3. Results and Discussion
3.1. Characteristics of Water Quality at Sampling Points
The statistical description of the water quality in the upstream, midstream, and downstream were presented in Table 3
, Table 4
and Table 5
. According to the surface water environmental quality standard of China (GB 3838-2002), a water quality of Grade I and II is considered to be clean or have a low pollution status; furthermore, Grade III corresponds to moderate pollution and Grade IV and V are considered to be high pollution. [35
]. The majority of water quality parameters in the study area ranged from Grade I–II, with the exception of TN in upstream (Table 3
) and both TN and TP in midstream (Table 4
) and downstream (Table 5
). The mean mass concentration of downstream TP was 0.16 mg/L, which corresponded to Grade II–III, and the mean mass concentration of TN was 2.62 mg/L, which corresponded to an inferior Grade V. The major pollutants in this study area are TN, followed by TP.
The coefficient of variation (CV) is used to describe the variation degree of samples over time or space, which can eliminate the influence caused by the difference of units and mean value [2
]. Specifically, CV < 20% means low variability, 20% ≤ CV ≤ 50% means moderate variability, CV > 50% means high variability, and CV > 100% means exceptionally high variability. Since the water quality indexes in this study are discussed separately from upstream, midstream, and downstream, they are primarily used to describe the variations in water quality indexes over time, and the coefficient of variation also reflects the temporal differences between various water quality indexes.
In the Xinjian River, the CV of DO, CODMn, and TN were less than 0.3, which belongs to weak variability, thereby indicating low variability and the resulting in the concentration steadily fluctuating around the mean value. In contrast, the CV of NH3-N was greater than 0.5, indicating that there is significant time variation. Combined with the skewness value, the skewness values of NH3-N in the upstream, midstream, and downstream were all greater than 1, showing a positive skewness, and indicating that there was a high peak concentration of NH3-N.
depicts the variation of the selected water quality at four sites during the study period, and Figure 3
presents the average concentration of water quality in the upstream, midstream, and downstream in the Xinjian River. It can be found that the water quality gradually declines from the upstream to the midstream and then to the downstream.
All parameters in the upstream were superior to those in the midstream and downstream because the river originates from the mountain area, which has less impact from human activities and pollution. Specifically, according to the DO concentration, the upstream belongs to Grade I water, and the midstream and downstream belong to Grade II water during most of the study period, and dropped to Grade II water during some period of 2019. According to the judgement of the CODMn concentration, the upstream water quality fluctuates between Grade I and Grade II water, whereas the midstream and downstream water quality fluctuate between Grade II water and Grade III water. According to the NH3-N concentration, the upstream water quality belongs to Grade I water and seldom belongs to Grade II water. According to the judgement of the TN concentration, the TN concentration in the upstream is generally significantly lower than that in the midstream and downstream. Similarly, the TP concentration in the upstream is generally significantly lower than that in the midstream and downstream. The midstream and downstream belong to Grade III water most of the time, with seldom Grade II. The F− concentration is lower in the upstream than that in the midstream and downstream. According to the classification criteria of F−, the water for all of the Xinjian River belongs to Grade I water (<1).
From the aforementioned analysis, we can find that the overall water quality of the Xinjian River is in good condition. Most of the water quality in the upstream is Grade I water, which meets the standards for drinking water sources, while the indicators exceeding the standard are TP and TN.
When the average concentrations of TN and TP in the Xinjian River and in Lishui City with 100 sites during the study period were compared, the former is evidently inferior to that of the latter (Figure 4
). The TP in the upstream is comparable to that of Lishui’s average condition, which is significantly lower than that in the midstream and downstream.
3.2. Source Apportionment
The source apportionment and various factor contributions were estimated by the EPA PMF software, which incorporate the concentration and uncertainty factors as model inputs. The resulting input parameters then generated a signal-to-noise ratio (S/N) for each W/Q parameter. The parameters were classified as Strong (S/N > 2), Weak (0.2 < S/N < 2), and Bad (S/N < 2), respectively, based on the S/N ratio. Based on the iterations for each factor obtaining the minimum Q (Robust)/Q(Exp) value, four factors were considered for the PMF analysis to assess each of their contributions to the water pollution of the Xinjian River. (Table 6
). The model was then simulated, considering all of the input parameters and the four factors, thereby providing the results shown in Figure 5
, Figure 6
and Figure 7
shows the goodness-of-fit parameters of the PMF model operations in the upstream, midstream, and downstream, respectively. Q is the critical parameter for PMF, where Q(true) is the goodness-of-fit parameter that is calculated including all points and Q(robust) is the goodness-of-fit parameter that is calculated excluding points not fit by the model, which are defined as samples for which the uncertainty-scaled residual is greater than 4 [25
, Figure 5
and Figure 6
show the fitting comparison diagrams of measurement and prediction of water quality parameters in the upstream, midstream, and downstream, respectively. The prediction is mainly evaluated using statistics such as the intercept, slope, and R2
. It can be found that the correlation coefficient between the simulated and predicted values of most water quality parameters is high, and R2
is close to 1 with the exception of CODMn
and TP in the upstream, and CODMn
and TN in both the midstream and downstream. It can be seen that the PMF has good fitting performance for these water quality parameters, and the selected factors can explain the information contained in the original data.
3.3. Identification of Pollution Sources
According to the water quality concentrations in the upstream, midstream, and downstream, the source profile and contribution rate of the corresponding pollution factors can be identified by the PMF model. The pollution source corresponding to the factor can be determined according to the significant identified pollutants in the source profile of each factor and the relative relationship between the factors.
3.3.1. Identification of Pollution Sources in Upstream
The PMF model resolved the component profile and contribution rates of four pollution factors in the upstream of the Xinjian River (Figure 8
). These factor profiles need to be interpreted to identify the source types that may be contributing to the samples.
The Factor (F1) showed high loading on F−
, which accounted for 44.1% of the total variance (Table 7
is an important identification element of F1, which can be inferred to industrial sources; however, there are no big factories in the upstream area. Considering the local conditions, rural family workshops and small factories are prevalent in Zhejiang Province, and rural family workshops for the processing industry might contribute to this source. TN and TP significantly contribute to Factor 2 (F2), accounting for 49.5% and 48.8% of the total variance, respectively. Specifically, the contribution to TP is significantly more than that of other factors, and potential sources include the use of chemical fertilizers, pesticides, livestock and poultry wastewater, and waste discharge. Therefore, F2 can be considered as an agricultural non-point source of wastewater. Factor 3 is primarily dominated by NH3
-N (74.2%) and TP (29.0). Consequently, NH3
-N can be considered as an essential identification element for F3. This factor might be interpreted as urban and rural domestic sewage. The concentration of water quality in Factor 4 (F4) is much lower than the other three factors—except DO; hence, F4 can be considered as the natural background source, such as soil.
3.3.2. Identification of Pollution Sources in Midstream
shows the identification results of pollution sources in the midstream. F1 showed high loading on F−
, accounting for 67.1% of the total variance (Table 8
is an important identifier for F1. Concurrently, CODMn
, TN, and TP substantially contributed to F1 with 29%, 27.6%, and 28.6%, respectively. Consequently, F1 can be inferred to industrial sources. The pollution level of this factor is relatively heavy, which is consistent with the fact that the existence of the factory and also the number of family workshops are more than that in the upstream (Figure 1
). F2 indicates a low pollution level. TP and TN have a slightly higher contribution rate in this factor, which are 14.2% and 24.6%, respectively. It is judged that the source is the local small livestock and poultry breeding. F3 has a high loading on NH3
-N, accounting for 99.6% of the total variance. NH3
-N can be considered as a crucial identifier for F3. This factor can be interpreted as nutrient pollution from strong anthropogenic impacts such as domestic sewage. TN and TP make substantial contributions to F4, accounting for 41.8% and 28.1%, respectively. In particular, their contribution to TN is much greater than that of other factors, and possible sources could be agricultural non-point sources.
3.3.3. Identification of Pollution Sources in Downstream
shows the identification results of pollution sources in the downstream. In the downstream, F3 showed a significant loading on NH3
-N, accounting for 76.4% of the total variance (Table 9
-N can be considered as a crucial identifier for F3. This factor can be interpreted as domestic sewage. F2 showed high loading on F−
, accounting for 55.6%. F−
is an important identifier for F1. Concurrently, CODMn
, TN, and TP substantially contributed to F1 with 30%, 19.3%, and 16.6%, respectively. Given this, F1 can be inferred to industrial sources. TN and TP significantly contribute to F3, accounting for 40.4% and 36.9% of the total variance, respectively. In particular, their contribution to TN and TP were much greater than that of other factors, and possible sources could be agricultural non-point sources. F4 showed relatively high loading on TN, TP, and F−, accounting for 24.2%, 36%, and 34.5% of the total variance. Concurrently, the contribution of CODMn
-N could not be ignored, which can be regarded as the integrated source from local emissions and the tributary inflow.
When the source apportionment results are compared, it is evident that F1 in the upstream, F1 in the midstream, and F2 in the downstream are similar (Figure 11
), which can be referred to as Matched Factor 1 (MF1) (Table 10
). The identification element for this factor is F−
, which can be identified as the industrial sources.
F2 in the upstream, F4 in the midstream, and F3 in the downstream are similar, which is referred to as Matched Factor 2 (MF2). This factor contributes significantly to CODMn, TN, and TP, which are indicative of nutrient pollution and can be identified as agricultural non-point pollution. The pollutants from agricultural planting directly enter the body of water as the rain runoff. Agricultural planting pollution cannot effectively be reduced due to the absence of effective interception or isolation facilities along the rivers. According to the survey, the planting mode of Zizania latifolia is extensive, and the degree of intensification and refinement is not high, with the number of pesticides being relatively large and a large number of organic pollutants being discharged from farmland runoff, thereby exerting significant pressure on the water ecological environment in this basin.
F3 in the upstream and midstream and F1 in the downstream are similar, which is referred to as Matched Factor 3 (MF3). This factor has a high contribution rate in NH3-N, which can be identified as domestic sewage. According to the local conditions, although a majority of the rural domestic sewage along the rivers has been piped for treatment, problems still exist in some villages, such as aging and deteriorated pipelines and direct discharge of sewage; moreover, most of the terminal treatment processes utilize anaerobic and artificial wetlands, so the efficiency of pollutant treatment is insufficient. Furthermore, there are numerous terminals that are widely distributed, making standardized operation and maintenance difficult.
In addition to the aforementioned pollution factors, there exists a group of factors without similar characteristics, which are called Unmatched Factor (UMF) and are F4 with a lower pollution degree in the upstream, F2 with a certain contribution to TN and TP in the midstream, and F4 with a higher pollution degree in the downstream.
Matched Factor 1 is identified as industrial sources. Its important identification element is F−. The contribution of the downstream is slightly greater than that in the midstream, and the contributions of the midstream and downstream are significantly greater than that of the upstream, which is consistent with the distribution of factories in the Xinjian River basin. The average concentrations of F− in the upstream, midstream, and downstream are 0.202 mg/L, 0.242 mg/L, and 0.320 mg/L, respectively, which are far lower than 1.0 mg/L—the national standard of Grade II water. The level of pollution is relatively low, and the pollution sources in the Xinjian River basin are mainly agricultural non-point sources and domestic sewage sources.
The receptor model can only identify the potential source but cannot locate the specific position of the source. Satellite remote sensing can compensate for the deficiency of the receptor model in space description. Therefore, the land cover and land use and the spatial distribution of the quantitative water quality information based on high-resolution satellite images can assist in locating the source and verifying the performance of the receptor model. To match the in situ measurements, we inversed the water quality concentration for every odd month from November 2020 to September 2022 based on Sentinel-2 images. Then, the accumulated concentration was derived by the total 12 months to reduce the errors caused by the inversion algorithm and image noise, and to reveal the general trend for the specific parameter in the past two years (Figure 12
). The BOD5
and COD present similar spatial patterns in different sections of the river. The tributary presents a low concentration compared with the main channel according to the BOD5
, COD, CODMn
-N, and TN, which is consistent with the distribution of DO, which has high concentrations in the tributaries. The concentration of CODMn
is high at the intersection of tributaries (Bihe River, Xinanxi River, and Qianyang River) and the Xinjian River.
In the upstream, the concentration of NH3-N of the tributary is higher than that of the Xinjian River, while the TN and TP concentration of the tributary is lower than that of the Xinjian River. This can be attributed to the domestic sewage discharge of residents on both banks of the river. The tributary aggravates the TN concentration in the Xinjian River.
The high concentration in the midstream is closely related to emissions from the factories in the south of river. TP in the tributary from the Xinanxi River and Qianyang River cannot be ignored, and the emissions from the farmland adjacent to the river contribute significantly. The TN concentration of the tributaries is low, which plays a role in diluting the Xinjian River. Currently, Xinbi Street has completed the construction project of “Zero Direct Sewage Discharge Area”. Nevertheless, satellite monitoring reveals that the TP and TN concentration in some downstream sections of the Xinjian River is still high (Figure 12
), thereby indicating the presence of pollution sources entering the river. Consequently, this project should be strengthened and improved.
As we know, metal elements have very good identifiability, but the body of water in Lishui City is in good condition as a whole. Among the tested indicators of water quality, metal elements are below the detection limit, so they are not used in this study.
To reduce the error caused by the image noise or inversion algorithm of a single period image, 12 inversed water quality concentrations were composed to reveal the overall water quality trend of the whole basin; however, due to the limited resolution (10 m) of the Sentinel-2 image, errors still exist in some areas for rivers with complicated conditions, including shallow terrain, mixed pixel phenomena, and shadows of the trees along the river, etc. High precision remote sensing for narrow rivers remains a challenge.
It should be noted that the amount of monitoring data in this study is not very large, and more reliable source apportionment results could be obtained if the data is accumulated for a longer period of time.
Strengthen the prevention and control of rural domestic pollution.
An emphasis should be placed on harnessing rural domestic waste by classification, harmlessness treatment, and resource utilization to improve the rural environment; moreover, the rural domestic sewage treatment facilities should be improved and supporting pipe network facilities should be constructed; the number of sewage treatment households should be increased; and the sewage treatment levels and coverage should be further enhanced.
Strengthen the prevention and control of agricultural non-point source pollution.
Technology should be adopted to promote the model of reducing fertilizer and pesticide consumption while increasing efficiency. Development of ecological ditches to intercept the nitrogen and phosphorus from farmland is a promising option. The combined use of physics, chemistry, and biology to strengthen the purification and advanced treatment of the total nitrogen and total phosphorus should be used.
Strengthen environmental management and ecological restoration
First, the comprehensive management of the river should be improved. Specific countermeasures for source control and sewage interception and ecological restoration include the development, restoration, and reinforcement of embankments, as well as the development of green land to protect the bank and improve the quality of the water and ecology and living environment. The second way is to upgrade the dam. The retaining dam at the intersection of the Bihe River and Xinjian River was built so long ago that an environmental assessment was not conducted. To improve the downstream water quality, it is proposed that an environmental assessment be undertaken to improve the dam, discharge water regularly, and increase the discharge water volume.
Check the area of direct sewage discharge
A mechanism for standardizing the operation and maintenance of sewage wells should be established. It is necessary to inspect the state of rainfall and sewage diversion in the residential community and commercial district in Xinjian Town and Xinbi Street. Whether all the sewage is intercepted and disposed of, and whether the rainwater and sewage pipe network are wrongly connected or not should be emphasized.
In this study, the PMF receptor model was used to identify the water pollution source in the Xinjian River, Lishui City, Zhejiang Province, China. Remote sensing technology was also utilized to aid in the analysis of source apportionment and to provide the spatial distribution of the water condition.
According to the location of the monitoring station, the whole basin is divided into three sections: upstream, midstream, and downstream. By comparing the source apportionment results in different sections, we can distinguish the local source and the extent input by tributary inflow. Accordingly, some anti-pollution measures are proposed. TN, followed by TP, are the major pollutants in the study area, and non-point source pollution is the major pollution source; they can be traced to tributary inflow, particularly in the Xinan River and Qianyang River. It is still essential to take prevention measures to control non-point source pollutants to preserve the natural ecology and environment. Some ecological measures should be implemented to mitigate the impact of tributaries’ pollution by TN and TP in the midstream and downstream. The industrial pollution resolved by F− exists from upstream to downstream. Therefore, it is imperative to strengthen the management of untreated pollutant discharge in family workshops and factories.