Introduction

Heavy metals are widely distributed in nature, in places such as water, soil, sediments, air and various forms of organisms (Zhang et al. 2009). Metals are non-biodegradable and accumulative in nature. Contamination of the aquatic environment by heavy metals occurs as the result of various human activities (Behra et al. 2002). Elevated emissions and their deposition over time can lead to anomalous enrichment, causing metal contamination of the surface environment (Wong et al. 2006). Metal contamination in these sediments could directly affect the river water quality, resulting in potential consequences to the sensitive lowest levels of the food chain and ultimately to human health (Kabir et al. 2011).

The sediments consist of inorganic and organic particles with complex physical, chemical and biological characteristics. They can scavenge some elements, thus acting as an adsorptive sink with metal concentrations in sediments often being many times greater than in the water column (Kalantzi et al. 2013). These metals may present both natural and anthropogenic origin. Main natural sources of heavy metals are weathering of soils and rocks and atmospheric deposition. Discharge of agricultural, municipal and industrial wastewaters into water bodies is dominant contribution from anthropogenic sources (Lin et al. 2011). To identify pollution problems, the anthropogenic contributions should be distinguished from the natural sources. Geochemical approaches, such as the enrichment factor (EF) and geochemical index methods, have been successfully used to estimate the impact of the human activities on sediments (Chabukdhara and Nema 2012) and many scientists have studied metal contamination in sediments using these approaches (Gao et al. 2013; Hui-na et al. 2012; Hu et al. 2011; Kabir et al. 2011; Harikumar et al. 2009; Abrahim and Parker 2008). Calculating pollution indices of heavy metal is a powerful tool for ecological geochemistry assessment. These indexes evaluate the degree to which the sediment-associated chemical status might adversely affect aquatic organisms and are designed to assist sediment assessors and managers responsible for the interpretation of sediment quality (Caeiro et al. 2005).

Multivariate chemometric techniques, such as principal component analysis (PCA) and cluster analysis (CA), are often used as additional methods of monitoring heavy metals (Wang et al. 2014; Jamshidi-Zanjani and Saeedi et al. 2013; Chabukdhara and Nema 2012; Iqbal et al. 2013; Relić et al. 2010, 2011; Varol 2011; Li and Zhang 2010; Yang et al. 2009). These statistical techniques appeared to be very useful in solving many environmental compartments and the identification of existing pollution pattern.

In this study, the evaluation of the metal pollution level and possible sources compared to background pollution was performed for river sediments in Serbia. The main objectives were (1) to determine the total content of As, Cd, Co, Cr, Cu, Fe, Mn, Ni, Pb, V and Zn; (2) to carry out calculations of the contamination factor (CF), EF, index of geoaccumulation (Igeo), ecological risk factor (Er), potential ecological risk index (RI), pollution load index (PLI), combined pollution index (CPI), modified degree of contamination (mCd) and toxic units to estimate the anthropogenic input of the elements and to assess the pollution status of the area; and (3) to identify the main sources of toxic elements. The authors consider that the combination of a wide range of indices as well as the use of multivariate statistical methods is a novel approach for assessing the distribution of metals in sediments which can be extended to other similar contaminated aquatic systems.

Materials and methods

Study area

European rivers are affected by a wide variety of pollution problems. European catchments are under pressure from ever-increasing water stresses and land-use change, especially those with a high conservation value in the Mediterranean area. In Western European rivers, very high levels of cadmium, mercury, lead and zinc have been recorded. Record levels of contaminations are observed in river basins that receive considerable discharges from industrial and mining activities and from large cities (Tockner et al. 2009). This is the case for many mining sites in the Western Balkans, where copper, lead, zinc and other elements are frequently found (Stuhlberger 2014). In Serbia, the primary extracted minerals include copper, coal, lead–zinc with associated gold, silver, bismuth and cadmium, red bauxite and modest quantities of oil and gas.

The largest and most important rivers that flow through Serbia are the Danube, Sava, Great Morava, West Morava, South Morava, Tisa, Ibar, Drina, Timok, Nišava, Tamiš and Begej Rivers. The Danube is the largest river in the country, flowing for 588 km inside Serbia, whilst forming the border with Romania, and is the second largest river in Europe. The basin drains parts of 19 countries. Some major tributaries of the Danube are still heavily polluted. Water quality in the Middle and Lower Danube remained relatively high (class II) between 1950 and the 1970s, but deteriorated afterwards due to rapid industrial development, poor pollution control and inputs from heavily polluted tributaries (Tockner et al. 2009).

The 945-km-long Sava River is the largest tributary of the Danube by volume (average discharge: 1,572 m3/s) and the second largest, after the Tisza, by catchment area (95,793 km2). In Serbia, it remains a typical lowland river. Until the 1990s, the Sava was affected by heavy pollution from the metallurgical, chemical, leather, textile, food, cellulose and paper industries, as well as from agricultural activities. The Sava is the main recipient of wastewater from many cities and is impacted by the polluted water of the tributaries (Tockner et al. 2009).

The River Tisza (966 km; 157,220 km2) is the largest Danube tributary in its length and catchment area. A small part of the Tisza watershed area lies within Serbian territory and covers only 6 % or 8,994 km2. This area belongs to the Autonomous Province of Vojvodina. It is predominantly lowland and part of the Pannonian basin, which is the largest of the sediment-filled, postorogenic basins of the Alpine region (Adriano et al. 2003). Ecological accidents that took place in 2000 in the superior watershed and their effects on the rivers upstream with resulting transboundary influences proved the necessity of a regional approach to environmental protection and the active involvement of riverside countries in establishing common strategies and programs for preventing and reducing the risk of accidental pollution (Adriano et al. 2003).

The 308-km-long West Morava drains south-western Serbia (catchment area: 15,567 km2). The largest tributary is the Ibar.

Description of the sampling site and sampling

As riverine pollution can cross national borders, it is important that levels of pollutants in water and sediment are assessed in all the countries through which the river flows. Due to this factor, sediment samples in this study were collected from all of the main rivers in Serbia. A total of 35 samples of river sediment from 15 rivers in Serbia were collected during the year 2008 (Table 1). For the larger rivers, sampling was conducted at several locations (Fig. 1; Table 1). The sediment samples were stored at 4 °C to prevent changes in the chemical composition of the sediments. The contents of the micro- and macro-elements were determined in the granulometric fraction <63 μm of the bottom sediment samples (“grab”—the samples) after air drying for 8 days (Sakan et al. 2011). Detailed information about natural variability and sediment and soil properties in studied region is shown in paper Sakan et al. (2014).

Table 1 Sampling locations
Full size table
Fig. 1
figure 1

Location map of the study area

Full size image

Analytical procedures and element analysis

The total content of elements in the sediments was determined by digestion with very strong acids: HNO3 + HCl + HF. A 0.5-g sediment sample was placed in a Teflon vessel and then 9 ml HCl (37 %), 3 ml HNO3 (70 %) and 3 ml HF (48 %) were added to the vessel. One Teflon vessel only contains the acid mixture for a blank. A microwave digestion system brings the sample to 165 °C in 10 min (holding time 0 s), then 175 °C in 3 min, where it was kept for 10 min (max power was 1200 W) (Rönkkömäki et al. 2008). Microwave digestion was performed in a pressurised microwave oven (Ethos 1, Advanced Microwave Digestion System, Milestone, Italy) equipped with a rotor holding 10 microwave vessels (PTFE).

After cooling the digestion system, 10 ml H3BO3 (5 g/100 ml water) is added. Again, the microwave digestion system brings the sample to 175 °C in 10 min at 1420 W and holds the temperature at 170 °C for 3 min. The digestate is then diluted to a final volume of 100 ml (Nam et al. 2001).

In this research, the following elements were determined in each sample: As, Cd, Co, Cr, Cu, Fe, Mn, Ni, Pb, V and Zn. The results are expressed in mg kg−1 dry sediment. The analytical determination of the studied elements was realised with an atomic emission spectrometer with an inductively coupled plasma iCAP-6500 Duo (Thermo Scientific, UK). The detector was a RACID86 charge injector device (CID). Analytical grade chemicals were used throughout the study without any further purification. The metal standards were prepared from a stock solution of 1,000 mg L−1 by successive dilutions. The concentrations obtained for all the elements in the blanks were close to the detection limit of the method, indicating that contamination was not a problem in the digestion.

Assessment of sediment contamination

The degree of metal enrichment in sediments was carried out using following formulas:

CF was calculated as the ratio between the metal content in the sediment at a given station and the normal concentration levels (background concentration), reflecting the degree of metal enrichment in the sediment: CF = Cs/Cb. CF was classified into four groups in Hakanson (1980) and Pekey et al. (2004). EF was calculated as: EF = (M/Y)sample/(M/Y)background, where M is the concentration of the potentially enriched element and Y is the concentration of the proxy element. Igeo was calculated as Igeo = log2 [C metal/1.5 C metal (control)], where C metal is the concentration of the heavy metal in sample and Cmetal (control) is the concentration of the metal in the unpolluted sample or control. The factor 1.5 is attributed to lithogenic variations in the sediment (Mendiola et al. 2008). Ecological risk factor (Eri): Er i = Tr i C i f, where Tr i is the toxic–response factor for a given substance [for Hg, Cd, As, Cr and Zn, they are 40, 30, 10, 2 and 1, respectively; and 5 for Pb, Cu and Ni (Yang et al. 2009)], and C i f is the contamination factor (Hakanson, 1980). Ecological risk index (RI): RI = ∑ (T i  × C i /C o), where n is the number of heavy metals, T i is the toxic–response factor for a given substance, C i represents the metal content in the sediment and C 0 is the regional background value of heavy metals (Hakanson, 1980; Yang et al. 2009). PLI: the PLI for a single site is the nth root of the number (n) of multiple CF multiplied together (Tomlinson et al. 1980; Mohiuddin et al. 2012): PLI = (CF1 × CF2 × CF3 × …..CF n )1/n. CPI: there are two steps in this assessment process. The first step is to determine the contamination coefficient of each heavy metal which is obtained by: C if  = C i/C in . In the second step: the CPI is obtained by CPI = ∑ C if /m, where C if is the contamination coefficient of heavy metal, C i is the concentration of each heavy metal in the sediment, C in is the background values, and m is number of heavy metals and CPI. mCd: this factor is similar to CPI according to the method of calculation: mC d = ΣC if (i = 1, i = n)/n, where n = the number of analysed elements, i = the ith element (or pollutant) and C f = contamination factor, but differs in the gradation degree of contamination (Abrahim and Parker 2008). TU is defined as the ratio of the determined content to the probable effect level (PEL) (Gao et al. 2013; Harikumar et al. 2009; Pedersen et al. 1998). The PEL is defined as the lower limit of the range of chemical concentrations that is usually or always associated with adverse biological effects (Gao et al. 2013). Sediments with measured chemical concentrations equal to or greater than the PEL are considered to represent significant and immediate hazards to exposed organisms. The PEL represents the lower limit of the range of chemical concentrations that is usually or always associated with adverse biological effects (CCME 1995). It was derived using a weight of evidence approach in which matching biological and chemical data from numerous modelling, laboratory, and field studies performed on freshwater sediments were compiled and analysed, i.e. the geometric mean of the 50th percentile of the effect data set and the 85th percentile of the no effect data set (Smith et al. 1996; CCME 1995). In this manuscript, for TU calculation used PEL values are published in Smith et al. (1996) and MacDonald et al. (2000).

Statistical analysis

Multivariate analysis was performed using PCA and CA. The statistical analyses were conducted using SPSS 11.0 using the log-transform normalised data set. The element used for sediment normalisation is aluminium. This element was chosen because Al is a conservative element and a major constituent of clay minerals, and has been successfully used in previous investigations (Rubio et al. 2000).

Results and discussion

Performance of the analytical procedure

The accuracy and precision of the obtained results were checked by analysing sediment reference material (BCR standards, 143 and 146R). The percentage recovery of each element was obtained as: [(measured concentration in mg kg−1)/mean certified value for CRM in mg kg−1) × 100]. The results indicate a good agreement between the certified and analytical values (Table 2). The recovery of elements being practically complete for most of them and the values were in the acceptable range (recovery 80–120 %) (Chang et al. 2009).

Table 2 The result of accuracy and precision of applied method
Full size table

The precision is expressed as relative standard deviations. The relative standard deviations of the means of duplicate measurement were less than 4 % (0.03–3.80 %) for all the measured elements.

Descriptive statistics of elements contents in sediments

The total contents and basic statistics (minimum and maximum, as well as the means and standard deviations) for the total contents of all the measured elements in the investigated river sediments are shown in Table 3. The distribution of heavy metals in studied sediments showed large spatial variations amongst the different rivers suggesting large natural variability of metal concentrations in sediments and different origins of these elements. The order of the total element content was Zn > Cu > Pb > Ni > Cr > V > Co > Cd > As (not taking into account Fe and Mn). The contents of all elements were higher than the background values for the investigated elements (Table 4). Also, the contents of Cd, Cr, Cu, Ni, Pb and Zn were above the PEL, which indicates that the investigated river systems are polluted with toxic elements and that the content of these elements in sediments poses a potential risk to aquatic life. Because of that, additional calculations were performed the severity of contamination by these elements in sediments.

Table 3 Total contents and descriptive statistics of elements in different river sediments, this study (mg kg−1)
Full size table
Table 4 Comparison between background values for elements given by other authors and calculated for the investigated river sediment (mg kg−1)
Full size table

Calculating the grade of contamination and risk category

Contamination of the investigated river sediments with the studied trace elements was assessed using: CF, EF, Igeo, Er, RI, PLI, CPI, mCd and TU. Tables 5 and 6 show the predicted severity of contamination determined by these indices and ecological risk based on the calculated factors.

Table 5 Grades of enrichment factor, index of geoaccumulation, and modified degree of contamination
Full size table
Table 6 Grades of ecological risk factor, contamination factor, and ecological risk index
Full size table

Table 4 shows the background values for the elements in sediments given by other authors (Rubio et al. 2000; Sakan et al. 2009) and calculated for the investigated river sediment, as well as values for the element content in continental crust (Wedepohl 1995) and the PEL values (Smith et al. 1996; MacDonald et al. 2000). The element contents in the DTD canal (station in Vrbas) were chosen as the background values for elements in this research because there are no significant anthropogenic sources of toxic elements at this locality and the sediment samples are similar to the other investigated river sediments in geochemical characteristics and composition. Many researchers recommend that it is better to use the regional background values (Rubio et al. 2000) than the element content in continental crust or the average values of the element contents in world soils, etc. In this research, the same background values (quoted in Table 4) are used for calculating all the pollution indices. For the background value of As, the content in the continental crust is used since the content of As extracted from the DTD canal (station in Vrbas) was below the detection limit.

Contamination factor (CF)

Values of the CF (CF, Table 6) were classified into four groups using the classification scheme given by Pekey et al. (2004) and Hakanson (1980). The results of the CF groupings are presented in Table 7 and Supplementary Table 1. The highest CF value, for Cu (very high contamination), was found at site 33 (the river Pek). CF values >6, which indicate “very high contamination”, were observed for Zn, Cu, Ni, Cd and As. For other elements, the CF values indicated: no metal enrichment to considerable contamination (Cr), no metal enrichment to very high contamination (As, Ni), moderate contamination to very high contamination (Cd, Cu, Zn); moderate contamination to considerable contamination (Fe, Mn, Co, Pb), and moderate contamination for V. From the CF values, contamination with the investigated elements can be observed, especially with Zn, Cu, Ni and Cd.

Table 7 CF, EF, Igeo, Er and TU values of elements in sediment samples
Full size table

Enrichment factor (EF)

The EF values were interpreted as suggested by Acevedo-Figueroa et al. (2006) (Table 5) and were calculated separately for all the sampling sites (Table 5, Supplementary Table 1). River sediments in Serbia showed a wide range of trace element enrichment. In general, the order of the average EF values was Cu > Zn > Cd > Co > Pb > Ni > Mn > Cr > As. According to the categories (Table 5), these findings indicate that Cu, Zn and Cd enrichment was high. From the pollution point of view, the EF of Cu in the river Pek (35.03) was the highest amongst the elements in the investigated sediments, suggesting significant contamination at this locality.

Index of geoaccumulation (Igeo)

Müller (1979) proposed seven grades or classes based on the increasing numerical value of the Igeo index (Table 5). The greatest number of samples and elements belong to Class 1 (Fig. 2; Table 7, Supplementary Table 3), i.e. unpolluted to moderately polluted sediment (with As, Cd, Co, Cr, Cu, Fe, Mn, Ni, Pb and V), Class 2 (Cd, Cu, Zn, and Pb in some samples), Class 3 (Cu in the river Pek and the West Morava 1, and Zn in the Ibar 1 and Ibar 2 and Pek) and the largest Igeo value –3.92 had the sediment from the Pek for Cu (Class 4, heavy pollution). These findings indicate that the primary contaminants in the rivers are Cu and Zn, like Cd. These results are consistent with EF values.

Fig. 2
figure 2

Index of geoaccumulation (Igeo) of analysed elements in the sediments

Full size image

Ecological risk factor (Er)

The ecological risk assessment results are summarised in Table 7 and the Supplementary Table 4. The potential Er indices of Cr, Ni, Pb, Zn and As were lower than 40 (except for arsenic at two locations in the river Ibar), which indicates a slight potential ecological risk of these elements in the main Serbian river. Amongst the studied elements, Cd and Cu present a higher ecological risk than any other elements because of their higher toxicity coefficient.

Ecological risk index (RI)

The RI was calculated as the sum of all the calculated risk factors (Table 8 and Supplementary Table 4). The potential RI values were generally lower than 300, which suggest that sediment samples from the river catchments exhibited low and moderate ecological risk for the investigated elements. However, three (two sediment samples from the Ibar and one from the Pek Rivers) of the 34 samples of sediments had large RI values (300–600), which indicates a high ecological risk of these elements.

Table 8 The average values of RI, PLI, CPI, mCd and TU in sediment samples
Full size table

The contribution percent of the individual element to the overall potential ecological risk revealed that the most toxic element, Cd, is the main contributor to the total potentially ecological risk. Because Cd pollution, in general, was the result of a long history of accumulation, there is a strong potential risk to the ecosystem and the health of the residents in this region (Gao et al. 2013).

Pollution load index (PLI)

Supplementary Table 1 presents the PLI values of the studied elements. When the PLI > 1, it means that pollution exists; otherwise, if the PLI < 1, there is no metal pollution (Varol 2011). The PLI values varied from 1.4 to 4.27 for the investigated sediments, suggesting that pollution exists. The highest PLI values were observed in the following sediments: Ibar, Pek, West Morava and Great Morava and these river systems are significantly polluted.

Combined pollution index (CPI)

The CPI values are shown in Supplementary Table 4. When the CPI < 1, the sediment is unpolluted; when the CPI ≥ 1, the sediment is contaminated by heavy metals (Hui-na et al. 2012; Jian-Min et al. 2007). The CPI values range from 1.65 to 10.19, which indicates that the investigated sediments are contaminated with the studied trace elements. A more detailed discussion will be presented through a discussion of the mCd, considering that these two factors are calculated in the same manner.

Modified degree of contamination (mCd)

The mCd is based on integrating and averaging all the available analytical data for a set of sediment samples. Based on the values of the mCd factors (Supplementary Table 4), it is possible to distinguish differing levels of metal contamination (Table 5). The mCd values for the investigated sediments generally lie in the range 1.65–10.19. Using the classification system proposed for this parameter, the overall range of mCd values indicates a low to very high degree of contamination. The mCd data indicate a significant anthropogenic impact in sediment from the Pek, Ibar and West Morava Rivers. The mCd values in the Danube sediments indicate a low degree of contamination, whilst in other mCd sediment values indicate a moderate degree of contamination.

Ecological toxicity (TU)

Figure 3 shows the TU value for each metal and the sum of the TU values of all the trace elements in the investigated sediments. The PEL values used for calculating the TU in this paper are shown in Table 4. The average trace element TU at the investigated localities followed the order: Ni > Pb > Cd > Cr > Zn > Cu > As. Based on the mean value of ΣTU, it can be concluded that potential toxicity existed for the quoted elements. All the values were higher than 4. From the toxic unit values, stations I1 and I2 (Ibar), Z2 (West Morava), V2 (Great Morava 2) and Pe (Pek) are more polluted compared to the others (ΣTU > 10).

Fig. 3
figure 3

The TU of each element and the sum of TUs of elements in the investigated sediments

Full size image

Trace elements and sample site grouping using principal component analysis (PCA) and hierarchical cluster analysis (HCA)

PCA

A PCA with Varimax normalised rotation was performed in this research. The scores and loadings of the principal components are presented in Table 9 and Fig. 4. PC1, explaining 60.86 % of the total variance, has strong positive loadings for Fe, Mn, Co, Cr, V, Pb, Cd, and Ni. PC2, explaining 15.67 % of the total variance, has strong loadings for As, Pb, Zn, Cd and Ni. PC3, explaining 10.58 % of the total variance, has a strong positive loading on Cu.

Table 9 Loadings of experimental variables on significant principal components for the river sediments in Serbia
Full size table
Fig. 4
figure 4

PCA of selected elements (R mode)

Full size image

Figure 5 displays isolated three groups of samples: 20 (I1), 21 (I2), 23 (V2) and 25 (Z2)—first group, 33 (Pe)—second group, and the other sediment samples constitute the third group. With regard to PCA (Q mode), sampling site 33 (Pek river) has the most polluted river sediments at the investigated localities. The contribution of Cu to the overall pollution is the most significant compared to other elements in this sediment. In the group with samples 20, 21, 23 and 25, the sediment with significant pollution is compared to other moderately polluted studied sediments.

Fig. 5
figure 5

PCA of selected elements (Q mode). Explanation: 1–9 (Tisa), 10–15 (Danube), 16–19 (Sava), 20, 21 (Ibar), 22, 23 (Great Morava), 24, 25 (West Morava), 26 (Južna Morava), 27, 28 (Nišava), 29 (Tamiš), 30 (Topčiderska River), 31 (Porečka River), 32 (Kolubara), 33 (Pek), 34 (Toplica)

Full size image

HCA

Hierarchical cluster analysis (HCA) was performed on the dataset on both variables (R mode) and samples (Q mode), to identify clusters of elements and sites. The hierarchical methods are used because they are not limited to pre-determined numbers of clusters and can display the similarity of samples across a wide range of scales.

The results of the CA (R mode) are shown in a dendrogram (Fig. 6) where all the elements were grouped into four significant clusters: cluster 1 includes Fe, V, Co, Cr and Mn; cluster 2 contains Pb, Zn, Cd, and Ni, cluster 3 contains Cu, and cluster 4 contains As.

Fig. 6
figure 6

Dendrogram showing clustering of the analysed elements

Full size image

The results of the spatial cluster analysis (Q mode) are shown in a dendrogram (Fig. 7), where all 34 river sediment sampling sites were grouped into four significant clusters: Cluster 1 (T2–T9, V1, D1, D3–D6, Ta, JM, N1, and N2, CPI mean is 2.50, TU mean is 6.26), Cluster 2 (V2, Ko, Z2, S1–S4, Tr, To, Z1, Pr, T1, and D2, CPI mean is 2.75, TU mean is 8.52), Cluster 3 (I1 and I2, CPI is 4.94 and TU mean is 17.35) and Cluster 4 (Pe, CPI is 10.18 and TU mean is 13.06). The sampling sites in Clusters 1, 2 and 3 represented sediments from relatively to moderate polluted regions. Cluster 4 site (river Pek) was located in high-pollution region.

Fig. 7
figure 7

Dendrogram showing clustering of sampling sites (HCA, Q mode)

Full size image

The HCA results were similar to those obtained with PCA. Elements with similar origins are (1) Fe, V, Mn, Co and Cr; (2) Pb, Zn, Cd, As and Ni; and (3) Cu. It is observed that the sources of Cu are different from the other elements.

Fe, V, Mn, Co and Cr have a combined source and these elements are derived from both lithogenic and anthropogenic sources. Iron is abundant in the Earth’s crust, though the pollution indices suggest the existence of anthropogenic sources of this element at some localities. Also, enrichment is observed for Mn in some of sediments. Cr is generally a marker of the paint and metal industries, V is greatly impacted by anthropogenic activities such as mining and agricultural processes, and Co is mainly from anthropogenic signatures, i.e., paint, fertiliser or agrochemical industries (Li and Zhang, 2010). The low content of V, Co and Cr in the investigated sediments and the values of the calculated pollution indices, which indicate insignificant pollution, confirmed the mixed sources of Fe, V, Mn, Co and Cr at the studied locality.

Pb, Zn, Cd, As and Ni are likely to be mainly derived from anthropogenic sources. The significant anthropogenic source of these elements is mining and smelting complexes in Serbia. These results are in accordance with Bird et al. (2010) that throughout the River Danube drainage basin, the mining and processing of metal ores lead to point-source releases of contaminant metals (Cd, Cu, Pb, Zn) and As to the local surface drainage network. The Trepca mining complex is the dominant source of lead and zinc. In addition to mining, anthropogenic sources of these elements may be direct discharges from local point sources, such as industrial and urban discharges carrying metal and metalloid contaminants. Contaminants mainly come from metallurgy and coal burning. Fertiliser and pesticides containing As can also lead to an increase in As. Ni is also mainly derived from anthropogenic activities, especially industrial sources (Gao et al. 2013).

Cu is derived from anthropogenic sources, most significantly from metallic wastewater discharges. The significant anthropogenic source of Cu is the open-pit mine in Majdanpek. Another source of Cu may be discharges from industries, as well as agricultural activities.

The obtained results indicate that the most polluted sediments at the studied localities are sediments from the Ibar and the Pek Rivers. Additionally, calculated indices suggest that the Great Morava and West Morava are also significantly polluted river systems. Pollution of these rivers is mainly caused by permanent and accidental pollution from industrial plants and mines that are located in the basins of these rivers.

One of the main sources of industrial pollution of the River Ibar is likely to be the Trepca mining complex. This huge industrial mining site had been in operation and engaged in large-scale exploitation since 1925. Since then, and until the year 2000, lead and zinc have been mined and smelted and the tailings still threaten the environment and keep polluting the Ibar (Pedersen and Klitgaard 2004). Results of Rexhepi et al. (2010) are shown that high zinc, lead and copper content in the sediment of Ibar River suggests that the lead and zinc smelter in Mitrovica still discharges strongly polluted effluents into this stream and that the high metal content could also reflect the accumulation of fine, strongly polluted particles in the river bed for a fairly long time.

The significant pollution with Cu of sediments from the Pek River is probably due to discharge from the Cu mining complex in Majdanpek. Bird et al. (2010) have shown that metal mining in Serbia is focused on the Bor and Madjanpek Cu deposits. The copper ore treatment process produced large amounts of ore waste and flotation tailing heaps, located in the vicinity of the towns of Bor and Majdanpek. In addition to permanent pollution, accidental pollution significantly affects the pollution of the Pek. The most serious incidents listed are the breaches of tailing dams at Majdanpek and Veliki Majdan causing direct, serious contamination of the river Pek by heavy metals (Stuhlberger 2014).

In general, the results obtained using multivariate statistical analysis are consistent with that obtained using the calculation of pollution indices. Sediments in rivers in the same statistical grouping appear to indicate the same source of pollution and a similar degree of contamination. The applied methods can classify sediments in the whole investigated region according to the degree of pollution and to identify the main sources of toxic elements. The combined application of PCA and pollution indices can effectively identify the comprehensive and single pollution levels of elements in sediments, thus important to the extent determination of heavy metals pollution around the world.

Conclusion

The impact of trace element pollution on the quality of the river sediments in Serbia was evaluated using CF, EF, Igeo, Er, RI, PLI, CPI, mCd, and TU. The source identification carried out using multivariate analyses (PCA and CA) shows that (1) Fe, V, Mn, Co and Cr are derived from lithogenic and anthropogenic sources, (2) Pb, Zn, Cd, As and Ni are mainly derived from anthropogenic sources, and (3) Cu is derived from anthropogenic sources, most significantly from metallic wastewater discharges.

Contamination was observed in the studied elements, most of all with Zn, Cu and Cd. Cd and Cu present a higher ecological risk than any other element because of their higher toxicity coefficient. The potential Er varied amongst the studied elements in the sediments as well as in the different areas. The values of the PLI and mCd suggest that the degree of contamination ranges from low to very high. The highest values of the pollution indices suggest a significant contribution of elements from anthropogenic sources in the river sediments from the Ibar, Pek, West Morava and Great Morava. Significant pollution indexes are also observed in the Tisa and Sava rivers. Pollution of these rivers is mainly caused by permanent and accidental pollution from industrial plants and mines that are located in the basins of these rivers. Significant influences are evident from mining complexes (mainly in Trepca and Majdanpek) on river contamination with toxic trace elements.

The above results confirmed the existence of pollution with trace elements in the studied river sediments. The calculated pollution indices and the appliance of statistical methods are recognised as useful for improving ecological risk assessment and the management of trace elements in sediments. A future fractionation sediment study is recommended to determine the proportion of the total metal pool that might be bioavailable.