The Impact of Various Types of Cultivation on Stream Water Quality in Central Poland

Abstract

Agricultural practices have a significant impact on stream water quality in rural landscapes; however, there is still little empirical evidence of how different types of cultivation alter the hydrochemistry of running water. Thus, the current study explored the spatial dynamics of selected ion concentrations and their land cover dependence in lowland agricultural catchments. From November 2021 to October 2022, water samples were collected from 30 sites located across small tributaries of the rivers Bzura, Pilica, and Radomka for chemical analysis of their NO₃, NO₂, NH₄, Ca, Mg, K, Na, As, Ba, Sr, and V concentrations. The results indicated a clear spatial heterogeneity of water quality, related to lithology and dominant land cover evaluated with the CORINE Land Cover 2018 dataset. Overall, sites representing agricultural land promoted increased concentrations of major and trace elements, while those with pepper cultivation were additionally contaminated with NO₃ and NO₂. The correlation performance for nitrogen compounds was the highest for narrower buffer zones, which was not documented for major and trace elements, which were linked more strongly with land cover at larger scales. Such new insights into the water quality dynamics of lowland agricultural catchments, being a simultaneous reflection of lithology, agricultural practices, and several municipal impacts, have significant implications for appropriate water management in rural landscapes.

1. Introduction

Water pollution due to various human impacts is considered a serious challenge faced by state governments in recent years [1,2]. It was documented that one of the most significant sources of such degradation is area-based water contamination from agricultural practices, such as plant cultivation, livestock production, and rural wastewater inflows [3,4,5]. In terms of agricultural practices, nitrogen pollution was documented to be a crucial problem [6], which, among others, leads to excessive eutrophication of lakes and seas [7,8]. Such a process, rapidly increasing after 1950 [9], is observed in the case of the Baltic Sea [10]. It was documented that more than 1/3 of the total nitrogen input transported from Polish watercourses to the Baltic Sea comes from agriculture, while slightly more than half of the nitrogen transport to the Baltic Sea is moved in the form of nitrates [7]. In consequence, excessive concentrations of phytoplankton have been found, along with oxygen deficiency in the bottom zone of the sea [11,12], causing negative effects and exerting spatially heterogeneous effects on the sea ecosystem [11].

In order to reduce agricultural nitrate pollution, the Nitrates Directive [13] was enacted to oblige EU Member States to designate vulnerable areas and to regularly monitor surface water as well as groundwater in these areas. The effects of its introduction have been generally positive, although largely dependent on the actions of member states. For example, in the Netherlands, following the implementation of the Nitrates Directive, nitrates in shallow groundwater were reduced by more than 50% [14], while for rivers located in Denmark the value was approximately 30% [15]. Unfortunately, studies conducted in other European catchments have not shown a similar decreasing trend; in some regions, such as in Northern Italy, the trend was the opposite [16], so future works are strongly recommended.

Eutrophication of freshwater ecosystems is also driven by other biogenic compounds, such as nitrites, phosphates, and ammonium ions [17]. Such contaminants, apart from inorganic fertilizers and livestock manure, could originate from infiltration from septic tanks [18,19]. According to [20], there were more than 1.8 million such installations in Poland in 2017, which could be related to the still-inappropriate level of sewage-system users, including only 42% of people in rural areas. Agricultural practices also result in additional potassium, calcium, and magnesium release from fertilizers, especially in the presence of non-balanced fertilization, in terms of inadequate component ratios [21] as well as the application of fertilizers at inappropriate times, particularly at the end of autumn or winter [22]. Organic fertilizers in the form of animal manures and biosolids may contain potentially toxic elements, such as arsenic, chromium, lead, mercury, nickel, and vanadium [23,24]. Such contaminants are also broadly found in pesticides [25], used for controlling weeds, bacteria, fungi, and insect infestations [26].

There is a broad spectrum of studies concerning the impact of land use/land cover on stream water quality, which use various analytical, statistical, and methodical approaches [27,28]. In recent years, the application of remote-sensing techniques and datasets can be very useful in such investigations, as some of the parameters can be measured directly from the air with the use of satellites, planes, and unmanned aerial vehicles (UAVs) [29]. Their undoubted advantage is fast, spatially extensive, and often cost-effective water quality assessment; for example, chlorophyll-a concentrations, turbidity, and trace metals can be assessed quite well with the use of hyperspectral data [30,31]. However, the most broad applications of remote-sensing techniques can be found in terms of land cover classifications, which are used for simple comparison of water quality within the catchment’s land use area [32], calculation of land use type contributions in buffer zones [33], and landscape metric computing [34,35]. These investigations have used datasets from authorial classifications of land cover [36], as well as broadly available and free datasets, such as CORINE Land Cover and Sentinel-2 Global Land Cover [37]. Such studies have focused mainly on comparing catchments with different land use types, e.g., forested, agricultural, and urbanized [38,39,40]. In turn, there is little evidence from investigations where different types of cultivations in the catchment area were linked with water quality parameters, which could be identified as an important knowledge gap.

To yield some new findings, this study focused on the hydrochemical characterization of lowland agricultural catchments in central Poland. The specific objectives were to explore the spatial dynamics of selected ion concentrations (a); assess the effects of the type of cultivation on their concentrations (b); and examine the impacts of different spatial scales on the relationships between ion concentrations and land cover (c).

2. Study Area

Investigations were carried out in catchments distributed across the Southern Masovia Hills [41], which could be considered as a lowland landscape with altitudes of approximately 150–200 m a.s.l. Superficial deposits in the study area are dominated by fluvioglacial sands, gravels, and glacial clays from the Riss (Odra and Warta) glaciation, which form ground and terminal moraines, eskers, and kames. In the western parts of the study area, there are locally exposed Jurassic rocks, such as sandstones and chalcedonites [42]. According to the Köppen–Geiger classification, the southern Masovia Hills have a warm-summer humid continental climate (Cfb), characterized by dry winters and humid and warm summers [43]. In consequence, the hydrological regime of streams can be considered nival, with high flows during early spring in March and April and low flows during summer and early autumn in July, August, and September [44]. According to state environmental monitoring data from 2016–2021 [45], the status of the majority of surface water bodies in the study area was qualified as bad, mainly due to excessive polycyclic aromatic hydrocarbon (benzo(a)pyrene, benzo(ghi)perylene) concentrations and some biological elements (mainly benthic invertebrates and fishes).

The land cover of the area in question is generally heterogeneous and specific. In the northeastern part of the region near Grójec, Mogielnica, and Biała Rawska, the dominance of orchards is evident, with mainly apple, sweet cherry, cherry, and pear tree cultivation [46,47]. In turn, the southeastern part around Przytyk and Potworów displays pepper cultivars under foil tunnels [46]. The western part is characterized by the highest contribution of forested area (the Spała Forests), with a domination of Scots pine (Pinus sylvestris L.), common oak (Quercus robur L.), and—especially in river valleys—black alder (Alnus glutinosa (L.) Gaertn.). Rural settlements, such as towns and villages, are relatively small and usually do not exceed 500–1000 inhabitants.

3. Materials and Methods

3.1. Field and Laboratory Investigations

Water quality measurements were conducted in 30 small streams—tributaries of the rivers Bzura (“B” sites), Pilica (“P” sites), and Radomka (“R” sites) (Figure 1 and Figure 2). Site selection used a purposive sampling scheme, based primarily on the land use of the upstream catchment area and its maximum variability in terms of different cultivation. In addition, site localization precluded direct anthropogenic modifications, such as point wastewater discharges, dams, and runoffs from motorways, which are documented to directly impact water quality [48,49,50,51]. It is worth noting that all sampling sites were spatially independent, while their upstream catchment areas ranged from 1.5 to 61.0 km² (Appendix A).

Measurements were carried out once a month during the hydrological year 2022, from November 2021 to October 2022. Electrical conductivity [µS/cm] was monitored in situ with the use of a portable Hanna Hi 991300 m, with an accuracy of ±2% µS/cm. Water samples were simultaneously collected in 0.35 L polyethylene bottles and immediately transported at 4 °C to the laboratory to conduct the measurements. After the filtration of samples with the use of membrane filters, biogenic compounds (NO₃, NO₂, NH₄, PO₄) were determined with the use of an LF 300 photometer and dedicated Testoval reagents (adopting the sulfanilic acid, indophenol blue, and molybdenum blue methods). Quality control of the photometric measurements was conducted using Hach standard buffers; for approximately 10% of random samples, additional control measurements were conducted. The deviations did not exceed 5% in any control samples. Major elements and trace elements (As, Ba, Ca, Cd, Co, Cr, Cu, Fe, K, Mg, Mn, Na, Ni, Pb, Sr, V, Zn) were determined with a ICP-OES PerkinElmer Optima 5300 DV; during the measurements, appropriate manufacturer standards were used for controlling the relationships between emission intensity and ion concentrations. Due to concentrations being below the detection limit (0.001 mg/L) in the case of most of the trace elements (in more than 50% of samples), only NO₃, NO₂, NH₄, Ca, Mg, K, Na, As, Ba, Sr, and V ions were analyzed in terms of their spatial heterogeneity. In fact, this set of water quality parameters could reflect the effects of lithology (the chemical composition of sedimentary rocks and soils) and varied types of agricultural practices [52,53]. Some of them, such as NH₄, Na, and K, can also act as indicators of sewage pollution [54,55].

3.2. Land Cover Evaluation

Land cover in the upstream catchment area of the monitoring sites was assessed on the basis of the CORINE Land Cover 2018 (CLC 2018) product. This vector dataset of 100 m spatial resolution is based predominantly on the visual interpretation of Landsat satellite imagery [56] and is considered as a reference source, as it covers European spatial and temporal variability. It is used in a lot of studies, such as of forestry changes and urban activities [57,58].

For the purposes of the study, selected land cover classes—artificial surfaces (sum of the classes 1.1, 1.2, 1.3, and 1.4), arable lands (Class 2.1.1), fruit trees and berry plantations (henceforth called “orchards”, Class 2.2.2), and forests (all subclasses of Class 3)—were distinguished with the use of vector-processing tools. Peatbogs, marshes, meadows, and water bodies were omitted from the analysis because of their sporadic occurrence in the catchment areas. Although both CLC 2018 classifications distinguish the various types of forests (coniferous, broadleaf, and mixed), they were joined together as one class “forests”. The percentages of land use in the investigated catchments were computed in ArcMap 10.7.1.

3.3. Statistical Analysis

Initially, all sampling sites were objectively divided according to the dominant type of land cover in the upstream catchment area from the CLC 2018 dataset, similar to [59]. The percentage contributions of artificial surfaces (sum of the classes 1.1, 1.2, 1.3, and 1.4), arable lands (Class 2.1.1), orchards (Class 2.2.2), pastures (Class 2.3.1), and forests (all subclasses of Class 3) were a basis for such a separation using hierarchical cluster analysis. For this purpose, the Ward agglomeration method was applied, while the Euclidean distance was set as a measure of the similarity between the land cover of catchments, represented by sampling sites. The optimal number of clusters was separated with the use of the q_j index [60]; however, an additional separation of one cluster was made, as the aggregated clusters should be subjected to a logical interpretation based on the author’s intuition and good knowledge of the research objects [61]. Then, the spatial variability of the measured water contaminants was compared across all sampling sites and divided into groups distinguished using the Ward agglomeration method. In this way, the distribution of EC values, as well as those of NO₃, NO₂, NH₄, Ca, Mg, K, Na, As, Ba, Sr, and V concentrations, was documented on standard box-and-whisker charts, presenting mean, median, interquartile range, and outlier values, similar to [59]. To check if there were statistically significant differences (p < 0.05) between the concentrations of measured water quality parameters in the distinguished groups, the Kruskal–Wallis non-parametric test was selected due to it working with non-normally distributed data. Additionally, the post-hoc Dunn test was used for multiple comparisons for such groups. To provide the spatial context of the water contaminants heterogeneity, annual mean concentrations of contaminants calculated from measurement values are presented graphically on the maps using circles with sequential color and scaling.

To evaluate the relationships between ion concentrations, principal component analysis (PCA) was applied, a method broadly used in recent water quality studies [62,63]. The parameters selected for PCA included EC and mean concentrations of NO₃, NO₂, NH₄, Ca, Mg, K, Na, As, Ba, Sr, and V ions. Initially, variables were tested to see if they met the assumptions of factor analysis. The normality of the distributions of all of the variables was checked with the use of the Shapiro–Wilk test, and the results indicated that some of them did not have a normal distribution (p < 0.05). They were transformed using a logarithmic function (Log₁₀) after [64], and their distribution became satisfactory (p > 0.05). In addition, Bartlett’s test of sphericity was significant (p < 0.001), which indicates that variables are related to each other and can be reduced to a smaller number of components, while the Kaiser–Meyer–Olkin test result (0.738) suggested that the sampling was adequate to a middling level [65]. After final standardization with the use of the common z-score formula, PCA was performed with a correlation matrix on the basis of a 12-variable dataset. The Kaiser empirical criterion was chosen to select significant components, so eigenvalues under 1.0 were removed from further interpretation. Component loadings equal to or greater than 0.50 were considered significant correlations, similar to in [66].

Finally, correlation analysis was applied to evaluate the performance of land cover metrics, calculated for several spatial scales, in explaining ion concentration heterogeneity. The Spearman rank correlation coefficient was used for this purpose, which is favorable in the case of a small sample size and the lack of a normal distribution of data, instead of the Pearson correlation coefficient. Correlation analysis was performed to link mean EC values, as well as mean concentrations of NO₃, NO₂, NH₄, Ca, Mg, K, Na, As, Ba, Sr, and V ions and land cover metrics, calculated for the total catchment area and buffer zones of 50, 100, 250, 500, and 1000 m widths, similar to in [53] and [34]. A statistically significant threshold for the correlation was assumed at p = 0.05, while the analysis (together with statistical testing and PCA) was performed using Statistica 13.5 software and is presented in the form of a heat map.

Meteorological background during the investigated period was provided using mean monthly air-temperature values and monthly precipitation sums from the Dąbrówka Stara meteorological station, operated by the Institute of Meteorology and Water Management, National Research Institute (IMWM-NRI). Such a station, although located in the north-eastern area, is the only one across the investigated area where systematic air temperature measurements are conducted. In the case of precipitation, the available reference period covers the years 2003–2021, while for air temperature the reference period is 2007–2021. Hydrological data, used for streamflow variability evaluation, were acquired from IMWM-NRI gauging stations on the rivers Pilica (station Nowe Miasto), Radomka (Rogożek), Drzewiczka (Odrzywół), and Rawka (Kęszyce).

4. Results

4.1. Hydrometeorological Background

The investigated period, covering the hydrological year 2022, was slightly warmer and drier than the reference period 2007–2021 (Figure 3). Mean air temperature on the Dąbrówka Stara meteorological station reached 9.0 °C, which was 0.2 °C higher than the mean value for the multi-year period 2007–2021. Air temperature followed a typical seasonal pattern, with the maximum monthly mean value occurring in August (20.2 °C) and the minimum in December (−1.6 °C). The sum of precipitation reached 596.1 mm, which accounted for 95.7% of the average sum of precipitation calculated for the reference period (622.7 mm). The sums of precipitation varied significantly in certain months; the highest precipitation sum was noted in July at 115 mm, while the lowest was documented in March (1.1 mm). However, the winter and summer half-years were characterized by similar precipitation sums in comparison with the multi-year period, representing 95.1 and 96.1% of the average sum from 2003–2021, respectively. Streamflow patterns observed across hydrological stations located on the Pilica, Radomka, Drzewiczka, and Rawka rivers were similar and reflected meteorological conditions (Figure 3). High streamflow peaks were observed in winter and spring as a result of snowmelt, while, from June to September, low flows appeared due to intensive evapotranspiration. Thus, water quality sampling took place in different hydrometeorological conditions, which makes the obtained results representative.

4.2. Land Cover Classification

Different contributions of CLC 2018 classes were the basis for site separation in terms of the most similar land cover structures in their upstream catchment areas. The cluster analysis method made it possible to distinguish three main groups of sites (Figure 4). Those with evident domination of orchards (more than 65% contribution to the total area) were linked into the first cluster (O). All sites with more than 50% forests in their upstream catchment areas were grouped into the second cluster (F); the only exception was P17 (the Kiełcznica River), where forested areas accounted for 42.9% while 24.1% of the upstream catchment area was occupied by pastures. Thirdly, the most extensive cluster was distinguished mainly by sites with more than 50% arable land in total in the upstream catchment area. The B1 and B4 sites did not meet these criteria, as the slightly lower contributions of arable lands (35.3% and 49.0%, respectively) were compensated by greater participation of orchards (38.5% and 19.7%, respectively). Nevertheless, due to the specific cultivations in some catchments (Figure 5) (foil tunnels with pepper) expected to influence water quality, arable sites were divided into two smaller groups—arable land (catchments B1, B4–B7, P1, P8, P11, and P12) and pepper sites (P18, P19, and R1–R4).

4.3. Spatial Variability of the Water Quality Parameters

Clear spatial variability in the selected water contaminants was observed in the investigated lowland catchments during the hydrological year 2022, as indicated by statistical analysis of the measurement data. The distribution of the concentrations, divided into agricultural and forest catchments, shows the impact of land cover on the concentrations of selected ions (Figure 6). The most distinctive group was sites representing forested catchments, such as the rivers Gać (P9), Luboczanka (P10), and Olszówka (P14), as well as the Ceteń Stream (P16), where the lowest mean and maximum concentration values were recorded, as well as the lowest EC (Figure 6). This was particularly visible in the case of the NO₃ ion, the concentrations of which did not exceed 5 mg/L, as well as for major (Ca, Mg, K, Na) and trace elements (As, Ba, Sr, V).

Agricultural areas, which represent sites with high participation of orchards, pepper cultivation, and arable land in their upstream catchment areas, exhibited different patterns of ion concentrations (Figure 6). Overall, most of the water quality parameters reached higher values in such sites, as documented by the distribution of measured concentrations, especially of macroelements, which mean concentrations were usually several dozen mg/L higher than in forested sites. Differences between the concentrations in forested and agricultural areas were statistically significant, as indicated by the Kruskal–Wallis test for most of the measured parameters (p < 0.05) (Appendix B). Only NH₄ concentrations, which were generally uniform (Figure 6) and did not follow a clear spatial pattern across arable, orchard, pepper, and forested catchments (Figure 7), did not exhibit significant differences (p > 0.05). Within the group of agricultural sites, notably smaller variability in ion concentrations was documented (Figure 6). As highlighted by the post-hoc Dunn statistical test (p > 0.05), arable lands, orchards, and pepper catchments were not significantly differentiated in terms of K, Na, As, or Sr concentrations. However, in the case of EC values and Mg and V concentrations, only arable lands and pepper sites did not vary significantly; in orchard catchments, relatively higher EC values, as well as Mg and V concentrations, were noted, as indicated by the value distribution (Figure 6). In terms of Ba concentrations, only orchard and pepper sites were significantly similar (p > 0.05). In the case of Ca the situation was completely different—only orchard and pepper sites were significantly different. A completely outstanding pattern was found in the case of biogenic compounds, such as NO₃ and NO₂ ions; these were significantly different across all agricultural lands, as streams draining catchments with pepper cultivation exhibited greatly increased NO₃ and NO₂ concentrations (Figure 6 and Figure 7). For example, in the Pierzchnianka River and Kozieniec Stream, mean concentrations of a major biogenic compound—NO₃—reached 30 mg/L. In contrast, arable lands, represented by sites such as the Krzemionka and Rylka rivers, exhibited the lowest nitrogen contamination of all of the agricultural sites (Figure 6 and Figure 7).

Additional insights into the relationships between certain water quality parameters, as well as into their spatial heterogeneity, were provided by PCA. The spatial variability of ion concentrations across streams was represented by two independent components (factors), the cumulative explained variance of which was 74.1% (

Table 1. Component loadings (Cload), eigenvalues, and explained variance obtained in the PCA analysis.

Measured Parameter	PC1	PC2
EC	−0.96	0.02
NO3	−0.51	−0.83
NO2	−0.62	−0.73
NH4	−0.42	0.32
Ca	−0.91	0.18
Mg	−0.93	0.02
K	−0.85	−0.14
Na	−0.84	0.11
As	−0.77	0.42
Ba	−0.78	0.15
Sr	−0.80	0.11
V	−0.91	0.05
Eigenvalue	7.5	1.6
Explained variance [%]	62.8	13.4

). The first principal component (PC1) explained 62.8% of the variability in the dataset. As indicated by the loadings of the first principal component (PC1), EC, Ca, Mg, K, Na, As, Ba, Sr, and V were negatively correlated on a very high to moderate level (Figure 8a). The second principal component (PC2) explained 13.4% of the variability in the dataset and was associated with biogenic compounds, such as NO₃ and NO₂ (Figure 8a). It was related negatively to NO₃ and NO₂ (C_load = −0.83 and −0.73, respectively). Projection of cases on the factor plane defined by PC1 and PC2 indicated a clear distinction between sampling sites, representing differences in the dominant land cover of the upstream catchment areas (Figure 8b). The first PC exhibited a transition from agricultural to forested sites, reflected by decreasing concentrations of major and trace elements and, in consequence, a decrease in EC. The second PC was primarily linked to NO₃ and NO₂ concentrations, which increased from forest catchments to reach their maximum values in the case of pepper sites, particularly the Kozieniec Stream (P18) and the Pierzchnianka River (P19). It is worth noting that while the pepper and forest catchments were grouped relatively closely, the catchments representing arable land had a greater spread of points on the projection plane, which indicate that the highest variability in ion concentrations was inside this group (Figure 8b).

4.4. Relationships between Land Use and Water Quality

The overall pattern of Spearman correlation coefficients, characterizing the relationships between the concentrations of selected ions and land cover from the CLC 2018 dataset in the upstream catchment areas representing sampling sites, is presented in

Table 2. Spearman rank correlation coefficients linking land cover metrics from CLC 2018 and the mean concentrations of the measured water quality parameters in lowland streams during the study period. Abbreviations: 50, 100, 250, 500, and 1000—buffer zone widths in meters, TCA—total catchment area. Bold type indicates a statistically significant correlation at p = 0.05.

Parameter	Land Cover Type	50	100	250	500	1000	TCA
EC	Artificial surfaces	0.32	0.30	0.35	0.40	0.41	0.29
	Arable lands	0.00	0.00	0.01	−0.01	−0.09	−0.11
	Orchards	0.49	0.51	0.61	0.60	0.63	0.64
	Forested areas	−0.50	−0.52	−0.56	−0.64	−0.76	−0.75
NO3	Artificial surfaces	0.45	0.44	0.40	0.38	0.47	0.49
	Arable lands	0.55	0.53	0.56	0.44	0.36	0.31
	Orchards	0.03	0.04	0.02	0.04	0.04	0.05
	Forested areas	−0.46	−0.45	−0.44	−0.39	−0.34	−0.40
NO2	Artificial surfaces	0.59	0.57	0.55	0.53	0.60	0.56
	Arable lands	0.39	0.38	0.43	0.32	0.23	0.15
	Orchards	0.16	0.18	0.20	0.22	0.20	0.22
	Forested areas	−0.53	−0.53	−0.53	−0.51	−0.48	−0.51
NH4	Artificial surfaces	0.22	0.23	0.25	0.22	0.24	0.28
	Arable lands	−0.30	−0.31	−0.29	−0.31	−0.33	−0.31
	Orchards	0.20	0.22	0.26	0.28	0.20	0.21
	Forested areas	−0.14	−0.16	−0.19	−0.21	−0.25	−0.15
Ca	Artificial surfaces	0.23	0.24	0.30	0.36	0.37	0.22
	Arable lands	0.03	0.04	0.07	0.09	0.03	0.01
	Orchards	0.44	0.45	0.55	0.53	0.57	0.58
	Forested areas	−0.52	−0.54	−0.58	−0.67	−0.75	−0.72
Mg	Artificial surfaces	0.30	0.28	0.30	0.37	0.35	0.19
	Arable lands	0.03	0.01	−0.05	−0.11	−0.19	−0.19
	Orchards	0.50	0.52	0.54	0.55	0.54	0.56
	Forested areas	−0.39	−0.41	−0.43	−0.52	−0.65	−0.68
K	Artificial surfaces	0.58	0.63	0.69	0.70	0.75	0.65
	Arable lands	0.36	0.37	0.42	0.38	0.32	0.27
	Orchards	0.16	0.16	0.23	0.24	0.21	0.22
	Forested areas	−0.61	−0.62	−0.65	−0.60	−0.51	−0.42
Na	Artificial surfaces	0.39	0.45	0.53	0.58	0.62	0.50
	Arable lands	0.26	0.28	0.28	0.24	0.17	0.17
	Orchards	0.32	0.32	0.39	0.39	0.37	0.37
	Forested areas	−0.49	−0.51	−0.53	−0.54	−0.52	−0.46
As	Artificial surfaces	0.16	0.18	0.28	0.43	0.49	0.45
	Arable lands	0.15	0.15	0.10	0.04	0.02	0.05
	Orchards	0.17	0.18	0.29	0.29	0.26	0.27
	Forested areas	−0.27	−0.29	−0.30	−0.34	−0.50	−0.47
Ba	Artificial surfaces	0.14	0.18	0.30	0.38	0.44	0.33
	Arable lands	0.11	0.08	0.08	0.03	−0.04	−0.02
	Orchards	0.44	0.44	0.47	0.44	0.44	0.43
	Forested areas	−0.43	−0.44	−0.49	−0.52	−0.56	−0.50
Sr	Artificial surfaces	0.48	0.51	0.55	0.56	0.62	0.48
	Arable lands	0.19	0.22	0.34	0.37	0.36	0.36
	Orchards	0.13	0.13	0.24	0.24	0.24	0.25
	Forested areas	−0.56	−0.58	−0.64	−0.66	−0.62	−0.52
V	Artificial surfaces	0.24	0.22	0.25	0.33	0.33	0.17
	Arable lands	0.02	0.00	−0.03	−0.08	−0.15	−0.16
	Orchards	0.47	0.49	0.55	0.54	0.55	0.57
	Forested areas	−0.39	−0.41	−0.43	−0.53	−0.67	−0.69

.

Generally, the presence of artificial surfaces in upstream catchment areas was positively related to concentrations of NO₃, NO₂, K, Na, and Sr at all spatial scales, with the strongest relationships found for K (0.75). Furthermore, significant relationships (p < 0.05) were calculated at some spatial scales for Ca, Mg, As, and Ba, but with notably lower performance. Only the concentrations of vanadium were not linked with artificial surfaces in upstream catchment areas. A similar pattern was exhibited by the arable land class, but for this land cover type a slightly lower number of statistically significant correlations was noted (Table 2). Arable lands clearly promoted increased mean NO₃ and NO₂ concentrations at the sampling sites, as indicated by positive, statistically significant correlation values at all spatial scales (p < 0.05). An analogous effect was observed in the case of K and Sr concentrations; however, the relationships were weaker and appeared only for selected buffer zones. The impact of arable lands on the concentrations of other major and trace elements was not documented (Table 2). Finally, positive, significant relationships with ion concentrations were also found in the case of orchards; the presence of this type of cultivation in buffer zones and upstream catchment areas was related to major elements (Ca, Mg, and Na) as well as trace elements (Ba and V). For biogenic compounds, the relationships, even if positive, were not significant. A notably different pattern was found for forested areas (both coniferous and deciduous), the effect of which was usually negative and led to a decrease in their concentrations (Table 2). Statistically significant correlations (p < 0.05) were documented for all measured ions with the exception of NH₄; the best performance was calculated for Ca, K, and Sr ions.

The performance of the correlation analysis was also dependent on the spatial scale, e.g., the width of the buffer zone. Overall, in the case of biogenic compounds, such as NO₃ and NO₂, slightly better performance was documented for narrower buffer zones of 50, 100, and 250 m widths, both for arable lands and forests. The influence of artificial surfaces did not exhibit such a pattern. The correlation values were the highest for the 1000-m-wide buffer zones; however, the differences in performance were generally insignificant (Table 2). The opposite pattern was observed for other contaminants, such as major and trace elements. For Ca, Mg, K, and Na, as well as for As, Ba, Sr, and V, the correlation performance usually increased with the width of the buffer zone, reaching its maximum values at 500 or 1000 m scales. In the case of NH₄ ions, there was no spatial tendency, as none of the land cover metrics were linked significantly with their concentration (Table 2).

5. Discussion

Spatial variability in selected macro and trace elements was explored in streams representing lowland agricultural catchments with similar hydrological and climatological properties. Although numerous studies characterizing the impact of land cover on water quality have recently appeared [67,68], the novelty of the present investigation is the inclusion of specific types of agricultural practices, such as fruit tree plantations and pepper cultivations, creating unique landscapes in rural areas.

5.1. Spatial Variability Evaluation

As expected, the type of dominant land cover in upstream catchment areas has a significant impact on water quality at monitoring sites over the year. Overall, sites representing agricultural catchments promote higher concentrations of measured ions in comparison with forested ones, which was previously reported in the literature [67,69]. Cultivated lands in the area in question were generally developed on sediments and soils naturally enriched in Ca and Mg, such as clays and postglacial tills [70]; in those deposits, As, Sr, and V concentrations are also higher than on fluvioglacial sands and sandstones [71,72,73], where forested areas dominate. However, apart from the chemical compositions of sedimentary rocks across the catchment areas, increased concentrations of major and trace elements across agricultural sites could also be related to agricultural activity, such as intensive fertilization, liming, and pesticides [74,75,76,77,78]; the last was particularly evidenced in orchard catchments, in terms of the highest As and V concentrations, where a lot of pesticides are applied to control numerous insects, fungi, and diseases [79]. In addition, trace elements, such as As and V, could be components of commercial phosphate fertilizers [23,73]. This overall pattern was revealed by PCA; all agricultural sites were spread relatively closely on the first principal component, which, similarly to in [80], could be related to lithology (chemical composition of sediments) and anthropogenic activity. The last could be seen, especially, within sites representing intensive fruit tree and pepper cultivation, which exhibited similar, increased Ba concentrations; this element was previously expected to be an indicator of such nonpoint agricultural sources of ions [81]. In consequence, the cooperation and mutual interpenetration of geological (lithology) and anthropogenic factors influencing water hydrochemistry is present in such agricultural catchments.

A specific, interesting pattern of ion concentrations was exhibited by sites representing pepper cultivation in foil tunnels, which was highlighted mainly by a second PC, related to the inputs of nitrogen fertilizers. Sites with such cultivation promoted increased concentrations of NO₃, which, during the cold half-year, regularly exceeded 30–40 mg/L, while maximum values even reached 97.1 mg/L at site P18 (Kozieniec Stream). Documented nitrate contamination is also relatively high in comparison with concentrations presented by [82,83] for other agricultural regions of Poland. The elevated concentrations of NO₃ in the current study should be strictly related to the intensive fertilization of pepper cultivation. According to [84,85], the demand of peppers is approximately 200–300 kg/ha of N; in contrast, fruit trees in full fruiting require only around 50 kg/ha of N [84,86]. Moreover, the inadequate ratio of nitrogen to potassium in fertilizers is expected to be a second reason for contamination; the appropriate ratio on agricultural lands should be between 1:1.2 and 1:1.5 [21], but in the case of the pepper areas the ratio of nitrogen to potassium was estimated at 1:0.4–1:0.7. This could result in limited potential for nitrogen uptake by plants and, consequently, leaching of nitrates from the soil to groundwater and streams. In contrast to the sites representing pepper cultivation, those draining arable lands were the most variable in terms of impact on water quality. In fact, arable lands could be managed differently in terms of fertilization and pesticide use according to the type of cultivation (e.g., wheat, barley, corn, vegetables) and its specific requirements [84].

The opposite impact was found in the case of forested areas, which were characterized by lower concentrations of all measured ions at the annual timescale, varying significantly from other types of catchments. Generally, it was confirmed that woodlands, both coniferous and deciduous, could accelerate purification effects on stream water quality, as reported previously by [87,88]. Forests provide an accumulation of nitrogen, phosphorus, and metallic elements [89], mainly due to tree uptake [90] and the presence of litter layer- and organic-rich soils, supporting complex and differentiated soil microbial communities [91]. However, it must be emphasized that several factors influence the purification effects of forests, such as management practices [92], tree species composition [93], and the spatial configuration of woodlands [94], which could be also a reason behind the relatively high variability of forested sites in terms of water quality.

Finally, a clear spatial tendency was not documented for NH₄, the concentrations of which were not significantly different despite differences in land cover in upstream catchment areas; an analogous pattern was previously documented by [95] using the example of springs in western Germany. In fact, NH₄ concentrations reflect contamination from various forms of human activity across the investigated catchments, such as inflows from sewage discharges, livestock production, the fruit-processing industry, and landfills [96]. Thus, the presence of rural and urban settlements contributes to increased NH₄ concentrations [35,97], which could be accelerated by the lack of a sewage network in such areas [20]. The slightly lower factor loading for K concentrations in the PCA suggests that this ion could also reflect sewage contamination. Potassium, considered the “neglected nutrient” [98], is broadly used in fertilizers; however, it has a different migration path in comparison with nitrates, as K is washed less deep into the soil profile, because it is better absorbed by clay minerals [98]. In this way, the case of the Bystrzyca River [99] suggests that the K concentration resulted from pollution with domestic sewage rather than from the use of fertilizers, which might be exhibited in the investigated catchments.

5.2. The Role of Spatial Scale for Water Quality

Studies focusing on land cover effects on water quality have been conducted at several spatial scales, mainly to evaluate different predictors of water contaminants for management purposes [94]. Catchment scale [100], buffer zones [101], circular buffers [28], and distance-weighted approaches [27] have been adopted in numerous studies, which have modeled the spatial variability of biogenic compounds, oxygen indices, and trace metals [53,94]. There are no unified conclusions about the most informative spatial scale, as in some studies the whole catchment is the key to understanding some of the water quality parameters [102,103]. There is also broad evidence that land use in the close proximity of streams may be the most important in terms of explaining water quality [37,104]. As suggested previously, the comparison of such results is difficult due to the heterogeneous land cover datasets used, different periods of interest, and factors such as topography and relief [33,105]. Nevertheless, in the current study, some interesting tendencies and patterns were noted on the basis of the CLC 2018 dataset. Generally, in the case of major (Ca, Mg, K, Na) and trace elements (As, Ba, Sr, and V), the best performance in explaining spatial variability was noted for wider buffer zones, particularly those 500 m and 1000 m wide. In turn, NO₃ and NO₂ concentrations were slightly better explained for narrower buffer zones (50, 100, and 250 m wide); however, higher correlation performance was also documented for artificial areas in the case of wider buffer zones. This corresponds with PCA and suggests that the concentrations of major elements and trace metals have lithological and anthropogenic origins, facilitated by a larger absolute area of orchards, forests, and artificial areas in catchments. In turn, the better correlation performance for nitrogen compounds in narrower buffer zones could be related to the high mobility of nitrate ions [106] and, simultaneously, their preferable uptake by vegetation even at scales of dozens of meters [107]. Additionally, buffer zones with natural vegetation in sites representing pepper cultivation are generally narrower in comparison with orchard areas. This is due to the unsuitability of wetlands for the cultivation of fruit trees [108], as well as the lack of cattle farming [109] in orchard-dominated catchments. Thus, buffer zones in pepper catchments do not provide an appropriate geochemical barrier, in contrast to orchards, where forests or agriculturally unused grasslands hinder the migration of nitrates into streams. Finally, it is worth noting that a clear distinction in predictors was observed for major and trace elements, suggesting their dominant sources. Orchards significantly increased Ca, Mg, Ba, and V concentrations, while artificial surfaces were better correlated with K and Na, confirming that such ions may be associated with urbanization and domestic activity. Different patternwas documented for NH₄, being a primary indicator of stream pollution [110] with no statistically significant relationship with land cover type. It must be emphasized that the impact of various land cover types on water quality is differentiated seasonally, which is broadly documented to mainly be the result of hydrometeorological conditions and vegetation cycles [111].

5.3. Recommendations and Limitations

Water quality evaluation seems to remain an important research issue, as the enrichment of lotic ecosystems in nutrients can lead to eutrophication and limited water suitability for drinking purposes [112]. Simultaneous with low flows and increased water temperature, nutrient enrichment can lead to unexpected ecological disasters, such as in the case of the Odra River in 2022, where toxins released by the haptophyte golden algae, Prymnesium parvum, resulted in mass mortality among fish, bivalves, and water snails [113]. Thus, such water quality studies are particularly significant in the initial parts of stream systems, which are greatly vulnerable in terms of hydrochemical alteration and therefore crucial in appropriate water management [114].

The current study revealed significant stream pollution by nitrates in catchments with intensive pepper cultivation. However, an analogous situation might also be observed in similar catchments with the cultivation of vegetables, which have relatively low nutrient use efficiency compared with arable crops [115]. Our recommendation to reduce nitrate pollution in such areas might be to build denitrification barriers to absorb leached nitrogen compounds. Such a barrier is an organic layer (containing carbon) mixed with soil, which achieves the reduction of nitrate to gaseous nitrogen by denitrifying bacteria with limited oxygen and abundant organic carbon [116]. In many cases, the reduction in nitrate concentration using barriers is over 50%, and studies by [117,118] show that this method can also be useful for the removal of pollutants from domestic wastewater. Another possibility is the use of highly effective buffer zones, which are strips of grass or woodlands between arable land and streams. [119] found that the application of such ‘saturated buffers’ using plants such as ryegrass, fescue, bluegrass, and timothy allowed nitrate–nitrogen concentrations to be reduced from 15 mg N-NO₃/L to less than 1.5 mg N-NO₃/L. The high effectiveness of such zones was also noted in a study by [120], in which nitrate concentrations exceeding 100 mg/L were reduced by 99% with a zone width of 45 m. Finally, an interesting pattern to reduce the nitrate contamination of streams could be the use of crop rotation with leguminous plants, which are known for their major role in nutrient capture in soils [121]. It was proven that crop rotation with leguminous plants generally produces lower nitrogen runoff to groundwater and further streams than monocultures [122,123]. Such solutions require careful overall consideration and attention by managers on a local scale, as agriculture nonpoint sources remain a significant problem in such areas. It is worth noting that the investigated period (hydrological year 2022) was extremely unstable in Europe, mainly due to the ongoing EU–Russia gas war, as well as EU embargos on fertilizers from Belarus and Russia, which resulted in significant increases in fertilizer prices [124]. This in consequence could have led to a decrease in fertilizer use and thus slightly less nitrogen pollution than during previous years. Economic motives have been a catalyst for reducing nitrate pollution, as evidenced previously across post-socialist countries in the 1990s, e.g., in Poland [125], Slovakia [126], and the Czech Republic [127].

Finally, the obtained results also have practical significance in terms of future investigations and approaches related to land cover–water quality interaction.Land use/land cover data derived from aerial or satellite data sources should take into account different types of agricultural practices due to their different impacts, as was confirmed for the example of orchards, pepper, and arable lands sites. Many studies have separated out only the class of agricultural land/crops [39,53,62], which is, by its nature, a great simplification if the cultivated areas represent significantly different crops in terms of nutrient demand and agricultural practices.