Content area
This study employs multivariate statistical techniques, water quality index (WQI), and a positive matrix factorization (PMF) receptor model to identify pollution sources entering a river, and evaluate the water quality. The study aims to establish strategies for effective water quality management in a watershed by identifying water quality characteristics using principal component analysis (PCA), and evaluating the effect of each pollution source using the PMF model. Through PCA, we identified organic matter and nutrients (e.g., nitrogen and phosphorus) as the primary sources of pollution with a significant impact on the target watershed. The PMF receptor model showed that the pollution sources included organic matter (29.61%), chlorophyll (22.52%), and nitrogen-based nutritive salts (19.80%). Furthermore, the WQI revealed a decrease in the calculated values in urban districts; site 1 (85.1) showed the highest value, whereas sites 5 (64.0) and 6 (63.8) showed lower values. The overall water quality remained safe above the moderate level. To maintain safe water quality and ensure effective management practices, it is imperative to consistently monitor the treated water flowing into the river from domestic sewage and industrial wastewater treatment facilities, and implement countermeasures against various non-point pollution sources. By selecting the sections affecting the target watershed and presenting the main factors and contributions of pollution sources, this study provides a range of methods for water quality management through scientific and precise analysis. The diverse analysis techniques employed in this study can be applied to future water quality evaluations.
Introduction
Population growth and increased economic activities within watersheds have contributed to socioeconomic development. Nevertheless, it is noteworthy that this has also led to significant pollution of the global environment. Environmental pollution has greatly affected human living conditions, necessitating the adoption of new environmental management strategies to address the resultant effects of regional climate change. Water, particularly, has always been a vital resource for human life. Surface water significantly influences human habitat, and access to safe drinking water is crucial for human survival (Kristanti et al. 2022; Rahman et al. 2023; Levin et al. 2024). Although three-quarters of the earth’s surface is covered with water (approximately 71%), 97% of the total distribution of water is seawater, and less than 1% of the total water is available for human use (e.g., rivers and lakes) (Li and Qian 2018; Unigwe and Egbueri 2023).
The growth of human civilization has led to encroachment on river ecosystems, leading to the development of cities and the concentration of populations around these valuable water resources. Population growth in these urban areas has resulted in excessive pollutant discharge, which has consequently contributed to increased pollution in urban rivers (Nong et al. 2020; Kumar et al. 2022; Latif et al. 2024). Urban rivers located near densely populated areas can experience high levels of pollution owing to the influx of point source pollutants, such as domestic sewage and industrial wastewater from various human activities, as well as non-point pollutants from various sources. The excessive influx of nutrients has caused serious environmental issues in rivers, including eutrophication, algal blooms, and fish kills (Tiwari and Pal 2022; Devlin and Brodie 2023; Tokatlı et al. 2023). Hence, the implementation of effective water quality management strategies is crucial for mitigating these environmental issues.
Comprehensive data collection through ongoing monitoring surveys and sampling analysis is essential for effective river water quality management; therefore, identifying the factors that affect water quality based on the interpretation of scientific parameters is necessary. As a tool to interpret these data, multivariate statistical techniques (MST) such as principal component, factor, discriminant, and cluster analyses (PCA, FA, DA, and CA, respectively) have been employed (Ahmad et al. 2022; Gani et al. 2023; Jo et al. 2024), to analyze surface water pollution contributors. Additionally, metrics such as water quality, trophic state, comprehensive pollution, organic pollution indices (WQI, TSI, and CPI, OPI, respectively), and health risk assessment (HRA) can be used to assess surface water pollution (Larrea-Murrell et al. 2022; Varol and Tokatlı 2023; Karadeniz et al. 2024). Furthermore, there have been continuous efforts to manage surface water pollution levels using various water quality management tools. The positive matrix factorization (PMF) model is among the methods employed to quantify pollution sources. It encompasses a mathematical model that utilizes multivariate statistical methods based on observation data, and effectively identifies pollution sources (Wang et al. 2022; Zhang et al. 2023, 2024). PMF models especially provide uncertainty weights for sample data, facilitating further refinement in identifying accurate pollution sources. Despite the various methods used, there have been numerous challenges in defining water quality and interpreting a large number of variables and datasets.
This study supports the development of effective strategies for efficient water quality management in the study area (the Geumho River), by providing an approach that diagnoses and evaluates pollution sources affecting water quality using various analytical techniques. This study aims to identify clusters according to sampling site concentrations through CA and pollution levels according to WQI results, extract the main variables per factor using PCA results, identify the spatiotemporal water quality characteristics of the target watershed through WQI calculations, and quantitatively interpret the contribution of each pollution source and variable utilizing the PMF model. The study provides a scientific method that contributes to decision-making for enhanced water quality management. Furthermore, it proposes a series of methods for identifying and evaluating the quality of pollution sources influencing rivers.
Data and methods
Target watershed
We selected the Geumho River—a tributary of the Nakdong River and one of the longest rivers in the Republic of Korea—as the research target watershed. The Geumho River is a typical urban river that flows through the cities of Yeongcheon and Gyeongsan, with more than 20 tributaries, and then through the city of Daegu (Fig. 1). The upper stream section of the target watershed comprises forests and agricultural land, whereas the middle and downstream sections comprise large industrial complexes and densely populated areas, resulting in increasing pollution levels downstream (Hidayaturrahman and Lee 2019; Lee et al. 2019; Jo et al. 2022). The average annual precipitation in the Geumho River watershed was 1143 mm in 2024, which is considerably less than the average precipitation of 1414 mm in the Republic of Korea. Owing to its geographical features, rainfall in the watershed is concentrated during the monsoon season between July and September; consequently, it is necessary to develop countermeasures for higher river pollution caused by the influx of various non-point pollution sources that occur during this period. In contrast, during the dry season, it is necessary to monitor and manage the influx of discharges from domestic sewage and industrial wastewater treatment plants located along the river (Table 1).
[See PDF for image]
Fig. 1
Land cover, water quality monitoring sites, and public treatment facility locality maps of the target watershed. Red dots: Monitoring sites. Green dots: Domestic sewage and industrial wastewater treatment facilities
Table 1. Treatment capacity of domestic sewage and industrial wastewater treatment plants (m3/day)
Abb | Treatment facility | Facility capacity | Inflow sewage | Outflow discharge |
|---|---|---|---|---|
T1 | Yeongcheon | 31,000 | 31,571 | 29,920 |
T2 | Geumho | 10,000 | 6314 | 6274 |
T3 | Gyeongsan | 40,000 | 38,355 | 36,958 |
T4 | Gyeongsan sandan | 100,000 | 90,262 | 89,345 |
T5 | Ansim | 47,000 | 38,738 | 35,282 |
T6 | Sincheon | 680,000 | 488,787 | 469,533 |
T7 | Jisan | 33,750 | 21,179 | 20,525 |
T8 | Bukbu | 170,000 | 106,063 | 101,264 |
T9 | Dalseoncheon | 400,000 | 190,487 | 185,620 |
The abbreviations T1–T9 represent domestic sewage and industrial wastewater monitoring sites as illustrated in Fig. 1
Abb, abbreviation
Water quality analysis methods
Water quality sampling was conducted monthly at the six sites located in the target watershed. Long-term water quality data are publicly available in the water environment information system’s water quality database (http://water.nier.go.kr, 2025.03). Samples collected from each site were stored, transported, and analyzed according to the standardized water pollution test method of the Korean Ministry of Environment. Water temperature (WT), pH, dissolved oxygen (DO), and electrical conductivity (EC) were measured in the field using a portable multimeter (YSI 6600 Extended Deployment System [EDS], YSI Incorporated, Yellow Springs, OH, USA). Additionally, biochemical oxygen demand (BOD), chemical oxygen demand (COD), total organic carbon (TOC), total nitrogen (TN), nitrate–nitrogen (NO3–N), ammonia–nitrogen (NH3–N), total phosphorus (TP), phosphate–phosphorus (PO4–P), and chlorophyll-a (Chl-a) were analyzed in a laboratory. Table 2 presents the analysis methods used for each variable.
Table 2. Water quality variables analysis methods
Variable | Abbreviation | Unit | Analysis method |
|---|---|---|---|
Biochemical oxygen demand | BOD | mg/L | Winkler azide method (5 days) |
Chemical oxygen demand | COD | mg/L | Potassium permanganate (KMnO4) |
Total organic carbon | TOC | mg/L | High-temperature combustion method |
Total nitrogen | TN | mg/L | Continuous flow analysis (UV/visible spectrometry) |
Nitrate–nitrogen | NO3–N | mg/L | Continuous flow analysis (ion chromatography) |
Ammonia–nitrogen | NH3-N | mg/L | Continuous flow analysis (UV/visible spectrometry) |
Total phosphorus | TP | mg/L | Continuous flow analysis (UV/visible spectrometry) |
Phosphate–phosphorus | PO4–P | mg/L | Continuous flow analysis (ion chromatography) |
Chlorophyll-a | Chl-a | mg/m3 | UV/visible spectrometry |
Data analysis methods
Water quality data were analyzed using techniques such as PCA, CA, WQI, and PMF. Monthly data collected during the period from 2014 through 2024 were analyzed for 13 water quality variables. The data used in the statistical analysis were converted to a standard score (z-score) to avoid errors caused by unit differences between variables. The z-score is a dimensionless value that distributes data normally, and indicates the degree of separation by a standard deviation based on the mean. The data were analyzed using SPSS 24.0, XL-STAT (IBM Corp., Armonk, NY, USA), and EPA PMF 5.0 (Environmental Protection Agency (EPA), Washington, D.C., WA, USA) software programs.
Principal component analysis (PCA)
PCA is among the commonly employed multivariate statistical analysis methods. It is a linear transformation method that reduces dimensionality from higher to lower dimensions through factors comprising multiple uncorrelated variables with maximum preservation of complex data (Masood et al. 2022; Ibrahim et al. 2023). We performed Kaiser–Meyer–Olkin (KMO) and Bartlett’s sphericity tests to confirm the goodness of fit of the data for PCA. Data are useful when the KMO value is higher than 0.5; the closer the index is to 1, the more appropriate it is for analysis. Additionally, Bartlett’s sphericity test should have a significance level of p < 0.05 (Zubaidi et al. 2022; Ali et al. 2024).
We conducted the KMO and Bartlett’s sphericity tests, followed by PCA. The varimax method of rotation was used to improve the interpretation of the variables (Nayak et al. 2023; Subba Rao et al. 2024). Loading values higher than 0.75 were classified as “strong,” those ranging between 0.5 and 0.75 were categorized as “moderate,” and those lower than 0.5 were classified as “weak” (Liu et al. 2003; Jo and Kwon 2023).
Real-time water quality index (RTWQI)
WQI has been continuously used to assess overall water quality conditions using datasets since the 1960s (Horton 1965; Chidiac et al. 2023). The index has been frequently used by water quality policy managers and the public because of its advantage of assessing water quality conditions solely based on index values. Furthermore, WQI has been widely used in water quality research because it can be used in planning and evaluation, to determine and manage water quality conditions in rivers (Uddin et al. 2022; Syeed et al. 2023).
The real-time WQI (RTWQI) is an index calculation method developed by modifying the water quality index of the Canadian Council of Ministers of the Environment to reflect the situation in Korea. RTWQI is calculated as Eq. (1):
1
where F1, F2, and F3 represent the scope, frequency, and amplitude, respectively. RTWQI was classified into five classes according to the index value. Table 3 summarizes the definition of water quality status for each class. An index value close to 100 indicates good water quality, and a value close to 0 indicates high pollution levels. We calculated the index value per site within the target watershed to determine spatiotemporal water quality conditions. The analysis period was from 2014 to 2024. Moreover, ten water quality variables were used for the assessment: WT, pH, DO, EC, BOD, COD, TOC, TN, TP, and Chl-a.Table 3. General description of RTWQI per class
Category | RTWQI value | Description |
|---|---|---|
Excellent | 80 ≤ value | The water is always appropriate for water-related activities owing to clean water status with few contaminants |
Good | 60 ≤ value < 80 | The water is appropriate for water-related activities as water quality is relatively good |
Moderate | 40 ≤ value < 60 | Generally good water quality, but with occasional contaminants that may affect water-related activities |
Caution | 20 ≤ value < 40 | The water is polluted owing to frequent pollutant inputs, which requires caution for water-related activities |
Poor | Value < 20 | The water is highly polluted and inappropriate for water-related activities |
RTWQI, real-time water quality index
Positive matrix factorization (PMF)
The PMF model is a multivariate factor analysis tool that decomposes a dataset matrix into two matrices: factor contributions and factor profiles. PMF is recommended by the USEPA for source distribution, and is a widely used receptor model (Reff et al. 2007; Mao et al. 2023; Rashid et al. 2024). The PMF receptor model is a mathematical approach to quantifying the contribution of a source to a sample based on the composition of the sources. Similar to a typical factor analysis model, a PMF model distributes source profiles and source contributions based on chemical composition data, and identifies sources through the extracted source profiles. The advantage of PMF is the ability to assign weights to individual data points, thereby enabling adjustments for uncertainties in the evaluated data (Haji Gholizadeh et al. 2016). The dataset is calculated using Eq. (2):
2
where Xij indicates i samples, and j monitoring variables, which can be decomposed into two matrices, Gik and Fkj. Gik are the source contribution matrix, and Fkj is the factor loading of the source components. eij is the residual representing the difference between the analysis results and the measured values, and p refers to the number of independent pollution sources (extracted factors). We analyzed the contribution of pollution sources using PMF 5.0 software recommended by the USEPA.Analysis results
Water quality features of the target watershed
Table 4 presents the descriptive statistics (e.g., mean and standard deviation) for the six target watershed sites. The monitoring sites showed an overall increasing concentration range from upstream to downstream. WT was higher in the summer and lower in the winter under the temperate climatic zone of Korea. pH was slightly alkaline, with mean values ranging between 8.0 and 8.5 across sites, with higher values in summer. The photosynthetic activity of chlorophyll can reduce carbon dioxide in water, which can increase the pH level. The average value of DO ranged between 9.9 and 12.1. EC had a mean value of 142.1 at Site 1, and increased downstream to 653.9 at Site 6. This pattern is attributable to the large amount of treated water flowing into the river from the domestic sewage and industrial wastewater treatment plants located within the target watershed.
Table 4. Descriptive statistics of water quality variables at the monitoring sites from 2014 to 2024 (range, mean, and standard deviation)
Parameter | Site 1 | Site 2 | Site 3 | Site 4 | Site 5 | Site 6 | |
|---|---|---|---|---|---|---|---|
WT | Range | 1.0–27.6 | 1.0–32.6 | 0.6–31.9 | 2.2–32.6 | 2.8–32.9 | 1.4–32.8 |
M ± SD | 15.0 ± 7.0 | 17.0 ± 8.5 | 16.8 ± 8.8 | 17.3 ± 8.9 | 17.8 ± 8.7 | 17.5 ± 8.1 | |
pH | Range | 6.7–9.7 | 6.8–10.0 | 6.8–9.3 | 7.0–10.0 | 6.8–10.1 | 6.8–9.2 |
M ± SD | 8.1 ± 0.5 | 8.4 ± 0.6 | 8.1 ± 0.6 | 8.5 ± 0.6 | 8.4 ± 0.6 | 8.0 ± 0.5 | |
DO | Range | 6.3–18.5 | 7.2–20.7 | 5.7–16.5 | 5.8–19.0 | 4.4–15.1 | 6.0–17.5 |
M ± SD | 11.2 ± 2.3 | 12.1 ± 2.9 | 10.0 ± 2.3 | 10.2 ± 2.4 | 9.9 ± 2.2 | 10.8 ± 2.5 | |
EC | Range | 58–263 | 145–777 | 186–736 | 144–723 | 160–775 | 205–1162 |
M ± SD | 142.1 ± 32.7 | 365.3 ± 106.5 | 433.3 ± 109.8 | 419.7 ± 115.1 | 448.6 ± 113 | 653.9 ± 185.5 | |
BOD | Range | 0.3–5.9 | 0.3–6.6 | 0.8–9.5 | 0.9–7.9 | 0.9–10.2 | 0.8–8.6 |
M ± SD | 0.9 ± 0.6 | 1.9 ± 1.1 | 2.4 ± 1.2 | 2.9 ± 1.4 | 3.2 ± 1.9 | 3.1 ± 1.7 | |
COD | Range | 2.0–10.2 | 3.8–10.3 | 4.3–9.4 | 4.2–11.7 | 4.0–14.2 | 5.3–14.5 |
M ± SD | 5.5 ± 1.3 | 6.7 ± 1.4 | 6.0 ± 1.1 | 6.2 ± 1.5 | 6.4 ± 1.7 | 8.0 ± 1.7 | |
TOC | Range | 1.1–5.4 | 2.2–9.8 | 1.7–9.5 | 1.2–9.7 | 1.0–9.8 | 4.0–11.5 |
M ± SD | 3.3 ± 0.7 | 3.9 ± 1.0 | 4.3 ± 1.2 | 4.5 ± 1.3 | 4.3 ± 1.2 | 5.9 ± 1.4 | |
TN | Range | 0.487–5.720 | 0.774–6.774 | 1.863–15.360 | 1.144–16.080 | 1.958–18.384 | 2.508–10.130 |
M ± SD | 1.459 ± 0.536 | 3.496 ± 1.407 | 5.543 ± 2.542 | 5.371 ± 2.970 | 6.514 ± 2.971 | 5.590 ± 1.453 | |
NO3–N | Range | 0.208–3.154 | 0.001–6.293 | 0.843–5.667 | 0.556–6.105 | 0.954–8.177 | 1.733–7.952 |
M ± SD | 1.020 ± 0.348 | 2.516 ± 1.335 | 3.245 ± 1.093 | 3.032 ± 1.199 | 4.150 ± 1.340 | 4.293 ± 1.255 | |
NH3–N | Range | 0.001–0.716 | 0.001–1.380 | 0.001–1.725 | 0.001–1.497 | 0.001–1.467 | 0.018–1.842 |
M ± SD | 0.067 ± 0.086 | 0.175 ± 0.250 | 0.266 ± 0.353 | 0.214 ± 0.322 | 0.213 ± 0.278 | 0.268 ± 0.321 | |
TP | Range | 0.003–0.080 | 0.009–0.312 | 0.017–0.202 | 0.014–0.226 | 0.020–0.422 | 0.023–0.218 |
M ± SD | 0.018 ± 0.014 | 0.052 ± 0.050 | 0.065 ± 0.038 | 0.067 ± 0.038 | 0.088 ± 0.058 | 0.076 ± 0.041 | |
PO4–P | Range | 0.001–0.074 | 0.001–0.194 | 0.001–0.155 | 0.001–0.173 | 0.001–0.213 | 0.001–0.186 |
M ± SD | 0.005 ± 0.008 | 0.021 ± 0.035 | 0.023 ± 0.029 | 0.021 ± 0.031 | 0.028 ± 0.038 | 0.027 ± 0.033 | |
Chl-a | Range | 0.2–24.9 | 1.6–45.9 | 1.6–57.9 | 1.5–194.6 | 2.8–161.3 | 3.8–236.6 |
M ± SD | 4.4 ± 3.3 | 9.4 ± 7.4 | 12.0 ± 11.1 | 23.0 ± 26.8 | 27.4 ± 26.4 | 35.5 ± 39.3 | |
M, mean; SD, standard deviation
The mean values of BOD ranged between 0.9 and 3.2, which is “very good” to “moderate” according to the Living Environment Standards for Water Quality in Korea, whereas the mean values of COD ranged between 5.5 and 8.0, implying “slightly poor” at Site 6 and “moderate” at the other sites. TOC ranged between 3.3 and 5.9, which is the same level as COD. Furthermore, as illustrated in Fig. 2, the time series data of BOD, COD, and TOC concentrations per site showed a decrease in BOD at all sites. COD increased slightly at Sites 1–4, and decreased at Sites 5 and 6. TOC increased slightly at Sites 1 and 2, and decreased at the other sites. Based on the concentrations of BOD, COD, and TOC, which indicate the degree of organic pollutants, it is crucial that biodegradable organic materials remain within acceptable limits as per the stringent regulation of sewage treatment facilities; however, the concentrations may be elevated owing to inadequate treatment of non-biodegradable organic substances. The Geumho River has garnered the attention of researchers because it flows through an urban area and covers a large industrial complex. Previous studies highlight the issue of non-biodegradable organic matter and microplastics from urban areas entering rivers, owing to being untreated during the sewage treatment process (Eo et al. 2019; Hidayaturrahman and Lee 2019). The mean values of TN, NO3–N, and NH3–N ranged between 1.459 and 6.514, 1.020–4.293, and 0.067 and 0.268, respectively, with higher concentrations generally observed downstream. This phenomenon is attributable to lower treatment efficiency that occurs during the winter denitrification process at various sewage and wastewater treatment plants located within the target watershed (Jo et al. 2022). The concentrations of TP, PO4–P, and Chl-a ranged between 0.018 and 0.088, 0.005and 0.028, and 4.4 and 35.5, respectively.
[See PDF for image]
Fig. 2
Time series data on the concentrations of water quality variables (BOD, COD, and TOC) per monitoring site. Each linear dotted line represents a linear trend line of BOD, COD, and TOC
Figure 3 illustrates the cluster analysis results according to water quality concentrations using the dataset of the target watershed. Spatial cluster analysis results revealed that the sites were divided into two groups, Cluster 1 (Sites 1 and 2) and Cluster 2 (Sites 3–6), which confirmed the differences in the concentrations of the water quality variables. The temporal cluster analysis results indicated that the sites were categorized into three groups: monsoon (July–October; with intensive rainfall), pre- (April–June) and post-monsoon, and winter (November–March) seasons. The homogeneity test through the analysis of variance (ANOVA) was performed to confirm the difference between the groups (p < 0.05).
[See PDF for image]
Fig. 3
Spatiotemporal cluster analysis results according to water quality variables concentrations using Ward’s method. a Spatial cluster analysis and b temporal cluster analysis (ANOVA, p < 0.05)
Derivation of principal factors of the target watershed (PCA)
PCA was performed using water quality monitoring data from 2014 to 2024, to identify the main factors influencing the target watershed. The KMO test for the goodness of fit of the PCA was high (0.732), and the Bartlett’s sphericity value was 0.000 (p < 0.05), which is appropriate for the analysis (Table 5).
Table 5. Loading matrix of water quality variables according to varimax rotation for the Geumho River watershed
Variable | PC 1 | PC 2 | PC 3 | PC 4 |
|---|---|---|---|---|
WT | 0.259 | 0.807 | − 0.214 | − 0.265 |
pH | 0.190 | − 0.047 | − 0.064 | − 0.800 |
DO | − 0.053 | − 0.844 | 0.071 | 0.013 |
EC | 0.373 | − 0.239 | 0.697 | − 0.143 |
BOD | 0.772 | 0.054 | 0.276 | − 0.097 |
COD | 0.880 | 0.141 | − 0.074 | 0.030 |
TOC | 0.781 | 0.169 | 0.125 | 0.184 |
TN | 0.089 | 0.029 | 0.880 | 0.150 |
NO3-N | 0.036 | − 0.036 | 0.917 | 0.154 |
NH3-N | 0.226 | − 0.48 | 0.248 | 0.563 |
TP | 0.508 | 0.641 | 0.227 | 0.325 |
PO4-P | 0.142 | 0.673 | 0.079 | 0.532 |
Chl-a | 0.789 | 0.074 | 0.138 | − 0.323 |
Eigen value | 3.186 | 2.576 | 2.390 | 1.629 |
Total variance | 24.510 | 19.817 | 18.383 | 12.534 |
KMO test | 0.732 | |||
Bartlett’s test | 0.000 |
Values in Bold and italics indicate “strong” and “moderate” loading, respectively
KMO, Kaiser–Meyer–Olkin (test)
Four principal components (PCs) were extracted with eigenvalues greater than 1.0 and a total variance of 75.244% (Table 6). As shown in Table 5, PC 1 accounted for 24.51% of the total variance, with BOD, COD, TOC, and Chl-a showing “strong” loading, and TP showing “moderate” loading. PC 2 accounted for 19.817% of the total variance, with WT and DO showing “strong” loading, and TP and PO4-P showing “moderate” loading. PC 1 and PC 2 can be defined as organic matter-related and nutrient (seasonal) factors in the phosphorus series, respectively. PC 3 accounted for 18.383% of the total variance, with TN and NO3–N showing “strong” loading, and EC showing “moderate” loading. PC 4 accounted for 12.534% of the total variance, with pH showing “strong” loading, and NH3–N and PO4-P showing “moderate” loading. PC 3 can be defined as nitrogen-related factors, and PC 4 can be defined as other factors with unclear causes. Furthermore, the relationship between the parameters can be confirmed, as illustrated in Fig. 4, based on the PCA results.
Table 6. Total variance explained for the Geumho River watershed
Component | Initial eigenvalues | Extraction sums of squared loadings | Rotation sums of squared loadings | ||||||
|---|---|---|---|---|---|---|---|---|---|
Total | Variance | Cumulative | Total | Variance | Cumulative | Total | Variance | Cumulative | |
1 | 3.987 | 30.671 | 30.671 | 3.987 | 30.671 | 30.671 | 3.186 | 24.510 | 24.510 |
2 | 2.820 | 21.692 | 52.362 | 2.820 | 21.692 | 52.362 | 2.576 | 19.817 | 44.327 |
3 | 1.900 | 14.615 | 66.977 | 1.900 | 14.615 | 66.977 | 2.390 | 18.383 | 62.710 |
4 | 1.075 | 8.267 | 75.244 | 1.075 | 8.267 | 75.244 | 1.629 | 12.534 | 75.244 |
5 | 0.729 | 5.606 | 80.849 | ||||||
6 | 0.619 | 4.764 | 85.613 | ||||||
7 | 0.536 | 4.121 | 89.734 | ||||||
8 | 0.339 | 2.609 | 92.344 | ||||||
9 | 0.263 | 2.021 | 94.364 | ||||||
10 | 0.238 | 1.833 | 96.198 | ||||||
11 | 0.190 | 1.462 | 97.659 | ||||||
12 | 0.156 | 1.203 | 98.862 | ||||||
13 | 0.148 | 1.138 | 100 | ||||||
[See PDF for image]
Fig. 4
Relationship charts of variables within a principal component (PC) based on principal component analysis (PCA) results: Relationship charts between a PC 1 and PC 2; b PC 1 and PC 3
Hypothetically, this result is attributable to the fact that the Geumho River flows through an urban area, leading to the influx of a large amount of organic matter and nutrients from human and industrial activities. Various pollution sources, including untreated non-biodegradable substances from domestic sewage and industrial wastewater treatment facilities located along the river, urban non-point pollution sources, and untreated influents, can contribute to the high pollution levels. Based on the PCA results, measures to prevent pollution sources from entering rivers, such as strengthening polluted water treatment processes for domestic sewage and industrial wastewater, managing the acceptable concentration of treated water, and monitoring illegal sewage and wastewater discharges, should be implemented through river management (monitoring/surveillance),.
Spatiotemporal evaluation of water quality using the water quality index
We calculated WQI to assess the spatiotemporal water quality of the target watershed; Table 7 presents the results. The WQI calculation for the six sites in the target watershed ranked Site 1 as “excellent” (85.1), indicating a clean water condition with few pollutants, Site 2 as “good” (73.5), showing relatively good water quality, and Sites 3, 4, 5, and 6 as “good” with ranges of 67.6, 66.1, 64.0, and 63.8, respectively. However, Sites 5 and 6 require continuous attention as their water quality was ranked as “moderate” at certain periods. Regarding the temporal aspect, the WQI values were relatively lower during the period from June through August, with values lower than 60, compared with other periods. This phenomenon is attributable to the influx of non-point pollution sources during the rainy season.
Table 7. Calculation results of the spatiotemporal water quality index (mean) in the target watershed
Month | Site 1 | Site 2 | Site 3 | Site 4 | Site 5 | Site 6 | Mean |
|---|---|---|---|---|---|---|---|
1 | 89.5 | 77.9 | 74.2 | 73.1 | 71.5 | 72.0 | 76.4 |
2 | 89.5 | 75.2 | 71.0 | 69.9 | 67.8 | 70.4 | 74.0 |
3 | 84.2 | 74.7 | 65.2 | 62.0 | 61.5 | 62.0 | 68.3 |
4 | 85.3 | 72.1 | 64.7 | 62.0 | 63.1 | 59.9 | 67.8 |
5 | 83.7 | 73.1 | 66.3 | 65.2 | 60.4 | 62.5 | 68.5 |
6 | 84.7 | 67.9 | 66.3 | 62.0 | 56.2 | 58.8 | 66.0 |
7 | 83.7 | 69.9 | 65.7 | 65.7 | 61.0 | 59.4 | 67.6 |
8 | 83.7 | 75.2 | 63.1 | 61.0 | 59.4 | 59.3 | 66.9 |
9 | 83.7 | 74.2 | 66.3 | 67.3 | 65.7 | 64.7 | 70.3 |
10 | 82.9 | 70.9 | 70.4 | 68.2 | 65.1 | 63.5 | 70.1 |
11 | 82.1 | 73.3 | 70.0 | 67.5 | 68.0 | 66.4 | 71.2 |
12 | 88.0 | 77.1 | 68.4 | 69.3 | 67.9 | 66.8 | 72.9 |
Mean | 85.1 | 73.5 | 67.6 | 66.1 | 64.0 | 63.8 |
The “Moderate (40 ≤ value < 60)” class of the WQI values are in bold
Figure 5 shows the results of running CA based on the WQI values. Specifically, Cluster 1 (Sites 1 and 2) maintained clean water quality with low pollution levels, whereas Cluster 2 (Sites 3–6) maintained good water quality despite occasional exposure to pollution. Homogeneity for the clusters results was confirmed using a t test to determine differences between groups (p < 0.05).
[See PDF for image]
Fig. 5
Spatial cluster analysis using water quality index (WQI) values (t test, p < 0.05)
Locally weighted scatter plot smoothing (LOWESS) analysis was also conducted with the WQI values for each site to examine changes in water quality conditions for the target watershed (Fig. 6). LOWESS complements the monotonic fluctuations of linear trend analysis, and facilitates the identification of trend fluctuations over the study period. Site 1 showed a decreasing trend in WQI values, which is attributable to urbanization and industrialization expansion. Sites 2, 3, and 4 did not show significant variability. Sites 5 and 6 showed an increasing trend in WQI values, which is attributable to the control of treated water concentrations in sewage and wastewater treatment facilities, and continuous water quality management measures.
[See PDF for image]
Fig. 6
Trend analysis of water quality index (WQI) values using locally weighted scatter plot smoothing. The dots represent individual water quality index values. The red line represents the locally weighted scatter plot smoothing (LOWESS) trend line
Factor contribution using the positive matrix factorization (PMF) receptor model
We executed a PMF model to identify and quantify pollution sources for Cluster 2 sites in the CA results, based on the WQI analyzed earlier. We also executed the model by generating concentration and uncertainty files for the nine laboratory variables, excluding the field variables. The PMF results indicated that the average R2 value of the variables was at a good level (0.806), with that of TN relatively lower (0.517) than that of other variables. Lastly, we identified five optimal factors by executing the base model (Figs. 7, 8 and 9).
[See PDF for image]
Fig. 7
Contributions of pollution sources to variables using the positive matrix factorization (PMF) model. Factor 1: Organic pollution sources; Factor 2: Chlorophyll growth; Factor 3: Nutritive salts (nitrogen group); Factor 4: Nutritive salts (phosphorus group); Factor 5: Diverse non-point pollution sources (unknown causes)
[See PDF for image]
Fig. 8
Average contribution of pollution sources to water quality using the positive matrix factorization (PMF) model
[See PDF for image]
Fig. 9
Factor-specific profiles of water pollution sources using the positive matrix factorization (PMF)
In Factor 1, the contributions were in the order of TOC (84.82%), COD (66.59%), and BOD (37.75%), implying that Factor 1 originated from organic pollutants from domestic sewage and industrial wastewater. In Factor 2, the contributions were in the order of Chl-a (80.19%), BOD (45.00%), and TP (35.93%). Moreover, Factor 2 was related to the growth of chlorophyll. High BOD concentrations are also attributable to population growth. In Factor 3, the contributions were in the order of NO3-N (66.29%), TN (61.05%), and Chl-a (15.59%). Factor 3 is associated with nitrogen-based nutritive salts. High concentrations of nitrogen inflows are attributable to various sources, including denitrification processes at public treatment plants in the winter and excessive application of nitrogen fertilizers on agricultural land. In Factor 4, the contributions were in the order of PO4–P (93.94%), TP (38.47%), and TN (5.03%). Factor 4 is phosphorus-based nutritive salts, and PO4–P may be highly related to wastewater from industrial complexes. The downstream section of the Geumho River has large industrial complexes and various industrial wastewater treatment facilities; therefore, managing them is crucial. In Factor 5, the contributions were in the order of NH3-N (88.94%), TN (9.80%), and BOD (6.92%). Factor 5 may be associated with various non-point pollution sources without distinct causes. There may be diverse causes, such as plastic film-house cultivation facilities located in the downstream section, non-point pollution sources inflowing into urban centers, and untreated domestic sewage and wastewater from factories. PMF model results are fully consistent with PCA results, indicating that the PMF receptor model effectively captures the distribution of pollution sources.
Discussion
This study seeks to develop effective strategies for efficient water quality management by applying various analytical techniques (e.g., MST, WQI, PMF) to provide methods for identifying and assessing pollution sources affecting water quality. We extracted four PCs from the water quality datasets of the six sites through PCA. Consequently, organic matter factors and factors related to phosphorus and nitrogen nutrients accounted for the majority of the total variance. We performed WQI and CA to confirm the pollution levels of the monitoring sites and categorize the groups. Subsequently, PMF was performed using the water quality data of the group with low WQI values. In addition to the current qualitative analysis conducted in water quality assessment and pollution source identification, we successfully incorporated quantitative analysis to verify the contribution of each pollution source. Because the outcomes were extracted as similar factors to PCA, it can be interpreted that the findings of this study were significant. Given that urban rivers receive pollution from multiple sources, it is imperative to analyze the root causes and implement appropriate mitigation measures. This study reveals that the influx of organic matter and nutrients in the target watershed is a major factor affecting the river water quality, and suggested the quantitative contribution rate of the pollution sources. PCA results indicated that organic matter was the crucial component, consistent with the findings of previous studies. WQI results showed similar results; however, the calculated values decreased with moving downstream sites (Jung et al. 2015, 2016; Jo et al. 2022).
We comprehensively assessed the water quality of the Geumho River watershed. This entailed the careful selection of monitoring sites with a substantial impact on water quality from a wide range of locations, and the identification and presentation of the main factors and contribution rates of pollution sources, while not limiting the assessment to the concentrations of individual variables. This study may benefit river management aspects, such as maintaining the health of river ecosystems and securing safe water quality by maintaining the various functions of rivers in urban areas (e.g., recreation, water supply, and urban aesthetics). Additionally, we hope that this study will serve as a foundational reference for relevant agencies in establishing water quality improvement plans and countermeasures for rivers that have been neglected owing to growth-oriented urban development driven by urbanization and industrialization. However, this study has limitations. As the study only targets and analyzes the monitoring sites of a single river in the target watershed, further studies should be conducted for a more accurate analysis of pollution causes, encompassing non-point sources. For instance, diverse studies should be conducted to assess the effects of inflowing rivers, sewage and wastewater treatment plants flowing into rivers, and various urban non-point pollution sources.
Conclusion
This study employs multivariate statistical analysis techniques, PMF receptor model analysis, and WQI calculations examine and evaluate the water quality of the Geumho River flowing through urban areas, and explores efficient water quality management measures through the distribution of contributions from diverse pollution sources. Specifically, 11-year (2014–2024) monitoring datasets of six sites in the Geumho River were used.
PCA was also performed using data from six sites, and four significant PCs were extracted. Organic matter and nutrients accounted for the majority of the total variance, confirming the need to control human and industrial pollution sources. Spatiotemporal water quality assessment results from WQI had the lowest value for Site 6 (63.8) and the highest value for Site 1 (85.1). All sites were generally rated as “good,” but Sites 5 and 6 had lower calculated values at certain periods, highlighting the importance of continuous water quality management. Furthermore, the sites were categorized into Clusters 1 (Sites 1 and 2) and 2 (Sites 3–6) through CA, and also divided into a group with clean water quality conditions and low pollution levels, and another with good water quality but occasional pollution sources exposure. After executing the PMF model to identify and quantify pollution sources at Sites 3 to 6, which were evaluated as relatively highly polluted areas, we identified five potential sources of pollution. Specifically, the contributions were high in the order of organic pollution sources (29.61%), chlorophyll growth (22.52%), and nitrogen-based nutritive salts (19.80%). The findings of this study highlight human activities by-products and industrial discharges as major pollution sources in the target watershed. Therefore, preventing domestic sewage and industrial wastewater discharges into rivers is crucial to ensure safe water quality in such a watershed. Continuous monitoring and management of effluent is essential because effluent discharged from public treatment facilities can impact the overall instream flow.
We further examined the validity and reliability of water quality assessment and the identification of potential pollution sources by integrating MSTs and PMF receptor models based on monitored river datasets. Central and local agencies/governments should consistently prioritize water quality management to improve and protect the water quality of rivers. As the findings of this study were obtained from a limited number of monitoring sites and water quality variables, they should not be considered absolute evaluation outcomes. However, the diverse analysis techniques employed in this study can be applied to future water quality assessments.
Acknowledgements
We thank the members of the Nakdong River Environment Research Center for their assistance.
Author contributions
All authors contributed to the planning and structuring of the study. Chang Dae Jo performed data collection and investigation, data analysis, visualization, manuscript writing, and manuscript review and editing. Heon Gak Kwon performed data collection and investigation and manuscript review and editing. All authors have read and agreed to the published version of the manuscript.
Funding
This work was supported by the National Institute of Environmental Research (NIER) funded by the Ministry of Environment (MOE) of the Republic of Korea [grant number NIER-2023-01-01-167]. The funding agency had no role in study design, in the collection, analysis, and interpretation of data, in the writing of the report, and in the decision to submit the article for publication.
Data availability
Not applicable.
Declarations
Conflict of interest
The authors declare that they have no conflicts of interest.
Peiyue Li
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
Ahmad, S; Umar, R; Ahmad, I. Assessment of groundwater quality using entropy-weighted quality index (EWQI) and multivariate statistical techniques in Central Ganga plain, India. Environ Dev Sustain; 2022; 26,
Ali, S; Verma, S; Agarwal, MB; Islam, R; Mehrotra, M; Deolia, RK; Kumar, J; Singh, S; Mohammadi, AA; Raj, D; Gupta, MK; Dang, P; Fattahi, M. Groundwater quality assessment using water quality index and principal component analysis in the Achnera block, Agra district, Uttar Pradesh, Northern India. Sci Rep; 2024; 14,
Chidiac, S; El Najjar, P; Ouaini, N; El Rayess, Y; El Azzi, D. A comprehensive review of water quality indices (WQIs): history, models, attempts and perspectives. Rev Environ Sci Biotechnol; 2023; 22,
Devlin, M; Brodie, J. Reichelt-Brushett, A. Nutrients and eutrophication. Marine pollution–monitoring, management and mitigation; 2023; Cham, Springer: pp. 75-100. [DOI: https://dx.doi.org/10.1007/978-3-031-10127-4_4]
Eo, S; Hong, SH; Song, YK; Han, GM; Shim, WJ. Spatiotemporal distribution and annual load of microplastics in the Nakdong River, South Korea. Water Res; 2019; 160, pp. 228-237.1:CAS:528:DC%2BC1MXhtVKitbzE [DOI: https://dx.doi.org/10.1016/j.watres.2019.05.053]
Gani, A; Pathak, S; Hussain, A; Ahmed, S; Singh, R; Khevariya, A; Banerjee, A; Ayyamperumal, R; Bahadur, A. Water quality index assessment of river Ganga at Haridwar stretch using multivariate statistical technique. Mol Biotechnol; 2023; [DOI: https://dx.doi.org/10.1007/s12033-023-00864-2]
Haji Gholizadeh, MH; Melesse, AM; Reddi, L. Water quality assessment and apportionment of pollution sources using APCS-MLR and PMF receptor modeling techniques in three major rivers of South Florida. Sci Total Environ; 2016; 566–567, pp. 1552-1567.1:CAS:528:DC%2BC28XhtVShu73I [DOI: https://dx.doi.org/10.1016/j.scitotenv.2016.06.046]
Hidayaturrahman, H; Lee, TG. A study on characteristics of microplastic in wastewater of South Korea: identification, quantification, and fate of microplastics during treatment process. Mar Pollut Bull; 2019; 146, pp. 696-702.1:CAS:528:DC%2BC1MXhsVartLbN [DOI: https://dx.doi.org/10.1016/j.marpolbul.2019.06.071]
Horton, RK. An index number system for rating water quality. J (Water Pollut Control Fed); 1965; 37,
Ibrahim, A; Ismail, A; Juahir, H; Iliyasu, AB; Wailare, BT; Mukhtar, M; Aminu, H. Water quality modelling using principal component analysis and artificial neural network. Mar Pollut Bull; 2023; 187, 1:CAS:528:DC%2BB38Xjt1SntrvI [DOI: https://dx.doi.org/10.1016/j.marpolbul.2022.114493] 114493.
Jo, CD; Kwon, HG. Temporal and spatial evaluation of the effect of river environment changes caused by climate change on water quality. Environ Technol Innov; 2023; 30, 1:CAS:528:DC%2BB3sXjs1ejtr4%3D [DOI: https://dx.doi.org/10.1016/j.eti.2023.103066] 103066.
Jo, C; Kwon, H; Kim, S. Temporal and spatial water quality assessment of the Geumho River, Korea, using multivariate statistics and water quality indices. Water; 2022; 14,
Jo, CD; Choi, SY; Kwon, HG. Statistical analysis of water quality change by total maximum daily load policy stage. Appl Water Sci; 2024; 14,
Jung, KY; Ahn, JM; Lee, KL; Lee, IJ; Yu, JJ; Cheon, SU; Kim, KS; Han, KY. Temporal and spatial analysis of non-biodegradable organic pollutants in the Geumho River system. J Environ Sci Int; 2015; 24,
Jung, KY; Ahn, JM; Kim, K; Lee, IJ; Yang, DS. Evaluation of water quality characteristics and water quality improvement grade classification of Geumho River tributaries. J Environ Sci Int; 2016; 25,
Karadeniz, S; Ustaoğlu, F; Aydın, H; Yüksel, B. Toxicological risk assessment using spring water quality indices in plateaus of Giresun Province/Türkiye: a holistic hydrogeochemical data analysis. Environ Geochem Health; 2024; 46,
Kristanti, RA; Hadibarata, T; Syafrudin, M; Yılmaz, M; Abdullah, S. Microbiological contaminants in drinking water: current status and challenges. Water Air Soil Pollut; 2022; 233,
Kumar, M; Gikas, P; Kuroda, K; Vithanage, M. Tackling water security: a global need of cross-cutting approaches. J Environ Manag; 2022; 306, 1:CAS:528:DC%2BB38XisFCntL8%3D [DOI: https://dx.doi.org/10.1016/j.jenvman.2022.114447] 114447.
Larrea-Murrell, JA; Romeu-Alvarez, B; Lugo-Moya, D; Rojas-Badía, MM. Acid phosphatase activity in freshwater ecosystems of western Cuba and its relationship with water quality. Water Air Soil Pollut; 2022; 233,
Latif, M; Nasir, N; Nawaz, R; Nasim, I; Sultan, K; Irshad, MA; Irfan, A; Dawoud, TM; Younous, YA; Ahmed, Z; Bourhia, M. Assessment of drinking water quality using water quality index and synthetic pollution index in urban areas of mega city Lahore: a GIS-based approach. Sci Rep; 2024; 14,
Lee, Mi-Hee; Lee, Yun Kyung; Derrien, Morgane; Choi, Kwangsoon; Shin, Kyung Hoon; Jang, Kyoung-Soon; Hur, Jin. Evaluating the contributions of different organic matter sources to urban river water during a storm event via optical indices and molecular composition. Water Res; 2019; 165, 115006.1:CAS:528:DC%2BC1MXhs1GitrrE [DOI: https://dx.doi.org/10.1016/j.watres.2019.115006]
Levin, R; Villanueva, CM; Beene, D; Cradock, AL; Donat-Vargas, C; Lewis, J; Martinez-Morata, I; Minovi, D; Nigra, AE; Olson, ED; Schaider, LA; Ward, MH; Deziel, NC. US drinking water quality: exposure risk profiles for seven legacy and emerging contaminants. J Expo Sci Environ Epidemiol; 2024; 34,
Li, P; Qian, H. Water resources research to support a sustainable China. Int J Water Resour Dev; 2018; 34,
Liu, CW; Lin, KH; Kuo, YM. Application of factor analysis in the assessment of groundwater quality in a Blackfoot disease area in Taiwan. Sci Total Environ; 2003; 313,
Mao, H; Wang, G; Liao, F; Shi, Z; Zhang, H; Chen, X; Qiao, Z; Li, B; Bai, Y. Spatial variability of source contributions to nitrate in regional groundwater based on the positive matrix factorization and Bayesian model. J Hazard Mater; 2023; 445, 1:CAS:528:DC%2BB38XjtVyhs7fE [DOI: https://dx.doi.org/10.1016/j.jhazmat.2022.130569] 130569.
Masood, A; Aslam, M; Pham, QB; Khan, W; Masood, S. Integrating water quality index, GIS and multivariate statistical techniques towards a better understanding of drinking water quality. Environ Sci Pollut Res Int; 2022; 29,
Nayak, A; Matta, G; Uniyal, DP. Hydrochemical characterization of groundwater quality using chemometric analysis and water quality indices in the foothills of Himalayas. Environ Dev Sustain; 2023; 25,
Nong, X; Shao, D; Zhong, H; Liang, J. Evaluation of water quality in the South-to-North Water Diversion Project of China using the water quality index (WQI) method. Water Res; 2020; 178, 1:CAS:528:DC%2BB3cXotVGms7Y%3D [DOI: https://dx.doi.org/10.1016/j.watres.2020.115781] 115781.
Rahman, MM; Haque, T; Mahmud, A; Al Amin, M; Hossain, MS; Hasan, MY; Shaibur, MR; Hossain, S; Hossain, MA; Bai, L. Drinking water quality assessment based on index values incorporating WHO guidelines and Bangladesh standards. Phys Chem Earth Parts A B C; 2023; 129, [DOI: https://dx.doi.org/10.1016/j.pce.2022.103353] 103353.
Rashid, A; Ayub, M; Gao, X; Khattak, SA; Ali, L; Li, C; Ahmad, A; Khan, S; Rinklebe, J; Ahmad, P. Hydrogeochemical characteristics, stable isotopes, positive matrix factorization, source apportionment, and health risk of high fluoride groundwater in semiarid region. J Hazard Mater; 2024; 469, 1:CAS:528:DC%2BB2cXlvFCitL8%3D [DOI: https://dx.doi.org/10.1016/j.jhazmat.2024.134023] 134023.
Reff, A; Eberly, SI; Bhave, PV. Receptor modeling of ambient particulate matter data using positive matrix factorization: review of existing methods. J Air Waste Manag Assoc; 2007; 57,
Subba Rao, NS; Das, R; Sahoo, HK; Gugulothu, S. Hydrochemical characterization and water quality perspectives for groundwater management for urban development. Groundw Sustain Dev; 2024; 24, [DOI: https://dx.doi.org/10.1016/j.gsd.2023.101071] 101071.
Syeed, MMM; Hossain, MS; Karim, MR; Uddin, MF; Hasan, M; Khan, RH. Surface water quality profiling using the water quality index, pollution index and statistical methods: a critical review. Environ Sustain Indic; 2023; 18, [DOI: https://dx.doi.org/10.1016/j.indic.2023.100247] 100247.
Tiwari, AK; Pal, DB. Nutrients contamination and eutrophication in the river ecosystem. Ecological significance of river ecosystems; 2022; Amsterdam, Elsevier: pp. 203-216. [DOI: https://dx.doi.org/10.1016/B978-0-323-85045-2.00001-7]
Tokatlı, C; Varol, M; Ustaoğlu, F; Muhammad, S. Pollution characteristics, sources and health risks assessment of potentially hazardous elements in sediments of ten ponds in the Saros Bay region (Türkiye). Chemosphere; 2023; 340, 1:CAS:528:DC%2BB3sXhvVSkurjK [DOI: https://dx.doi.org/10.1016/j.chemosphere.2023.139977] 139977.
Uddin, MG; Nash, S; Rahman, A; Olbert, AI. A comprehensive method for improvement of water quality index (WQI) models for coastal water quality assessment. Water Res; 2022; 219, 1:CAS:528:DC%2BB38Xht1ansLrF [DOI: https://dx.doi.org/10.1016/j.watres.2022.118532] 118532.
Unigwe, CO; Egbueri, JC. Drinking water quality assessment based on statistical analysis and three water quality indices (MWQI, IWQI and EWQI): a case study. Environ Dev Sustain; 2023; 25,
Varol, M; Tokatlı, C. Evaluation of the water quality of a highly polluted stream with water quality indices and health risk assessment methods. Chemosphere; 2023; 311,
Wang, X; Zhang, M; Liu, L; Wang, Z; Lin, K. Using EEM-PARAFAC to identify and trace the pollution sources of surface water with receptor models in Taihu Lake Basin, China. J Environ Manage; 2022; 321, 1:CAS:528:DC%2BB38XitlahsbbF [DOI: https://dx.doi.org/10.1016/j.jenvman.2022.115925] 115925.
Zhang, Q; Zhang, J; Wang, H; Zhai, T; Liu, L; Li, G; Xu, Z. Spatial patterns in water quality and source apportionment in a typical cascade development river southwestern China using PMF modeling and multivariate statistical techniques. Chemosphere; 2023; 311,
Zhang, H; Ren, X; Chen, S; Xie, G; Hu, Y; Gao, D; Tian, X; Xiao, J; Wang, H. Deep optimization of water quality index and positive matrix factorization models for water quality evaluation and pollution source apportionment using a random forest model. Environ Pollut; 2024; 347, 1:CAS:528:DC%2BB2cXmt1Grsbg%3D [DOI: https://dx.doi.org/10.1016/j.envpol.2024.123771] 123771.
Zubaidi, SL; Hashim, K; Ethaib, S; Al-Bdairi, NSS; Al-Bugharbee, H; Gharghan, SK. A novel methodology to predict monthly municipal water demand based on weather variables scenario. J King Saud Univ Eng Sci; 2022; 34,
© The Author(s) 2025. This work is published under http://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.