Linking Riverbank Morphodynamics to Water Contamination: A Long-Term Evaluation of the Global Pollution Index in the Timiș River, Romania

Abstract

Riverbank height plays a potentially important role in hydrological dynamics and pollutant transport, yet its influence on long-term water quality trends remains insufficiently documented. This study explores possible relationships between riverbank height variations and the Global Pollution Index ($I_{GP}^{*}$) in the Timiș River, Romania, over eleven (11) years (2013–2023). A dataset of 17 physicochemical parameters—including BOD₅, COD-Cr, dissolved oxygen, nutrients (N and P species), heavy metals (As, Cr, Cu, and Zn), detergents, and phenols—was used to tentatively assess ecological status. The results suggest that, despite a maximum riverbank elevation change of ~11 cm between 2020 and 2025, $I_{GP}^{*}$ values remained within a relatively narrow range (1.98–2.56, mean 2.19), pointing to persistent but moderate anthropogenic pressure. The highest index value (2.56, in 2016) coincided with a transient pollution event, whereas subsequent years stabilized around 2.0–2.3, which may reflect chronic diffuse pollution. Correlation analysis revealed strong associations between BOD₅ and conductivity (r = 0.76, linked to organic loads), COD-Cr and heavy metals (r = 0.79, suggestive of industrial influence), and total nitrogen and nitrate (r = 0.97, related to agricultural inputs), appear to outline distinct source-related signatures. This study offers preliminary evidence that even modest riverbank fluctuations may influence hydrodynamics and the fate of pollutants, while basin-scale water quality seems to remain largely governed by diffuse pollution sources. By integrating long-term geomorphological monitoring with multi-parameter water quality data into a composite index ($I_{GP}^{*}$), our work sketches a potentially innovative framework for diagnosing pollution drivers. The findings underscore the importance of incorporating riverbank morphology into EU Water Framework Directive monitoring, alongside GIS, IoT, and machine learning tools, could contribute to more adaptive river basin management.

1. Introduction

Rivers are dynamic systems that serve as critical water sources, ecological corridors, and indicators of environmental health. The quality of river water is influenced by a complex interplay of natural and anthropogenic factors, including climate variability, land use, hydrology, and geomorphology [1,2]. Among these, topography plays a particularly crucial yet underexplored role.

At larger basin scales, natural drivers such as tectonic uplift, monsoon variability, and glacial melt interact with dams, irrigation, and land clearance to reshape sediment production and delivery, with downstream consequences for both pollutant transport and ecological stability [3]. Likewise, case studies of industrialized aquifers, such as the Thriassion Plain in Greece [4], demonstrate how hydro-geological features and seasonal dynamics strongly mediate the spread, dilution, and persistence of organic contaminants, underscoring the coupled role of geology and human pressure in shaping water quality. Importantly, geomorphological processes regulate both conventional water quality parameters and the mobility of hazardous substances—including heavy metals (As, Cr, Cu, and Zn), phenols, and detergents—by controlling their adsorption to sediments, potential accumulation, and subsequent release into the water column, with implications for ecological and human health.

Topographic features—such as slope, elevation, and relief—profoundly influence river water quality by altering hydraulic dynamics, runoff, erosion, and sedimentation, which in turn affect pollutant transport, retention, and key parameters such as dissolved oxygen (DO), biological oxygen demand (BOD₅), total nitrogen (T-N), and total phosphorus (T-P) [5,6,7,8]. For example, steeper slopes are associated with greater runoff and sediment transport, leading to increased pollutant loading. In contrast, high-relief forested areas may buffer pollutant flow, enhancing water quality. Land use intensification, climate change, and infrastructure development—such as weir construction and channel dredging—further complicate this dynamic by modifying flow regimes, water stagnation, and nutrient cycling [5,6,9,10].

Several studies underscore these effects. In the Nakdong River, the construction of large weirs increased BOD₅ by 9.4% and chlorophyll-a by 12.2%, while decreasing T-N and T-P levels due to enhanced sedimentation [6]. Similarly, in the Chishui River Basin, better water quality was associated with forest cover in high-relief areas [8], while flatter terrain was more vulnerable to pollution accumulation. Topography also explained the variation in water quality in the Xiangxi River [10], and further studies highlight the joint role of slope and soil type in determining surface water quality [11,12]. While these studies highlight the topographic effects on nutrient dynamics and organic loads, fewer investigations address how river morphodynamics interact with hazardous substances of mineralogical and industrial origin, leaving a gap in chemical characterization for effective river management.

Despite growing evidence, topographic factors remain underrepresented in many water quality models. Understanding the mechanisms through which topography regulates aquatic systems—particularly pollutant mobilization and spatial distribution—is crucial for developing effective, topography-informed strategies for riverine water quality assessment and management [13,14].

In parallel with topographic concerns, broader environmental and regulatory contexts shape how water quality is evaluated and managed. Natural variables such as hydrological, atmospheric, meteorological, and geological conditions also contribute to variations in water quality. These natural influences are often compounded by anthropogenic stressors—urbanization, agricultural runoff, and industrial discharge—which have caused persistent physicochemical changes in water bodies [15,16,17,18,19,20].

In the Timiș River basin, water management is overseen by the Banat Water Basin Administration, which operates within Romania’s national and EU legal frameworks; however, the acceptance and enforcement of these policies by local farmers and residents is uneven. While some comply with permit requirements and participate in nitrate-action plans or water-user associations, many smallholders face financial and technical constraints that limit the adoption of best practices, resulting in persistent diffuse pollution from agriculture and rural settlements despite regulatory measures.

This uneven enforcement highlights the need for monitoring tools that can detect and attribute pollution pressures, even when formal compliance data are incomplete. By explicitly integrating riverbank morphodynamics with long-term physicochemical monitoring, the present study’s $I_{GP}^{*}$ approach offers a means to capture such diffuse and point-source interactions more effectively.

To address these challenges, structured water quality assessment methods are employed. A common approach involves delineating control sections within river basins to monitor pollution sources and regulate pollutant discharge [21,22,23]. A key tool in these assessments is the Water Quality Index (WQI), which aggregates multiple water quality parameters into a single numerical value, simplifying the classification of water safety levels and supporting comprehensive environmental monitoring [24].

Water quality assessment frameworks are increasingly aligned with legal and policy instruments such as the European Union Water Framework Directive [25] and the Flood Risk Assessment and Management Directive [26]. These directives aim to ensure the sustainable use of water resources, reduce pollution, and protect aquatic ecosystems by promoting integrated basin-wide management and public engagement.

Against this backdrop, the Timiș River, another significant tributary of the Danube Basin, presents a valuable case study for examining topography-informed water quality analysis. The Timiș River basin supports diverse uses—including agriculture, aquaculture, industry, and drinking water supply—but faces increasing pressure from urban development, land-use change, and wastewater discharge.

The basin hosts growing urban, agricultural, and industrial activities, including several wastewater treatment plants, all of which affect water quality. The monitoring network in the Timiș River basin includes multiple control sections, with varying frequencies of data collection, enabling both high-resolution and long-term trend analysis.

Water management in the Timiș River basin aligns with European environmental policies and incorporates principles such as sustainability, fairness, cost recovery, the “user pays” model, and equal access to resources. These values are embedded in regional strategies to manage water quantity and quality efficiently and equitably.

The Timiș River, which flows through southwestern Romania into northern Serbia, is representative of a complex system. Its ecological and hydrological dynamics are influenced by a mix of mountainous and lowland terrains, varied land use, and differing national environmental policies. As a water body that supports multiple functions—including irrigation, biodiversity conservation, and potential potable use—the Timiș River basin presents both opportunities and challenges in the context of sustainable water management.

While numerous studies have examined the influence of slope, land cover, and channel morphology on water quality [1,2,3,4,5,6,7,8,9,10], the specific role of riverbank height in regulating hydrological dynamics and pollutant transport remains largely underexplored. In particular, little is known about how bank height may influence erosion, sediment retention, and the mobilization of particle-bound contaminants under various flow regimes. Previous research has primarily emphasized hydrodynamic drivers such as flow velocity, sediment transport, and channel cross-sectional shape, demonstrating how bank erosion contributes to suspended sediment loads and how altered hydraulics affect nutrient cycling and turbidity [23,27,28,29,30,31,32,33]. However, the role of riverbank height as a distinct geomorphic parameter has received little systematic attention. By explicitly integrating riverbank morphodynamics with long-term water quality monitoring in the Timiș River basin, this study seeks to explore whether bank height could be linked to nutrient cycling and the fate of hazardous substances within a management-relevant framework.

Despite the recognized importance of the river, comprehensive long-term assessments that integrate multiple water quality indicators and evaluate spatiotemporal trends remain limited. Most existing studies have focused on short-term monitoring or isolated parameters, offering only a partial understanding of ecological conditions and pollution sources. Furthermore, the influence of topography and landscape transitions on pollutant distribution and retention has often been overlooked, limiting the applicability of findings to basin-wide management strategies.

This study aims to investigate possible associations between variation in riverbank height and hydrological dynamics, as well as the transport, retention, and remobilization of pollutants in a transboundary river system. Using an 11-year dataset of 17 physicochemical parameters (BOD₅, COD-Cr, DO, nutrients, detergents, phenols, and heavy metals) integrated with field-measured riverbank height changes, we apply descriptive statistics, Pearson correlations, and principal component analysis to identify dominant pollutant groupings and their potential relation to geomorphological changes. This combined geomorphology–water quality framework ($I_{GP}^{*}$) is intended to explore whether subtle bank elevation fluctuations (≈approximately 11 cm between 2020 and 2025) modulate pollutant fate and differentiate between diffuse and point-source contributions under persistent anthropogenic pressure. The findings are expected to provide preliminary insights for topography-informed water quality assessments in line with EU policies.

Given these gaps, this study assesses the overall water quality of the Timiș River over an eleven-year period (2013–2023) across selected monitoring sections. By integrating riverbank morphodynamics with descriptive and multivariate statistical analyses, including composite pollution indices and Pearson correlation, the research aims to provide a tentative evaluation of water quality trends, considering both natural and anthropogenic drivers. The study also contributes to the mineralogical and chemical characterization of transboundary river waters, supporting the development of locally adapted, science-based strategies for monitoring and managing pollutants in line with EU water policies, thereby promoting environmental sustainability, ecosystem resilience, and regional cooperation.

2. Materials and Methods

2.1. Study Areas

The Timiș River basin is situated in the southwestern part of Romania and northern Serbia, extending from the Western Carpathians into the Pannonian Basin. The basin is bounded to the north by the Bega–Tisa hydrographic system, to the east by the Caraș and Nera basins, to the south by the Danube drainage area, and to the west by the Mureș and Tisza River basins. The upper part of the river flows through a mountainous landscape (the Semenic Mountains), while the lower segment lies in a lowland, alluvial plain (Figure 1).

The Timiș River drains a basin of approximately 10,280 km² (of which about 7310 km² lies in Romania) and spans ~359 km from source to its mouth into the Danube near Pančevo, Serbia, with an average discharge of 47 m³/s. It traverses industrial and urban centers such as Caransebeș, Lugoj, and Pančevo, exposing it to industrial point-source loading. Field surveys in the basin have recorded chloride up to 18.3 mg/L and sulfate up to 68.8 mg/L, consistent with anthropogenic inputs. The region is also subject to water stress during dry seasons, driven by irrigation demands and climate-driven reductions in flow and precipitation. On the Romanian side, it is monitored by the Banat Water Basin Administration (ABA Banat), which oversees quantitative and qualitative parameters using an integrated water information system (ABA Banat, 2021) [35]. The Banat watershed area (including Timiș and adjacent counties) has a population density of about 61 inhabitants/km².

In the present study, it was considered a water body (Timis–Timisana border) that has a length of 90.21 km and is characterized by two sections, the Sag and Graniceri localities. The Sag section was chosen to represent upstream influences from urban and agricultural activities, while Grăniceri reflects downstream, transboundary conditions influenced by cumulative pressures. Geographic coordinates of the two sections (Sag: 45.69° N, 21.06° E; Grăniceri: 45.55° N, 20.85° E) are provided for reproducibility.

2.2. Data Collection

This study analyzed water samples from the designated section of the Timiș River, focusing on the following parameters: chemical oxygen demand (COD-Cr), biochemical oxygen demand (BOD₅), dissolved oxygen, electrical conductivity, pH, nutrients (nitrogen and phosphorus species), detergents, phenols, and selected heavy metals (As, Cr, Cu, and Zn)—sampling procedures adhered to the methodology established by Romanian national legislation [36]. Samples were collected annually in April at mid-depth (0.5 m) and mid-channel, following standardized Romanian and EU protocols to ensure consistency across years. For each station, triplicate samples were taken and transported in cooled containers (4 °C) to accredited laboratories. Data were collected annually over eleven years (2013–2023), with all analyses conducted in accredited laboratories in compliance with the applicable legal standards [37]. All targeted physicochemical parameters—COD-Cr, BOD₅, dissolved oxygen, electrical conductivity, pH, nutrients (N and P species), detergents, phenols, and selected heavy metals (As, Cr, Cu, and Zn)—were consistently measured in every year from 2013 to 2023 in accredited laboratories, ensuring a complete dataset for the calculation of the Global Pollution Index.

Analytical methods included Winkler titration for DO, dichromate reflux for COD-Cr, manometric respirometry for BOD₅, spectrophotometry for nutrients, flame atomic absorption spectroscopy (AAS) for Cr, Cu, Zn, and hydride generation AAS for As. Detection limits ranged from 0.002 to 0.01 mg/L, depending on the analyte. Quality assurance/quality control (QA/QC) included the use of blanks, duplicates, calibration standards, and inter-laboratory comparisons. A full description of the instruments, reference standards, detection and quantification limits, precision/accuracy, as well as the sampling frequency and cross-section coordinates for all monitored parameters is available in Supplementary Table S1.

The topographic measurements were performed using a Trimble R8s GNSS receiver (Trimble Inc., Westminster, CO, USA). Measurements were taken at six different points in 2020 and 2025, revealing an average difference of approximately 11 cm.

The GNSS receiver has an accuracy of ~8 mm + 1 ppm (horizontal) and ~15 mm + 1 ppm (vertical). Cross-sections were selected to represent stable (forested) and unstable (agricultural/urbanized) banks, allowing evaluation of bank height as a driver of pollutant retention and release. Maximum elevation change (~11 cm) calculated as the difference in mean bank height between 2020 and 2025 surveys, allowing evaluation of bank erosion/deposition and links to pollutant retention. The standard deviation of each elevation difference is approximately 2.2 cm, and the standard error of the mean difference is approximately ±0.9 cm. Thus, although absolute elevations may carry minor errors, the relative change of ≈11 cm remains robust and well above measurement uncertainty.

The investigated segments of the Timiș River are of particular interest due to their potential use as a source of irrigation water. However, the presence of possible pollution sources along the river course may necessitate specific water treatment measures. Additionally, the transboundary nature of the river requires compliance with international legal frameworks, especially considering its potential use for agricultural irrigation and drinking water supply by downstream communities.

2.3. Data Analysis

To assess water quality and identify relationships between the analyzed physicochemical parameters, a complex statistical methodology was employed, integrating both descriptive statistics and multivariate correlation methods.

For each variable, descriptive statistics parameters (mean, median, standard deviation, extreme values, and mode) were calculated to characterize the centrality and dispersion of the data. Pearson correlation analysis was applied to determine the degree of linear association between variables, considering standard interpretation thresholds: negligible correlation (r < 0.3), low positive (negative) (r = 0.3–0.5), moderate positive (negative) (r = 0.5–0.7), high positive (negative) (r = 0.7–0.9) and very high positive (negative) (r = 0.9–1.0) [38].

Pearson’s correlation coefficient was selected because all variables were measured on a continuous scale and preliminary inspection of their distributions showed no extreme skewness that would preclude linear analysis. Our primary objective was to quantify linear relationships between parameters, for which Pearson is appropriate. Although environmental data can deviate from normality, trial comparisons with rank-based methods indicated that Spearman’s or Kendall’s coefficients would not materially alter the direction or strength of the observed associations; therefore, the conclusions presented here are robust to the choice of correlation metric.

For each correlation coefficient, the associated two-tailed p-value was computed to test the null hypothesis of no association. A significance threshold of p < 0.05 was adopted, with p < 0.01 considered highly significant. Given the limited number of annual observations (n = 11), only correlations meeting these thresholds are interpreted as meaningful in the discussion.

The results were synthesized into a correlation matrix, facilitating the identification of direct or inverse relationships between physicochemical parameters [39]. This approach enabled the identification of potential sources of pollution (organic, nutritional, or industrial), as well as the assessment of their combined effects on water quality.

While the dataset comprises annual, single-point measurements (n = 11), the consistent, long-term collection of multiple physicochemical parameters across key monitoring sites provides a robust basis for assessing long-term trends and inter-parameter relationships. The integration of multivariate statistics and composite pollution indices further strengthens the reliability of our findings, informing river management and complementing higher-frequency studies rather than replacing them.

The analyses were performed using specialized Excel add-in software. The datasets were used to obtain relevant graphical representations (stacked column, boxplots, correlation maps).

To identify potential sources of pollution and mechanistic linkages between parameters, pairwise Pearson correlations were calculated for all physicochemical and metal variables measured at each sampling campaign. This allowed for the detection of associations between key parameters.

Given the size and completeness of the dataset (17 parameters measured annually across eleven years), principal component analysis (PCA) was applied to reduce dimensionality and reveal dominant pollution structures. PCA was performed only on attributes with values in all years (n = 11), using the Kaiser criterion (Eigenvalue > 1) to retain significant components.

Another essential aspect considered in this study was the calculation of the overall pollution index based on the methods described by Zaharia [40,41]. The Zaharia equation was selected because it integrates individual pollutant concentrations into a single index value that accounts not only for exceedances relative to regulatory thresholds, but also for the cumulative and combined effect of multiple pollutants. This makes the method particularly suitable for evaluating complex, multi-parameter water quality datasets such as the one analyzed in the present study.

Initially, the quality index was determined using the following formula:

$${EQ}_i={\displaystyle \frac{C_i}{MACi}}$$

(1)

where EQ_i denotes the parameter’s quality index, C_i is the parameter’s measured value, and MAC_i is its maximum allowable value for parameter i. For each parameter, the maximum allowable concentration used in Equation (1) was taken from the legally binding Romanian national standards for surface waters (Law no. 107/1996 ‘Water Law’; Government Decision 351/2005) and, where relevant, aligned with the limits specified in the EU Water Framework Directive (2000/60/EC) and its daughter directives.

Each parameter’s quality index (EQ_i), calculated as the ratio of the measured concentration to its maximum allowable concentration (MAC_i), was subsequently converted into an evaluation score (ES_i) to standardize the assessment of water quality across multiple indicators. The evaluation scores are presented in

Table 1. Evaluation of Water Quality Based on Quality Index (EQ_i), Evaluation Score (ES_i), and Environmental Impact.

Quality Index (EQi)	Evaluation Score (ESi)	Environmental Impact
EQi = 0	10	Water bodies are unaffected by industrial activity.
0.00 < EQi ≤ 0.20	9	Industrial influence is present but not quantifiable.
0.20 < EQi ≤ 0.70	8	Impact detected, but below first alert threshold.
0.70 < EQi ≤ 1.00	7	Alert level: Possible consequences.
1.00 < EQi ≤ 2.00	6	Impact within second-level threshold limits. Intervention level: Potential outcomes expected.
2.00 < EQi ≤ 4.00	5	Pollution exceeds the first legal limit. Effect: Noticeably strong impact.
4.00 < EQi ≤ 8.00	4	Pollution surpasses the second limit. Effect: Environmentally detrimental.
8.00 < EQi ≤ 12.00	3	Third limit exceeded. Effect: Clear negative consequences.
12.00 < EQi ≤ 20.00	2	Severe degradation—Level 1. Impact: Fatal effects over average exposure duration.
EQi > 20.00	1	Severe degradation—Level 2. Impact: Rapid onset of fatal effects; environment rendered unsuitable for life

, showing thresholds and corresponding classifications for pristine to severely degraded conditions.

Equation (2) was used to determine the overall pollution index for each sample in this investigation after the assessment score was assigned, considering that each parameter is assumed to contribute equally to the overall pollution index. Moreover, the formula considers both individual exceedances (via EQ_i and ES_i) and cumulative effects across adjacent parameters.

$$I_{GP}^{*}={\displaystyle \frac{100\times n}{{ES}_1\times {ES}_n+{\sum }_{i=1}^{n-1}{ES}_i\times {ES}_{i+1}}}$$

(2)

where $I_{GP}^{*}$ is the Global Pollution Index, n is the number of parameters, and ES_i is the attributed evaluation score.

The advantage of Zaharia’s formulation is that it accounts for both the severity of individual exceedances (through EQ_i) and the interaction of adjacent parameters (through cumulative ES_i values). Thus, it provides a synthetic and management-relevant representation of ecological pressure, which complements correlation and PCA results.

The impact of human activity on water bodies can be assessed using the $I_{GP}^{*}$ index, with its values corresponding to various levels of ecological consequence. A value of 1 indicates that the water body is unaffected by anthropogenic influence, representing a pristine state. When the index falls between 1 and 2, minor human influence is present, though activities remain within environmentally acceptable limits. As the index increases to the 2–3 range, signs of noticeable stress on aquatic life begin to emerge, reflecting growing anthropogenic pressure. Values between 3 and 4 suggest a significant disruption of natural biological processes due to pollution. An index in the 4–6 range reflects serious ecological damage, where the survival of aquatic organisms is increasingly endangered. When the $I_{GP}^{*}$ value exceeds 6, the water body is considered severely degraded and unfit to support any form of life.

To explore geomorphological influences, correlations were also computed between $I_{GP}^{*}$ values, individual pollutants, and riverbank height data measured at Sag and Grăniceri, thereby linking physicochemical and geomorphic dynamics.

3. Results and Discussion

3.1. Evolution of Target Parameters

Over the eleven (11) years, most parameters remained within national and EU threshold limits (Table 1). BOD₅ averaged 4.6 mg O₂ L⁻¹ (range 2.98–5.28) with only three exceedances of the 5 mg O₂ L⁻¹ limit. COD-Cr averaged 24.3 mg O₂ L⁻¹ (range 14.5–36.1), with a single year (2020) below the 15 mg O₂ L⁻¹ “very good” quality threshold and peaks in 2016 that may reflect episodic discharges. DO remained high (mean 7.6 mg O₂ L⁻¹), with an unusual maximum of 9.4 mg O₂ L⁻¹ in 2020. Conductivity stayed well below the 1000 µS cm⁻¹ limit (mean 406 µS cm⁻¹). Total nitrogen occasionally exceeded 2.0 mg L⁻¹, likely reflecting diffuse agricultural runoff, whereas total phosphorus surpassed 0.1 mg L⁻¹ only in 2016, indicating a potential temporary risk of eutrophication. Heavy metals (As, Cr, Cu, Zn) remained below regulatory limits but displayed slight downward trends, with Cu showing the highest interannual variability. Active anionic detergents were an order of magnitude lower than the legal limit (max. 0.09 mg L⁻¹), and phenols remained close to 1 µg L⁻¹ except for minor peaks in 2018 and 2022.

These condensed patterns suggest that the Timiș River generally maintains moderate water quality but may be subject to episodic events associated with urban wastewater, agricultural inputs, and potentially geomorphological changes in bank height (Section 3.2). The whole time series, anomaly analyses, and boxplots supporting this summary are provided in Supplementary Figures S1–S27.

Table 2. Descriptive statistical analysis for all parameters discussed during the eleven years investigated.

Characteristic		Indicator Value *
	Threshold Limit [42]	Mean (μ)	Median	Mode	Minimum	Maximum	Standard Deviation
BOD5	5 [mg O2/L]	4.62	4.83	5.28 (2023, 2019, 2013)	2.98 (2020)	5.28 (2023, 2019, 2013)	0.77
COD-Cr	25 [mg O2/L]	24.31	20.72	29.89 (2023, 2019, 2013)	14.49 (2020)	36.11 (2016)	6.70
DO	8 [mg O2/L]	7.57	7.36	7.89 (2023, 2019, 2013)	6.59 (2016)	9.42 (2020)	0.75
Conductivity	1000 [µS/cm]	405.66	448.50	517 (2023, 2019, 2013)	206 (2016)	517 (2023, 2019, 2013)	108.89
pH	8.5 pH units	7.86	7.91	8.01 (2023, 2019, 2013)	7.29 (2014)	8.01 (2023, 2019, 2013)	0.22
Total N	1.5 [mg/L N]	2.57	2.15	4.23 (2023, 2019, 2013)	1.57 (2014)	4.23 (2023, 2019, 2013)	1.10
N-NH4	0.4 [mg/L N]	0.25	0.25	0.25 (2023, 2019, 2013)	0.13 (2020)	0.44 (2014)	0.08
N-NO2	0.01 [mg/L N]	0.04	0.04	0.04 (2023, 2019, 2013)	0.02 (2016, 2014)	0.06 (2020, 2018)	0.01
N-NO3	1 [mg/L N]	1.27	1.09	2.05 (2023, 2019, 2013)	0.69 (2014)	2.05 (2023, 2019, 2013)	0.51
Total P	0.02 [mg/L P]	0.21	0.19	0.19 (2023, 2019, 2013	0.16 (2020, 2014)	0.40 (2016)	0.07
P-PO4	0.05 [mg/L P]	0.08	0.07	0.07 (2023, 2020, 2019, 2013	0.06 (2022, 2018)	0.14 (2016)	0.03
detergents	500 [µg/L]	55.44	50.00	50.00 (2023, 2021, 2020, 2019, 2013)	15.00 (2015)	99.00 (2017)	21.07
Total phenols	1.00 [µg/L]	1.03	1.00	1.00	1.00 (the rest of the considered years)	1.13 (2022, 2018)	0.05
As	5.00 [µg/L]	0.87	1.09	1.09 (2023, 2019, 2013)	0.50 (2017)	1.10 (2022, 2018)	0.29
Cr3+ + Cr6+	50.00 [µg/L]	1.12	1.00	1.00 (2023–2018, 2013)	1.00 (2023–2018, 2013)	1.35 (2016–2014)	0.17
Cu	20.00 [µg/L]	4.52	4.28	4.28 (2023, 2019)	1.10 (2014)	8.82 (2020)	2.25
Zn	100.00 [µg/L]	19.41	23.46	23.46 (2023, 2019, 2013)	12.50 (2021, 2020)	25.00 (2016–2014)	5.66

* Each mean, median, mode, minimum, maximum, and standard deviation value is calculated from n = 11 annual measurements (2013–2023) at the same monitoring point.

summarizes the leading indicators for each of the characteristics considered in this study. The highest frequency of occurrence in the dataset was observed for the same three years (2013, 2019, and 2023). In many cases, these are the periods with the highest values registered. Analyzing the relationship between the mean and median of the attributes, it can be considered that the datasets have a negatively skewed distribution to the left in the cases of BOD, conductivity, pH, and Zn, since the average value is lower compared to the median. The central tendency tends to be towards the higher end of possible scores. A symmetrical distribution of the data, with the average and median values being the same, is observed in the cases of N-NH₄ and N-NO₂. The other parameters are characterized by a mean rate that is higher than the median. In this situation, the data are positively skewed or right-skewed, meaning there are more lower values in the dataset. There could be exceptionally high values that drag the mean upward.

The evolution of each target parameter is graphically presented, along with the observed anomalies (Supplementary Figures S1–S28).

3.2. Statistical Analysis

Unless otherwise noted, only correlations significant at p < 0.05 are discussed below.

To simplify interpretation and provide clearer insights, water quality parameters were categorized based on the nature of their correlations. Positively correlated variables—those that tend to increase or decrease together—are illustrated in Figure 2, which presents a correlation matrix of the selected parameters.

The matrix utilizes the Pearson correlation coefficient (r), a statistical measure that indicates the degree and direction of the linear association between two variables [43]. In our multi-year analysis of Timiș water samples, we employed this method to identify potential familiar sources of pollution. The array was helpful in this investigation to determine the possible interdependencies between pollutants. Its interpretation enables us to identify potential dominant causes of pollution and the most relevant parameters for monitoring. It also allows us to visualize the relationships between all the analyzed parameters considered simultaneously. Based on its interpretation, we could suggest the dominant causes of pollution and the most relevant parameters for monitoring.

A very high positive correlation of 0.97 was found between the total nitrogen and its oxidized form (the nitrate ion). Such a relationship suggests that total nitrogen is quantitatively dominated by its oxidized form of nitrogen (NO₃⁻). The other forms (ammonium ion (NH₄⁺), nitrites (NO₂⁻), and organic nitrogen) are present in small quantities compared to nitrates.

The situation may be linked to the flow path of the Timiș River, which crosses the Western Romanian Plain and passes through important agricultural areas. There may be a potential agricultural influence through the use of fertilizers. A high concentration of nitrates correlated with total N may indicate infiltration or runoff from fertilized agricultural land, or a diffuse pollution of a non-point source nature (not from a single source). If nitrates dominate the total nitrogen quantity, it can be inferred that the water has a stable oxidative profile, generally not recently polluted with fresh organic matter, which would give higher concentrations of NH₄⁺. The situation could also be the result of a complete mineralization process where the organic nitrogen has been completely converted to NO₃⁻, a good aeration of the river contributing to an efficient nitrification, or the fact that the pollution process is an old one and all the unstable forms of nitrogen have already been converted to nitrates.

Three highly positive correlations were found between the data sets used in this study. The value of 0.89 was between total phosphorus and orthophosphates, with the ionic form being the main component of the total one, compared to other forms such as organic phosphorus, particulate phosphorus, and polyphosphates. The situation indicates a high availability of phosphorus in a form that is biologically assimilable. Possible sources include detergents based on phosphates, which are directly assimilable from household wastewater containing soluble phosphates, such as those from detergents, urine, feces, industrial sewage, or from biological decomposition. Given that the anion active detergent level in our case was very low, we suspect that deficiencies in various wastewater treatment processes may have contributed to the situation.

The second correlation of increased importance, with a value of 0.79, was found between Zn and COD-Cr. Zinc and oxidizable organic substances, which contribute to COD-Cr, originate from the same sources, primarily industrial wastewater, urban runoff, or untreated domestic wastewater discharges. Considering the localization of the Timiș River, the most likely source of contamination with Zn could be household water due to domestic activities and the road runoffs that wash the roads with tire wear marks.

The third significant correlation, with a value of 0.76, was found between BOD5 and conductivity. This correlation has multiple meanings, depending on the ecological context and the sources of pollution. A high correlation may indicate that biodegradable organic matter (BOD₅) comes from the same sources as mineral ions (Na⁺, K⁺, Cl⁻, NO₃⁻, SO₄²⁻, etc.) that increase conductivity. The situation could be determined by domestic and industrial wastewater, manure seepage, or livestock platforms.

Correlations between biological (BOD₅) and chemical (COD-Cr, DO) indicators show the oxygen balance in the water. According to the Pearson correlation presented in Figure 2, there is a low correspondence of only 0.43 between the two indicators, COD-Cr and BOD₅. The result suggests fewer sources of organic pollution. The observation is also sustained by the low inverse relation between DO and COD-Cr (−0.32). Such a result shows decreased oxygen consumption due to organic load.

Negatively correlated variables—those that tend to increase or decrease in opposite directions—are illustrated in Figure 3, which presents a correlation matrix of the selected parameters.

A very high negative correlation with a value of −0.9 was found between Cu and N-NH₄⁺. An inverse relationship may reflect a toxic influence on microbiological processes. The higher the Cu concentration, the lower the accumulation of NH₄⁺, either because its oxidation is temporarily stimulated or because the NH₄⁺ supply is reduced due to the decrease in degradable organic matter. This situation also explains the high contribution of NO₃⁻ to the total nitrogen content. Copper, being a toxic heavy metal even in low concentrations, affects microorganisms involved in the nitrogen cycle, especially those responsible for nitrification.

A moderate negative correlation of −0.61 was detected between Cu and P-PO₄³⁻. The significance of the situation highlights that the areas or times where urban industrial pollution dominates do not show a corresponding increase in Cu, unlike those where domestic/agricultural pollution dominates orthophosphate. It is also known that copper is toxic to aquatic microorganisms that recycle or release phosphates from organic matter, or participate in the mineralization of organic phosphorus into phosphate (PO₄³⁻). Thus, in the presence of copper, biological activity may be affected, leading to a reduction in the release of PO₄³⁻, and its accumulation in organic or sediment-bound forms.

In our case, there is a moderately harmful dependency of −0.558 (Figure 3) between the two parameters DO and COD-Cr. A mild negative correlation between the two attributes indicates a general trend of water quality degradation when the level of organic pollution increases. It is a signal of moderate ecological stress, but it can become critical under certain conditions (summer, stagnant waters, other concomitant pollutants).

A very high correlation between total N and N-NO₃⁻ is an indicator that the system is oxidatively stable, but also that it should be monitored for risks of eutrophication, nitrate, or phosphate contamination. A high correlation between total phosphorus (P) and orthophosphate (P-PO₄³⁻) indicates recent and active pollution, with a high risk of ecological imbalance [27,44].

A high correlation between Zn and COD-Cr indicates a mixed pollution, comprising both organic and metallic components, with a common source [45]. It is a warning signal regarding the toxic potential and anthropogenic pressure on the aquatic ecosystem. The high correlation between BOD₅ and conductivity also supports the pressure exerted by human activities on the environment [46]. It requires continuous monitoring and identification of sources to enable effective intervention.

While the correlation patterns in Figure 3 and Figure 4 highlight plausible mechanistic linkages between water quality parameters, these interpretations are supported by both our spatial observations (e.g., proximity of stations to industrial or agricultural zones) and prior studies documenting similar processes in European rivers [33,46,47,48,49,50]. The proposed mechanisms—such as the mobilization of nutrients and metals during bank erosion or the attenuation of pollutants under higher banks—remain hypotheses that require site-specific confirmation. Future work could employ complementary approaches, such as additional laboratory analyses, satellite observations coupled with runoff or inflow measurements, hydrogeochemical modelling of pollutant fluxes, mesocosm tests, or isotopic labelling of eroded soils to trace sources and processes directly. These methods would strengthen the causal understanding of the relationships identified in this study.

The results of the study underscore the need to improve the performance of wastewater plants. There is an increased need for effective methods to remove contaminants that could pose significant environmental hazards.

The correlations between water quality parameters provide essential insights into the sources and processes driving pollution in the Timiș River. The strong relationship between BOD₅ and conductivity reflects how organic-rich wastewater inputs (e.g., municipal discharges) often contain dissolved ions (such as chloride, bicarbonate, and sulfate) that increase conductivity while simultaneously raising oxygen demand due to the microbial decomposition of organic matter. Similarly, the association of COD-Cr with heavy metals (As, Cr, Cu, and Zn) suggests a common industrial origin, as wastewater streams from industrial activities often release both organic chemicals and trace metals, thereby increasing chemical oxygen demand. The close correlation between total nitrogen and nitrate (NO₃⁻) highlights diffuse agricultural runoff as a dominant source, given that fertilizers predominantly leach as nitrates [28,49]. Additionally, weaker but relevant associations exist between detergents, phosphorus species (total P and PO₄³⁻), and BOD₅, suggesting inputs from household effluents and runoff from intensively farmed lands. These mechanistic linkages demonstrate how different pollutant groups co-occur, reflecting the interplay of diffuse agricultural pollution and point-source discharges in shaping water quality trends.

Starting from the data set dimension, we considered the feasibility of applying the principal component analysis (PCA). Based on this method, a smaller number of principal components was generated. These resulted in linear combinations of the original variables. All constituents aim to maintain as much variation in the data as possible, enabling more precise visualization and interpretation. The efficacy of the methodology was underscored in several research studies published [51,52,53,54,55]. The present PCA was based only on the attributes that had values in all the periods considered (a total of eleven). Figure S28 shows the variance contribution of each characteristic examined. Based on the Kaiser criterion (Eigenvalue > 1), four of the aspects could be of interest in our study. Their contributions were 40.67%, 28.44%, 13.43%, and 10.67%. These four items point out 93.23% of the variation in the inputs.

Table 3. PCA outputs for the four main components.

Feature\Component	Component 1	Component 2	Component 3	Component 4
BOD5 (mgO2/L	−0.092	0.429	−0.336	−0.024
COD-Cr (mgO2/L)	0.185	0.384	0.199	−0.170
DO (mgO2/L)	−0.184	−0.184	0.553	−0.134
Conductivity (µS/cm)	−0.238	0.330	−0.117	0.329
pH (pH unit)	−0.339	0.177	−0.283	−0.140
Total N (mg/L N)	−0.217	0.338	0.352	−0.157
N-NH4 (mg/L N)	0.278	0.122	0.156	0.540
N-NO2 (mg/L N)	−0.275	0.050	−0.264	0.273
N-NO3 (mg/L N)	−0.201	0.407	0.215	−0.115
Total P (mg/L P)	0.275	0.162	−0.290	−0.450
P-PO4 (mg/L P)	0.365	0.069	−0.112	−0.303
Dissolved Cr (Cr3+ + Cr6+) (µg/L)	0.396	−0.020	−0.081	0.069
Dissolved Cu (µg/L)	−0.315	−0.233	−0.019	−0.348
Dissolved Zn (µg/L)	0.215	0.335	0.283	0.002

presents the proportions of the first four principal components (PC1, PC2, PC3, and PC4) in relation to the total variability of the data.

Principal component analysis (PCA), applied in our case to the set of physicochemical and metal parameters determined for the river water, explains distinct pollution structures and relevant ecological processes.

The concentrations of dissolved chromium, total phosphorus, orthophosphates, and ammonia nitrogen mainly define the PC1. In contrast, variables such as pH and dissolved oxygen present negative loadings. This distribution suggests a predominantly industrial and urban source, where the input of heavy metals and nutrients is associated with decreased oxygenation and changes in pH, typical of anthropogenically polluted waters.

The second component is characterized by high values of BOD₅, COD-Cr, total nitrogen, and nitrates, while dissolved oxygen has a negative charge. This reflects a classic pattern of organic and agricultural pollution, originating from urban wastewater and agricultural runoff. High oxygen consumption and increased nutrients are accompanied by decreased dissolved oxygen, suggesting an increased risk of eutrophication.

The third component is defined by positive values for dissolved oxygen, total nitrogen, zinc, and nitrates, negatively correlated with BOD₅, pH, and nitrites. This structure reflects an ecological gradient, where well-oxygenated waters are associated with higher concentrations of nitrate and zinc, but with lower organic load and nitrites. PC3 thus captures the dynamics between self-purification processes and the pressure of organic pollution.

The last component is dominated by total phosphorus, orthophosphates, dissolved copper, with significant negative charges for ammonia nitrogen, conductivity, and nitrite. This combination indicates a mixed agricultural and industrial type of pollution, where nutrients (especially phosphorus) are correlated with heavy metals (such as copper), suggesting that wastewater discharges from both agricultural and industrial or urban activities are present.

The results confirm the complexity of the pollution sources and reveal that the river water is simultaneously influenced by industrial, agricultural, and urban processes, each having a distinct imprint on the structure of the analyzed variables.

A scientifically relevant principal component analysis assumes that the data set considered is continuous. Hydrotechnical works were carried out with a two-year periodicity compared to water quality assessment plans (every year), focusing on specific parameters. These aspects make it difficult to carry out a PCA-type analysis involving the qualitative attributes of the annually assessed waters and the changes brought to the watercourses. Such a perspective could lead to erroneous conclusions.

In the PCA analysis, water quality parameters were used exclusively, avoiding the inclusion of geomorphological data, such as bank height. This methodological decision is justified by the fundamentally different nature of the two categories of variables. Chemical and biological water parameters reflect dynamic processes that are dependent on both anthropogenic and natural factors. At the same time, geomorphological variables are relatively stable and describe the physical and geographical characteristics of the riverbed. Integrating them into a standard analysis would have led to a distortion of the principal components structure, as geomorphological variables, measured in units incompatible with those of chemicals, could have artificially dominated the explained variability. Consequently, to avoid forced interpretations and loss of relevance in identifying pollution patterns and correlations between water parameters, it was decided to exclude these data from the PCA analysis.

To scientifically sustain our investigation, we considered the data obtained through an advanced statistical investigation. It includes the factor loadings matrix and a biplot graph. The higher values (independent of sign) indicate which variables have the most significant influence on the component,

Table 4. Correlations between variables for the first four components.

Feature\Component	Correlation Variables-Components
	C1	C2	C3	C4
BOD5 (mgO2/L)	−0.219	0.856	−0.461	−0.029
COD-Cr (mgO2/L)	0.442	0.767	0.273	−0.208
DO (mgO2/L)	−0.440	−0.367	0.759	−0.163
Conductivity (µS/cm)	−0.567	0.659	−0.161	0.403
pH (pH unit)	−0.810	0.353	−0.389	−0.171
Total N (mg/L N)	−0.518	0.674	0.483	−0.192
N-NH4 (mg/L N)	0.663	0.244	0.214	0.660
N-NO2 (mg/L N)	−0.655	0.100	−0.362	0.333
N-NO3 (mg/L N)	−0.479	0.811	0.294	−0.141
Total P (mg/L P)	0.657	0.322	−0.398	−0.551
P-PO4 (mg/L P)	0.871	0.138	−0.153	−0.370
Dissolved Cr (Cr3+ + Cr6+) (µg/L)	0.945	−0.039	−0.110	0.084
Dissolved Cu (µg/L)	−0.753	−0.465	−0.027	−0.426
Dissolved Zn (µg/L)	0.512	0.668	0.388	0.003

.

Analyzing the data from Table 4, we observe that PC1 is dominated by variables such as P-PO₄, dissolved chromium, N-NH₄, total P, COD-Cr, and dissolved Zn, in contrast to pH, dissolved copper, and N-NO₂. This suggests that PC1 reflects the impact of industrial and agricultural pollution. The correlation parameters also support the observation. The component mainly reflects the effects of anthropogenic pollution with nutrients and heavy metals.

PC2 is strongly positively correlated with BOD₅, COD-Cr, total N, N-NO₃, dissolved Zn, and negatively correlated with dissolved oxygen and dissolved copper. This component highlights the contrast between organic pollution and the river’s ability to self-purify. Samples with high values on PC2 exhibit a high organic load and oxygen deficiency, which is characteristic of sectors heavily impacted by anthropogenic pollution. The relationship between high organic and nitrogen loading, associated with increased biochemical and chemical oxygen demand, as well as high concentrations of total nitrogen and dissolved zinc, in contrast to higher levels of dissolved oxygen, suggests an intense oxygen consumption process in the presence of organic pollution.

PC3 is characterized by DO, total N, dissolved Zn, in opposition to BOD₅, pH, and total P. The situation reflects the balance between dissolved oxygen and the presence of organic pollutants and nutrients. A high score indicates samples with better oxygenation status, while negative scores signal the impact of phosphorus and organic matter loading. This result suggests the essential role of oxygen in regulating water quality and counteracting organic pollution.

PC4 separates the variables N-NH₄, conductivity, and N-NO₂ from total P and dissolved copper. This highlights the differentiation between ammonia pollution sources (possibly household discharges) and those of phosphorus or metals (industrial sources). This component may reflect the biochemical processes of nitrogen transformation (nitrification/denitrification).

Figure 4 includes the variable distribution in cases of PC1 and PC2, as these components explain most of the variance and describe the main pollution factors (nutrient and heavy metal loading, organic pollution, and oxygen deficiency, respectively). It is a biplot graph type that overlays the vectors of variables and the observation points (Obs 1–11 corresponding to the data set 2023–2013).

PC3 and PC4 provide additional details related to the oxygen-pollution balance and the differentiation between sources of ammonia and phosphorus pollution.

The PCA confirms that variables such as COD-Cr, BOD₅, total N, P-PO₄, and dissolved chromium are key parameters for water quality monitoring and sample differentiation. PCA allowed the identification of the primary sources of variability in river water quality, specifically nutrients and heavy metals (F1), as well as organic and nitrogen loading (F2). Secondary factors highlighted the role of dissolved oxygen, transformations of nitrogen forms, and the influence of metals on the ecological balance. This approach enables a clear differentiation of pollution sources and provides a solid foundation for effective water quality management.

As a complementary statistical method to the PCA, we also used the k-means clustering. The aim was to address one PCA limitation, as it reveals how variables correlate with each other and which ones are most important for variability; however, it does not automatically separate the samples into distinct groups. Mahanty et al. considered that excluding only one water parameter from the PCA analysis could significantly influence the quality marks [56]. The situation was also encountered in our case due to the lack of results for some attributes considered. The K-means was chosen since it focuses on clustering water samples and directly shows where there are differences between them, while PCA only reduces the data and highlights the important variables. Celestino et al. considered in their study the opportunity to link these two statistical approaches [57]. Other researches underline the feasibility of different clustering methodologies for environmental conditions studies [58,59,60].

Initially, the simulation considered two to six possible groups. Based on Silhouette scores (with the highest value), the optimal number of clusters was determined to be three. Such a result leads to the conclusion that the water samples have been divided into three distinct groups, each representing a different type of quality or ecological status.

Table 5. Distances between the central objects.

	2023	2021	2016
2023	0	68.088	249.600
2021	68.088	0	184.080
2016	249.600	184.080	0

contains the matrix of distances between three observations (Obs1-2023, Obs3-2021, Obs8-2016) clustered.

Between the observations made in 2023 and 2021, the smallest distance indicates that the samples are most similar, forming a typical cluster. The values of the water parameters (BOD₅, COD-Cr, N, P, etc.) are comparable, likely originating from areas with similar environmental conditions. The same type of pollution could have characterized the considered years.

The year 2016 is very different from both 2023 and 2021. The situation could be considered as a separate cluster, with a different set of conditions (perhaps industrial influence, high metal concentrations, or pronounced nutrient changes).

The characterization of “pervasive diffuse pollution” in the Timiș River primarily reflects nutrient and detergent inputs from agriculture and rural settlements dispersed across the basin. Fertilizer application, livestock farming, and unregulated septic systems are known drivers of elevated nitrate (NO₃⁻), ammonium (NH₄⁺), and phosphate (PO₄³⁻) concentrations, which were consistently observed during the monitoring period. Unlike point-source discharges (e.g., municipal wastewater or industrial effluents), which are spatially concentrated and often episodic, diffuse pollution exerts a chronic and spatially widespread influence, maintaining pollutant levels across years. Riverbank height modulates the impact of these pollution types differently: lowered or eroded banks enhance lateral connectivity with adjacent fields, promoting the transfer of nutrients and agrochemicals during rainfall and runoff events, thereby amplifying diffuse pollution. In contrast, higher or stabilized banks reduce such exchanges but accentuate the influence of point sources, as effluents remain concentrated in the main channel with limited dilution in the floodplain. This dual role highlights how subtle morphological adjustments—such as the ~11 cm elevation fluctuation recorded between 2020 and 2025—can alter the balance between diffuse and point-source contributions to pollutant loads. Thus, integrating riverbank morphology into pollution assessments is essential for disentangling the mechanisms that sustain chronic contamination despite regulatory interventions.

Being aware of the method limitations, it was considered due to its advantages. It enables us to create a quick visualization of relationships between multiple parameters over a medium period and the dynamism of the attributes [61]. The feasibility of correlation methods in water quality assessment was previously presented in other relevant investigations [62,63]. In our case, the various interrelationships were established. Based on the results obtained, we can highlight groups of dependent variables, including nutrients, organic pollution, and metals. The outcomes presented could serve as a basis for more advanced methods in future investigations. As Attia et al. underlined in their investigation, the simultaneous analysis of water characteristics can be valorized through integration into predictive models that may serve as a basis for aquatic quality control, supervision [64,65], and hazard analysis.

The observed riverbank elevation changes of approximately 11 cm influence hydrological dynamics by modifying channel confinement, flow velocity, and sediment transport. Lowered banks reduce flow velocity, enhancing deposition of fine sediments and associated pollutants, whereas elevated or stable banks maintain higher flow energy, promoting downstream transport. These morphodynamic changes also affect floodplain interactions, with overbank flow providing temporary storage and natural attenuation of contaminants.

Between 2020 and 2025, the Timiș River experienced an average change in riverbank elevation of approximately 11 cm, reflecting increased erosion, sediment resuspension, and altered hydrodynamic conditions. Such geomorphological changes can directly influence nutrient mobilization, heavy metal release, and redox dynamics, thereby shaping the correlation patterns observed among water quality parameters [5,27,28,33,66].

Positive correlations align closely with the geomorphological context. The robust association between total nitrogen and nitrate (r = 0.97) suggests oxidative stability and nitrate dominance, a condition that may be intensified by enhanced surface–groundwater connectivity and accelerated leaching from fertilized soils following bank lowering and erosion events [32]. Similarly, the strong correlation between total phosphorus and orthophosphate (r = 0.89) indicates a predominance of active and bioavailable phosphorus pollution. Bank erosion likely exposes phosphorus-rich soils and increases sediment–water interactions, contributing to elevated PO₄³⁻ concentrations [28]. The Zn–COD-Cr correlation (r = 0.79) further supports the role of sediment resuspension, as destabilized banks can release legacy Zn and organic matter into the water column, generating a mixed organic–metallic pollution signal [27,33]. The association between BOD₅ and conductivity (r = 0.76) may also be linked to geomorphological change, as bank collapse mobilizes both dissolved ions and organic matter into the river system.

Negative correlations can likewise be interpreted through the lens of riverbank change. The strong inverse relationship (r = –0.90) between Cu and NH₄⁺ may reflect that Cu released from eroding sediments exerts toxicity on microbial nitrogen cycling, accelerating the conversion of NH₄⁺ to NO₃⁻ under oxygenated conditions [48]. The moderate negative correlation between Cu and orthophosphate (r = –0.61) suggests that microbial mineralization of phosphorus is inhibited in Cu-enriched zones, resulting in P retention in sediment-bound forms rather than its release into the water column [27,66]. Finally, the DO–COD-Cr relationship (r = –0.56) illustrates the combined effect of increased organic matter input from eroded banks and associated oxygen depletion, a trend likely exacerbated during low-flow summer conditions when hydraulic retention times increase [67].

In summary, the observed 11 cm decline in riverbank elevation is not an isolated geomorphological observation, but rather a driver of biogeochemical responses. It amplifies the mobilization of nutrients (NO₃⁻ and PO₄³⁻) and metals (Cu and Zn), reshapes microbial pathways, and reinforces the oxidative stability of the system while simultaneously stressing oxygen availability. These findings underscore the importance of integrating geomorphological and biogeochemical monitoring to capture the combined physical and chemical processes that shape water quality in the Timiș River.

3.3. Spatiotemporal Variation of the Global Pollution Index in the Șag–Grăniceri Section

Considering the method for the calculation of the overall pollution index based on the methodology described by Zaharia [40], the variation of the calculated general pollution indexes over the studied period for the Timiș River, and section Șag–Grăniceri is presented in Figure 5.

The index values exhibit moderate variability, highlighting differences in anthropogenic pressure along the watercourse and its tributaries.

Several sampling points consistently display $I_{GP}^{*}$ values in the range of 2.0–2.5, indicating a moderate degree of pollution. These sites are likely influenced by constant pollution sources, such as municipal discharges, industrial effluents, and agricultural runoff, particularly in more densely populated or intensively farmed sub-basins.

The observed variability of $I_{GP}^{*}$ values across sites likely reflects a combination of broadscale and local drivers. Localized variations, such as effluent discharges from municipal or industrial sources in urbanized reaches, and small-scale differences in watershed characteristics (land use, bank stability, drainage density), can produce strong local departures from the overall trend.

3.4. Longitudinal Assessment of Pollution Trends in the Timiș River—Șag–Grăniceri Section (2013–2023) with Consideration of Morphological and Hydrological Influences

An eleven-year evaluation (2013–2023) of the general pollution index ($I_{GP}^{*}$) for the Timiș River at the Șag–Grăniceri section reveals a relatively stable range of values between 1.98 and 2.56, with an average of 2.19. These values indicate moderate but persistent anthropogenic pressure on the water body. The index fluctuated around a threshold that suggests notable ecological stress without reaching critical degradation.

The maximum $I_{GP}^{*}$ value of 2.56 (recorded in 2016) may be explained by the occurrence of a transient pollution event, likely related to a combination of industrial discharges, municipal wastewater inputs, and increased diffuse pollution from agriculture during high-flow periods. This spike could be attributed to temporary lapses in treatment infrastructure or rainfall-driven runoff that exceeded dilution capacity. The following year (2017) marked a notable improvement, with the index dropping to 1.98, the lowest value in the series. This suggests a short-term recovery, possibly due to administrative enforcement actions, reduced pollutant loading, or favorable hydrological conditions.

From 2018 to 2023, the $I_{GP}^{*}$ values exhibit low inter-annual variability, oscillating narrowly between 2.02 and 2.26. Such stability implies the presence of steady pollution sources, most likely diffuse inputs (e.g., nutrients, detergents, and micro-pollutants) and insufficiently treated wastewater, which persist despite regulatory frameworks. This constancy reflects a state of chronic pollution, insufficiently mitigated by either natural river self-purification or infrastructural upgrades.

The observed 11 cm fluctuation in riverbank elevation between 2020 and 2025 could have influenced hydrodynamics and pollutant distribution through multiple interrelated mechanisms. Lowered or eroded bank sections reduce channel confinement, slow flow, and enhance sediment deposition, thereby increasing the retention of sediment-bound pollutants, such as nutrients (N, P species), organic matter, and trace metals (As, Cr, Cu, Zn) [6,7,8,28]. This can elevate BOD₅ and COD-Cr levels locally. Elevated or stabilized banks increase flow velocity, reduce sediment deposition, and promote downstream transport of contaminants, resulting in lower localized pollutant concentrations. Additionally, micro-pools and backwater zones adjacent to variable banks create localized deposition areas, concentrating pollutants even when basin-scale indices, such as $I_{GP}^{*}$, remain stable. Morphological changes also impact oxygen transfer, turbulence, and overbank interactions, thereby modulating biogeochemical processes that regulate nutrient cycling and organic matter decomposition, as evident in dissolved oxygen dynamics. While the 11 cm change is modest at the basin scale, these mechanisms tentatively suggest that topographic variations can locally modulate flow patterns, sediment transport, and pollutant fate, highlighting the importance of incorporating riverbank morphology into water quality assessments and management strategies.

Notably, the observed 11 cm difference in bank altitude (relative to sea level) between 2020 and 2025 (Figure 5) did not coincide with a measurable change in the general polluting index, suggesting that long-term pollutant loads from diffuse and point sources dominate water quality trends, while the influence of modest geomorphological fluctuations is not readily detectable at the basin scale. These observations reinforce the exploratory nature of the bank height–pollution relationship and the need for higher-resolution, co-located measurements to test coevolutionary dynamics more robustly.

3.5. Prognostic Interpretation and Retrospective Dynamics

Topographic measurements carried out in 2020 and 2025 at the Șag section of the Timiș River indicated riverbed elevation variations of up to 11 cm. These morphological changes were analyzed in relation to the general pollution index ($I_{GP}^{*}$) to explore potential correlations between physical alterations in the riverbed and water quality dynamics [6,8,28].

Fluctuations in riverbed elevation of this magnitude can potentially influence the $I_{GP}^{*}$ indirectly by modifying hydraulic parameters, such as flow velocity and turbulence, which in turn affect the dilution capacity of pollutants, the sedimentation and remobilization of contaminated particulates, and the oxygenation potential of the water column through processes of aeration or stagnation.

Despite these plausible mechanisms, the recorded $I_{GP}^{*}$ values remained relatively stable across the studied period. This consistency suggests that the observed 11 cm morphological fluctuation did not exert a measurable immediate effect on the overall pollution status of the water body. It is therefore likely that persistent pollution sources (e.g., untreated or partially treated effluents, diffuse agricultural inputs) exert a dominant influence on water quality, overshadowing the effects of localized topographic variability.

From a retrospective standpoint, the year 2013 already reflects a moderately impacted state, with an $I_{GP}^{*}$ of 2.26, suggesting that chronic stress was already established at the beginning of the observation period. Given the stable trend that followed, it is plausible that the river has been subjected to long-term, low-to-moderate intensity pollution for well over a decade.

The absence of a decreasing trend in $I_{GP}^{*}$, despite infrastructural and policy developments during the EU Water Framework Directive implementation cycle, may indicate one or more of the following:

• Incomplete or uneven modernization of wastewater treatment facilities in rural and peri-urban settlements;
• Persistent agricultural runoff, especially under intense rainfall or irrigation events.
• Lack of enforcement or limited monitoring of industrial discharges in some segments.
• Sediment legacy contamination slowly releases pollutants during flow events or morphological changes.

Suppose current patterns persist, and no additional measures are taken to reduce nutrient and pollutant loads from point and non-point sources. In that case, the prognosis suggests that the $I_{GP}^{*}$ will likely remain in the 2.0–2.3 range, reflecting ongoing ecological stress and limited river resilience. Conversely, the introduction of targeted interventions, such as nature-based solutions, sediment remediation, or stricter effluent control, could contribute to reductions, improving ecological status, and aligning with the objectives of the Water Framework Directive (WFD).

To mitigate both point and non-point source pollution in the Timiș River Basin—particularly in areas with limited or outdated wastewater treatment infrastructure—this study recommends the implementation of nature-based solutions specifically designed to stabilize riverbanks and reduce the mobility of pollutants. Hybrid constructed wetland systems strategically placed along critical tributaries and drainage channels would not only treat incoming waters but also limit lateral erosion and bank destabilization, thereby reducing sediment resuspension and the release of legacy contaminants stored in bank sediments.

Unlike conventional systems, these wetlands directly address the geomorphological drivers of pollution identified in this study. By buffering runoff before it reaches the main channel and enhancing bank stability, they decrease the influx of nutrients (especially NO₃⁻ and PO₄³⁻) and trace metals (Cu and Zn) that are otherwise mobilized during high-flow or erosion events.

The hybrid configuration—combining vertical and surface flow wetlands—maximizes both aerobic and anaerobic pathways for pollutant removal (sedimentation, adsorption, microbial degradation, and plant uptake). The use of native macrophytes, such as Phragmites australis and Typha latifolia, not only enhances nutrient and heavy metal uptake but also reinforces riverbank soils through root networks, thereby further reducing bank erosion.

Integrating these systems into buffer strips or ecological corridors adjacent to the riverbanks simultaneously strengthens multiple ecosystem functions: controlling erosion, trapping sediments, attenuating peak flows, and creating conditions for pollutant transformation and retention before contaminants enter the river.

In this context, river resilience is understood as the capacity of the Timiș River to maintain or recover acceptable water quality (e.g., $I_{GP}^{*}$ < 2.5, DO ≥ 6 mg L⁻¹) following disturbances such as diffuse runoff events, local effluent discharges, or bank morphological changes. This resilience was assessed indirectly by analyzing the temporal stability of key parameters (BOD₅, DO, nutrients, and heavy metals) over the 11 (eleven) years, the speed of recovery after anomalous years (2014 and 2020), and the coherence between bank elevation changes and water quality trends. Although no formal resilience index was computed, this operational definition highlights sub-basins where water quality recovered quickly compared to areas where it remained degraded, guiding priority zones for intervention.

4. Conclusions, Limitations, and Future Perspectives

The longitudinal assessment of the general pollution index ($I_{GP}^{*}$) for the Timiș River at the Șag–Grăniceri section over the 2013–2023 period suggests a persistently moderate level of pollution, with index values ranging from 1.98 to 2.56 and an average of 2.19. This stability may indicate that the river is subject to chronic and diffuse anthropogenic pressures, primarily from municipal effluents, agricultural runoff, and possibly underregulated industrial discharges.

Despite minor fluctuations in riverbed elevation—up to 11 cm between 2020 and 2025—pollution levels remained largely unaffected, which could underscore the dominance of consistent pollutant sources over transient hydromorphological factors.

Correlation analysis revealed associations among several physicochemical parameters. The strong relationship between BOD₅ and conductivity appears to point to organic pollution of urban or industrial origin. COD-Cr was highly correlated with Zn and Cu concentrations, which may suggest industrial discharges. The significant association between total nitrogen and nitrate (N-NO₃⁻) further supports the likelihood of agricultural influences on nutrient loading.

This study is based on annual water quality measurements (n = 11) and a discrete morphometric dataset comprising riverbank height measurements at six cross-sections in 2020 and 2025. We acknowledge that this represents limited temporal and spatial resolution. However, this dataset provides a useful basis for exploring long-term trends in riverine pollution, detecting inter-annual variability, and tentatively assessing the possible influence of riverbank morphology on pollutant retention. The consistent sampling methodology, replication across two monitoring sections, and inclusion of 17 physicochemical parameters support a preliminary but informative evaluation of basin-scale water quality and geomorphological controls.

Future studies could strengthen these findings by enhancing temporal and spatial resolution by incorporating more frequent water sampling (e.g., monthly or seasonal) and additional morphometric surveys using high-resolution techniques such as LiDAR or UAV-based DEMs. These enhancements would enable a finer-scale assessment of pollutant mobilization, capture short-term hydrological and sediment dynamics, and improve the predictive modelling for river management.

Despite the insights gained, the study has several limitations. The dataset comprises a relatively small number of observations, collected annually over eleven years, which restricts the statistical robustness of trend analysis. In some years, certain parameters (e.g., detergents, phenols, and arsenic) were not determined, limiting the comprehensiveness of the evaluation. Furthermore, the reliance on linear correlation methods means that they may not fully capture complex or nonlinear relationships among pollutants. Due to the study’s localized scope and limited temporal resolution, the findings should not be generalized without further validation.

Future research should aim to expand the spatial and temporal resolution of water quality data. This includes multi-point sampling along the river, seasonal measurements, and integration of biological indicators such as macroinvertebrates or phytoplankton to validate the ecological impact of chemical pollutants. Incorporating advanced statistical and machine learning models—such as neural networks, Random Forest, or XGBoost—could help improve the prediction and interpretation of water quality dynamics.

Emerging technologies offer promising opportunities for enhancing monitoring and management. The use of GIS tools can enable spatial mapping of pollution sources and their correlation with land use patterns, while IoT-based sensor networks can support real-time data acquisition and automatic alerts for regulatory exceedances. AI-assisted platforms could facilitate early detection of pollution trends and adaptive response strategies.

Furthermore, knowledge from state-of-the-art interventions—such as micro-nano bubble technology, as reported by Wu et al. [68]—can inform the development of effective local remediation solutions. Intervention studies should also assess the efficacy of vegetated buffer zones, construct wetlands, and upgraded wastewater treatment facilities in reducing pollutant loads.

While this study identifies several advanced investigative approaches, its scope was deliberately focused on establishing a long-term baseline and identifying dominant pollution patterns, which will serve as a foundation for future in-depth analyses.

Despite efforts under the EU Water Framework Directive, the lack of a clear downward trend in $I_{GP}^{*}$ over the study period suggests possible gaps in implementation and enforcement. Moving forward, a strategic shift from reactive to preventive management appears essential. This includes targeted investments in pollution control infrastructure, regulation of non-point agricultural sources, and restoration of riparian ecosystems. Although several advanced investigative tools and techniques—such as high-frequency and multi-site sampling, biological indicators, real-time IoT sensors, machine learning models, and riparian restoration measures—are identified in this paper, they were intentionally discussed as future directions rather than applied within the present work. The primary aim of the present study was to establish a robust long-term baseline of water quality and geomorphological conditions for the Timiș River and to identify dominant pollution patterns using the currently available data. By clarifying this baseline and highlighting possible interdependencies between bank elevation and pollutant dynamics, the study offers preliminary groundwork upon which these more sophisticated approaches can be meaningfully implemented and evaluated in subsequent research or management programs.