Enhanced Assessment of Water Quality and Pollutant Source Apportionment Using APCS-MLR and PMF Models in the Upper Reaches of the Tarim River

Abstract

To evaluate the pollution sources and dynamics of the upper reaches of the Tarim River, 10 typical sampling points were selected, and 23 water quality parameters from 2020 to 2022 were analyzed using one-way analysis of variance, the comprehensive Water Quality Identification Index (WQI), and Principal Component Analysis (PCA). The pollution status, sources, and contribution rates of water quality were investigated using the Absolute Principal Component-Multiple Linear Regression Model (APCS-MLR) and Positive Matrix Factorization (PMF). The results indicated that the water quality parameters of dissolved oxygen (DO), chemical oxygen demand (COD_Mn), biochemical oxygen demand after 5 days (BOD₅), total nitrogen (TN), total phosphorus (TP), fluoride ions (F⁻), and ammonia-nitrogen (NH₃-N) in the upper reaches of the Tarim River exceed standards, with noticeable spatial variations observed for each parameter. The water quality evaluation grades in the upper reaches of Tarim River primarily indicate “moderate” and “good” levels, with DO, TN, NH₃-N, and electrical conductivity (EC) being the key parameters influencing variations in water quality. The source analysis results from APCS-MLR and PMF yielded similar outcomes, identifying six potential pollution sources. Among these, soil weathering, livestock and poultry breeding, and agricultural activities exhibited higher contribution rates. Specifically, the contribution rates for these sources according to APCS-MLR were 44.11%, 19.63%, and 11.67%, respectively; while according to PMF they are 24.08%, 17.88%, and 27.54%, respectively. Furthermore, industrial pollution sources contributed at a rate of 6.01% according to APCS-MLR, while urban living sources contributed at a rate of 2.13%. However, based on PMF analysis, the contribution rates for industrial pollution sources increased significantly to 16.71%. Additionally, APCS-MLR identified natural sources as contributing at a rate of 16.45%, whereas PMF suggested that a combination of agricultural activities and natural sources contributed at a lower rate of only 9.52%. In conclusion, the water quality within the upper reaches of the Tarim River is predominantly satisfactory. Nonetheless, localized pollution, primarily attributable to human activities, presents a substantial challenge. These observations provide critical insights into improving and protecting the fragile water quality of the Tarim River.

1. Introduction

The water resources of river basins play a crucial role in supporting the production and livelihoods of residents, particularly in arid and semi-arid regions [1]. With the rapid industrialization and urbanization, the deterioration of water quality has emerged as a significant global environmental issue, seriously impacting human and ecological health as well as the sustainable development of socio-economies [2,3]. The quality of river water is primarily influenced by natural factors such as climate, landform, soil use type, and anthropogenic factors including domestic sewage, agricultural activities, industrial wastewater, and transportation [4,5]. Accurately identifying the main pollution sources is essential for effectively investigating and managing the temporal and spatial changes in river water quality and improving the quality of the water environment.

The spatial-temporal variation characteristics in water quality and the identification of potential pollution sources pose challenges due to the complex nature of monitoring water quality parameters [6]. Multivariate statistical techniques offer distinct advantages in interpreting extensive and complex water quality datasets. They are widely used to evaluate spatial-temporal variations in water quality and to identify sources of water pollution, employing methods such as cluster analysis (CA), factor analysis (FA), and principal component analysis (PCA) [7,8,9,10]. CA has been effectively utilized to group sampling points or water quality indicators into clusters based on their similarities [7,8]. For example, in the case of the Shuangji River in China, CA was applied to classify 14 sampling points collected over 12 months into three distinct categories. This classification was based on the similarity of water quality characteristics and pollution levels, providing a solid foundation for the categorization of surface water in the study area [7]. PCA/FA, a method for reducing the dimensionality of complex datasets, is widely used to analyze seasonal variations in extensive water quality data [9,10]. This technique helps in identifying the principal components or factors that explain the majority of the variance in the data, simplifying the analysis of water quality trends and enabling more focused environmental management efforts. In the North Anhui Plain, PCA/FA analysis identified the main sources of surface water pollution and assessed the impact of these major pollution sources on water quality [11]. While PCA/FA are commonly used to qualitatively identify pollution sources, they lack the capability to quantitatively assess the contribution of these sources [11,12,13]. Howladar et al. found that PCA was used only to identify river water quality primarily affected by industrial waste, sewage pollution, surface runoff, organic compounds, and agricultural activities [12]. Bimol et al. used the PCA to effectively reduce the number of parameters and derived the critical parameters, which were pH, temperature, DO, BOD, total solids, total alkalinity, and turbidity [13]. The receptor model of absolute principal component-multiple linear regression (APCS-MLR) enhances this by incorporating the qualitative recognition results from PCA, enabling a quantitative analysis of pollution source contributions [14,15]. Memet et al. found that source apportionment in the APCS-MLR model revealed that seasonal and anthropogenic sources contributed 35.2% and 25.5% to river water quality parameters, respectively, followed by phytoplankton (21.4%) and natural sources (17.9%) [16]. Peng et al. found that the APCS-MLR model was used to analyze the pollutants in urban agricultural land from animal manure and natural sources (31.17%), agricultural activities (19.6%), and industrial activities (2.72%), respectively [17]. Positive Matrix Factorization (PMF) considers the uncertainty associated with each data point and imposes non-negativity constraints to ensure that the contributions from pollution sources are positive. However, it does not provide a method to determine an optimal number of pollution sources [18]. Mohammad et al. used the PMF model to assess the water quality and apportion the contributions of different potential pollution sources to each water quality variable in three major rivers of South Florida [19]. PMF and APCS-MLR models, as the two main tools in the field of exogenous analysis, have shown extensive application value. However, due to geographical differences and the complexity of different media (such as rivers, lakes, reservoirs, etc.), a single model is often difficult to fully and accurately analyze pollution sources. Additional information is needed in order to make full use of the complementarity of the two models, improve the reliability and accuracy of analytical results, meet the needs of practical applications, and provide readers with more comprehensive and accurate information support. By integrating PCA/FA with the APCS-MLR and PMF models, the capacity to identify potential pollution sources in rivers is enhanced, providing a robust scientific basis for controlling water environmental pollution.

The Tarim River is located in the heart of southern Xinjiang and faces drought conditions and a fragile ecological environment. It plays a crucial role in ecological protection and socio-economic development under the “Belt and Road” initiative. Despite rapid socio-economic advances, increasing environmental pressures have led to significant degradation of the river’s water environment function. Current research on water pollution in the Tarim River primarily focuses on water quality assessment and lacks comprehensive analysis of pollution sources. This study evaluates the water quality in the upper reaches of the Tarim River, specifically from Alar to Shaya section, analyzing 23 water quality parameters from March to April in 2020–2022 across 10 monitoring points. The employed methods include one-way analysis of variance, a comprehensive Water Quality Identification Index (WQI), and correlation analysis. These methods pinpoint key factors in the inland river’s water quality changes in extremely arid regions. The APCS-MLR and PMF methods identified six potential pollution sources, ranked differently, and the effective mechanisms, rules, and causes were explained. Although the Positive Matrix Factorization (PMF) method is extensively utilized in air pollution research, its application in pollution source allocation within the field of water resources remains sparsely reported. Specifically, comprehensive traceability analysis of pollution sources in river basins is seldom conducted. The results can be used to compare the applicability of the APCS-MLR and PMF models in source analysis, thus providing a scientific basis for water pollution prevention and control, promoting sustainable development of water environments, and facilitating regional sustainability in the Tarim River. This study is also an important supplement to the ecological security assessment of inland river basins in extreme arid areas and has great significance for ecological security regulation and sustainable socio-economic development of inland river basins in extreme arid areas.

2. Materials and Methods

2.1. Study Area

The Tarim River, the largest inland river in China, lacks its own natural flow and extends over approximately 1321 km as its main channel. Its drainage basin forms a relatively autonomous hydrological region, characterized by a closed inland water cycle and water balance [20]. Adjacent to the Tarim River, a pronounced evaporation and concentration effect is observed, leading to the sequential precipitation of salts with low solubility, such as calcite. Concurrently, ions from more readily soluble salts, including those found in rock salt and gypsum, increasingly dominate the water’s composition. This process results in the establishment of a Ca-Mg-SO₄-HCO₃ hydrochemical type, characterized by a complex ionic composition in the surface water. This phenomenon significantly influences the hydrochemical dynamics and environmental characteristics of the Tarim River’s vicinity. The distribution area of brackish water covers 2163.45 km², accounting for 77.26% of the total area [21]. The Tarim River possesses the dual attributes of abundant natural resources and a delicate ecosystem, serving as a vital lifeline for sustaining both the economy and natural ecology of the oasis in the Tarim Basin. It also supports the livelihoods of all ethnic groups in Xinjiang and holds significant representativeness in studies on water resources in arid regions [22,23,24]. The study area, ranging from 40°28′57″ to 40°54′19″ N and 80°56′50″ to 82°24′04″ E, extends from Taheyuan, the confluence of Hetian River and Aksu River in the west, to No. 3 Bridge in Shaya County, Aksu, in the east (Figure 1). The primary land use types in the region include arable land, forested areas, grasslands, and various surface water bodies. The arable land is predominantly cultivated with crops such as cotton, dates, and apples, with drip irrigation being the principal irrigation method. The population in the basin has been steadily increasing due to the rapid development of the socio-economy. Agriculture and animal husbandry have also experienced steady growth. However, this has led to the continuous discharge of domestic sewage, farmland drainage, livestock breeding wastewater, and industrial wastewater into the river throughout the year, exacerbating pollution within the basin and deteriorating environmental water quality.

2.2. Sampling Collection and Analysis

Water samples were systematically collected from ten designated monitoring sites (P1–P10) in the basin, which varied by land use type and hydrotopographic conditions. From 2020 to 2022, the sampling was conducted at an average annual frequency of once per year. In total, 90 representative samples were successfully obtained. The entire process, including sample collection, storage, and pretreatment, was meticulously carried out in strict accordance with the standard [25]. The 23 selected water quality parameters included pH, electrical conductivity (EC), total hardness (TH), dissolved oxygen (DO), chemical oxygen demand (COD_Mn), 5-day biochemical oxygen demand (BOD₅), turbidity (TU), total phosphorus (TP), total nitrogen (TN), ammonia nitrogen (NH₃-N), fluoride (F⁻), suspended solids (SS), nitrate nitrogen (NO₃-N), nitrite nitrogen (NO₂-N), arsenic (As), cadmium (Cd), chromium (Cr), Cuprum (Cu), iron (Fe), manganese (Mn), nickel (Ni), lead (Pb), and zinc(Zn), which were analyzed at the Tarim Oasis Agriculture Key Laboratory of Ministry of Education in Tarim University. The water quality indicators were tested and analyzed according to the standard method outlined in the “Water and Wastewater Monitoring and Analysis Method” (fourth edition) [26]. A reagent blank was included for all test items, and each determination was repeated three times to ensure the reliability of the experimental results.

2.3. Water Quality Assessment

2.3.1. The Nemero Pollution Index

This method can reflect the comprehensive harm degree of the water body in the study area. The calculation formula can be expressed as:

$$P_n=\sqrt{{\displaystyle \frac{p_{imax}^2+P_{iave}^2}{2}}}$$

(1)

where $P_n$ is Nemero index, $P_{imax}$ is the maximum value of single factor index of water quality parameters $i$, and $P_{iave}$ is the average value of the single factor index of water quality parameters $i$.

2.3.2. WQI

The PCA is used for objective screening of representative water quality evaluation indices. The WQI proposed by Pesce and Wunderlin [27] was selected for water quality assessment using 23 water quality parameters combined with “Environmental quality standards for surface water” [28]. The WQI calculation formula can be expressed as:

$$WQI={\displaystyle \frac{\sum _{i=1}^n{C}_i{P}_i}{\sum _{i=1}^n{P}_i}}$$

$$\left\{\begin{array}{c}{C}_i=100-\left[{\displaystyle \frac{\left(X_i-S_{i,k}\right)}{\left(S_{i,k+n}-S_{i,k}\right)}}\times 20m+I_{i,k}\right] S_{i,k}\le X_i<S_{i,k+n}\\ C_i=100-{\displaystyle \frac{X_i}{S_{i,k+n}}}\times 20m 0\le X_i<S_{i,k}\end{array}\right.$$

(2)

where $n$ is the number of water quality parameters, $C_i$ is the normalized value of $i$, $P_i$ is the weight of $i$ (1 < $P_i$ < 4) (

Table 1. Normalized values (C_i) and weights (P_i) of water quality parameters.

Index	Pi	Ci
		100	90	80	70	60	50	40	30	20	10	0
DO (mg·L−1)	4	≥7.5	>7	>6.5	>6	>5	>4	>3.5	>3	>2	≥1	<1
CODMn (mg·L−1)	3	<5.0	<10	<20	<30	<40	<50	<60	<80	<100	≤150	>150
BOD5 (mg·L−1)	3	<0.5	<2	<3	<4	<5	<6	<8	<10	<12	≤15	>15
TN (mg·L−1)	3	<0.1	<0.2	<0.35	<0.5	<0.75	<1	<1.25	<1.5	<1.75	≤2	>2
TP (mg·L−1)	4	<0.5	<2	<3	<4	<5	<6	<8	<10	<12	≤15	>15
NH3-N (mg·L−1)	3	<0.01	<0.05	<0.1	<0.2	<0.3	<0.4	<0.5	<0.75	<1	≤1.25	>1.25
NO₃-N (mg·L−1)	2	<0.5	<2	<4	<6	<8	<10	<15	<20	<50	≤100	>100
NO2-N (mg·L−1)	2	<0.005	<0.01	<0.03	<0.05	<0.1	<0.15	<0.2	<0.25	<0.5	≤1	>1
pH	1	7	>7 ≤8	>8 ≤8.5	>8.5 ≤9	≥6.5 <7	≥6, <6.5 >9, ≤9.5	≥5, <6 >9.5, ≤10	≥4, <5 >10, ≤11	≥3, <4 >11, ≤12	≥2, <3 >12, ≤13	≥1, <2 >13, ≤14
TU (NTU)	2	<5	<10	<15	<20	<25	<30	<40	<60	<80	≤100	>100
EC (μS/cm)	1	<750	<1000	<1250	<1500	<2000	<2500	<3000	<5000	<8000	≤12,000	>12,000
SS (mg·L−1)	3	<20	<40	<60	<80	<100	<120	<160	<240	<320	≤400	>400
Index	Pi	Environmental quality standards for surface water [28]
		Ⅰ		Ⅱ		Ⅲ		Ⅳ		Ⅴ
		Ii,1 = 20		Ii,2 = 40		Ii,3 = 60		Ii,4 = 80		Ii,5 = 100
As (μg·L−1)	4	50		50		50		100		100
Pb (μg·L−1)	4	10		10		50		50		100
Cd (μg·L−1)	3	1		5		5		5		10
Cr (μg·L−1)	3	10		50		50		50		100
Cu (μg·L−1)	2	10		1000		1000		1000		1000
Zn (μg·L−1)	2	50		1000		1000		2000		2000
Fe (μg·L−1)	1	300
Mn (μg·L−1)	1	100
Ni (μg·L−1)	1	20
TH (μg·L−1)	2	150		300		450		550		550
F− (μg·L−1)	2	1		1		1		1.5		1.5

) and it builds upon the work of other scholars [27,28,29,30,31,32], $X_i$ is the actual concentration of $i$, $S_{i,k}$ and $S_{i,k+n}$ are the threshold concentration for water quality standards in class $k$ and $k+n$, respectively, $I_{i,k}$ is the normalized threshold concentration for water quality standards, and m is the number of equal threshold concentration for the water quality standard (if no equal threshold, m = 1). Based on the WQI, water quality is classified into five categories: excellent (91–100), good (71–90), medium (51–70), poor (26–50), and very poor (0–25), corresponding to Class I to V water bodies, respectively, as defined in environmental quality standard for surface water [28].

2.4. Pollution Source Resolution

2.4.1. APCS-MLR

The APCS-MLR model is designed to establish a multiple linear regression model that correlates absolute principal factor scores (APCSs), transformed by PCA factor scores, with water quality parameters. It quantitatively analyzes the contribution of identified pollution sources to the monitoring indicators in the receptor. The APCS-MLR method is extensively employed in the analysis of pollution sources across various environmental media [33,34,35]. The calculation formula can be expressed as:

$$C_i=b_{\mathrm{o}i}+{\sum }_{p=1}^n\left(b_{pi}\times \mathrm{A}\mathrm{P}\mathrm{C}{\mathrm{S}}_P\right)$$

(3)

where $C_i$ is the concentration contribution of water quality parameters $i$, $b_{\mathrm{o}i}$ is the multiple regression constant of MLR on $i$, $b_{pi}$ is the multiple regression coefficient of pollution source $p$ on $i$, and $\mathrm{A}\mathrm{P}\mathrm{C}{\mathrm{S}}_P$ is the absolute principal factor score of the pollution source $p$.

2.4.2. PMF

The PMF model, recommended by the US Environmental Protection Agency for source apportionment, identifies and quantifies pollution sources based on a dataset of receptor chemical composition [19]. It decomposes concentration matrix (X) into a source contribution matrix (G), a source composition moment matrix (F), and a residual matrix (E); the calculation formula can be expressed as:

$$X_{ij}={\sum }_{k=1}^p\left(g_{ik}{f}_{kj}\right)+e_{ij}$$

(4)

where $X_{ij}$ is the content of parameter $j$ in sample $i$, $g_{ik}$ is the contribution of the pollution source $k$ to the parameter $i$, $f_{kj}$ is the concentration of $j$ in the pollution source $k$, $e_{ij}$ is the residual matrix, and it is computed from the minimum value of the objective function Q. PMF is capable of incorporating the uncertainty associated with various water quality parameters in the sample.

2.5. Data Processing and Analysis

Data processing was performed using Excel 2019. The APCS-MLR model was constructed with SPSS 21.0, and the PMF model calculations were carried out using EPA PMF 5.0. Graphs, and maps were produced using Origin 2021 and ArcGIS 10.8.

3. Results and Discussion

3.1. Concentration and Spatio-Temporal Distribution Characteristics of Water Quality Parameters

3.1.1. Statistical Analysis of Water Quality Parameters

The statistical analysis of the water quality parameters in the study area is presented in

Table 2. Descriptive statistics of water quality parameters in the upper reaches of Tarim River.

Parameters	Range	Average ± SD	Coefficient of Variation/%	Standard	Rate of Excess/%
DO (mg·L−1)	5.04–11.62	8.68 ± 1.15	13.28	6	3.33
CODMn (mg·L−1)	3.14–17.30	8.72 ± 3.35	38.43	15	10.00
BOD5 (mg·L−1)	2.06–8.20	3.74 ± 1.21	32.48	3	66.67
TN (mg·L−1)	0.29–1.48	0.64 ± 0.29	45.43	0.5	60.00
TP (mg·L−1)	0.097–0.34	0.18 ± 0.06	34.22	0.1	93.33
NH3-N (mg·L−1)	0.17–1.35	0.51 ± 0.22	43.90	0.5	53.33
F− (mg·L−1)	0.35–1.07	0.58 ± 0.26	45.35	1	6.67
NO3-N (mg·L−1)	0.47–3.13	1.01 ± 0.79	78.39	10	0
NO2-N (mg·L−1)	0.005–0.29	0.07 ± 0.06	88.33	-	-
pH	6.88–8.60	7.97 ± 0.42	5.26	6–9	-
SS (mg·L−1)	241.31–850.47	473.85 ± 172.61	36.43	-	-
TH (mg·L−1)	152.31–493.50	303.04 ± 64.63	21.33	300	46.67
TU (mg·L−1)	157.36–1657.24	661.47 ± 284.11	42.95	-	-
EC (mg·L−1)	0.29–1.29	0.71 ± 2.67	37.43	-	-
As (μg·L−1)	1.04–9.42	3.63 ± 2.25	62.00	50	0
Pb (μg·L−1)	1.04–9.87	3.88 ± 2.54	65.43	10	0
Cd (μg·L−1)	0.07–0.92	0.34 ± 0.25	75.15	5	0
Cr (μg·L−1)	0.40–2.70	1.40 ± 0.58	41.73	50	0
Cu (μg·L−1)	1.19–7.78	3.69 ± 1.59	43.12	1000	0
Zn (μg·L−1)	3.01–27.06	11.89 ± 6.24	52.50	1000	0
Fe (μg·L−1)	6.00–49.70	18.41 ± 8.58	46.58	300	0
Mn (μg·L−1)	1.20–14.53	5.19 ± 3.13	60.40	100	0
Ni (μg·L−1)	0.11–7.91	4.45 ± 1.78	39.91	20	0

Note: The standard is “Environmental quality standards for surface water” [28].

. The average of pH is 7.97, which indicates a neutral to slightly alkaline water condition, which is related to the alkaline environment in arid areas of China. In arid regions, soils typically exhibit alkaline or strongly alkaline properties, characterized by pH values exceeding 7.0, and in some instances, surpassing 8.5. Research has demonstrated that the majority of surface water samples collected from the southern margin of the Junggar Basin in Xinjiang fell within the range of weak alkalinity, with pH levels spanning from 7.72 to 9.36 and an average pH of 8.39. This evidence highlights the prevalent alkaline conditions of both soil and surface water in these arid environments [36]. The average of SS is 473.85 mg·L⁻¹, suggesting a high level of turbidity in the river. The observed yellow tint of the Tarim River’s water suggests that the predominant form of suspended particulate matter is silt. This silt is primarily derived from soil erosion and weathering processes, as well as from human activities occurring within the mountainous regions upstream of the river and along its course. These factors collectively contribute to the presence of silt in the river, influencing its coloration and indicating the interaction between natural processes and human interventions in the river’s ecosystem. The average of TH is 303.04 mg·L⁻¹, indicating relatively high hardness. The average of TU is 661.47 NTU, indicating high turbidity due to the significant sediment from meltwater carrying sand through the desert areas of the Tarim River. Heavy metals concentrations are significantly below the surface water quality standard I [28], with Fe exhibiting the highest average concentration, while the concentrations of Zn, Mn, Ni, Pb, Cu, As, Cr, and Cd decrease successively. Concentrations of DO, COD_Mn, BOD₅, TN, TP, F⁻, and NH₃-N in certain monitoring sites exceeded the class II water quality standard [28]. Specifically, DO is a pivotal indicator reflecting the ecological status of riverine ecosystems [37,38]; DO ranged from 5.04 to 11.62 mg·L⁻¹, with an average of 8.68 mg·L⁻¹, with 3.33% of the samples exceeded the class II water quality standard. COD_Mn serves as a crucial metric for assessing the presence of organic pollutants within aquatic bodies [39]; COD_Mn varied between 3.14 and 17.30 mg·L⁻¹, averaging 8.72 mg·L⁻¹, with 10.00% of the samples having COD_Mn that exceeded the water quality standard for class II of China, indicating that the impact of human activities on the organic pollution of water bodies is not pronounced. The average of F⁻ is 0.58 mg·L⁻¹, with 6.67% exceeded the standard for class II. The mean values of NO₃-N, NO₂-N, and EC were 1.01 mg·L⁻¹, 0.07 mg·L⁻¹, and 0.71 mS·cm⁻¹, respectively. The concentration of NO₃-N did not exceed the limit for Class II water quality. However, there is no specified concentration limit for NO₂-N and EC in the standard. EC directly affects the quality of drinking water and irrigation water. In the context of potable water, an excessively high EC value may indicate that the mineral concentration surpasses acceptable standards, rendering prolonged consumption potentially harmful to health. Within the agricultural sector, EC values that are either significantly high or low can adversely affect the nutrient uptake and physiological operations of crops. This underscores the critical importance of monitoring and managing EC levels to safeguard human health and ensure optimal crop growth. The EC of clean river water is generally within the range of 0.1 mS·cm⁻¹−0.3 mS·cm⁻¹, and it should be noted that the upper reaches of Tarim River have significantly exceeded this standard. In addition, the mean concentrations of BOD₅, TN, TP, and NH₃-N were 3.74 mg·L⁻¹, 0.64 mg·L⁻¹, 0.18 mg·L⁻¹, and 0.51 mg·L⁻¹, while the rate of non-compliance with class II water quality standards was 66.67%, 60.00%, 93.33%, and 53.33%, respectively. Nutrients (TP, TN) are limiting factors for the growth of freshwater algae and are the main reasons for the eutrophication of rivers [40]. When TP exceeds 75 µg·L⁻¹ and TN exceeds 1.5 mg·L⁻¹, rivers experience eutrophication [41]. The instances where samples of Biochemical BOD₅ and TP exceed the class IV water quality standards signal a significant human influence on the water quality of the Tarim River’s upper reaches [30]. This situation highlights the impact of anthropogenic activities on the river’s condition. That this is the case, results in Table 2 show that BOD₅, TN, TP, and NH₃-N occur in relatively high concentrations in this water which parameters are associated with effluents from urban, industrial, agricultural, and livestock wastewater. These findings underscore the need for comprehensive management and mitigation strategies to address the diverse sources of pollution affecting the river’s ecosystem [40,42].

The coefficient of variation (CV) (Table 2) reflects the degree of data dispersion and is commonly used to describe the spatial variability of pollutant concentrations. When the CV exceeds 50%, it indicates high variability [10]. Figure 2 shows the variation trend of water quality index in the upper reaches of the Tarim River, which not only shows the degree of data dispersion, but also reveals the spatiotemporal heterogeneity of water quality index. Strong exogenous inputs or complex driving factors are usually the main reasons for high variability [43]. The coefficients of variation for NO₃-N, NO₂-N, Cd, Pb, Zn, As, and Mn were relatively high (>50%) and highly discrete (Figure 2a), suggesting uneven spatial and temporal distribution largely influenced by human activities [44]. TN, TP, and NH₃-N have high coefficient of variation and overstandard rate. The degree of data dispersion is weak (Figure 2b). Studies have shown that the concentrations of NH₃-N, TP, and TN in rivers are often associated with human activities (such as irrigation runoff, industrial wastewater, and domestic sewage) [9]. The CV values for COD_Mn, F⁻, and BOD₅ are relatively lower (32.48–45.35%), but their exceedance rates are higher, and the data are concentrated (Figure 2c), which may also be related to the region’s continuous and stable point source pollution and geological background [9]. The pH and DO show the lowest coefficient of variation, indicating a highly uniform distribution and relative insensitivity to anthropogenic influence [4]. Some studies have shown that the pH value of rivers is not susceptible to short-term human activities due to the buffering effect and self-purification ability of water bodies. In summary, the water quality in the upper reaches of the Tarim River exhibits significant spatial and temporal differences, consistent with previous research findings [45]. This spatial and temporal distribution also indicates that the water quality in the study area is closely related to local policy adjustments and changes in the intensity of human activities [46,47]. The implementation of the river chief system, the introduction of local watershed water resource management regulations, and the execution of water pollution control plans have played a positive role in the management and protection of the Tarim River basin. However, with the rapid economic development, the proportion of construction land, agricultural land, and residential land in the upper reaches of the Tarim River basin has significantly increased, intensifying human activities. According to previous studies on the response relationship between water quality and land use, the increase in human land use area and the resultant discharge of industrial, agricultural, and domestic wastewater are likely key factors affecting regional water quality [48]. Although localized pollution incidents occur, the overall water quality in the Tarim River basin is generally satisfactory.

3.1.2. Temporal-Spatial Distribution Characteristics of Water Quality Parameters

The ANOVA in

Table 3. Variance analysis of influence of years on water quality index.

Parameters	DO	CODMn	BOD5	TN	TP	NH3-N	F−	NO₃-N	NO2-N	pH	SS	TH
p values	0.412	0.081	0.105	0.003	0.412	0.000	0.022	0.247	0.060	0.000	0.071	0.024
Parameters	TU	EC	As	Pb	Cd	Cr	Cu	Zn	Fe	Mn	Ni
p values	0.010	0.059	0.006	0.312	0.01	0.01	0.009	0.61	0.41	0.634	0.863

presents the impact of different years on water quality parameters in the upper reaches of the Tarim River. The results revealed significant variations in the concentrations of water quality parameters TN, NH₃-N, F⁻, pH, TH, TU, As, Cd, Cr, and Cu across different years. Conversely, no significant differences were observed for other water quality parameters (p > 0.05). Therefore, it can be inferred that either the pollutant input remained constant or the pollutant concentration was consistently low.

The Nemero pollution index at each sampling point in the upper Tarim River is shown in Figure 3. In 2020, The upper reaches of the Tarim River exhibited slight pollution, while in 2021 and 2022, the levels were predominantly mild to moderate. These findings suggest no significant deterioration in water quality over time within the upper reaches of the Tarim River. The Nemero pollution index of the sampling points follows the following order: TH2 (1.16) < TH1 (1.17) < TH9 (1.19) = TH3 (1.19) < TH5 (1.22) < TH7 (1.24) < TH4 (1.26) = TH8 (1.26) < TH6 (1.30) < TH10 (1.43). With the increase of the Nemero pollution index, the water quality gradually deteriorates. Among them, the TH4, TH8, and TH6 are primarily affected by agricultural activities and small aquaculture, and the increase in the application amount of fertilizers and pesticides and the discharge of aquaculture wastewater leads to pollutants entering rivers with surface runoff. Point TH10, located downstream of Alar City, faces increased pollutant levels due to various human industries and activities and extensive farming of cotton and jujube in the city. Over the past three decades, the Tarim River basin has experienced a continuous population growth, accompanied by a rapid expansion of arable and residential lands. This expansion has posed significant threats to water quality, with wastewater discharge increasing annually, thereby exerting a profound impact on the river’s water quality and aquatic ecology [49,50]. Research has discovered that natural environmental factors, along with agricultural nonpoint source pollution from various prefectures and production and construction corps, cause the water quality at the Aksu, Alar, and Shaya sites of the Tarim River during the dry season to occasionally fall to Category IV. This results in the monitored sites’ water quality not consistently meeting the Category III standard [51]. This seasonal pattern of water quality degradation is corroborated by the Environmental Status Bulletin of the Xinjiang Uygur Autonomous Region, indicating a precise circannual occurrence during the dry season. In recent years, the influence of human activities on global warming has led to climatic shifts in northwest China’s arid and semi-arid regions from warm and dry to warm and wet conditions. The Tarim Basin, historically characterized by extreme aridity, has witnessed multiple instances of large-scale rainfall events. Consequently, there has been a noticeable increase in the surface water area of the Tarim River Basin, reflecting significant hydrological and environmental changes in the region [52]. Figure 4 shows the distribution of population density in the upper reaches of the Tarim River. It can be seen that sampling sites with high pollution levels are predominantly situated in the middle and lower reaches of the area, likely due to their proximity to densely populated areas with intense human activities. This observation suggests a strong correlation between pollution and anthropogenic factors [46,53]. While all sampling sites exhibited slight pollution, suggesting a certain degree of contamination, the overall water quality remains satisfactory.

3.2. Water Quality Assessment Using WQI

WQI can provide a comprehensive assessment of the condition of river water quality [3,54,55]. The changes in WQI value and the annual distribution of each grade in the upper reaches of Tarim River from 2020 to 2022 are illustrated in Figure 5. The results indicate that throughout the three-year monitoring period, the WQI values consistently were within the range of 68 to 80, indicating good water quality [32]. However, a declining trend in WQI was observed. During the three years, the WQI values were, respectively, 76.68, 74.96, and 72.91 (Figure 5a). In 2021, the proportion of “medium” category increased significantly by 10%, while the proportion of “good” category decreased correspondingly to 90%. This shift could be attributed to the rapid pace of economic development and population growth in the Alar section located upstream of the Tarim River. The economic incentives and supportive national policies have significantly contributed to the expansion of cotton cultivation in Xinjiang, with the planted area soaring from 869,000 hectares in 2019 to 2.4787 million hectares by 2021. Concurrently, the animal husbandry sector has also witnessed a steady growth, registering a year-on-year increase of 6.5 percent as of 2022. Studies indicate that fertilizer nitrogen runoff is responsible for 22% to 57% of the total non-point source pollution across China’s major river basins. With the expansion of the scale of agriculture and livestock and poultry farming, the total amount of pollutants discharged within the basin increased. This assertion is supported by the “Ecological and Environmental Status Bulletin of Xinjiang Uygur Autonomous Region for 2021 and 2022”, which notes that while the proportion of Class V water quality sections rose by 1.4% from 2020 to 2021, there was no further deterioration between 2021 and 2022. This stability in water quality aligns with the findings of the study, indicating that the measures taken have successfully controlled the phenomenon, maintaining the proportion of water quality classified as “medium” and “good” without any decline. Previous studies have reported that most of the surface water quality in the Tarim River Basin is classified as “excellent” or “good” [22].

The Pearson correlation coefficient between 23 calculated indexes and WQI value is shown in

Table 4. Correlation analysis between water quality parameters and WQI values.

DO

CODMn

BOD5

TN

TP

NH3-N

F−

NO₃-N

NO2-N

pH

SS

TH

WQI

0.435 **

−0.147 *

−0.155

−0.608 **

−0.017

−0.582 **

−0.405 *

−0.023

−0.317

0.448 **

0.157

−0.132

TU

EC

As

Pb

Cd

Cr

Cu

Zn

Fe

Mn

Ni

WQI

−0.240

−0.366 **

0.199

−0.473 **

−0.460 *

−0.367 *

−0.316 *

−0.262

−0.036

0.171

−0.253

Note: * Significant correlation, ** Highly significant correlation.

. There was a significantly positive correlation between DO and WQI, indicating that higher levels of DO in water corresponded to higher WQI values and better overall water quality. Research indicates that dissolved oxygen (DO) levels in the Minjiang River are negatively correlated with most pollution indicators. This correlation may stem from the fact that organic wastes and nutrients contribute to eutrophication, fostering excessive algae growth. This growth, in turn, consumes substantial amounts of DO, thereby deteriorating water quality [56]. pH levels also display a positive correlation with the Water Quality Index (WQI), suggesting that within the neutral range, an increase in pH values facilitates the natural degradation process of pollutants. This alteration in pH affects the form, migration, and transformation of pollutants, thereby reducing their concentration and toxicity in water. Concentrations of TN, NH₃-N, EC, and Pb exhibited significant negative correlations with the WQI, indicating that as nutrient levels (TN, NH₃-N) and EC in water increased, there was a corresponding decrease in WQI values and a deterioration in water quality. The COD_Mn, F⁻, Cd, Cr, and Cu exhibited a negative correlation with the WQI, indicating that an increase in toxic substances led to a decline in WQI values [54,57]. Studies have indicated that high concentrations of TN and NH₃-N can lead to excessive algae growth and subsequent deterioration of water quality [46]. Additionally, an increase in the COD_Mn value signifies a heightened level of organic pollution in the water [58]. The accumulation of heavy metals in water, particularly at high levels, can pose significant risks to human health [59]. Consequently, these indicators—TN, NH₃-N, COD_Mn, and heavy metal concentrations—are typically negatively correlated with WQI. Among them, DO, TN, NH₃-N, and EC exhibit a strong correlation with WQI value, indicating that these four factors significantly influence the water quality in the upper reaches of Tarim River. Therefore, DO, TN, NH₃-N, and EC can be considered as key indicators affecting water quality.

The spatial variations in annual WQI values for 10 monitoring sites between 2020 and 2022 are shown in Figure 6. Over the last three years, the average WQI of the 10 monitoring sites ranged from 72.80 to 76.74, corresponding to a “good” water quality grade. Among these, the WQI values of TH2, TH5, TH6, and TH8 were all above 75.00, indicating favorable water quality in the cross sites. Both TH2 and TH8 were located far from areas with significant human activities, characterized by low population density and absence of farmland and industrial facilities. The land use types surrounding TH5 and TH6 primarily consisted of grassland and forest land, resulting in superior water quality in the rivers near these sampling sites compared to other locations. The water quality in sites TH4 and TH10 is generally inferior to that of the remaining eight sites. The poor water quality in TH4 section, near the 10 Group Sixth Company of Alar City, is related to various industrial, farming, and residential activities. Section TH10 is situated downstream of Tuanjie Farm Village in Shaya County, which is mainly agricultural. The spatial variation of WQI values at monitoring sites is uneven, and water quality in the monitoring section exhibits variations, with poor water quality primarily observed in areas characterized by intense human activities and agricultural practices [9]. This suggests that human disturbances and land use patterns significantly influence the spatial distribution of water quality in the upper reaches of the Tarim River basin [3,35]. According to the classification of WQI values, the water quality assessment grades in the upper reaches of Tarim River predominantly fall under “moderate” and “good”, with an overall average value of 74.85, indicating a favorable water quality status. The Spearman correlation test revealed that DO, TN, NH₃-N, and EC were identified as the primary indicators influencing upstream water quality.

3.3. Source Analysis of Water Environmental Pollutants

3.3.1. Correlation Analysis of Pollutants

Beyond water quality assessment, the quantitative evaluation of pollution sources is a crucial step towards improving water quality [56]. The closeness and similarity of pollutants in the upper reaches of Tarim River were investigated by employing Pearson correlation analysis to explore their homology relationship [60]. A significant positive correlation indicates that these elements may have similar sources or geochemical behaviors [61,62]. The thermal map of the correlation coefficients between pollutants (Figure 7) reveals that, at a significance level of 0.001, there exists a significant positive correlation between the factor pairs of COD_Mn-BOD₅ and TN-NH₃-N. The correlation between pH-NH₃-N was found to be significantly negative. At a significance level of 0.01, the factor pairs pH-TH, Pb-COD_Mn, Pb-BOD₅, and Cu-Pb exhibited positive correlations. A significant negative correlation was observed between pH-TN and TH-TP. At a significance level of 0.05, the factors F⁻-TN, NO₃-N-NO₂-N, TH-BOD₅, EC-TP, EC-NO₃-N, As-pH, As-TH, Pb-TH, Cd-TU, Cr-TN, Cr-NH₃-N, Cr-F⁻, Cu-EC, and Cu-Fe were significantly positively correlated. TN-DO, NH₃-N-DO, NH₃-N-SS, NH₃-N-TH, NO₂-N-TH, NH₃-N-As, and As-NO₂-N showed significant negative correlations. The stronger the phase relationship, the more significant the correlation becomes, indicating a higher likelihood of homology between the two pollutants [63,64]. This can also serve as a basis for selecting water quality indicators. The negative correlation between DO, pH, and most nutrient indicators such as NH₃-N, TN may be due to organic waste and nutrients causing eutrophication and excessive algae growth, consuming a large amount of DO, leading to an increase in pH value and nutrient accumulation [65,66]. SS and NH₃-N also show a negative correlation. The higher the content of suspended particulate matter, the faster the degradation of water ammonia nitrogen [67], because suspended particles can provide attachment points for nitrifying bacteria, increase their number in the water body, and then promote the degradation of ammonia nitrogen. Correlation analysis can only determine that some elements might be affected by the same pollution source. To further identify and quantify pollution sources, the APCS-MLR and PMF models are used to fit the sample data [56].

3.3.2. Pollutant Source Apportionment by APCS-MLR Model

(1) PCA/APCS analysis

The original data of selected water quality indices (excluding TU, Zn, Mn, and Ni) from all monitoring sites between 2020 and 2022 were standardized. The KMO−Bartlett sphericity test revealed a significant correlation between the indices (KMO = 0.55, p < 0.001), indicating suitability for conducting PCA analysis [19,60,68]. According to the Kaiser rule, six principal components were extracted based on eigen values greater than 1. The cumulative variance contribution rate reached 71.43%, indicating that a significant portion of the information contained in the pollutant index data within the basin has been captured (

Table 5. Principal component analysis of pollutants in the Tarim River Basin.

Project		PCA
		PC1	PC2	PC3	PC4	PC5	PC6
pollutant	DO	−0.475	0.303	−0.264	0.498	0.334	−0.102
	CODMn	0.222	0.830	−0.13	0.133	0.113	−0.241
	BOD5	−0.127	0.834	−0.231	−0.094	0.072	−0.079
	TN	0.801	−0.008	0.302	−0.106	0.016	0.067
	TP	−0.055	−0.278	0.106	0.105	0.723	−0.332
	NH3-N	0.146	−0.095	0.885	0.093	0.014	−0.093
	F−	0.798	0.011	0.034	0.183	0.024	0.018
	NO3-N	−0.168	0.134	0.181	0.252	0.707	−0.092
	NO2-N	−0.400	0.094	0.523	0.270	0.041	0.089
	pH	−0.216	0.172	−0.844	−0.104	−0.213	0.082
	SS	−0.111	−0.057	−0.186	−0.748	−0.088	−0.107
	TH	−0.164	0.305	−0.144	−0.796	−0.165	0.079
	EC	0.289	0.271	0.054	0.039	0.697	0.235
	As	−0.270	−0.11	−0.461	−0.106	−0.112	0.527
	Pb	−0.039	0.758	0.191	−0.167	−0.059	0.424
	Cd	0.062	−0.063	0.100	0.091	0.034	−0.079
	Cr	0.656	0.115	0.072	0.459	0.028	0.163
	Cu	0.309	0.109	−0.082	−0.132	0.400	0.538
	Fe	0.203	−0.055	−0.061	0.179	−0.115	0.818
Eigenvalue		3.78	2.70	2.52	1.70	1.59	1.28
Variance contribution rate %		19.91	14.22	13.25	8.96	8.38	6.72
Cumulative contribution rate %		19.91	34.13	47.38	56.33	64.71	71.43

). To enhance the prominence of typical index variables for each principal component, the rotation factor load moment obtained from PCA was transformed into an absolute principal factor load matrix (APCS) in Figure 8. The PCA/APCS model was employed to analyze pollutants in the upper reaches of Tarim River.

The results depicted in Figure 6 demonstrate that the variance contribution rate of APCS1 is 27.99%, indicating a significant impact, and it exhibits a strong positive load on TN, F⁻, and Cr (0.867, 0.837, and 0.751), respectively. Furthermore, the pairwise positive correlation between these variables is statistically significant, suggesting potential homology [19,69]. TN primarily originates from point sources (such as domestic sewage and industrial wastewater) and agricultural non-point sources [70]. The weathering of fluorine-rich minerals, the use of phosphorus-containing apatite fertilizers, and fluorine-containing pesticides are the main sources of fluoride in surface water and groundwater [10,71,72]. Natural weathering, chemical fertilizers, and industrial activities are important sources of heavy metals in water [73]. The rapid development of agriculture in the Tarim River Basin has led to an increase in the use of chemical fertilizers and pesticides, which not only introduce nutrients, F⁻ into rivers but also affect the content of heavy metals in the water, such as Cr, Cu, and Zn. The study area is an important production base for cotton and special fruit and forestry industries in China. Combined with the spatial distribution of land use and elements, the areas near the pollution-severe sampling points have a lot of farmland and forest land planted with crops such as cotton, walnuts, and jujubes, belonging to areas of concentrated agricultural activities. Since TN, F⁻, and Cr are closely related to fertilization and pesticide application in agricultural production activities, source 1 likely represents an agricultural activity source. Research indicates that frequent agricultural activities may cause fertilizers and pesticides to enter rivers through surface runoff, leading to high nitrogen, F⁻ concentrations due to non-point source agricultural pollution [74,75,76].

The APCS2 exhibits a strong positive loading on Pb, BOD₅, and COD_Mn (0.826, 0.810, and 0.772, respectively), contributing to a variance rate of 14.05%. This finding suggests a significant positive correlation among these variables, indicating the presence of a potential common source. COD_Mn and BOD₅ reflect the degree of water pollution by organic substances [77], primarily originating from the discharge of domestic sewage and industrial wastewater, as well as the decomposition of dead plants and animals. Areas with high BOD₅ concentrations have high levels of industrialization and urbanization, leading to an increase in organic materials in water due to the discharge of substantial amounts of domestic sewage and industrial wastewater [9]. With the rapid economic development in the upper reaches of the Tarim River, more and more domestic sewage and industrial wastewater are flowing into the Tarim River. Pb represents emissions from vehicle exhaust and the washing of fine particulates from parking lots and urban road areas [74]. The parameters Pb, BOD₅, and COD_Mn are primarily associated with anthropogenic activities and predominantly found in urban areas characterized by high population density. They serve as indicators of the extent of organic pollution in water bodies, thus suggesting that source 2 may represent the urban life source. Studies have found that the discharge of large amounts of industrial wastewater, urban sewage, and livestock and poultry waste due to industrial production, population growth, and livestock breeding is the main reason for the high concentrations of COD_Mn, COD_Cr, and BOD₅ in the water quality of rivers [4,9].

The variance contribution rate of APCS3 is 10.37%, exhibiting a robust positive loading on SS and TH (0.873 and 0.823, respectively), as well as a moderate positive loading on pH (0.674). Furthermore, there exists a statistically significant positive correlation between the two variables. The concentration of SS in water bodies can affect the propagation of light and is an important factor affecting the health of lake aquatic ecosystems [78]. SS and TH are associated with soil erosion and runoff activities, with wind speed and water dynamics being the main factors affecting SS and TH concentrations [10]. The water of the Tarim River primarily originates from mountain precipitation and alpine snowmelt, and the SS and TH in the river are part of the natural erosion and sediment transport processes. There is a large amount of salinized soil in the Tarim River Basin. The pH can be influenced by salinized soils. Active chemical exchanges between surface water, groundwater, and soil lead to an increase in water body pH, indicating that source 3 may be responsible for soil weathering.

The APCS4 exhibits a significant positive loading on NO₃-N and TP (0.798 and 0.769, respectively), as well as a moderate positive loading on DO (0.656). However, the correlation between these variables is not considered strong. Nitrogen and phosphorus are key factors causing eutrophication in water bodies [10]. Phosphorus mainly comes from fertilizers, agricultural waste, and livestock and poultry manure [79]. NO₃-N levels in water bodies are influenced by both natural and anthropogenic factors. Natural sources of nitrate nitrogen include nitrogen fixation by atmospheric processes and the nitrogen cycling within soils. Through these natural processes, nitrogen-containing compounds are generated and can enter aquatic systems via atmospheric deposition and the cycling of nitrogen in the soil [80]. Human activities significantly contribute to the levels of nitrate nitrogen found in water bodies, often surpassing the contributions from natural sources, especially in regions with intensive agricultural practice. The primary economic development models in the Tarim River Basin are agriculture and animal husbandry. The presence of NO₃-N and TP in the water is commonly associated with livestock farming activities and agricultural runoff. The research findings by Gao Jing [51] in the Tarim River Basin indicate that there has been a recent expansion in the scale of livestock and poultry farming within this basin. Additionally, a portion of wastewater is being discharged directly without undergoing any treatment. The anaerobic fermentation process occurring in DO aquaculture wastewater results in elevated levels of nitrate nitrogen content. Therefore, it is possible that source 4 may be attributed to livestock and poultry farming.

The APCS5 exhibits a variance contribution rate of 6.80% and displays a strong positive loading on Fe and As (0.756 and 0.737, respectively), as well as a moderate positive loading on Cu (0.600). Furthermore, there exists a significant positive correlation between Fe and Cu. Iron (Fe) is widely distributed in rock layers rich in Fe₂O₃, and through water-rock interaction, runoff formed by mountain snowmelt, precipitation, and fissure water in bedrock contains a large amount of Fe [36]. Arsenic (As) is a common element in the Earth’s crust, and almost all soils, water bodies, and sediments contain As, with the main sources of As in water including soil erosion and mineral smelting [81]. Copper (Cu) mainly originates from the weathering of rocks such as granite and carbonate rocks due to climatic and other non-anthropogenic factors. The presence of Fe, As, and Cu in the Earth’s crust is closely associated with the deposition and weathering processes of crustal rocks. The spatial distribution reveals more pronounced pollution at river confluences influenced by natural weathering along their course, suggesting that source 5 may represent a natural contributor.

At last, APCS6 exhibits a robust positive loading on Cd and NO₂-N (0.768 and 0.720) as well as a moderate positive loading on NH₃-N (0.603), although the correlation is not particularly strong. Wastewater discharge from industries such as mining, paper mills, and chemical plants may be a major contributor to the content of Cd (Cadmium) in rivers [82]. NH₃-N mainly originates from agricultural non-point sources, domestic sources, livestock and poultry breeding sources, and nearby industrial point sources [80]. The biochemical transformation of a large amount of organic matter in water bodies can easily lead to anoxic conditions, promoting the decomposition of organic matter to produce NO₂-N, resulting in an increase in the content of NO₂-N in the water. In recent years, with the rapid economic development in the upper reaches of the Tarim River and the increasing number of industries, the discharge of industrial wastewater has gradually increased. Previous reports have indicated that relatively high nitrogen concentrations in river water are positively correlated with GDP and the volume of industrial wastewater discharge [9], reflecting the impact of industrial production on river water quality. Given that Cd, NO₂-N, and NH₃-N are primarily associated with the discharge of diverse industrial wastewater in urban areas, it is plausible that source 6 represents an industrial pollution source.

(2) APCS-MLR analysis

Based on the qualitative identification of potential sources of pollution, the APCS-MLR model was developed to quantitatively calculate the contribution value and rate of each pollution source to the selected water quality index. The analytical value of the water quality index was then linearly fitted with the measured value, as presented in

Table 6. Source contributions of water quality indicators in the sediment based on APCS-MLR (mg·kg⁻¹).

Water Quality Indicators	S1 a	S2 b	S3 c	S4 d	S5 e	S6 f	Estimated Mean Concentration	Observed Mean Concentration	Estimated/Observed	R2
DO	1.06	0.35	1.21	4.91	0.37	0.38	8.27	8.68	0.95	0.888
CODMn	0.98	3.88	0.11	0.61	0.38	0.59	6.55	8.72	0.75	0.867
BOD5	0.23	2.85	0.22	0.27	0.38	0.03	3.98	3.74	1.07	0.929
TN	0.57	0.13	0.02	0.09	0.13	0.02	0.96	0.64	1.49	0.888
TP	0.00	0.01	0.01	0.13	0.01	0.00	0.17	0.18	0.92	0.858
NH3-N	0.10	0.02	0.08	0.03	0.03	0.44	0.70	0.50	1.39	0.919
F−	0.37	0.02	0.02	0.12	0.01	0.01	0.56	0.58	0.97	0.810
NO3-N	0.01	0.01	0.10	0.95	0.01	0.03	1.11	1.00	1.11	0.726
NO2-N	0.01	0.01	0.01	0.01	0.01	0.04	0.09	0.07	1.30	0.804
pH	0.71	0.68	2.68	1.57	0.60	0.08	6.31	7.97	0.79	0.884
SS	20.96	6.48	257.71	96.86	41.54	9.37	432.92	473.85	0.91	0.784
TH	22.57	12.07	175.58	49.94	17.14	12.61	289.92	303.04	0.96	0.91
EC	147.65	7.26	316.58	180.18	210.08	77.77	939.52	714.12	1.32	0.790
As	0.15	0.03	1.12	0.22	2.73	0.27	4.51	3.63	1.24	0.768
Pb	0.18	2.16	0.20	0.41	0.28	0.15	3.38	3.88	0.87	0.903
Cd	0.04	0.05	0.03	0.01	0.04	0.25	0.42	0.34	1.24	0.801
Cr	1.15	0.08	0.07	0.13	0.14	0.09	1.66	1.40	1.19	0.809
Cu	0.81	0.31	0.29	0.17	1.82	0.23	3.62	3.69	0.98	0.858
Fe	3.05	0.26	2.36	0.85	7.17	1.00	14.69	18.41	0.80	0.827

Note: ^a: Agricultural activity sources, ^b: Urban subsistence source, ^c: Soil weathering sources, ^d: Livestock breeding source, ^e: Natural sources, ^f: Industrial sources.

and Figure 6. Table 6 demonstrates that both the linear fitting R² of the water quality index between analytical and measured values exceed 0.70, while the p value is less than 0.05, indicating high accuracy and applicability of the APCS-MLR model [4,9,53]. Moreover, most water quality indexes exhibit a nearly perfect ratio of analytical to measured values close to 1, confirming excellent fit through multiple linear regression analysis and thus validating the capability of the APCS-MLR model in identifying major pollutant sources in the upper reaches of Tarim River [10,60,83].

The results depicted in Figure 9 demonstrate that S1 (agricultural activity source) significantly contributes to TN, F⁻, and Cr with rates of 45.16%, 58.15%, and 60.78%, respectively. Moreover, it contributes to Cu and Fe at rates of 21.04% and 20.06%, respectively, while its contribution to other water quality indexes ranges from 0.82% to 15.02%; this indicates that most of the water quality indexes are affected by agricultural living sources. The Second National Pollution Census Bulletin of China shows that agricultural sources of nitrogen discharge account for about 47% of the total nitrogen discharge in the country. Studies have found that the contribution rate of agricultural nitrogen emission sources to surface water nitrogen pollution is 19–61% globally, and the widespread application of nitrogen and organic fertilizers is the main source [84]. The research area is an important high-quality cotton production base in China and a demonstration base for the transformation and value-added of Xinjiang’s characteristic agricultural and sideline products. For many years, under the influence of the concept of high input and high yield, the application amount of nitrogen fertilizer, phosphate fertilizer, compound fertilizer and pesticide was generally high, which significantly increases the pollution load of nutrient elements, organic matter, and heavy metals in river water.

The contribution rate of S2 (urban domestic source) to COD_Mn, BOD₅, and Pb is 49.87%, 61.30%, and 47.16%, respectively. Additionally, the contribution rate of S2 (urban domestic source) to other water quality indexes ranges from 0.48% to 10.77%. These findings indicate the significant impact of this factor on various aspects such as domestic sewage, domestic garbage, transportation, and other daily activities in the main area, leading to abnormal levels of COD_Mn, BOD₅, and Pb in river water. In the upstream area of the Tarim River Basin, there are two main urban sewage discharge outlets: the Aksu City Sewage Treatment Plant discharge outlet (about 28 million tons year⁻¹) and the Alar City Sewage Treatment Plant discharge outlet (about 7 million tons year⁻¹) (data source: Environmental Protection Department’s big data on environmental statistics), which cause pollution to the water quality of the Tarim River during the dry season [51]. The Tarim River, characteristic of seasonal rivers, exhibits ample water flow and comparatively high water quality during the wet season. Conversely, in the dry season, the diminution of water sources and runoff, coupled with a decrease in the river’s intrinsic self-purification capabilities, can lead to an elevated concentration of pollutants per unit volume, even when the rate of pollutant discharge remains constant. Consequently, the water quality frequently deteriorates during these periods. The contribution of S3 (soil weathering source) to pH, SS, TH, and EC was 33.72%, 42.31%, 45.70%, and 32.20%, respectively. Additionally, the contribution rate of S3 to As and Fe reached 18.2% and 15.51%. Furthermore, the contribution rate to other water quality indexes ranged from 1.32% to 11.69%. The water in Tarim River primarily originates from alpine ice and snow meltwater while soil along its path is transported into the river through runoff after undergoing weathering processes. The alkaline nature of Xinjiang’s soil leads to an increase in pH, SS, TH, and EC levels in the water.

The contribution rate of S4 (livestock breeding source) to DO, TP, and NO₃-N was 47.21%, 68.69%, and 73.58%, respectively, with an average value of 63.16%. Its impact on other water quality indicators ranged from 3.25% to 19.77%. Previous studies have shown that livestock and poultry manure often contains large amounts of nitrogen and phosphorus [85]; regions with high nutrient content in the basin are mainly concentrated in areas with a high proportion of farmland and large-scale animal husbandry [9]. As the core area of the Silk Road Economic Belt, Xinjiang has increased its investment in animal husbandry in recent years, gradually expanding the scale of livestock production, and the pollution of river water bodies caused by the discharge of breeding wastewater has gradually attracted attention. In addition to agricultural runoff, livestock breeding serves as a significant contributor to nutrient pollution in the study area due to the continuous expansion of the livestock and poultry industry scale and the discharge of aquaculture wastewater.

The contribution rate of S5 (a natural source) to As, Cu, and Fe is significant, accounting for 44.08%, 47.43%, and 47.15%, respectively. Additionally, the contribution rate of S5 (a natural source) to EC can reach up to 21.37%. Moreover, its contribution rate to other water quality indexes ranges from 0.56% to 10.10%. In addition to agricultural irrigation, urban sewage, and livestock breeding sources, the region’s own geological environment and natural conditions are also factors causing water pollution. The Tarim River basin is characterized by an arid climate and limited precipitation, factors that contribute to the water body’s diminished self-purification capacity. Additionally, the geological conditions of the region result in a naturally high salt content within the water body, further complicating and increasing the challenges associated with water quality management. The average concentrations of As, Cu, and Fe in the study area are 3.63 μg L⁻¹, 3.69 μg L⁻¹, and 18.41 μg L⁻¹, respectively, which are far below the surface water quality standard of 50 μg L⁻¹ for As, 10 μg L⁻¹ for Cu, and 300 μg L⁻¹ for Fe, indicating that human activities have little impact on the concentrations of As, Cu, and Fe in the river. The upstream region of the Tarim River Basin generally has low vegetation coverage, strong sunlight, and accelerated weathering of rocks and soil, which is influenced by the unique environmental geochemical characteristics of the study area. Furthermore, the levels of As, Cu, and Fe in river bodies within the region are relatively high.

The contribution of S6 (an industrial pollution source) to NH₃-N, NO₂-N, and Cd was 47.10%, 36.03%, and 41.86%, respectively. The contribution rate of S6 (an industrial pollution source) to other water quality indicators ranged from 0.65% to 7.91%, indicating that this factor was primarily influenced by various industrial pollutions in the study area. Research has found that the economic characteristics of a basin directly affect water pollution in the basin, with NH₃-N, NO₂-N, and Cd showing a positive correlation with GDP and the volume of industrial wastewater discharge [9]. To improve river water quality, it is necessary to strengthen the control of NH₃-N emissions from industrial enterprises within the basin [46]. In recent years, the Tarim River Basin has vigorously developed its secondary industry, with steady growth in industrial output value and an annual increase in pollutant emissions, putting considerable pressure on the ecological environment of the upper reaches of the Tarim River. In addition, the six factors accounted for 73.88% of the dataset, while 26.12% of the unknown sources still remained, with a contribution rate to the water quality index ranging from 3.42% to 30.48%. This could be primarily attributed to multi-factor mixed pollution. The results of APCS-MLR analysis indicate that water quality can be influenced by many factors, with the water quality in the upper reaches of the Tarim River mainly affected by natural and anthropogenic factors. A total of 60.56% of water pollution can be attributed to natural factors, while the impact of human activities on water pollution accounts for 39.44%.

3.3.3. PMF Source Resolution

Based on the EPA PMF model, a quantitative source analysis of water pollutants in the upper reaches of Tarim River was conducted. The concentration data and associated uncertainties for 19 selected water quality indices were input into the model for factor iteration. When the number of factors was set to six, the Q value reached its minimum (74.5), and each water quality index exhibited sample residuals between −3 and 3 with a normal distribution, indicating satisfactory results from the PMF simulation [53,56,86]. Figure 10 illustrates both the concentration values and contribution rates of these six factors to the analyzed water quality indices as determined by the PMF model.

The contribution rates of the six factors are as follows: 27.54%, 4.27%, 16.71%, 24.08%, 9.52%, and 17.88%. F1 exhibits significant contributions to Cd, NH₃-N, NO₂-N, and TN with respective rates of 45.93%, 40.05%, 35.59%, and 33.18%. The presence of Cd may be attributed to the application of nitrogen and phosphorus fertilizers in farmlands [43], while nitrogen can originate from both point source pollution (such as industrial wastewater and domestic sewage) and non-point source pollution (including agricultural activities) [53]. However, considering the higher concentration levels observed in samples from surrounding farmlands, it is likely that F1 represents the primary source associated with agricultural activities. The contribution rate of F2 to NO₂-N is 63.99%, and it contributes 23.66% and 24.53% to NH₃-N and Cd, respectively. NO₂-N, NH₃-N, and Cd are all related to industrial activities [22,73], indicating that F2 is a significant source of industrial wastewater pollution. The high contribution rates of F3 to Pb and COD_Mn are 59.20% and 30.34%, respectively. Pb and COD are significantly correlated, indicating that these two elements have a common source. Pb can be produced by mineral resource extraction, mechanical processing, and traffic emissions, leading to river heavy metal pollution [43]. From the analysis mentioned above, F3 is likely to be a source related to urban domestic life. The contribution rate of F4 to As, SS, and TH is 55.82%, 50.58%, and 38.33%, respectively. The sources of As include both natural processes such as rock weathering and soil runoff, as well as human activities. As exists in the form of sulfides in arsenic-rich coal seams, and changes in redox conditions can induce the release of As adsorbed on sulfides into the water [36]. It is worth noting that the presence of F4 in the area may contribute to soil weathering, given its association with meltwater and snow mountain runoff at the head of Tarim River. The contribution rates of F5 to F⁻, TN, and Cr are 53.00%, 34.85%, and 31.84%, respectively, while the contribution rates to Fe and Cu are 23.90% and 20.14%, respectively. The interaction between water and rocks is one of the main factors affecting water quality in arid areas [87]. Studies have shown that elements such as F⁻, Cr, Fe, and Cu generally originate from parent rocks of soil formation [60]; TN is often related to domestic, industrial, and agricultural wastewater being washed into rivers by rain [65]. The surrounding area of the study site primarily consists of agricultural and forested land, suggesting that F5 may serve as a composite source encompassing both agricultural activities and natural sources. The contribution rates of F6 to NO₃-N and TP are 56.68% and 40.80%, respectively. NO₃-N and TP may be related to breeding activities [10], with numerous livestock and poultry farms in the research area contributing to nitrogen and phosphorus pollution in the water bodies. Therefore, F6 represents a significant source associated with livestock and poultry farming.

3.3.4. Analysis Results of Water Environmental Pollution Sources

The ratio of APCS-MLR and PMF to the measured values of pollutants in the upper reaches of Tarim River (R² > 0.80) is approximately 1, indicating a strong agreement between the analytical and measured concentrations from both models. This suggests that the source analysis results are reliable, with PMF exhibiting minimal variation and a closer approximation to unity, resulting in an improved fitting effect as shown in Figure 11a. However, the correlation between water quality indexes in PMF pollution source identification is comparatively weaker than that in APCS-MLR, leading to a less connected identification of pollution sources. Figure 11b illustrates the relative contribution rates of sources identified by both APCS-MLR and PMF models. It can be observed that the identified pollution sources are largely consistent, with soil weathering sources, livestock and poultry breeding sources, and agricultural activities sources exhibiting high contribution rates. Nevertheless, there exists a slight disparity in the analysis of contribution rates between the two models. The contribution rates obtained from APCS-MLR and PMF analyses are 44.11%, 19.63%, and 11.67% for the former, and 24.08%, 17.88%, and 27.54% for the latter, respectively, which may be attributed to parameter selection and different factorization processes during model operation. Generally, the analytical results of APCS-MLR and PMF models exhibit a high level of agreement, accurately reflecting the majority of river pollution source identification cases. This suggests that the utilization of these two models in water pollutant source analysis is scientifically justified.

3.4. Implications

Both APCS-MLR and PMF receptor models are capable of analyzing and interpreting complex water quality datasets, identifying and analyzing pollution sources along with their contribution rates. However, the PMF model exhibits a better fitting effect, while the correlation between water quality indicators in the identification process of PMF pollution sources is lower compared to that of APCS-MLR. If the correlations among water quality indicators are underestimated or overlooked, there’s a risk that identified pollution sources could be mistakenly interconnected, causing multiple sources to be inaccurately consolidated into one or a few categories. This oversight can result in a less precise and comprehensive identification of pollution sources, potentially hindering effective water quality management and remediation efforts. Consequently, this leads to a less connected identification of pollution sources. Therefore, for the current study, the APCS-MLR approach appears to be more physically reasonable.

In summary, the water quality in the upper reaches of the Tarim River is generally satisfactory, yet there are areas of localized pollution, mainly due to human activities. This issue is not unique to the Tarim River basin but is common across many arid and semi-arid regions where human reliance on river systems is high. Similar ecological pressures have been observed in the ecosystems of the Syr Darya River and Amu Darya River, which share ecological characteristics with the Tarim River basin. The primary sources of water pollution in these areas include agricultural runoff from irrigation, industrial effluents from sectors such as mining, construction, manufacturing, and petrochemicals, inadequate treatment and management of urban domestic wastewater, and the overarching impacts of climate change [88]. Integrated basin management stands as the cornerstone of water environment management. Currently, the management of the basin water environment is conducted scientifically through the establishment of international conventions, as well as regional and national laws and regulations. For instance, the Water Convention mandates countries to devise and enact policies and measures aimed at safeguarding and enhancing the quality of water in transboundary waters and minimizing pollutant discharges. Similarly, the European Union’s Water Framework Directive stipulates the maintenance of a good qualitative and quantitative status for all water bodies, necessitating ongoing monitoring and enhancement of water quality. In the United States, the Environmental Protection Agency (EPA) introduced the Watershed Protection Approach Framework, which focuses on fostering cooperation among communities and basins to tackle water pollution through multidisciplinary and inter-agency collaboration. These policies serve as vital references for the development of an integrated water treatment system in China, particularly in addressing the pollution issues of the Tarim River, it is recommended to implement stricter and more comprehensive pollution control measures for untreated livestock and poultry breeding sewage, modify fertilization methods or enhance fertilizer utilization efficiency, and establish dynamic monitoring for major pollutants. These actions will provide a theoretical basis and technical support for water environment assessment and research on pollution characteristics in typical industrial and agricultural interlacing areas as well as long-distance, trans-regional watersheds in arid regions.

4. Conclusions

The water quality in the upper reaches of the Tarim River is characterized by high turbidity, ranging from neutral to slightly alkaline conditions and exhibits significant total hardness. This assessment is based on the average pH value of 7.97, SS concentration of 473.85 mg L⁻¹, and TH of 303.04 mg L⁻¹. Heavy metals, NO₃-N, and NO₂-N all comply with Class I water quality standards. However, some monitoring points exceeded Class II water quality standards for DO, COD_Mn, BOD₅, TN, TP, F⁻, and NH₃-N, indicating distinct spatial variations in the water quality that are largely attributed to anthropogenic activities in the upper Tarim River basin. The results of the comprehensive water quality evaluation indicate that there was no degradation in the upstream water quality from 2020 to 2022. Based on the classification of WQI, the upper reaches of the Tarim River predominantly exhibit “moderate” and “good” grades, with an overall average value of 74.85, indicating good water quality status. DO, TN, NH₃-N, and EC were identified as key parameters influencing upstream water quality. The pollution source analysis results of APCS-MLR and PMF exhibit a high degree of similarity, leading to the identification of six potential pollution sources in the upper reaches of Tarim River, including agricultural activity, urban domestic, soil weathering, livestock breeding, natural, and industrial sources. Additionally, it distinguished between pure agricultural activities and a mixed source of agricultural activities and natural sources. Among these sources, soil weathering, livestock and poultry breeding, and agricultural activities demonstrate higher contribution rates. The contribution rates to pollution from these sources were analyzed using two models, revealing the following distribution: 44.11%, 19.63%, and 11.67% for the first set of sources and 24.08%, 17.88%, and 27.54% for the second set, respectively. For industrial pollution and urban domestic sources, one model showed contribution rates of 6.01%, 2.13%, and 4.27%, whereas another model found a higher rate of 16.71% for these sources combined. The APCS-MLR analysis highlighted that natural sources were responsible for a contribution rate of approximately 16.45%. Furthermore, the PMF analysis indicated that the combined impact of agricultural activities and natural sources contributed at a rate of about 9.52%. These findings underscore the varied and significant impact of different pollution sources on the environment, emphasizing the need for targeted interventions.