values recorded during inversion events are not included in the calculation of means for respective wind directions considering the set-up function of the sampling device (see Section 2.1). For the PM₁₀ concentration synopsis, see Table S1 the Supplementary materials.

The lowest PM₁₀ values were observed from the SW and NW directions regardless of the season. The concentrations were below average in the warm part of the year, with the exception of the west direction due to the peak concentration in August 2018 (Figure 6). The highest concentrations were sampled from the east and north during the cold season, though prevailing wind direction in the season was SW Figure 5b. This wind direction in the cold season was the prevailing wind direction for the whole period of observation, as Figure 5b shows (expressed as a sampled air volume). Calm was recorded for 12% of the sampling time.

Despite the peak concentration in August 2018, there was, in general, no directionality of the pollution in the warm season, as the concentration rose in Figure 6a shows. The peak August concentration originated most likely from the metallurgical complex in the west of the sampling site (see Figure 1), as the elemental composition of this sample also suggests (see Section 3.2). According to the meteorological data, this high concentration occurred in the period of steady cyclonic airflow (wind speed from calm to 2 m/s) preceding an upcoming cold front (see Figure 7). The meteorological situation for this event is illustrated via the ICON EU model data 10 m AGL and the Skew-T diagram as the global model (ICON, NOAA) in this situation was in total disagreement with recorded wind speed and wind direction data on the sampling site.

The April concentration rose has a similar course to the one for March (Figure 6). Considering the elements found in samples of these two months and the results of PCA (see Section 3.3), a similar origin of the pollution is expected. Thus, the April concentrations should be considered to appertain rather to the cold-season-related pollution sources.

During the winter season, the pollution came predominantly from the north, northeast and east directions, as the monthly average PM₁₀ concentrations show in Figure 6b. This clearly confirms the importance of the PM inflow from the Polish borderland. The increase of PM₁₀ concentrations from these directions in the winter season average 14 µg·m⁻³ though the prevailing airflow in winter is from the inverse direction (Figure 5b).

High PM₁₀ concentration was sampled in this season also from the south direction in February 2019. As there is no important pollution source in this direction (mountainous rural area), this peak was investigated closer. According to the meteorological data, the airflow was steady (wind speed 1–2 m/s), coming from the southwest via the Moravian Gate for more than one day, accompanied by a radiation inversion (see Figure 8). This indicates that the peak concentration originated from an important pollution source in the south part of the Moravian Gate, directing to the cement works near the city of Hranice (about 50 km from the sampling site), as the elemental composition of this sample confirms (see below in Section 3.2).

During the winter season, there were two smog events sampled, the first on 19 and 23 November 2018 (sampled on the same filter), with the PM₁₀ concentration reaching 60.5 µg·m⁻³, and the second on 23 March 2019, with a PM₁₀ concentration of 57.4 µg·m⁻³. In both cases, there was a temperature inversion connected with a steady airflow (measured wind speed < 1 m/s). In November, the prevailing airflow was from the NE, E and SE directions. Most likely representing the inflow of the pollution from the metallurgical complex on the southeast of the sampling site as confirmed by models (see Figure 9). For the March smog situation, the modelled airflow indicates the direction from the NE and E, suggesting the origin of pollution in the Polish borderland (see Figure 10). The elemental composition of these samples is stated in the next chapter.

3.2. Element Content Characterization

The basis of the study was to characterise the sampled PM₁₀ using NAA. Though this method enables the determination of a wide range of elements, it has specific limitations. Therefore, some of the important markers that would facilitate the source identification are lacking in the obtained dataset of element composition. Thus, the data of Pb (impossible to determine), Cd, Cu, Ni and Ti (determinable only if high concentrations are present) are absent. However, Cd, Ni and Ti were determined in a few samples, but the concentrations were below the minimal detectable concentrations and hence not included in further assessment.

Minimum and maximum levels, mean concentration values, medians and standard deviations of the elements determined in the PM₁₀ are shown in

Table 2. Synoptic table (minimum, maximum value, mean, median and standard deviation) of elements concentration (ng·m⁻³) determined in the PM₁₀ using INAA.

Element	Min	Max	Mean	Median	Stand. Dev.
Al	1.15 × 10−2	873.83	15.35	0.73	83.97
As	5.84 × 10−4	5.17	0.34	0.04	0.94
Ba	4.63 × 10−2	350.98	18.76	0.52	45.96
Br	3.08 × 10−4	1.72	0.18	0.12	0.22
Ca	8.78 × 10−1	87.38	13.47	7.33	15.53
Ce	6.47 × 10−3	1.01	0.12	0.07	0.15
Cl	5.39 × 10−2	150.66	15.73	2.12	29.37
Co	1.71 × 10−4	0.247	0.019	0.011	0.029
Cr	4.30 × 10−3	3.22	0.39	0.25	0.49
Cs	9.64 × 10−5	0.023	0.005	0.003	0.004
Eu	3.39 × 10−5	1.90 × 10−2	3.26 × 10−3	1.60 × 10−3	3.67 × 10−3
Fe	2.15 × 10−1	92.77	19.32	13.42	21.12
Hf	5.49 × 10−4	0.121	0.013	0.004	0.020
I	7.19 × 10−4	0.64	0.09	0.05	0.11
K	5.42	524.77	74.20	46.06	84.08
La	4.37 × 10−4	0.67	0.04	0.01	0.12
Mg	3.65 × 10−1	163.72	9.54	5.77	17.61
Mn	1.18 × 10−3	11.77	1.18	0.62	1.67
Na	1.30 × 10−1	1074.73	62.80	1.58	139.69
Rb	3.02 × 10−3	0.666	0.057	0.028	0.085
Sb	4.73 × 10−5	0.198	0.050	0.035	0.047
Sc	3.18 × 10−4	0.021	0.003	0.002	0.004
Se	1.80 × 10−4	0.176	0.022	0.012	0.027
Si	1.69 × 101	6423.41	1529.79	1077.04	1418.16
Sm	2.51 × 10−5	6.35 × 10−2	2.65 × 10−3	7.44 × 10−4	7.21 × 10−3
Sr	6.01 × 10−2	5.428	0.847	0.486	0.967
Ta	1.55 × 10−6	1.94 × 10−3	2.36 × 10−4	1.23 × 10−4	3.21 × 10−4
Tb	3.06 × 10−5	6.54 × 10−3	8.90 × 10−4	5.75 × 10−4	1.03 × 10−3
Th	2.57 × 10−4	2.90 × 10−2	3.67 × 10−3	1.97 × 10−3	4.81 × 10−3
U	9.78 × 10−5	1.51 × 10−1	4.96 × 10−3	7.31 × 10−4	1.74 × 10−2
V	2.42 × 10−4	0.191	0.027	0.018	0.030
W	8.41 × 10−4	0.202	0.012	0.007	0.022
Zn	1.39 × 10−2	729.39	40.84	0.16	99.27
Zr	2.28 × 10−2	7.14	1.04	0.80	1.03

; the correlation of the analysed elements can be found in

Table 3. Elements with Pearson correlation coefficients >0.7 determined for the respective seasons.

Warm Season						Cold Season
As	Cl	Cr	I	Mg		As	La	Sm	Sr	U
Ba	Zn					Ba	Ce	Rb	Sc	Sm	Ta	U
Br	Fe	La	Mn	Sm	Th	Br	I
Ca	K					Ce	Rb	Sc	Ta	Th	U	Zr
Ce	K	Rb	Th			Fe	Sc
Cl	Cr	I	Mg			I	Sb	Se
Cr	I	Mg				K	Na
Fe	Sm	Th				La	Sr
Hf	Si					Rb	Sc	Ta	U
I	Mg					Sb	Se
La	Sm	Th				Sc	Ta
Rb	Sm					Sm	U
Sm	Th					Ta	U
Ta	V

. Element contents and concentrations vary depending on the season; thus, the data were investigated separately. To see the element concentrations for the respective directions and seasons, see Table S2 and Figures S1 and S2 in the Supplementary Materials.

The correlation between the analysed element contents was also season-dependent, thus confirming the variability of the sources and the different origin of the pollution. A strong positive correlation (R > 0.7) was determined between As and Cr, Mn, Br and I in the warm season, and As correlated with La, Sm, Sr and U in winter. This suggests—together with the directionality of the concentrations of these elements—that As occurrence in the warm season is connected to the metallurgical processes [52,53,54], while in winter samples, its source appears to be black coal burning [52,55,56,57].

The element content of coal combustion products, or more precisely, fly ash plays an important role in this study for the determination of pollution origin mainly during the cold season. The distribution of elements during this process is well described in many works [56,58,59,60], and numerous factors need to be taken into account. First are the element content and its bonding in coal, as well as boiling points of elements and their compounds (in connection to a combustion temperature). Other important factors affecting the resulting emissions are also the type of furnace, rated capacity, combustion temperature, exposure time, the type of separator and its operating temperature, physic-chemical reactions with other substances (additives, sulphur or halogens) and others. Depending on these conditions, different elemental compositions of fly ash emission are described in the literature [56,61,62,63,64]. The elements presented in bituminous coal fly ash in the majority of information sources are As, Cd, Se, Pb and Hg; other elements vary. Thus, named elements can be considered as strong markers of this process. Regarding the limitations of NAA mentioned above, other less common elements also need to be investigated. Consequently, while determining the main sources of pollution depending on weaker element markers, the element composition of emissions appertaining to particular sources in the region was taken into consideration [52,57]. Thus, a presence or absence of an element can indicate or exclude an origin in a certain source.

In the warm season, Fe was strongly correlated with Br, Sm and Th, while in winter, it was correlated only with Sc. Considering the direction where the highest concentrations of these elements originated (E, NE), it can be assumed that their occurrence is associated with the primary metallurgy [52,53]. Cr correlated with As, I and Mg in the warm season, while in the winter samples, no significant correlation was found. The highest concentrations of these elements in the warm season came from the west direction, suggesting the relation with a steel and iron production [54,65,66] as a relatively high concentration of Br and I was found also in a local coking plant emission [52]. For more correlations, see Table 3. It should be noted that the Rare Earth Element (REE) levels were in good correlation, which is important for both evaluating the data quality and understanding the mass transport events [67,68].

Special attention was paid to the samples with the PM₁₀ peak concentrations above those mentioned. The sample collected in August 2018 from the west sector was characterized by high concentrations of Cr, Mg and I (the highest of all the set) and relatively high concentrations of Mn and Co. Cr and Co are important solutes for steel alloying in order to obtain special properties of steel; both Cr and Mg form important refractories used as lining in metallurgical facilities [65,69]. Moreover, Mg (with Ca) constitutes a base additive used in almost every step of the steel-making process from agglomeration and blast furnaces (dolomitic limestone, dolomite) to final steel-making (magnesite). Mn is a common element in austenitic steels produced in local steelworks [52,54,65,69]. The presence of iodine can be also connected with steel alloying as it is used for the production of certain transition metals, among them Co, Cr or Ti. Another possible source is coking [52]. Considering all these facts, the steelworks west of the sampling site are the most likely origin of the peak PM₁₀ concentration (see Figure 1 above).

The elemental composition of the south sample in February 2019 was characterized by high concentrations of Ca, Se and V (the highest of all the set) and relatively high concentrations of I and Sb. The high concentration of Ca is, in all probability, connected with a cement plant [52,70,71,72], and regarding the meteorological conditions discussed above, this peak concentration most likely originated from the cement plant quite far (about 50 km) to the southwest of the sampling tower. This hypothesis can be confirmed by the fact that, in this factory, alternative fuels based on waste and tar are used during the process of clinker firing besides coal, explaining the high concentrations of the other elements determined.

The two samples collected during the smog events significantly differed in their elemental composition. The PM₁₀ sampled in November 2018 was characterized by high concentrations of Ba, Ce, Fe, Hf, Rb, Sc, Ta, Th, U and Zr (the highest of all the set), while the sample in March 2019 by high concentrations of Si, Sr, Zn and Eu. The elemental composition of the first sample indicates two sources of pollution: coal burning and metallurgy [28,52,54,57]. These high concentrations were sampled during steady airflow from the NE, E and SE directions, suggesting the origin of the pollution in the steel plant southeast to the sampling site (see Figure 1 above) together with residential coal burning. The origin of the pollution in the case of the second inversion situation is not as clear; nevertheless, the modelled airflow points to the Polish borderland (see Figure 10) [52,54,65,69].

3.3. Principal Component Analysis (PCA)

The samples were investigated using a PCA (see Section 2.4) to understand the variability of the element composition and thus also the pollution origin. The results of PCA performed on the untransformed dataset with the variables season and sector taken as supplementary qualitative variables indicate that the factor season separates the plane and the individual measurements distinctively. This separation of the plane was confirmed by the Wilks test (p-value of 9.108346 × 10⁻¹¹). Similar separation also occurs when PCA is performed on the whole dataset with the clr-transformed concentration data (p-value of 5.219250 × 10⁻⁸), only it is realized alongside the second dimension. Thus, PCA results confirm that the season is a determining factor both for the elemental composition of the entirety of the measurements and for the directionality of the elements expressed as the change in the sectors with similar analysed concentrations. For more details, see Section 3.3.1 and Section 3.3.2. According to the analysis (chi-squared goodness-of-fit test), the elements varying the most depending on the season are Sr (p < 0.001), Cl (p < 0.001), Th (p < 0.001), Na (p = 0.001), Zn (p = 0.01), Cs (p = 0.015), Rb (p = 0.02) and Co (p = 0.04) with higher concentrations in the cold season and Mg (p < 0.001), Hf (p < 0.001), Sm (p < 0.001) and Ba (p = 0.015) in the warm season.

3.3.1. PCA of The Warm Season Measurements

According to the PCA results, the first two dimensions express over 81% of the dataset inertia, meaning that the first factor plane represents a major part of the total variability and that no other dimensions need to be considered. Moreover, the first axis itself represents almost 75% of the data; hence, it can be concluded that only this axis carries real information. The variability explained by this axis alone exceeds the reference value of 10.26%; this value is equal to 0.95-quantile of inertia percentages derived from the simulation of 10,000 data tables of equivalent size (normal distribution). The percentages of explained variance of the first five axes were 74.93, 6.52, 4.74, 2.46 and 2.1, respectively. A visualization of the PCA results of the warm season data is presented in Figure 11.

The elements most positively correlated with this dimension were Ba, Zn, Ca, Rb, La, Sm, Na, Hf and Al (Pearson correlation coefficient of 0.98, 0.98, 0.98, 0.98, 0.98, 0.98, 0.97, 0.97 and 0.96, respectively), while Ce was negatively correlated (Pearson correlation coefficient of −0.98). Thus, can be assumed that these elements describe this dimension well. Figure 8b shows that the sectors (from where the particulate matter was sampled) are divided alongside this axis mostly to the east and south group on its right and west and north group on its left. The calm has a high score on this axis, while having a high score also on the second axis.

3.3.2. PCA of the Cold Season Measurements

In the case of the cold season measurements, the PCA results show that the first dimensions express over 66% of the dataset inertia, meaning that while the first-factor plane represents a significant part of the total variability. The variability explained by this plane is higher than the reference value of 18.75%, and the increase in explained inertia for the second axis is also higher than the reference value for this axis (8.61%); hence, both dimensions carry real information. The percentages of explained variance of the first five axes were 54.87, 11.67, 7.2, 5.8 and 3.3, respectively. A visualization of the PCA results of the cold season data is presented in Figure 12.

The elements most positively correlated with the first dimension are Ca, Al, Sr, Ba, Zn, Rb, Na, Hf and Th (Pearson correlation coefficient of 0.98, 0.98, 0.97, 0.97, 0.96, 0.95, 0.93, 0.92 and 0.84, respectively), while concentrations of Ce are, once again, correlated negatively with this dimension (Pearson correlation coefficient of −0.95). As and Sm are the elements most positively correlated with the second dimension (Pearson correlation coefficients of 0.88 and 0.75, respectively). Figure 9b shows that the sectors are divided alongside these axes in a different manner than in the case of the warm season. Samples from the southeast, south, and southwest directions form a cluster with calm on the right of the first axis and the northwest, west, east and northeast located on its left. The inversion event (INV) measurements simultaneously have a high positive score on the first and high negative scores on the second axis.

4. Discussion

Within the study, the elemental composition of sampled PM₁₀ was determined with NAA, and the origin of the pollution was interpreted using specific pollutants as markers of pollution sources, multivariate statistics and meteorological models. More commonly, the pollution origin is investigated by means of receptor modelling [25,28,71,73,74]. However, specifics of the dataset concerned (long sampling period, small number of samples within each sector), an irregular time resolution of samples and a questionable construction of uncertainty matrix (different weights of respective samples) make this method inappropriate.

Despite the limitations of the methods used, the study confirmed that the region has specific types of pollution sources, including two metallurgical complexes (west and southeast to the sampling site), which, in specific meteorological conditions, increase the pollution load in the region and contribute to the pollution transfer in the higher atmospheric levels. However, in certain situations, this transfer is not detected by the ground pollution monitoring stations (in comparison to [9]). This corroborates locating the sampling device at 90 m AGL. By sampling in such height, the contribution of the local sources (residential emission, transport, constructions, autumnal waste biomass burning and others) [28,75] is excluded from the sampling, and the pollution transfer within the region can be more properly investigated.

The other specific type of pollution in the region is the pollution connected to the trans-boundary transfer coming from Poland during the winter season originating from coal burning in domestic boilers. In the cold part of the year, the PM₁₀ concentrations originating in the Polish borderland (the north, northeast and east direction) increased by almost 50%, despite the prevailing wind in this season blowing in the opposite direction. This fact has already been stated in many previous studies [3,21,26,27,29] and confirms once again the importance of the cross-border PM pollution in this region, emphasizing that it is not just a problem of directly adjacent areas but of the region as a whole.

For the determination of elemental content in the samples, neutron activation analysis was applied, and a wide spectrum of elements was analysed. This, on the one hand, brought specific limitations, but on the other hand, helped to identify the origin of element concentrations in samples using data on less commonly determined elements. Though it should be noted that NAA does not provide information on important elements such as Cd, Cu, Hg or Pb, the obtained information is, in the majority of cases, sufficient to identify the pollution source.

To collect more data on the pollution transfer in the region during meteorologically different years and to render the assessment more precise, the monitoring on the tower continues. In addition to the sampling device used in this study, the PM continual monitoring is now operated both on the top of the tower and on the ground level.

5. Conclusions

A specially designed high-volume sampler (SAM Hi 30 AUTO WIND) was used to collect PM₁₀ samples depending on airflow conditions. The sampler was located on the top of a former mining tower in 90 m AGL. This allowed the elimination of the influence of local sources and investigation of the regional pollution transport. The sampled particulate matter was analysed using the neutron activation analysis, which provided information on the content of 34 elements. This information—together with the PM₁₀ concentrations and meteorological data (measured and modelled)—was used to characterize the pollution origin in the region. A significant difference in the element composition was observed: elemental concentrations were dependent on both the season and the sampling direction. Contribution of three industrial sources, two ironworks (in the west and in the southeast) and a cement plant (southwest from the sampling site) was identified, showing that—though not detected by ground air pollution monitoring—these sources have a significant impact on the pollution transfer in the region. The measurements also confirmed that the PM₁₀ cross-border pollution inflow from Poland plays a crucial role during the winter season and contributes significantly to the air pollution in the whole studied region.

Currently, the air quality management and decision-making in the region are performed just on a local level without considering the cross-border impacts. However, to assure that the air quality meets the air pollution limit values, international cooperation and the legitimacy of the joint interregional approach to the air quality management is imperative, as proven herein [9].