The determination of nine heavy metals was performed in 45 moss samples collected throughout the Kosovo territory. To better present and understand the concentrations which resulted from the chemical analysis, the crude data were processed by statistical methods. In table 1 descriptive statistics quantities are presented. In the table high concentrations of Al and Fe stand out, which are followed by Mn. High concentrations of these heavy metals are always expected as they are the prevailing heavy metals of earth’s crust. Zn and Pb also occur in high concentrations but less than Mn. The order of median concentration of the analysed metals is, Al > Fe > Mn > Zn > Pb > Cu > Cr > Ni > Cd. A striking difference between the maximum concentration and the 90th percentile can be observed for Ni and Pb. Although the maximum concentrations of these two elements are very high - 79 mg/kg for Ni and 38 mg/kg for Pb - their 90th percentiles are only 6.1 mg/kg and 10 mg/kg correspondingly. A big difference between the maximum concentration of 150 mg/kg and the P90 49 mg/kg was found also for Zn. The big difference between these two statistical quantities, indicates distinguishable zones over the study area with the presence of particular heavy metals, which can also mean presence of pollution. The high values of skewness and kurtosis, indicate that the distribution of concentrations of heavy metals in moss samples is not normal. Except for Pb and Ni the concentration distributions of all other heavy metals fit more closely lognormal distribution. Pb and Ni have also high coefficient of variation, most probably arising from artificial introduction of these two elements in the environment.
Spatial distribution
Spatial distribution maps of studied heavy metals are presented in figure 3. Al and Fe show very similar distribution patterns. Their highest concentrations are found in the east of Kosovo as well as around the centre. Maximum concentration of these two metals is found in Siboc (28) municipality of Obiliq, 2700 mg/kg for Al and 2000 mg/kg for Fe. High concentrations of Al and Fe are expected as they are the most abundant metals naturally present in soil, whereby under the effects of wind and precipitation they reach mosses. However the anthropogenic origin cannot be fully excluded (Ötvös et al. 2003) . The maximum Fe concentration in the present moss survey (2000 mg/kg) is lower than that of 2010 survey which was 3082 mg/kg, but the median is higher, 820 mg/kg vs 288 mg/kg (Maxhuni et al. 2016). Noteworthy is the fact that the difference between the present heavy metals concentration in mosses and that of 2010 survey, cannot be taken as completely valid as none of the sampling locations is the same, and the number of sampling points is larger in the present study. This should be considered every time the 2010 survey is referred to in this work. As can be seen in the distribution map, the lowest concentration of both elements are mostly found in the south and in the most north of sampling area. Both metals are present mostly in the districts of Prishtina and Gjilani. Geologically they are mainly present in the areas of clastites of Neogene-Paleogene and flysch of Mesozoic, where they occur in fine soil particles owing to rock weathering under atmospheric conditions and geological processes, whereby transferred in moss by wind and precipitations.
Manganese concentrations are also high in all moss samples with a maximum of 360 mg/kg in Petrovë (12), then Reqan (9) 329 mg/kg and Zveqan (41) 312 mg/kg. The highest concentrations of Mn lie almost in a diagonal of the Kosovo map from the north to the south. Mn also is a heavy metal with mostly geogenic origin, but according to its distribution map some anthropogenic contribution may be possible, as in the north of Kosovo Trepça mining and Coal powered electricity plant operate. Compared with 2010 survey, Mn concentration in the present survey is slightly higher (Maxhuni et al. 2016). Its concentration is comparable to values found in Albania and North Macedonia 2015 survey (Stafilov et al. 2018; Lazo et al. 2019), and very much smaller than the values found in Norway in 2015 (Steinnes et al. 2016). Prishtina, Prizreni and Mitrovica are the districts where Mn is present in highest concentrations, whereas according to geological areas it has the highest concentration in deluvium/proluvium of Quaternary period, Magmatic rocks of Paleogene – Neogene, and flysch of Mesozoic era. It can be seen that the highest concentrations of Mn relate mostly to particular geological formations which are found in the Prizren district, where there are not any heavy industrial sites. This area in the south, lies close to the border with North Macedonia where Mn was also found at high concentration in mosses (Stafilov et al. 2018). High presence of Mn in the Mitrovica district in Zveqan (41) corresponds to the mining area, but not only, as there is Lupç (37) where its concentration is also high (although lower than in Zveqan). Both, Zveqan and Lupç are found on or very close to Magmatic formations of Neogene-Paleogene and Clastic of Mesozoic. Lupç (37) is a relatively remote area where heavy pollution emissions are not expected - unlike Zveqan is - however Mn concentration in moss is high, probably indicating that Mn is mostly of geogenic origin even in Zveqan (41).
Zinc, lead, and copper come next as the most concentrated heavy metals in mosses samples in the territory of Kosovo. All of the three are found around the pollution sources, such as lead and zinc Trepça mines and ore processing units, ferro-nickel smelter in Drenas, coal power plant in Obiliq, and cement production plant in Hani i Elezit. Clearly these elements in the just mentioned areas are of anthropogenic origin. Zn is found at highest concentrations in the north of Kosovo at sampling points Zveqan (41) 86 mg/kg and Stanterg (42) 146 mg/kg, then in Çikatovë (27) 64 mg/kg and Shalc (34) 61 mg/kg. All those points in the district of Mitrovica where Zn concentration is the highest, are shown in figure 3. The median of Zn is lower than in the 2010 survey, the maximum is around twice as high. The median and maximum of Pb is slightly lower than in 2010, whereas the median of Cu is higher and the maximum lower than in 2010. The median concentration of Zn and Pb are higher than in Albania, North Macedonia and Norway, except for Zn which median is the same as in Norway (Steinnes et al. 2016; Lazo et al. 2018; Stafilov et al. 2018). In the contrary, median concentration of Cu was lower than in above mentioned countries. The highest concentration of Zn corresponds with geologic formation of Magmatic rocks of Neogene-Paleogene which are located around the same sampling sites too. Lead is more present in the North, in the South-West, and in the South-East, with concentrations lower than Zn and more heterogeneously distributed. Districts with the highest concentration of Pb are Mitrovica and Prizren, whereas according to geological composition Pb reaches the highest concentrations in Magmatic rocks of Paleogene-Neogene. Pb shows a similar pattern to Zn because of the Lead/Zinc mines and smelter in Mitrovica which clearly is the main source of these two heavy metals. Relatively high concentrations of Pb were found in the district of Prizren (7.9 – 10.08 mg/kg), despite the fact that there is no industry involving Pb particularly. In this regard it can be said that, apart from the traffic emissions coming from the roads and the highway, which pass just beside or join in the Prizren city, long range atmospheric transport may also contribute to the overall Pb concentration in moss samples (Steinnes et al. 1997a; Steinnes 2001). Cu is mostly present From the North to the South in the districts Mitrovica, Prishtina, Prizren, and Gjilan. Its highest concentration is found at sampling site Shalc (34), it is 8 mg/kg. Cu concentrations are lower than those of Zn and Pb, whereas they are more evenly spread than Pb. The districts with highest concentrations lie on the areas of geologic formations where Cu concentration is also high. Considering the fact that geologic areas where Cu concentration is higher, do not always correspond with pollution sources, it may be thought that Cu occurs in mosses due to natural processes. Particularly high concentrations of these heavy metals in the north have been reported also for soil samples (Šajn et al. 2013; Kerolli-Mustafa et al. 2015b; Kastrati et al. 2021), which partly explains their concentration in mosses. Fine dust particles contaminated with heavy metals, can become airborne during dry seasons and spread by wind in larger areas. But emissions during ore processing also can be spread and settle over a vast territory during time under gravity or by precipitations.
Chromium and nickel have a spatial distribution somewhat similar. Their highest concentrations appear in the west, centre and in the east of sampling area. The highest concentrations of Cr are those in sampling points Prapashticë (30) 10 mg/kg, Reqan (9) 6.6 mg/kg, and Lupç (37) 5.9 mg/kg. Cr and Ni medians are only slightly smaller than the 2010 survey in Kosovo, whereas medians of these two heavy metals are lower than those of Albania and North Macedonia but higher than Norway (always 2015 surveys). Prishtina is the district where Cr is at highest concentration, followed by Prizreni, and Peja. Geological formations that contain mostly Cr are clastic sediments of Neogene-Paleogene, whereas it is found at lowest concentrations in carbonate and metamorfic formations of Paleozoic. Ni represents an outstanding peak in sampling point Harilaq (22) 79 mg/kg, then in Llapushnik (21) 19.8 mg/kg, and Çikataovë (27) 11.6 mg/kg. These high concentrations compared to other samples are expected as the three sampling points are located just close by the Ferro-Nickel facilities, and they are as e result of direct pollution. Ni is most present in samples of the District of Prishtina, whereas Magmatic rocks of Mesozoic are the geologic areas where it is present mostly, with e very high difference compared to other geologic formations. The three sampling points with peak Ni concentration fall also in the areas of Mesozoic rocks, but outstandingly higher concentrations than in other Mesozoic rocks areas indicate the presence of pollution.
Cadmium is the heavy metal with the lowest concentration in mosses samples in the territory of Kosovo. Cd concentrations are the highest in sampling points Stanterg (42) 2 mg/kg and Zveqan (41) 1.7 mg/kg, Siboc (28) 1.15 mg/kg, Te Kalaja (25) 1 mg/kg, Kaçanik (6) 0.8 mg/kg. All these sampling points correspond to sites where pollution is expected because of industrial activities. Cd median is greater than that of 2010 survey in Kosovo, as well as it is greater than the medians found in Albania, North Macedonia, and Norway (2015 surveys). Districts with highest concentrations of Cd are Mitrovica and Gjilani, whereas according to geological formations it is Magmatic rocks of Paleogene-Neogene the formation with the highest concentration, and Carbonates of Paleozoic come second.
Multivariate analysis
In order to identify the pollution or geogenic origin of heavy metals studied, multivariate statistical analysis was carried out. In table 2, Pearson correlation coefficients are given for each combination of elements, where six significant correlations can be seen, all with p value under 0.05. The strongest correlation was observed for Al and Fe with a value of 0.936, then Zn and Cd with a value of 0.856. Strong correlation coefficients showed the pairs Fe – Cr (0.751) and Al - Cr (0.656), whereas there was only one moderate correlation Pb – Cd (0.437). Other correlations were weak or very weak.
To identify possible patterns of heavy metals distribution in mosses samples, and from it possible polluted areas and pollution sources, principal component analysis (PCA) was performed. The results obtained are shown in figure 4. The PCA analysis produced two main components (PCs), the first PC accounted for 29.75% of the total variance and the second PC for 26.25% of it. Sampling sites could be visualized in the plot of scores of PC1 and PC2. In the PCs score plot can be easily noticed the two sampling sites, Zveqan (41) and Stanterg (42), which score highly on PC2 but also considerably on PC1 and Shalc (34) score is relatively high in PC2. This result can be explained by the loading plot of the variables shown just under the samples plot. It can be seen that Zn, Cd, and Pb also load highly on PC2 and to some value on PC1 also, and the direction of the vectors in PCs loading plot are similar as it is the placement of the sampling points Zveqan (41) and Stanerg (42) in the PCs score plot. Similarly can be argued for Shalc (34). The angel between Zn, Pb, and Cd vectors is also small, particularly for Zn and Cd, which indicates a high correlation between these heavy metals. A high correlation of these elements as well as the vectors’ directions, show that Stanterg and Zveqan have high scores mostly as a result of these elements’ high presence in that area. This is obviously as a consequence of the Trepça mining and ore enrichment processes, as well as mineral tailing dump just in the south of Mitrovica city (Šajn et al. 2013; Kerolli-Mustafa et al. 2015a).
Sampling point Siboc (28) scored highly in both PCs, and Prapashticë (30), Lupç (37), and Reqan (9) scored highly in PC1. The high scores of these sampling sites can be explained by high values of the variables Al, Fe, and Cr. Although Al and Fe are mostly of geogenic origin, their particularly high concentrations in the moss sample of Siboc (28) clearly reveal the atmospheric pollution effect of coal digging and burning in electricity plant in Obiliq. Cr not being a crustal element but however present in this group of elements, also may be an indicator of the pollution effect of electricity plant. Moreover Cr was found at relatively high concentrations in the fly ash of lignite used to power the electricity plant (Kittner et al. 2018). Harilq (22) scores relatively high in PC1 and PC2 mostly due to high presence of Ni, and it reflects the atmospheric deposition of dust emitted majorly by ferro-nickel mine in Magure which is located just nearby, and probably also the geochemical composition of the surface soil of the area. As can be seen in variables loading plot, Cu, Mn, and Ni do not show a significant correlation between them or to any of the two already discussed groups. The score plot of the sampling points also shows that most of them score around the origin or in the upper left quadrant of the PCs axes, which means that most of the samples are not strongly influenced by the variables under consideration.
Pollution indices
To further estimate the atmospheric pollution level, the values of CF and PLI were calculated and are presented in table 3. According to the categorization of the CF values given by Fernandez (Fernández and Carballeira 2001), only Mn CF values fall in the no contamination interval since they never reach 1. Most of the CF of Cu also do not exceed 1 as well as those for Zn, except for Zn in sampling points Çikatovë (27) with a CF = 2.1 which corresponds to slightly contaminated, and Zveqan (41) CF = 2.8 and Stanterg (42) CF = 4.7 as moderately contaminated. Ni shows a variable CF values, among those four sampling sites correspond to moderate contamination, two severe and one has extreme contamination factor. Severe CF was found in Llapushnik (21) and Çikatovë (27), whereas extreme CF resulted Harilaq (22). In case of Al and Fe all sampling sites lie on moderate contamination category or lower. Cr has three sampling sites with severe CF and one with extreme contamination. Cd CF values are the highest for 10 moss samples and fall in the extreme contamination category. All moss samples with high CF are located nearby industrial facilities discussed previously, except Kukljan (1) where no industrial activities are performed but also the CF value of this site is in the lower limit of the sever category (8.9). The highest CF appear to be for Pb, the most of the sampling sites fall in the extreme contamination factor with values over 100, and the rest of samples corresponds to severe contamination category.
The PLIsite values for nine heavy metals throughout the sampling area, and then only for four of them are shown in figure 6. When taking all the nine heavy metals in calculation - according to Zhang (Zhang et al. 2011) PLI categorization - none of the moss sampling sites falls in the category unpolluted site (figure 6a). Ten samples fall in the unpolluted to moderately category, 24 of them are moderately polluted, nine are moderately to highly polluted, and two correspond with highly polluted category. The PLIzone for the whole territory of Kosovo when all nine elements were included in the calculation was 2.5, which corresponds to the moderately polluted. However when only Cd, Cr, Ni, and Pb were included in PLI calculation three sampling sites were put in category moderately polluted, only one moderate to highly polluted, three highly polluted, and 38 samples corresponded to very highly polluted category (figure 6b). The PLIzone of the entire study area now is 5.1, which is just in the threshold of very highly polluted category.