Spatial and Temporal Variability of Water Quality in the Bystrzyca River Basin, Poland

The aim of the study was to analyze the results of surface water quality tests carried out in the Bystrzyca river basin. The study was conducted over four years in four seasons. The following chemometric techniques were used for the purposes of statistical analyses: the principal component analysis with factor analysis (PCA/FA), the hierarchical cluster analysis (HCA), and the discriminant analysis (DA). The analyses allowed for determining the temporal variability in water quality between the seasons. The best water quality was recorded in summer and the worst in autumn. The analyses did not provide a clear assessment of the spatial variability of water quality in the river basin. Pollution from wastewater treatment plants and soil tillage had a similar effect on water quality. The tested samples were characterized by very high electrolytic conductivity, suspended solids and P-PO4 concentrations and the water quality did not meet the standards of good ecological status.


Introduction
In Poland, rivers are characterized by relatively low runoffs. Small water resources often overlap with eutrophication processes resulting from low dilution of pollutants. In the twentieth century, in Poland, often untreated municipal sewage and industrial wastewater was discharged directly into rivers. After Poland's accession to the European Union (EU), the implementation of environmental monitoring programs commenced [1,2].
Due to their location, rivers carry pollutants to lakes and the sea, contributing to the eutrophication of these water bodies. Water from rivers, lakes, and groundwater is used for human consumption needs. Water resources play an important role in drinking water supply, crop irrigation, industrial production, hydropower generation, and fish farming [3][4][5][6]. The increase in water demand caused by the development of civilization has contributed to the reduction in the amount of water and the deterioration of its quality. Water quality is affected by both natural processes and anthropogenic factors. Natural processes include, but are not limited to, topography, geological structure, seasonal temperature and rainfall changes, and land use. Many studies confirm that natural processes have a significant impact on water quality [7][8][9][10]. However, anthropogenic factors are much more important for the deterioration of water quality [11,12]. The anthropogenic factors include: industrial pollution, domestic sewage, agricultural drainage, as well as agriculture and urbanization intensity. Pollutants go to water from point sources (industry, municipal) and area sources, which are identified with agricultural land [13][14][15]. Both natural and anthropogenic factors influence the amount of nutrients eluted from the basin [16,17]. Excessive content in surface waters can lead to eutrophication, oxygen deficiency, as well as the development of organisms posing a danger to human health [18,19]. Studies on by five tributaries. The following tributaries have an outlet into the Bystrzyca River: Kosarzewka at 47 km, Kreżniczanka at 40 km, Czerniejówka at 26 km, Czechówka at 25 km, and Ciemięga at 12 km (from the outlet of the Bystrzyca river). From kilometer 18 to 38, the Bystrzyca River flows through the city of Lublin [51,52]. For this reason, the causes of eutrophication of surface water can be: intensive suburbanization, industrial activity, recreation, and agriculture. Suburbanization occurs in a ring-like system around the city of Lublin. Industry is mainly based in the eastern districts of the city. Recreational areas are concentrated in the southern part of the city, around the Zemborzyce reservoir. Agricultural areas are located downstream and upstream of the river. In 2014, five municipal sewage treatment plants and five industrial wastewater plants operated in the study area. In its catchment basin, there is the Zemborzyce reservoir with an area of 280 ha and three fishpond complexes with a total area of 70 ha. The basin includes agricultural land (78%): arable land constitutes 71%, grassland 4%, and orchards account for 3%. The land use structure of the basin is complemented by forestland (10%), urban areas (11%), and wasteland (1%). The soil cover consists mainly of Podzisols occurring in the top parts of the upland and on slopes, as well as Cambisols in the lowland parts [53][54][55]. In the years 2011-2014, the average annual air temperature was 9 • C, and the precipitation level was 570 mm. In the summer half-year, the precipitation was 370 mm, while in the winter half-year it was 200 mm. The average water runoff on the Sobianowice section was 4.1 m 3 s −1 , which was lower than the multi-year average.

Sample Collection
Surface water samples were collected for physical and chemical analyses over four years, 2011-2014, in four seasons (winter, spring, summer, autumn). A single 1 L sample was taken at each time at the depth of half of the water level in the river. The tests were carried out on five measuring sections along the Bystrzyca River and five on its tributaries (Table 1, Figure 1). Direct measurements with a Multi 340i multi-parameter meter (WTW, Weilheim, Germany) included: pH, dissolved oxygen (DO), and the electrolytic conductivity (EC) of water. Using the sampler, water samples were collected into PE bottles for laboratory testing. Physical and chemical analyses were carried out using a PC spectrophotometer (AQUALYTIC, Langen, Germany) and verified the following parameters: total phosphorus (P), phosphates (P-PO 4 ), total nitrogen (N), and ammonium nitrogen (N-NH 4 ), and by means of a LF300 photometer (SLANDI, Michałowice, Poland): sulphates (SO 4 ) chloride (Cl), nitrate nitrogen (N-NO 3 ), and Kjeldahl nitrogen (KN). Biochemical oxygen demand (BOD) was determined by the Winkler method, chemical oxygen demand (COD) by the bichromate method, suspended solids (SS) by the gravimetric method, and total organic carbon (TOC) using a TOC1200 analyzer (Trace Elemental Instruments, Delft, Netherlands).

Statistical Analysis
The assessment of the physical and chemical composition of the water in the Bystrzyca catchment basin was based on a set of data consisting of 15 water quality parameters. The study was carried out at ten measuring stations during the years 2011-2014 in four seasons (winter, spring, summer, autumn). Prior to the statistical analysis, the data was collated. Then, using the W test (Shapiro-Wilk), the compliance of the distribution of the physical and chemical parameters of water with normal distribution was checked. Environmental data was transformed and standardized to meet the normality assumption. In the case of chloride and sulphate concentrations, their distribution after transformation differed significantly from normal; therefore, these parameters were not included in chemometric analyses. In order to characterize the temporal and spatial variability of the river basin station Hydrometric measurements were also carried out at each measuring station, which included the measurement of water flow rate and cross-section parameters (river depth and width). This was done using a HEGA-1 hydrometric meter (Biomix, Poland) and a Leica Nova MS 50 total station (Leica, Switzerland).

Statistical Analysis
The assessment of the physical and chemical composition of the water in the Bystrzyca catchment basin was based on a set of data consisting of 15 water quality parameters. The study was carried out at ten measuring stations during the years 2011-2014 in four seasons (winter, spring, summer, autumn). Prior to the statistical analysis, the data was collated. Then, using the W test (Shapiro-Wilk), the compliance of the distribution of the physical and chemical parameters of water with normal distribution was checked. Environmental data was transformed and standardized to meet the normality assumption. In the case of chloride and sulphate concentrations, their distribution after transformation differed significantly from normal; therefore, these parameters were not included in chemometric analyses. In order to characterize the temporal and spatial variability of the remaining 13 water quality parameters, the multivariate analysis methods of classification and ordination were used. The hierarchical cluster analyses (HCA) were developed based on the monitoring stations' measurements using the Ward's minimum variance classification algorithm with Euclidean distance as a similarity measure. The principal component analysis with factor analysis (PCA/FA) was used to determine the relationships between water quality parameters at the measuring stations and individual test dates. Finally, the discriminant analysis (DA) was carried out, using the season as a discriminating variable. DA was applied to raw data, whereas CA and PCA were applied to standardized data to avoid misclassification arising from different parameter units.

Characteristics of Water Quality in the Bystrzyca River Basin
During the study period, the surface water in the Bystrzyca river basin showed an alkaline reaction ranging from 7.5 to 8.25. Nitrate nitrogen concentrations were low and ranged from 0.7 to 3.5 mg/L, ammonium nitrogen from 0.02 to 0.34 mg/L, and Kjeldahl nitrogen from 0.7 to 1.88 mg/L, while total nitrogen concentrations ranged from 1.5 to 5.2 mg/L. Low biochemical oxygen demand (BOD) was observed from 1.4 to 4.5 mg/L. Chemical oxygen demand (COD) ranged from 8.0 to 28.0 mg/L, and the concentrations of total organic carbon (TOC) ranged from 1.1 to 7.4 mg/L. Therefore, high DO concentrations above 7.6 mg/L could be observed. All the test samples were characterized by a very high content of suspended solids (SS), which ranged from 261 to 522 mg/L and the associated high electrolytic conductivity (EC) ranged from 393 to 802 µS/cm. In addition, high phosphorus concentrations ranging from 0.10 to 0.37 mg/L and very high phosphate concentrations (P-PO 4 ) ranging from 0.05 to 0.24 mg/L were found in the test samples. The statistical parameters of water quality indicators for the testing seasons are presented in Table 2.
Based on the concentration of nitrogen, ammonium nitrogen, pH, DO, COD, and TOC, the water quality corresponded to a very good ecological status (class I). Based on the concentrations of total phosphorus, Kjeldahl nitrogen, nitrate nitrogen, and BOD, the water quality corresponded to good ecological status (class II). The standards of good ecological status are met for these oxygen and nutrient indicators. Achieving this goal was associated with the implementation of a program to protect the aquatic environment against degradation in the EU [2,51]. The test samples were characterized by very high EC, SS, and P-PO 4 concentrations and the water quality did not meet the standards of good ecological status. Very high EC and SS concentrations are associated with the ionic composition of water and the runoff of soil and mineral salts from the slopes. Very high levels of orthophosphates are associated with the use of detergents and waste storage.
The analyzed surface waters had the best quality in the summer season, and the worst in the autumn season. The analysis of variance revealed fluctuations between the seasons in the study period in dissolved oxygen concentrations, biochemical oxygen demand, chemical oxygen demand, phosphates, electrolytic conductivity, and total phosphorus (statistically significant differences at the level α = 0.05). However, no differences in water quality parameters were found between the measuring stations.

Assessment of the Temporal and Spatial Variability
A hierarchical agglomerative cluster analysis was conducted to identify the temporal and spatial variability of water quality parameters at the monitored stations. Ward's algorithm of minimum variance was used for clustering and the Euclidean distance was applied as a measure of similarity. The obtained results of classification are presented graphically as dendrograms (Figure 2a,b). The measuring sections were grouped into two statistically significant clusters (D link /D max ) × 100 < 60. The first cluster consisted of the stations Pliszczyn, Krężnica, Zamek, and Hajdów (moderate pollution level). The stations Pliszczyn (10) and Krężnica (7) are located in eroded agricultural land, while the stations Zamek (9) and Hajdów (4) are located in an urban area. It follows that both intensive agriculture and urbanization contribute to water pollution. The second cluster includes the stations Kiełczewice (1), Osmolice (2), Zemborzyce (3), Spiczyn (5), Iżyce (6), and Głusk (8), with low water pollution (low pollution level). The investigated basin has a high capacity to retain dissolved chemical compounds. Therefore, the water quality parameters on the outlet stretch Spiczyn (5) have low values. In addition, despite the discharge of sewage from an industrial treatment plant, the Osmolice station was classified in the second group (low pollution level). It is similar in the case of the Hajdów station (moderate pollution level). This is where domestic sewage from the sewage treatment plant for the Lublin urban area (380,000 residents) is discharged. The cluster analysis detected no point source pollution in the form of industrial and domestic sewage discharges. This results from very good management of water resources and the use of the best available technology for wastewater treatment. In addition, point pollution is superimposed on area pollution from surface runoff. Stations located both in agricultural and urbanized areas were classified into one cluster. This indicates a comparable level of area and point pollution.
The CA was repeated taking into account both the stations and the date of sampling (Figure 2b). Four clusters were obtained and their analysis indicates that the date of sampling may be a factor determining that they should be included in a particular group. In particular, it can be seen that the samples taken in the summer season formed the first cluster (marked in pink). The second cluster comprised four samples from the autumn season (marked in olive green), while the third one comprised three samples from the winter season (marked in blue). The fourth cluster (marked in steel blue) includes the remaining samples from three different seasons. Samples taken during the summer and partly in the spring season show the lowest levels of pollution. This may be a result of highly diluted impurities and the uptake of nutrients by plants.

Assessment of the Temporal and Spatial Variability
A hierarchical agglomerative cluster analysis was conducted to identify the temporal and spatial variability of water quality parameters at the monitored stations. Ward's algorithm of minimum variance was used for clustering and the Euclidean distance was applied as a measure of similarity. The obtained results of classification are presented graphically as dendrograms (Figure 2a,b). The measuring sections were grouped into two statistically significant clusters (Dlink/Dmax) × 100 < 60. The first cluster consisted of the stations Pliszczyn, Krężnica, Zamek, and Hajdów (moderate pollution level). The stations Pliszczyn (10) and Krężnica (7) are located in eroded agricultural land, while the stations Zamek (9) and Hajdów (4) are located in an urban area. It follows that both intensive agriculture and urbanization contribute to water pollution. The second cluster includes the stations Kiełczewice (1), Osmolice (2), Zemborzyce (3), Spiczyn (5), Iżyce (6), and Głusk (8), with low water pollution (low pollution level). The investigated basin has a high capacity to retain dissolved chemical compounds. Therefore, the water quality parameters on the outlet stretch Spiczyn (5) have low values. In addition, despite the discharge of sewage from an industrial treatment plant, the Osmolice station was classified in the second group (low pollution level). It is similar in the case of the Hajdów station (moderate pollution level). This is where domestic sewage from the sewage treatment plant for the Lublin urban area (380,000 residents) is discharged. The cluster analysis detected no point source pollution in the form of industrial and domestic sewage discharges. This results from very good management of water resources and the use of the best available technology for wastewater treatment. In addition, point pollution is superimposed on area pollution from surface runoff. Stations located both in agricultural and urbanized areas were classified into one cluster. This indicates a comparable level of area and point pollution.
The CA was repeated taking into account both the stations and the date of sampling (Figure 2b). Four clusters were obtained and their analysis indicates that the date of sampling may be a factor determining that they should be included in a particular group. In particular, it can be seen that the samples taken in the summer season formed the first cluster (marked in pink). The second cluster comprised four samples from the autumn season (marked in olive green), while the third one comprised three samples from the winter season (marked in blue). The fourth cluster (marked in steel blue) includes the remaining samples from three different seasons. Samples taken during the summer and partly in the spring season show the lowest levels of pollution. This may be a result of highly diluted impurities and the uptake of nutrients by plants. The next stage of the statistical analysis was the principal component analysis (PCA). Significant PCA axes were selected with the Kaiser-Gutman criterion [59]. The first three components (PC1, PC2, PC3) had eigenvalues greater than 1 (Table 3), which allowed 74.14% of the total variation to be explained. Assuming the assessment of the relationship between factor loadings for water quality parameters and individual components according to Reference [9], the following conclusions can be formulated. The factor analysis reduced the set of 13 parameters initially used to characterize water quality to three VF (Variations Factor) variations necessary for the identification of the river pollution sources. If the factor loadings between water quality parameters and VF coefficients are 0.75-1.00, the values are strongly correlated, while at 0.50-0.75, they are moderately correlated. The first factor, VF1 (corresponds to 68% of total variance), was strongly correlated with NK, N-NH4, EC, SS, pH, and P concentrations. The next two factors with a cumulative variance of 28% were, on average, correlated with the concentration of N-NO3 and P-PO4 (VF2), with organic carbon TOC (VF3). A negative factor correlation with the concentrations of nitrogen and phosphorus compounds suggests an impact of organic pollutants. These pollutants can be associated with intensive land use (fertilizers and pesticides) as well as industrial production (waste and sludge). Nitrogen and phosphorus compounds contribute to the eutrophication of water and the deterioration of the quality of aquatic ecosystems [60][61][62]. The negative correlation of the VF1 factor with pH level indicates that when the parameter has low values, carbon and calcium can be released from carbonate rocks. High concentrations of calcium carbonate occur in arable fields due to intensive soil erosion. A graphical representation of the PCA analysis for the first two components is shown in the graphs in Figure 3.
The principal component analysis did not provide an unambiguous assessment of the spatial variability of water composition in the catchment basin and its quality. Based on the PCA analysis, insignificant differences were identified in the physical and chemical parameters of the water analyzed on the sections Pliszczyn and Hajdów (Figure 3a), which is clearly influenced by both point The next stage of the statistical analysis was the principal component analysis (PCA). Significant PCA axes were selected with the Kaiser-Gutman criterion [59]. The first three components (PC1, PC2, PC3) had eigenvalues greater than 1 (Table 3), which allowed 74.14% of the total variation to be explained. Assuming the assessment of the relationship between factor loadings for water quality parameters and individual components according to Reference [9], the following conclusions can be formulated. The factor analysis reduced the set of 13 parameters initially used to characterize water quality to three VF (Variations Factor) variations necessary for the identification of the river pollution sources. If the factor loadings between water quality parameters and VF coefficients are 0.75-1.00, the values are strongly correlated, while at 0.50-0.75, they are moderately correlated. The first factor, VF1 (corresponds to 68% of total variance), was strongly correlated with NK, N-NH 4 , EC, SS, pH, and P concentrations. The next two factors with a cumulative variance of 28% were, on average, correlated with the concentration of N-NO 3 and P-PO 4 (VF2), with organic carbon TOC (VF3). A negative factor correlation with the concentrations of nitrogen and phosphorus compounds suggests an impact of organic pollutants. These pollutants can be associated with intensive land use (fertilizers and pesticides) as well as industrial production (waste and sludge). Nitrogen and phosphorus compounds contribute to the eutrophication of water and the deterioration of the quality of aquatic ecosystems [60][61][62]. The negative correlation of the VF1 factor with pH level indicates that when the parameter has low values, carbon and calcium can be released from carbonate rocks. High concentrations of calcium carbonate occur in arable fields due to intensive soil erosion. A graphical representation of the PCA analysis for the first two components is shown in the graphs in Figure 3. sources of pollution (municipal sewage treatment plant) and area sources (soil erosion). No significant differentiation was found between the remaining points (the stations are mixed together).  However, the PCA analysis allowed significant differences in water quality to be determined between the testing seasons. With datasets for different periods, the PCA can also be used to investigate the temporal variations in water quality and find out the most important pollution sources for each period. By considering the deadline for sampling, the PCA can also be used to investigate the temporal variations in the water quality (Figure 3b). It can be seen that the first principal component is strongly correlated with the seasons. In particular, we can observe the clusters of samples in the summer and autumn seasons. The samples in the autumn season are characterized by a higher than average concentration of water quality parameters. Samples in the summer season are characterized by a lower than average concentration of the tested parameters. The lowest values of water pollution in the summer may be due to heavy rainfall and nutrient uptake by plants. In turn, in the autumn, rainfall is low, the vegetation period comes to an end, and the source of pollution is plant residues.
At the final stage of calculations, the discriminant analysis (DA) was performed on the data using a standard stepwise method. This made it possible to build a model containing 13 water quality parameters that were used to characterize the temporal variability of the physical and chemical composition of water in the basin ( Table 3). The discriminant analysis allows for building orthogonal functions with a cumulative variance of 97% in the event of temporal variability. The DA was performed on raw data after splitting the dataset into four groups (spring, summer, autumn, and winter) based on the results of the CA and PCA. Since the grouping variable had four categories, three discriminant functions (DF) were obtained (Table 4). Only two of them were statistically significant (p < 0.01). The DA identified six variables (DO, SS, pH, N-K, N, and P-PO4) as the most important discriminating variables ( Table 5). The first discriminant function is weighted most heavily by the SS and N ( Table 6). The second function seems to be marked mostly by P-PO4 and EC. The graphical presentation of DA results is shown in Figure 4. It can be stated that we obtain discrimination between the summer and autumn seasons by means of the first discriminant function. However, the second discriminant function seems to distinguish between winter and autumn seasons. The principal component analysis did not provide an unambiguous assessment of the spatial variability of water composition in the catchment basin and its quality. Based on the PCA analysis, insignificant differences were identified in the physical and chemical parameters of the water analyzed on the sections Pliszczyn and Hajdów (Figure 3a), which is clearly influenced by both point sources of pollution (municipal sewage treatment plant) and area sources (soil erosion). No significant differentiation was found between the remaining points (the stations are mixed together).
However, the PCA analysis allowed significant differences in water quality to be determined between the testing seasons. With datasets for different periods, the PCA can also be used to investigate the temporal variations in water quality and find out the most important pollution sources for each period. By considering the deadline for sampling, the PCA can also be used to investigate the temporal variations in the water quality (Figure 3b). It can be seen that the first principal component is strongly correlated with the seasons. In particular, we can observe the clusters of samples in the summer and autumn seasons. The samples in the autumn season are characterized by a higher than average concentration of water quality parameters. Samples in the summer season are characterized by a lower than average concentration of the tested parameters. The lowest values of water pollution in the summer may be due to heavy rainfall and nutrient uptake by plants. In turn, in the autumn, rainfall is low, the vegetation period comes to an end, and the source of pollution is plant residues.
At the final stage of calculations, the discriminant analysis (DA) was performed on the data using a standard stepwise method. This made it possible to build a model containing 13 water quality parameters that were used to characterize the temporal variability of the physical and chemical composition of water in the basin ( Table 3). The discriminant analysis allows for building orthogonal functions with a cumulative variance of 97% in the event of temporal variability. The DA was performed on raw data after splitting the dataset into four groups (spring, summer, autumn, and winter) based on the results of the CA and PCA. Since the grouping variable had four categories, three discriminant functions (DF) were obtained (Table 4). Only two of them were statistically significant (p < 0.01).
The DA identified six variables (DO, SS, pH, N-K, N, and P-PO 4 ) as the most important discriminating variables ( Table 5). The first discriminant function is weighted most heavily by the SS and N ( Table 6). The second function seems to be marked mostly by P-PO 4 and EC. The graphical presentation of DA results is shown in Figure 4. It can be stated that we obtain discrimination between the summer and autumn seasons by means of the first discriminant function. However, the second discriminant function seems to distinguish between winter and autumn seasons.

Discussion
The HCA analysis allowed for the identification of two clusters represented as low pollution and moderate pollution levels. Cluster 1 was formed by sites 4, 7, 9, and 10, while the other six sites were assigned to cluster 2. Stations 7 and 10 were located on the tributaries of the river (Figure 1) in areas with no industrial activity and with dispersed single-family housing. Various factors influence the water quality in these stations: agricultural practices, animal husbandry, and domestic sewage [10,20]. Stations 4 and 9 were located in the city of Lublin in the area of intensive road infrastructure and clustered housing development. In these stations, increased levels of pollution can result from surface runoff and domestic sewage discharge. Increased levels of river pollution in these locations are due to seasonal human activity, which can lead to eutrophication. In stations located in the Bystrzyca river basin, increases are observed in pollution concentrations; however, they are not constant. Cyclical changes in rainfall intensity and human agricultural activity contribute to the formation of clear differences in water quality depending on the season of the year [63][64][65].
The low level of pollution in spring is probably due to dilution of melting snow by runoff. The relatively low concentration values of the tested parameters in spring are mainly caused by dilution. In turn, the very low level of pollution in summer is due to heavy rainfall and the uptake of nutrients by plants. The use of large amounts of fertilizers in agricultural areas and the use of salt for de-icing of roads in urban areas may contribute to increased concentrations of pollutants. Strong seasonal variation in pH, DO, BOD, and COD may result from photosynthesis, hydrological pollution, and natural chemical processes [22,23,[66][67][68]. High TOC concentrations should be considered a result of the dissolution of minerals containing calcium carbonate, which occurs in loess soils [51,55]. The high SS content causes a reduction in DO, which affects the quality of water in the river. The high level of water quality parameters points to causes of pollution, such as withered and decaying plants, mineral acids, or agricultural and industrial waste discharged into the river [69,70]. For this reason, the highest values of pollution parameters were observed in autumn. Other studies show that the highest pollution values occurred in spring as a result of surface runoff. This indicates that, as a result of soil erosion processes, pollutants are accumulated in rivers [71]. Intensive erosion in highlands occurs during tillage and heavy rainfall. The frequency of pollution and soil erosion intensity depend on the land use and the type and manner of performing agricultural works [72,73].

Discussion
The HCA analysis allowed for the identification of two clusters represented as low pollution and moderate pollution levels. Cluster 1 was formed by sites 4, 7, 9, and 10, while the other six sites were assigned to cluster 2. Stations 7 and 10 were located on the tributaries of the river (Figure 1) in areas with no industrial activity and with dispersed single-family housing. Various factors influence the water quality in these stations: agricultural practices, animal husbandry, and domestic sewage [10,20]. Stations 4 and 9 were located in the city of Lublin in the area of intensive road infrastructure and clustered housing development. In these stations, increased levels of pollution can result from surface runoff and domestic sewage discharge. Increased levels of river pollution in these locations are due to seasonal human activity, which can lead to eutrophication. In stations located in the Bystrzyca river basin, increases are observed in pollution concentrations; however, they are not constant. Cyclical changes in rainfall intensity and human agricultural activity contribute to the formation of clear differences in water quality depending on the season of the year [63][64][65].
The low level of pollution in spring is probably due to dilution of melting snow by runoff. The relatively low concentration values of the tested parameters in spring are mainly caused by dilution. In turn, the very low level of pollution in summer is due to heavy rainfall and the uptake of nutrients by plants. The use of large amounts of fertilizers in agricultural areas and the use of salt for de-icing of roads in urban areas may contribute to increased concentrations of pollutants. Strong seasonal variation in pH, DO, BOD, and COD may result from photosynthesis, hydrological pollution, and natural chemical processes [22,23,[66][67][68]. High TOC concentrations should be considered a result of the dissolution of minerals containing calcium carbonate, which occurs in loess soils [51,55]. The high SS content causes a reduction in DO, which affects the quality of water in the river. The high level of water quality parameters points to causes of pollution, such as withered and decaying plants, mineral acids, or agricultural and industrial waste discharged into the river [69,70]. For this reason, the highest values of pollution parameters were observed in autumn. Other studies show that the highest pollution values occurred in spring as a result of surface runoff. This indicates that, as a result of soil erosion processes, pollutants are accumulated in rivers [71]. Intensive erosion in highlands occurs during Water 2020, 12,190 13 of 17 tillage and heavy rainfall. The frequency of pollution and soil erosion intensity depend on the land use and the type and manner of performing agricultural works [72,73].
The sources of pollution in a river basin can be identified on the basis of factor analysis [28,69,71]. The PCA used in the work did not allow for determining the spatial variability of the tested water quality parameters. No differences were found between stations located at sewage discharge points and in other places. This is due to the fact that point pollution is often superimposed on area pollution. The HCA and PCA/FA analysis showed that both sources of pollution are at a comparably high level. This suggests that potentially harmful substances may be of natural or anthropogenic origin, or both [74]. In addition, some man-made compounds also occur in natural conditions, e.g., gypsum [20,64,66]. In the studied Bystrzyca river catchment, water is only moderately polluted by agricultural practices or municipal sewage. The lowest quality water was found at station 4. This was due to surface runoff from the city and discharge from the municipal sewage treatment plant. At station 5, however, a significant improvement in water quality was observed, which reveals the river's self-cleaning ability [69]. Previous studies show that heavy metal concentrations were characterized by very low values [57].

Conclusions
Chemometric techniques are a useful tool to describe water quality and its spatial and temporal variability caused by natural and anthropogenic factors. The cluster analysis allowed for the identification of two clusters. One included four stations with a moderate pollution level, while the other comprised six stations with a low level of water pollution. The principal component analysis did not provide an unambiguous assessment of the spatial variability of water quality in the basin and the identification of hot spots. This is due to the simultaneous occurrence of area and point source pollution in the analyzed basin with a similar effect on water quality. Analyses carried out using PCA and DA showed statistically significant temporal variability between the study seasons. In the Bystrzyca river basin, we get differences between the summer and autumn seasons using the first discriminant function. With the second discriminant function, we get differences between winter and autumn seasons. The differences between the seasons are due to human activity, the type of agricultural treatment, and agricultural runoff. The main task of managers is to limit the frequency and intensity of erosion. This can be achieved by introducing tree cover and good agricultural practices, and by limiting surface sealing. This study can help the water resources management in the region.
The analyzed surface waters had the best quality in the summer season, and the worst in the autumn season. The samples were characterized by very high EC, SS, and P-PO 4 concentrations and the water quality did not meet the standards of good ecological status. Excess ortophosphate (P-PO 4 ) from feed, fertilizers, and industrial waste, is an underlying factor contributing to the deterioration of water ecosystems. In turn, electrical conductivity (EC) and suspended solids (SS) as water salinity indicators reflect the impact of urbanization and soil erosion.