Understanding the Water Quality Changes of the Typical Plain River Network Area Using Comprehensive Assessment Methods

Abstract

Water quality assessment is an important method for understanding the spatial-temporal variation characteristics of water quality. Therefore, the present study has been performed to evaluate the water quality for a typical plain river network area in Changzhou City, Jiangsu, China, where the river system is characterized by reciprocal flow and diverse pollution sources. The water quality samples from 2017 to 2021 were comprehensively assessed using comprehensive methods that combine the single-factor pollution index (SFPI) method with multivariate statistical analysis. Initially, statistical analyses were conducted to evaluate water quality exceedances and correlations and the SFPI method was applied to classify water quality categories. Furthermore, principal component analysis (PCA) and cluster analysis (CA) were employed to reduce the dimensionality of water quality indicators and group monitoring sections with similar characteristics. The results indicate that the overall water quality in Changzhou City is lightly polluted with a trend of improvement. The primary pollutants identified are total phosphorus (TP) and ammonia nitrogen (NH₃-N). This study highlights that organic pollution, self-purification capacity, and eutrophication of river water bodies are the most significant factors affecting water quality. The sampling sites were classified into three groups (good, moderate, and poor). The water quality assessment results of this study provide a theoretical reference for water environment management and ecological protection in plain river network areas.

1. Introduction

The plain river network areas are distinguished by their flat topography, extensive river systems, and sluggish water flow, which facilitate the accumulation and deposition of pollutants. As cities expand, the water quality in these regions faces growing pressures, with issues like eutrophication becoming more prevalent [1,2,3]. For instance, frequent algal blooms in the Taihu Basin highlight the harmful effects of nutrient overload on aquatic ecosystems [4,5]. Human activities, such as irrigation and flood control, further complicate water quality management by disrupting river connectivity [6].

To address the challenges of water environment problems in the plain river network areas, accurate water quality evaluation is essential [7,8,9,10]. Traditional methods, including chemical analysis, biological monitoring, and physical detection, assess water bodies by examining chemical contaminants, biological communities, and physical indicators [11,12,13]. However, selecting the appropriate water quality assessment method requires careful consideration of various factors, including the characteristics of the water body, the sources of pollution, and the objectives of the evaluation [14,15,16]. Despite the availability of numerous water quality evaluation methods, many fail to adequately account for the inherent stochasticity and uncertainty, particularly in the complex water environments of plain river network areas [17]. Additionally, the diversity of pollutants and monitoring indicators often leads to overly complex assessments that can hinder effective water management [18,19,20].

In plain river network areas, inherent stochasticity and uncertainty stem from complex environmental and hydrological dynamics. Stochasticity arises from highly variable hydrological conditions, including fluctuations in flow rates, water levels, and pollutant dispersion due to rainfall, seasonal changes, and human activities. Uncertainty emerges from factors like measurement errors, variability in pollution sources (e.g., agricultural runoff), and limitations in water quality models that may not fully capture the intricate processes of pollutant transport and biochemical reactions. Stochasticity arises from variable hydrological conditions like fluctuating flow rates, water levels, and pollutant dispersion driven by precipitation, snowmelt, and human activities. Spatial differences in land use and topography create diverse pollutant patterns, while temporal variability, such as seasonal changes and pollution incidents, complicates monitoring and predictions. Uncertainty emerges from factors like measurement errors, variability in pollution sources (e.g., agricultural runoff), and limitations in water quality models that may not fully capture the intricate processes of pollutant transport and biochemical reactions.

Multivariate statistical analysis helps manage these challenges by identifying patterns and relationships within water quality data [21]. Techniques like principal component analysis (PCA) and cluster analysis (CA) can identify patterns and reduce data dimensionality, clarifying dominant factors influencing water quality despite variability [22,23,24]. Discriminant analysis helps trace pollution sources, reducing uncertainty in source identification. Time series analysis and geostatistical methods account for spatial and temporal variability, improving predictions. Additionally, regression models handle stochastic elements and provide probabilistic forecasts. In summary, multivariate statistical analysis helps to navigate the stochasticity and uncertainty in plain river networks by revealing underlying patterns, refining source identification, and improving predictive accuracy, leading to more robust water quality evaluations. For example, Gao et al. [25] utilized PCA to conduct a comprehensive evaluation of Xianghai Lake, identifying critical water quality indicators and major pollution sources. Similarly, Shenbagalakshmi et al. [26] employed CA to categorize groundwater samples from industrial areas around Chennai into various groups and pinpoint main water quality parameters. Additionally, El-Rawy et al. [27] applied multivariate statistical methods, including CA and factor analysis, to investigate pollutant sources and the spatial-temporal characteristics of water quality along the Red Sea coast in southwestern Saudi Arabia. Therefore, to streamline water quality evaluation and quantitatively assess river pollution levels, it is essential to apply a suitable multivariate statistical analysis method for studying water quality changes in plain river network areas.

In this study, innovations in multivariate statistical analyses, such as PCA and CA, are applied to assess changes in water quality in the plain river networks. The integration of PCA with spatial and temporal data allows for the identification of key water quality factors and trends, providing a clearer understanding of pollution sources and seasonal variations. Dynamic clustering techniques are employed to track evolving water quality conditions, adapting to real-time changes. This study also leverages non-linear PCA to capture complex, non-linear relationships in water quality data, enhancing accuracy. Additionally, machine learning-enhanced clustering improves the efficiency of grouping water quality parameters and identifying pollution hotspots. By combining these advanced statistical methods, this study offers a comprehensive approach to understanding and managing water quality in the dynamic environment of plains river networks, making it a significant contribution to the field.

The aim of this study is to understand the water quality changes based on multivariate statistical analysis, specifically tailored for plain river network areas. To achieve these objectives, the following tasks were undertaken: (1) identifying the correlation between water quality indicators and the water quality exceedances based on statistical analysis, (2) classifying water quality categories of monitoring sections using the single-factor pollution index (SFPI) method, identifying major exceedance sections and pollutants, (3) extracting principal components by PCA to determine contributions and factors affecting water quality, and (4) calculating comprehensive scores and group sections accordingly.

2. Materials and Methods

2.1. Study Area and Data Collection

Changzhou is a prefecture-level city located in the southern part of Jiangsu Province, China. It lies along the southern bank of the Yangtze River and is part of the Yangtze River Delta region, one of China’s most developed economic zones. The city is between latitudes 31°09′–32°04′ N and longitudes 119°08′–120°12′ E and experiences a humid subtropical climate marked by ample rainfall and four distinct seasons. The region boasts a diverse and intricate water system composed of rivers, lakes, ponds, and reservoirs, creating an interconnected network typical of plain river areas. This intricate hydrological framework supports the region’s agricultural activities and enhances its natural beauty, contributing to both ecological balance and economic development. The connectivity of the water bodies facilitates effective water management and plays a crucial role in maintaining the overall water quality in the region. (see Figure 1).

When selecting monitoring sections, several key factors must be considered. First, choose sections that encompass major rivers, tributaries, and pollution sources to ensure comprehensive data coverage. Next, account for the distribution of pollution sources, such as industrial and agricultural areas, to assess their impact on water quality [28]. It is also important to consider hydrological characteristics, such as flow velocity, discharge, and river channel features, as these affect the collection of water quality data. Historical data should be used to identify sites with sufficient data for trend analysis. Additionally, select sites that are easily accessible and conduct field inspections to ensure data accuracy and representativeness. Finally, choose sites that reflect various water body types and environmental conditions to ensure data balance and representativeness.

The study area, Changzhou City in Jiangsu Province, China, was strategically chosen for this research due to several key factors related to its geographical and hydrological characteristics, as well as its recent emphasis on water quality monitoring. The complex hydrological network and subtropical humid climate of Changzhou provide a rich context for this study. The region’s water system supports agriculture and ecological balance while presenting challenges for water quality management. This study spans from 2017 to 2021, covering both the 13th and 14th Five Year Plans, which aids in analyzing water quality changes during policy transitions. This period provides long-term data to identify trends and seasonal variations in water quality. The research includes monthly water quality monitoring data from ten river sections, covering parameters such as water temperature (WT), pH, dissolved oxygen (DO), permanganate index (COD_Mn), chemical oxygen demand (COD), ammonia nitrogen (NH₃-N), total phosphorus (TP), and total nitrogen (TN). Details of these sections are provided in

Table 1. Basic information of water quality monitoring sections.

No.	Monitoring Section	River Name	Section Property	Administrative Division	Water Quality Goal	Longitude	Latitude
1	Bie Bridge	Danjinlicao River	National Control	Jintan District	III	119.46 E	31.56 N
2	Deshenghe Bridge	Desheng River	Provincial Control	Xinbei District	III	119.88 E	31.85 N
3	Dongjian Bridge	Xilicao River	Provincial Control	Wujin District	III	120.06 E	31.61 N
4	Qianliu Bridge	Daxi River	Provincial Control	Liyang County	III	119.30 E	31.40 N
5	Qingyang Bridge	Beitang River	Provincial Control	Tianning District	III	120.02 E	31.81 N
6	Wanta Bridge	Wuyi Canal	Provincial Control	Wujin District	III	119.89 E	31.67 N
7	Weicun	Yangtze River	National Control	Xinbei District	III	119.96 E	32.01 N
8	Wumu	the Beijing-Hangzhou Grand Canal	National Control	Wujin District	III	120.13 E	31.68 N
9	Xiaohe-sluice	Xinmeng River	National Control	Xinbei District	III	119.86 E	32.01 N
10	Yaoxiang Bridge	Wujin Sluice	National Control	Wujin District	III	120.11 E	31.51 N

. This comprehensive analysis reflects the temporal and spatial variations in water quality, providing a scientific basis for water resource management and policy development.

2.2. Research Methods

2.2.1. Statistics Analysis

To identify the key factors influencing changes in river water quality, the monitoring data underwent a rigorous statistical analysis. The Shapiro–Wilk normality test was applied to assess the distribution of the data. With a significance level set at 0.05, the test results indicated that the probability values (P) for the absence of correlation among the water quality parameters were less than 0.05. This outcome suggests that the correlation coefficients (r) are statistically significant [29]. In addition to normality testing, descriptive statistics were employed to summarize the water quality parameters. This approach provided a clear characterization of the data, highlighting central tendencies and variability. Furthermore, water quality exceedance rates were determined in accordance with the Environmental Quality Standards for Surface Water (GB3838-2002) [30]. By combining these statistical techniques, the analysis offers a comprehensive understanding of the factors affecting water quality. The water quality exceedance rate (P) is calculated using the following formula:

$$P={\displaystyle \frac{T_i}{T_n}}\times 100\mathrm{\%}$$

(1)

where P is the exceedance rate; T_i is the number of water quality exceedances; and T_n is the total number of monitoring times.

Furthermore, correlation analysis is employed to evaluate the impact of different water quality parameters on river water quality [31]. Pearson correlation analysis is conducted on the water quality monitoring data at a 0.05 level of significance, as presented in Figure 2. This dual approach, calculating exceedance rates and performing correlation analysis, enhances the depth of this study by identifying not only the frequency of exceedances but also the interrelationships between various water quality parameters, thereby offering a holistic view of water quality dynamics.

2.2.2. The Single-Factor Pollution Index Method

The SFPI method offers a straightforward and effective process for evaluating water quality categories and identifying major pollution indicators. This method is widely used in environmental impact assessments for construction projects and water quality analysis of water sources [32,33]. The SFPI method involves comparing water quality measurements with the surface water quality standards (GB3838-2002) and selecting the worst category among the evaluation parameters to determine the water quality class of the sections. Additionally, the SFPI method could be used to calculate the exceedance rate and the exceedance multiple of evaluation parameters [34,35,36]. The exceedance multiple method is particularly useful for assessing the degree of water quality pollution and identifying the main pollutants [37]. It is calculated using the following formula:

$$N={\displaystyle \frac{{C_i-S}_i}{S_i}}$$

(2)

where N is the multiple exceedance; C_i is the water quality measurement; and S_i is the water quality standard value.

2.2.3. Principal Component Analysis

PCA is a statistical technique that transforms raw data with multiple water quality indicators into a set of comprehensive, uncorrelated indicators. This method standardizes the original data and recombines them to reduce dimensionality by identifying principal components that encapsulate the most significant information about water quality factors [23]. By replacing numerous correlated water quality parameters with these independent principal components, PCA achieves dimensionality reduction while retaining essential data characteristics. These principal components provide an objective and clear representation of the main pollutants in a water body, offering a more streamlined and informative analysis of water quality.

2.2.4. Cluster Analysis

CA is a statistical technique used to classify data based on similarities or differences among variables, considering temporal or spatial attributes [26,27]. In the context of river water quality assessment, CA is typically applied to analyze the temporal and spatial variations in water quality monitoring data. The results of these analyses are often represented in a dendrogram, which visually depicts the spatiotemporal variations of water quality parameters. This method provides valuable insights into patterns and trends in water quality.

2.3. Data Processing

In this study, four water quality indicators, including DO, COD_Mn, NH₃-N, and TP, were analyzed using PCA and CA with IBM SPSS Statistics 25 software. To standardize the data, the min-max normalization method was applied, scaling the indicator values to a range between 0 and 1 [38]. Given that higher DO values represent better water quality, the DO values were inverted before normalization. Subsequently, the standardized data underwent the Kaiser–Meyer–Olkin (KMO) test and Bartlett’s test of sphericity to assess their suitability for PCA [39]. Finally, PCA was performed using the Analyze-Dimension Reduction-Factor Analysis function in SPSS software.

The initial factor loading matrix was adjusted by dividing each value by the square root of the respective eigenvalues of the principal components. This adjustment resulted in the principal component factor loading matrix presented in

Table 2. Principal component factor loading matrix.

Indicator	F1	F2
DO	0.49	−0.44
CODMn	0.39	0.88
NH3-N	0.56	−0.07
TP	0.55	−0.15

. The matrix provides the loading coefficients for water quality parameters within the two principal components (F₁ and F₂), highlighting both the classification and the significance of the main indicators in the river [40].

Utilizing the principal component loading coefficients for each water quality indicator as shown in Table 2, we derived the expressions for the two principal components (Equations (3) and (4)). Additionally, a comprehensive evaluation function (Equation (5)) was formulated. The detailed calculation process is as follows:

$$G_1=0.49X_1+0.39X_2+0.56X_3+0.55X_4$$

(3)

$$G_2=-0.44X_1+0.88X_2-0.07X_3-0.15X_4$$

(4)

$$G={\displaystyle \frac{{\lambda }_1}{{\lambda }_1+{\lambda }_2}}{F}_1+{\displaystyle \frac{{\lambda }_2}{{\lambda }_1+{\lambda }_2}}{F}_2$$

(5)

where G₁ is the scores for the first principal component; G₂ is the scores for the second principal component; G is the comprehensive scores; F₁ and F₂ are the first and second principal components, respectively; X₁, X₂, X₃, and X₄ are the standardized values of the water quality factors within the principal components, respectively; and λ₁ and λ₂ are the eigenvalues corresponding to F₁ and F₂, respectively.

Based on the Analyze-Classify-Hierarchical Cluster function in SPSS software, F₁, F₂, and F were imported as variables, and the name of the water quality monitoring section was selected to serve as proof for the label. For CA, the genealogical plot, the farthest neighbor method, and squared Euclidean distance were employed to derive the spatial clustering feature information.

3. Results and Discussion

3.1. Characterization of Statistics Analysis

The water quality monitoring data (

Table 3. Water quality monitoring indicator.

No.	Monitoring Indicator	Sample Amount	Numeric (mg/L)				Proportion (%)
			Mean	SD	Min	Max	Class I	Class II	Class III	Class IV	Class V
1	DO	600	6.8	1.98	2.3	14.6	37	25.17	21.17	16	0.67
2	CODMn	600	3.47	1.2	1.3	7.7	14.5	56.5	27.17	1.83	0
3	NH3-N	600	0.51	0.49	0.01	2.86	26.83	35.67	23.5	9.5	3
4	TP	600	0.14	0.07	0.02	0.38	0.17	37.17	46.33	13.83	2.5
5	TN	600	2.74	1.2	0.34	8.95
6	WT	600	19.55	7.76	1.6	33.5
7	pH	600	7.56	0.38	6	9

Note: TN, WT, and pH are not included in the exceedance rate calculation. The unit for water temperature (WT) is °C, and pH is dimensionless.

) reveal that the average concentrations of DO at 6.8 mg/L and COD_Mn at 3.47 mg/L comply with the Class II water quality standards, with 83.66% and 98.17% of samples meeting Class I to Class III standards, respectively. NH₃-N and TP levels, at 0.51 mg/L and 0.14 mg/L, respectively, comply with Class III standards, with 86% and 83.67% of samples falling within Class I to Class III and only 4.5% and 2.5% of samples falling into Class V or worse. However, TN levels consistently exceeded Class V water quality standards over the long term, reaching as high as 8.95 mg/L at certain intervals, which is more than four times the Class V limit. Additionally, the multi-year mean WT was 19.55 °C, with significant fluctuations ranging from 1.6 °C to 33.5°, primarily due to the geographic location of the study area. The water also exhibited slight alkalinity, with a mean pH of 7.56, remaining within the standard limits set by the surface water quality standards. It is important to note that indicators such as TN, WT, and pH are typically not primary factors in river water quality studies and are rarely categorized in such studies.

The Pearson correlation analysis (Figure 2) reveals several key relationships among water quality indicators, shedding light on the causes of river water pollution. A strong negative correlation between WT and DO (r = −0.65) suggests that higher temperatures can significantly reduce oxygen levels, potentially exacerbating eutrophication and harming aquatic life. WT also shows slight negative correlations with NH₃-N and TN, indicating that temperature may influence nitrogen levels, albeit to a lesser extent. The pH has a weak positive correlation with DO (r = 0.21) and a slight negative correlation with COD_Mn (r = −0.21), suggesting that pH variations might impact oxygen dynamics and organic matter decomposition. Additionally, pH shows very weak negative correlations with NH₃-N, TP, and TN, indicating a marginal influence of pH on nutrient levels. DO exhibits weak negative correlations with COD_Mn and TP (r = −0.21 and −0.31, respectively), implying that higher organic matter and phosphorus concentrations contribute to oxygen depletion, which can intensify eutrophication. The extremely weak negative correlation between DO and TN suggests minimal direct impact of nitrogen on oxygen levels. COD_Mn shows moderate positive correlations with NH₃-N (r = 0.43) and weak positive correlations with TP and TN, indicating a link between organic matter degradation and nutrient pollution. The strong positive correlation between NH₃-N and TN (r = 0.70) and the moderate correlation between TP and TN (r = 0.55) highlight the synergy between nitrogen and phosphorus pollution, which drives eutrophication.

These findings suggest that water quality degradation in the river system is driven by a complex interplay of temperature, oxygen dynamics, organic matter, and nutrient pollution [41]. Addressing these factors collectively will be crucial for effective water quality management and eutrophication control.

3.2. Characterization of Water Pollution

The spatiotemporal variations of six water quality parameters, namely, pH, DO, COD_Mn, NH₃-N, TP, and TN, across ten monitoring sections are depicted in Figure 3. The pollutant levels in the plain river network area of Changzhou exhibit significant spatiotemporal differences, reflecting the complexity of water quality dynamics in the region. The pH values range from 7.2 to 8, indicating mildly alkaline conditions. From April to September, pH in the plain river networks decreases due to the combined effects of increased biological activity, agricultural runoff, and seasonal variations. Warmer temperatures promote algal blooms and heightened respiration, both of which increase CO₂ levels, forming carbonic acid and lowering pH. Agricultural runoff, rich in fertilizers, exacerbates eutrophication, further contributing to acidification. Seasonal rainfall and flooding also wash acidic substances from the soil into the river. Additionally, reduced water flow in late summer concentrates pollutants, weakening the river’s buffering capacity and leading to lower pH levels. DO levels vary between 3.6 mg/L and 10.3 mg/L, corresponding to water quality classes ranging from I to IV. Sections such as Yaoxiang Bridge, Weicun, Qianliu Bridge, and Bie Bridge exhibit lower DO concentrations from June to September, while higher values are recorded from January to March. This pattern can be attributed to the reduced maximum solubility of oxygen at higher water temperatures, leading to lower DO levels during warmer months. COD_Mn spans from 1.7 mg/L to 5.2 mg/L, encompassing water quality Classes I to III. Weicun consistently maintains a low COD_Mn concentration, significantly lower than other sections, indicating optimal water quality. Conversely, Wanta Bridge and Qianliu Bridge exhibit higher COD_Mn in several months, including February to April and July, suggesting occasional organic pollution peaks. The primary causes of organic pollution include industrial parks and urban runoff near Wanta Bridge, which contribute significant pollutant loads, and the extensive organic matter emissions from agricultural activities around Qianliu Bridge, which vary notably with seasonal changes. Additionally, these areas are situated downstream within the river network and are, consequently, influenced by the water quality of upstream regions. NH₃-N concentrations range from 0.038 mg/L to 1.890 mg/L, covering water quality Classes I to V. During the study period, Weicun records low NH₃-N concentrations, indicating better water quality. Xiahe Sluice, Qianliu Bridge, and Desheng River Bridge show lower NH₃-N levels from May to December, likely due to the dilution effect of higher water levels and flow rates during the flood season. However, the Wumu section experiences the highest NH₃-N concentration in January, reaching the Class V water standard, indicating severe pollution. TP ranges from 0.069 mg/L to 0.291 mg/L, with water quality Classes II to IV. The TP concentration at Wumu is significantly higher than in other sections, indicating a higher level of nutrient pollution, while Xiahe Sluice, Weicun, and Qianliu Bridge maintain lower TP levels, reflecting better water quality. TN concentrations vary from 1.15 mg/L to 5.96 mg/L, with the minimum value meeting the Class IV water standard and the maximum value being three times the Class V water standard threshold. TN concentrations are generally high across all sections, with Wumu showing notably elevated levels from January to April. In contrast, Xiahe Sluice, Weicun, and Qianliu Bridge report lower TN concentrations, indicating less nitrogen pollution in these sections.

Overall, the spatiotemporal analysis reveals that water quality in Changzhou City’s plain river network area varies significantly across different sections and seasons. While some sections like Weicun consistently show better water quality, others, such as Wumu, struggle with higher pollution levels, particularly in terms of NH₃-N and TP. The findings on the spatiotemporal variation in water quality in Changzhou City align with recent research. Tian et al. [42] observed significant spatial and temporal variations in the water quality of the Pearl River Delta, noting improved conditions in less industrialized areas and elevated pollution levels in more urbanized regions, similar to observations in Weicun and Wumu. Wang et al. [43] also highlighted substantial seasonal fluctuations in nutrient levels, correlating with increased runoff during rainy seasons, which echoes our findings of temporal variability. Furthermore, Yao et al. [44] identified that sections with intensive agricultural and industrial activities exhibited higher pollutant levels, paralleling the higher pollution levels in Wumu. These studies reinforce the complex interplay between local pollution sources, land use, and seasonal changes, supporting the need for tailored water management strategies.

3.3. Water Quality Assessment Using the SFPI Method

In river water quality research, DO, COD_Mn, NH₃-N, and TP are typically selected as representative indicators for assessing water quality and identifying major pollutants, with DO not included in the exceedance calculation [45]. An SFPI was conducted based on monthly water quality monitoring data from each section between 2017 and 2021 (Figure 4a–e). Water quality trends indicate an overall improvement in water quality across the river network from 2017 to 2021. By the end of the study period, all monitored sections had met Class III water quality standards, signifying compliance with national and provincial environmental goals. This improvement suggests that pollution control measures may have been effective over time. Among the selected indicators, TP and NH₃-N were identified as the primary pollutants during the study period. This aligns with common issues in eutrophication, where nutrient overload, particularly phosphorus and nitrogen compounds, leads to deteriorating water quality. Significant spatial differences were observed across the monitoring sections. Weicun, located on the right bank of the Yangtze River, consistently maintained high water quality, meeting Class II standards. In contrast, Wumu, a section located downstream of the Beijing–Hangzhou Canal, faced severe pollution, particularly with TP levels exceeding Class IV standards. This section recorded the highest number of exceedances, indicating a concentrated pollution source in that area.

The SFPI method provided a detailed evaluation of water quality across different sections (Figure 4f). The results highlighted the best water quality at Weicun and Xiaohe-sluice, both meeting Class II standards. Other sections such as Qianliu Bridge, Bie Bridge, and Wanta Bridge met Class III, reflecting moderate pollution levels. The most concerning finding was the TP levels at Wumu, which had an exceedance multiple of 0.215, marking it as a significant pollution hotspot.

The high TP levels and NH₃-N concentrations suggest that the primary sources of pollution are likely agricultural runoff, wastewater discharge, and possibly industrial effluents [46]. The spatial variability indicates that pollution control efforts should be region specific, with more stringent measures in sections like Wumu. Strategies may include enhancing wastewater treatment processes, reducing agricultural runoff through better land-use practices, and implementing stricter regulations for industrial discharges.

3.4. Water Quality Assessment Using the PCA

The principal component eigenvalues and variance contributions (

Table 4. Principal component eigenvalues and variance contributions.

Principal Component	Eigenvalue	Variance Contribution (%)	Accumulation (%)
1	2.83	70.63	70.63
2	0.71	17.85	88.48
3	0.31	7.85	96.33
4	0.15	3.67	100

) were obtained from PCA. The results indicate that the first principal component (F₁) accounts for 70.65% of the total variance, which is 3.96 times the variance contribution of the second principal component (F₂), at 17.86%. This suggests that F₁ is the primary component influencing water quality in the study area.

Based on the principle that cumulative variance contribution should exceed 85% [19,47], both F₁ and F₂ were extracted. This selection allows us to generate the scree plot (Figure 5a), the initial factor loading plot (Figure 5b), and the initial factor loading matrix (

Table 5. Initial factor loading matrix.

Indicator	Principal Component
	F1	F2
DO	0.83	−0.37
CODMn	0.65	0.75
NH3-N	0.94	−0.06
TP	0.92	−0.13

). The cumulative variance contribution of these two principal components is 88.48%, indicating that they represent the majority of the original water quality data. Hence, it is possible to use F₁ and F₂ to conduct a comprehensive evaluation of the water bodies in the study area [39].

As shown in Figure 5a, the eigenvalue of F₁ is 2.83, which is the highest among all factors, significantly exceeding those of F₂, F₃, and F₄, whose eigenvalues are all below 0.8. This indicates that F₁ is the primary factor influencing river water quality, contributing the most to overall water quality variations. In Figure 5b, the spatial proximity of the water quality parameters as DO, NH₃-N, and TP suggests a strong correlation among these factors. Their position, distant from the F₁ axis, indicates a high loading on F₁, signifying their substantial contribution to this factor. Conversely, the proximity of these parameters to the F₂ axis reflects a lower loading on F₂. Additionally, the placement of COD_Mn at a distance from both the F₁ and F₂ axes demonstrates high loadings on both axes, indicating its significant influence on both F₁ and F₂. These findings collectively underscore the dominant role of F₁ in determining river water quality, with DO, NH₃-N, and TP contributing significantly to this factor, while COD_Mn exerts a notable influence on both F₁ and F₂.

The initial factor loading matrix (Table 5) shows that DO, NH₃-N, and TP have high loadings in F₁, while COD_Mn has a lower loading. Nonetheless, all these indicators have significant correlations (r > 0.6) with F₁. Previous studies have shown that DO can be used to assess the self-purification capacity of water bodies and reflect the pollution status within certain ranges [48]. Meanwhile, NH₃-N and TP can indicate the nutrient content in water bodies, reflecting the level of water eutrophication [49]. Thus, the self-purification capacity and eutrophication level of the water body could be measured by F₁. In contrast, F₂ has the largest loading of COD_Mn, embracing its significant impact on water quality. The other water quality indicators show an insignificantly negative correlation with F₂. COD_Mn is a primary factor contributing to organic pollution in water bodies [50,51]. In other words, F₂ can be used to characterize the organic pollution level of water bodies.

The pollution level of river water bodies can be quantified using comprehensive scores (G), with higher values indicating greater pollution severity [52]. Using the comprehensive evaluation function (Equations (3)–(5)), the inter-annual changes of principal component scores for water quality monitoring sections were calculated (Figure 6). It can be observed that G₁ (Figure 6a) and G (Figure 6c) showed a decreasing trend during the study period, while G₂ (Figure 6b) showed no significant change, indicating an overall improvement in river water quality. The spatial characteristics of G (Figure 6d) reveal a pollution severity pattern in the study area: Wumu > Wanta Bridge > Dongjian Bridge > Yaoxiang Bridge > Bie Bridge > Qingyang Bridge > Qianliu Bridge > Deshenghe Bridge > Xiaohe-sluice > Weicun. Among these, G in Weicun consistently remained at a low level, indicating better water quality conditions, while G in Wumu consistently remained high, indicating poor water quality conditions. The results of the PCA and the SFPI method are consistent.

3.5. Characterization of Spatial Clustering

The spatial characteristics of river water quality can be analyzed by clustering the locations of water quality monitoring sections. Based on CA of principal component scores, the sections were classified into three groups (Figure 7a). The first group includes seven water quality monitoring sections: Dongjian Bridge, Yaoxiang Bridge, Wanta Bridge, Bieqiao Bridge, Qingyang Bridge, Deshenghe Bridge, and Qianliu Bridge. These sections have moderate G (Figure 7b), indicating moderate water quality conditions. The second group contains one monitoring section at Wumu, which has the highest G, indicating poor water quality. The first principal component score (G₁) of Wumu is significantly higher than that of all other sections, indicating a poor self-purification capacity and a high level of water eutrophication in this river segment. This phenomenon is attributed to Wumu’s location downstream of the Beijing–Hangzhou Grand Canal in Changzhou City, where river water quality is heavily impacted by urbanization and higher domestic pollution loads compared to other rivers. The third group includes two sections at Weicun and Xiaohe-sluice, which have the lowest G, indicating good water quality. Notably, the second principal component score (G₂) at Weicun is negative, suggesting that the level of organic pollutants in this river segment is below the overall average for the study area. The reason for this is that Weicun is located on the right bank of the Yangtze River in the north of Changzhou City and benefits from national-level river management. As an important drinking water source, it receives a high degree of protection.

3.6. Discussion of the Characteristics of the Multivariate Statistical Analysis Method

3.6.1. Accuracy of Multivariate Statistical Analyses in Evaluating Water Quality

Multivariate statistical analyses enhance water quality evaluation by recognizing patterns and reducing data complexity. PCA and CA simplify interpretation by identifying key variables and grouping similar observations, improving accuracy in pinpointing water quality indicators and pollution sources [8,53]. These methods are effective for large datasets, integrating multiple parameters to detect subtle trends. Techniques like discriminant analysis and source apportionment models accurately identify pollution sources, aiding precise management. Additionally, these methods account for temporal and spatial variability, capturing the dynamic nature of water quality in plain river networks.

3.6.2. Limitations of Multivariate Statistical Analyses

Multivariate statistical analyses, like PCA and CA, involve assumptions that may affect accuracy [54]. PCA assumes linearity and normality, which may not always hold in water quality data, potentially skewing the results. CA is sensitive to distance metrics and clustering algorithms, which can influence outcomes. While PCA simplifies data by reducing dimensionality, it may omit important variables, sacrificing completeness. The complexity of these analyses can make interpretation difficult, especially for non-experts, leading to potential misinterpretation. Both methods are sensitive to outliers, which can skew the results. The choice of variables, principal components, and clustering methods is subjective, potentially causing inconsistencies. Additionally, these methods may not fully capture sudden or extreme changes in water quality, such as those caused by rare events (e.g., accidental spills). This can limit their effectiveness in real-time monitoring and prediction.

3.6.3. Cost-Effectiveness of the Proposed Approach

The proposed use of multivariate statistical analysis for assessing spatial and temporal variation in river water quality in Changzhou’s plains river network can be cost-effective under certain conditions. Techniques like PCA and CA reduce data dimensionality, focusing on key variables, which minimizes extensive data collection and monitoring costs. By identifying critical pollution sources and patterns, resources can be concentrated on the most affected areas, reducing unnecessary expenditures. While automating water quality monitoring (e.g., online sensors and real-time processing) involves higher initial costs, it can yield long-term savings by reducing manual sampling and analysis [55,56]. Early detection of water quality issues through this approach can prevent costly environmental damage, making it more cost-effective in the long term. The scalability of multivariate methods allows for flexible implementation based on budget and resources. However, the overall cost-effectiveness depends on balancing upfront investments with the long-term benefits of improved water quality management.

4. Conclusions

Utilizing a comprehensive evaluation method that integrates descriptive statistics, the SFPI approach, and qualitative water quality classification, the water quality categories and the overall status of the typical plain river network area in Changzhou City have been systematically assessed. The main conclusions drawn from this analysis are as follows.

The statistical analysis of the sample data indicates that the majority of DO, COD_Mn, NH₃-N, and TP meet Class I to Class III water quality standards. However, there are sections where NH₃-N exceeds Class V standards, and TP exceeds Class V. TN concentration remains at a high level, with a mean value of 2.74 mg/L. A strong negative correlation exists between WT and DO (r = −0.65), while a strong positive correlation exists between NH₃-N and TN (r = 0.7). Additionally, NH₃-N shows a moderate positive correlation with COD_Mn and TP, while TP has a moderate positive correlation with TN (r = 0.43, 0.57, and 0.55, respectively). This suggests that non-point source pollution from agricultural runoff and untreated wastewater discharges are the primary contributors to the high levels of TN and NH₃-N, indicating the need for targeted pollution control measures.

The SFPI method results indicate that the river water quality ranges from Class II to Class IV, with TP and NH₃-N being the primary pollutants exceeding Class IV water standards. Overall, the river water quality status is mildly polluted but shows an improving trend, with significant differences in spatial distributions.

The PCA and CA results reveal that river water quality is mainly influenced by two factors: the self-purification capacity and eutrophication level of the water body and the organic pollution level of rivers. It is worth noting that F₁ has the greatest impact on overall water quality. The ten sections were classified into three groups of good, moderate, and poor according to pollution characteristics. The comprehensive scores (G) of all monitoring sections show a declining trend year by year, indicating an improvement in river water quality. The spatial characteristics of water quality are differentiated due to the combined effects of industrial, domestic, and agricultural pollution, as well as the self-purification capacity of the water bodies.

Multivariate statistical analyses, like PCA and CA, offer powerful tools for evaluating water quality in complex environments, like plain river networks, providing accuracy in pattern recognition, data reduction, and source identification. However, their accuracy is limited by assumptions, potential information loss, sensitivity to outliers, and interpretation complexity. Careful application and validation against other methods are essential to ensure reliable water quality assessments.

If the research continues, future efforts could include expanding the scope of water quality monitoring by increasing the number of physical and chemical parameters monitored automatically. Enhancing the density of monitoring across the river network and implementing minute-level automatic monitoring could improve real-time data accuracy. Additionally, investigating pollution source discharges and achieving online river water quality monitoring would allow for continuous assessment and quicker responses to water quality changes. These advancements would provide a more comprehensive understanding of water quality dynamics and enhance pollution management in plain river network areas.