Introduction

As a result of the direct passive effects of urbanization, water shortage and water pollution have become a serious cause for concern and thus pose a significant threat to water habitats and human well-being. Reclaimed water (RW) reuse has shown a remarkable advantage in protecting urban water and expanding water sources and has thus received considerable attention worldwide (Crook and Surampalli 1996). The landscape reuse of RW has recently begun to take shape. The volume of RW reused for landscape environment is approximately 300,000 m3/day. In Beijing, China, a portion of the water from the Gaobeidian Sewage Treatment serves as a supplement to different parks around the city. However, abundant nitrogen and phosphorus (N and P) may easily result in water eutrophication and thus degrade the entertainment and ornamental value. Given their low management cost and outstanding purification capability, constructed wetlands (CWs) are used to improve the quality of urban RW (Rousseau et al. 2008; Cui et al. 2011). Thus, the temporal–spatial variations in water quality must be recognized.

An urban CW is a landscape ecologic environment system that serves view-and-admire, entertainment, and wastewater purification functions. This system has a significant landscape effect and is capable of repairing environmental pollution, including effectively reducing BOD and TSS and removing nitrogen, phosphorus, organic matters, heavy metal oxide, and pathogenic bacteria (Thurston et al. 2001; Zhao et al. 2009). Urban CWs are widely applied to improve the quality of urban landscape water (He et al. 2007; Cui et al. 2011). However, previous research on RW supply, particularly for CW reuse, concentrates on the estimation of the efficiency of pollutant removal and on the purification conditions of plants or functional areas; it rarely focuses on the spatial–temporal variations in the water quality of CWs under the landscape process.

Previous research indicates that multivariate statistical methods, such as discriminant analysis (DA) and principal component analysis/factor analysis (PCA/FA), are highly valuable in extracting significant information from a complicated dataset and in identifying major sources that affect water quality; hence, these methods provide accurate information on ecosystems and water quality conditions within study areas (Morales et al. 1999; Alberto et al. 2001; Reghunath et al. 2002). In view of the above considerations, we performed a fuzzy comprehensive assessment (FCA) to analyze pollution conditions after a basic understanding of water quality. Then, we employed DA with the Spearman correlation analysis (SCA) to identify significant parameters that affect spatial and temporal variations in water quality. The results were presented in a box-and-whisker plot, which provides significant insights into the spatial and temporal variations of remarkable pollution factors. Finally, the origin of different pollution sources and the variations between two periods were examined using PCA/FA combined with on-site investigation. The results may prove highly valuable for application in network optimization as well as in screening monitoring indicators and future river restoration projects.

Materials and methods

Study areas

The Olympic Park CW is located north of Beijing. This wetland has been in operation since 2007. Located in a continental monsoon climate zone, the wetland has an average temperature of 11.2 °C, a mean annual precipitation of 600 mm, and an average surface evaporation of 1200 mm. The rainy season (June to September) accounts for 85 % of the total annual precipitation. The CW is a multilayered purification system with an oxidation pond as a supplement. It receives RW from the Qinghe River and is situated between two composite vertical flow constructed wetlands (CVW; go-up flow wetland and go-down flow wetland). The water plants cover 70 to 90 % of the wetland system and mainly include reed, cattail, Scirpus tabernaemontani, and loosestrife. The functional area distribution and flow direction in the purification system are shown in Fig. 1. RW first enters the CVW, travels through the plant oxidation pond (POP) and the mixed oxidation pond (MOP), and finally fills the main lake (ML). After a short layover, water in the ML is then delivered by a high-pressure pump to the CVW for the next water quality cycle.

Fig. 1
figure 1

Location and distribution of the purification system in the constructed wetland

Full size image

Sampling and parameters

In 2011, samples were collected from ten sites of four functional areas in the Olympic Park wetland in spring (May 5, 13, and 22), summer (August 6, 15, and 23), and fall (October 5, 16, and 28). The first six sites (1–6) are located in the CVW. The other two sites (7 and 8) are located in the POP. The last two sites (9 and 10) are located in the MOP and ML (Fig. 1). Water sample collection, preservation, and transportation were strictly executed by “water and wastewater monitoring analysis methods” (4th edition). The following 13 parameters were measured: Chl-a, Cl−, chemical oxygen demand (CODMn), NO3 −-N, DO, NH4 +-N, oxidation–reduction potential (ORP), cyanobacteria (PCY), pH, salinity (Sal), total dissolved solids (TDS), total nitrogen (TN), and total phosphorus (TP). Among these parameters, pH, Sal, TDS, ORP, DO, PCY, and Chl-a were measured on site (Hydrolab Datasonde5 5X, Hach, USA). TN was determined in a laboratory (Multi N/C, Analytik-jena, Germany). NH4 +-N, NO3 −-N, and Cl– were determined using an ion chromatograph (ICS-1500, ICS-90, Dionex, USA). Other measurements were taken based on the Chinese guidelines in the Environmental Quality Standards for Surface Water (GB 3838—2002).

Analytical methods

Fuzzy comprehensive assessment

Fuzzy comprehensive assessment is a comprehensive evaluation based on fuzzy mathematics that estimates complex matter. It has been proven effective in solving problems presented by fuzzy boundaries and in avoiding the influence of monitoring errors. The accuracy of FSE is considerably higher than that of other index evaluation techniques (Onkal-Engin et al. 2004). In this work, the level of water quality evaluation was determined according to the Chinese guidelines in the Environmental Quality Standards for Surface Water (GB 3838—2002). A single-weight matrix was calculated using the clustering weight method (Wu et al. 2006). The membership degree of water quality was determined by the multiplication and addition methods (Fu et al. 2011).

Multivariate statistical methods

Multivariate statistical methods (DA, PCA/FA, etc.) were carried out to process the water quality data of the Olympic Park wetland. All calculations were performed using Office Excel 2007 and SPSS 18.0.

DA is often used to determine the most significant parameters (Singh et al. 2005). Based on different discrimination functions, DA is categorized into three modes: standard, protrusive, and stepwise, with the last mode having an advantage in terms of dimension reduction and discrimination capability (Alberto et al. 2001; Zhou et al. 2007). The clusters demonstrated by the functional areas/seasons were considered as grouping variables while the measured parameters were considered as independent variables. DA was used to determine the most significant parameters responsible for the spatial and temporal variations, even when such variations only appear in a specific space–time. SCA was performed to estimate the parameters related to the spatial and temporal variations. The combination and comparison of both methods can identify elements existing in space–time variations or those only occurring in specific space–times; thus, natural/artificial factors influencing the spatial/temporal variation in water quality could be determined (Zhang et al. 2012).

PCA/FA is a pattern recognition technique that attempts to explain the variations in a large set of intercorrelated variables by transforming them into a small set of independent variables (Simeonov et al. 2003; Singh et al. 2005). PCA is executed to extract significant principal components (PCs). Subsequently, FA is used to reduce the contribution of the less significant variables obtained from PCA. The new group of variables, known as varifactors (VFs), is extracted by rotating the axis as defined by PCA. The VFs can include unobservable, hypothetical, and latent variables, whereas PCA can only be a linear combination of observable water quality variables (Kannel et al. 2007). When applied to water quality analysis, PCA/FA is mainly used to extract and recognize pollution sources (Pekey et al. 2004).

Result and analysis

Overall status of water quality characteristics

The characteristics of each parameter versus the national guidelines for surface water are presented in Table 1. The comparison of the values of NH4 +-N, TN, and TP in the CW and in the RW shows that the values in the CW were significantly lower than those in the RW. Thus, the CW has high nitrogen removal efficiency and phosphorus nutrition. By comparing the mean value of each parameter and the national standard, we infer that the overall water quality in the Olympic Park wetland was relatively satisfactory during the sample period. In particular, only the TN in spring and autumn (1.66 and 4.01 mg L−1, respectively) and the TP in autumn of the 13 parameters exceeded national standard IV in terms of the seasonal scale. Only the value of CODMn (34.62 mg L−1) in the MOP and the value of TN in the POP, MOP, and ML (1.63, 2.36, and 2.57 mg L−1, respectively) exceeded national standard IV in terms of the spatial scale. NH4 +-N (0.21 mg L−1) and NO3 −-N (0.38 mg L−1) were inferior to national standard II. These results indicate that the Olympic Park CW is suitable for landscape reuse. As for the variable coefficients, NO3 −-N had the highest values in the temporal–spatial variations scale, followed by NH4 +-N, TN, and PCY. TP had the highest values in the specific spatial scale, whereas Chl-a and CODMn had the highest values in the specific seasonal scale. The high variation values indicate a high degree of dispersion and an uneven distribution among these parameters across different periods and sites.

Table 1 Summary statistics of measured parameters and the national standards for surface water quality
Full size table

An integrated evaluation of water quality could be obtained through FCA (Table 1). In the temporal aspect, the best water quality was observed in summer (degree 0.58, level III). This result was largely influenced by the strong growth of plants and frequent rhizosphere microbial metabolism that enhanced the nutrient assimilation effect. The spring water was of poor quality and belonged to class IV (0.41). In fall, the withering and decay of plants and the decreasing water mobility caused water quality to drop to level V (0.59). In the spatial aspect, the water environmental quality of the CW and POP was the best and matched water quality standard III (0.459 and 0.39, respectively). This result is related to the high vegetation coverage and rapid water circulation. Furthermore, the water quality of the CVW was classified as level IV (0.53). The MOP was easily disturbed by people because of its small area and hydrophilic landscape and thus belonged to standard V (0.55). The wide receiving area and long hydraulic retention time in the ML possibly reduced its water quality to standard V (0.54).

Temporal and spatial variations in water quality

Temporal variations

The temporal distribution of water quality could be achieved through temporal DA and verification (Table 2). In temporal DA, the standard mode assigned 91.8 % of the cases correctly. The assessments of temporal variations in water quality can be deemed highly accurate for the 14 selected parameters. In the stepwise mode, 91.8 % of the cases were assessed correctly using only CODMn, NO3 −-N, ORP, and TN; thus, the use of these four parameters should be sufficient to illustrate temporal variations. This finding suggests that future works should strengthen the monitoring of these four parameters.

Table 2 Classification functions for HDA and correlation coefficient for SCA of seasonal variations
Full size table

Both seasons and points were non-numeric parameters. Thus, spring, summer, and autumn were labeled as 1, 2, and 3, respectively. The results of SCA are also shown in Table 2. Cl−, NH4 +-N, ORP, PCY, TN, and TP were found to be significantly correlated with the seasonal changes (sig. < 0.01), whereas DO and NO3 −-N were found to be correlated (sig. < 0.05). The findings suggest that these parameters were easily affected by temporal factors, such as water temperature, water yield, and flow rate. A comparison of SCA and DA shows that Cl−, NH4 +-N, PCY, DO, and TP were significantly related to seasons but had no discriminant capability. The loss of discriminant capability might be caused by hydraulic characteristics (such as water level and hydraulic retention) and wetland plants. That is, the random drainage moment of the dams between different functional areas considerably interfered with the water flow. Meanwhile, wetland water quality was closely associated with plant growth cycle. In spring and summer, high water temperatures caused an increase in nutrient consumption (N and P) (Liang et al. 2004) and DO concentration, thus restraining algal growth; in autumn and winter (Zhang and Hong 2006), low temperature resulted in the withering and decay of the plants that influenced purification. Additionally, CODMn indicated discriminant capability but had no seasonal relation; thus, CODMn only occurred in a specific season and thus requires further explanation through box-and-whisker figures.

The box-and-whisker plots of the selected parameters showing evident seasonal relation are given in Fig. 2. The previous study indicated that temperature, as a result of seasonal changes, mainly affects nutrient removal efficiency in wetlands (Peng et al. 2014; Ávila et al. 2013), and TN and NO3 −-N had a similar trend in terms of season; both were steady in spring and summer because of the rapid growth and fast anabolism of the wetland plants and the strong microbial degradation rate of nitrogen. This has been verified by Wang, and microorganisms are sensitive factors that affect TN (Wang et al. 2012). In autumn, high concentrations of TN and NO3 −-N were caused by two factors: (1) the wilting of plants and the weakened metabolism that increased nitrogen accumulation and (2) the reverse flow of nutrient elements during the latter period of plant growth as a result of seed drops as well as the nutrient transmission from stem leaf to root that resulted in a decrease in the nitrogen nutrition removal rate (Meulemanvan et al. 2003; Thoren et al. 2004; Kyambadde et al. 2005). CODMn in summer was the lowest (19.79 mg L−1) but was slightly high in autumn (27.95 mg L−1). This result is related to temperature and to the alternation between high and low water levels (Liu et al. 2004). In autumn, the low temperature and small water fluctuations decreased the organic pollutant removal rate.

Fig. 2
figure 2

Seasonal and spatial variations of significant water quality parameters

Full size image

Based on the variation range, the largest range of CODMn, TN, and NO3 −-N occurred in spring, thus indicating large fluctuations in organic and nutrient removal efficiency. ORP affected microbial activity directly and further influenced the overall denitrification in the wetland (Wang et al. 2011). Therefore, the dramatic variation in ORP suggests frequent microorganism activities in summer; Xiongwei’s research about Olympic Park represented community density, and evenness of root-associated aerobic bacteria in summer increased 76.37—96.70 % than those in autumn (Xiong et al. 2013).

Spatial variations

The spatial distribution of water quality could be obtained through spatial DA and verification (Table 3). In spatial DA, the standard mode assigned 89.6 % of the cases correctly. The assessments of spatial variations in water quality can be deemed highly accurate for the 13 selected parameters. In addition, the spatial variations were significantly lower than the seasonal variations. In the stepwise mode, 70.8 % of the cases were assessed correctly using only DO, pH, and PCY; thus, the use of these three parameters should be sufficient to illustrate spatial variations. According to the accuracy of a specific functional area alone, the accuracy of any discriminant method applied in the MOP would be the lowest. This condition indicates that water quality in the MOP was unstable. This finding suggests that future works should strengthen the monitoring of these three parameters. The CVW, POP, MOP, and ML were labeled as 1 to 4. The SCA results are shown in Table 3. DO and pH were found to be significantly (sig. < 0.01) correlated with the spatial variations, whereas Cl−, TN, and Chl-a were related to the spatial variations (sig. < 0.05). A comparison of the SCA and DA results shows that Cl−, Chl-a, and TN were significant in SCA but not in DA, possibly because each functional area was controlled and connected by a water valve; that is, pollutant concentration and distribution changed along with the open–close moment and the valve opening frequency. Large fluctuations in Cl−, Chl-a, and TN resulted in a loss of discriminant capability, thus showing that the three indicators were significantly influenced by human activities. Meanwhile, PCY could be determined by DA but not by SCA. Hence, the variations in the two parameters only occurred in specific functional zones and not in the entire CW. Specific variation conditions require further explanation by box-and-whisker figures.

Table 3 Classification functions for HDA and correlation coefficient for SCA of spatial variations
Full size table

The box-and-whisker plots of the selected parameters showing spatial variation are given in Fig. 2. The significant variations in PCY only occurred in a single functional area, thus verifying the DA and SCA results. PCY content in the ML was higher than that in other functional zones because of the hydrological environment, land runoff, and exogenous input (Hao et al. 2012). In addition, PCY level in water could reflect alga biomass, with a high concentration indicating that the water eutrophication trend in the ML was more evident than that in other functional areas. DO concentration was higher in the POP and ML than in the CVW. The research of Avila indicated DO dramatically decreased in its passage through horizontal and vertical subsurface wetland (Ávila et al. 2013), possibly because of the high density growth of emerged plants and submerged plants. For example, weeds, which are the most proficient in the production/transfer of oxygen among wetland plants, favor oxygen supplement in wetland ecosystems (Huang et al. 2006). Previous studies showed that pH is significantly related to lake pollution and may cause lake pollution (Mo et al. 2007). Pollutants aggravate eutrophication and accelerate the progenitivity of phycophyta, thus leading to high assimilation and increases in pH in water. In the present work, high pH in the ML denotes a high eutrophication level.

Source identification

Eigenvalues greater than 1 were taken as criteria to extract the factors required to explain the sources of the variances in the data (Pekey et al. 2004; Kannel et al. 2007). Each extraction factor contained complex, comprehensive information about the original variables. Therefore, a rotation based on the extraction factors must be created to clarify the meaning of each factor. In this study, varimax rotation was adopted to minimize the variables with the highest load in each factor (Li and Luo 2010). To recognize spatial and temporal variations, three seasons (spring, summer, and autumn), and four functional areas (CVW, POP, MOP, and ML) were analyzed by FA.

Seasonal source identification

The factor loading value of seasonal FA is shown in Table 4. For the spring dataset, four VFs were found; these VFs accounted for 85 % of the total variations. VF1 accounted for 22.58 % of the total variations, thereby demonstrating significant loadings on Sal, TDS, ORP, CODMn, and NO3 −-N. VF1 was thus identified as the source of contamination by internal impurity and organic pollutant. VF2 accounted for 25.68 % of the total variation and had a positive correlation with Chl-a, PCY, NO3 −-N, TN, DO, and pH. VF2 was thus found to be typical of eutrophication and nitrogen pollution. VF3 accounted for 17.43 % of the total variation and was correlated with TP, CODMn, ORP, and PCY. Organic pollutants within the supplement water (RW) already matched the national standard. The positive correlation of CODMn suggests that the organic pollutants originated not from the RW but from a non-point pollution source, such as agricultural runoff and atmospheric conditions. Therefore, VF3 was identified as a combination of point source and non-point source pollution (Liu et al. 2003). VF4 accounted for 13.42 % of the total variation and had a positive correlation with NH4 +-N and Cl−, thus indicating ammonia nitrogen pollution in the RW.

Table 4 Factor loading values and explained variance of water quality parameters across three seasons
Full size table

For the summer dataset, five VFs were found to account for 74.56 % of the total variations. VF1 accounted for 28.47 % of the total variations, thus demonstrating significantly positive loadings on physical and chemical indicators such as K and TDS. Hence, VF1 demonstrates the influence of internal impurity, such as salt and other impurities, on the wetland environment. VF2 accounted for 14.63 % of the total variations and was correlated with NO3 −-N, TN, and TP, thus indicating nutritive salt pollution (N and P). VF3 accounted for 14.45 % of the total variations and had a significant association with CODMn and moderate correlation with Cl−, DO, and ORP. Hence, VF3 can be interpreted as organic matter pollution. VF4 accounted for 14.18 % of the total variations, and the significant loadings on Chl-a and PCY indicated eutrophication in the RW. VF5 accounted for 8.72 % of the total variations, thus representing the ammonia nitrogen pollution from the RW supplement.

For the autumn dataset, five VFs were found; these VFs accounted for 90.22 % of the total variations. VF1 (variation 30.91 %; characterization factors Cl−, pH, K, TDS, and Sal), VF2 (variation 19.82 %; characterization factors Chl-a, PCY, and DO), VF3 (variation 17.14 %; characterization factors NO3 −-N, TN, TP, and Cl−), VF4 (variation 11.39 %; characterization factor CODMn), and VF5 (variation 10.95 %; characterization factors NH4 +-N and ORP), respectively, indicated the impact of internal impurity, eutrophication, nutrients (N and P), organic pollutant, and ammonia nitrogen pollution on the CW.

The analysis of the primary influential indicators across three seasons reveals that internal impurity was the main source that affected water quality in each season. Unlike other water bodies, the wetland could not be easily affected by point (industrial wastewater and domestic sewage) and non-point pollution (overland runoff agricultural wastewater) because of comprehensive treatment techniques and sufficient purification. Organic pollution was the severest in spring (VF1), probably because the low temperature was sufficient to restrain the metabolism of rhizospheric microorganisms. Nutrients (N and P) were higher in summer than in other seasons, but eutrophication was lower. Research has shown that the increase of N and P concentrations does not necessarily result in eutrophication (Fu et al. 2005). The low eutrophication levels in summer may increase water mobility and shorten the renewable cycle, both of which could damage algae growth conditions. Ammonia nitrogen pollution was the least influential factor across three seasons, thus indicating that the CW had strong nitrogen removal efficiency. The research also indicated that the removal of ammonia nitrogen in hybrid constructed wetland was effective, amounting to 70∼94 % (Jia et al. 2014; Ávila et al. 2013).

Spatial source analysis

The factor loading value of spatial FA is shown in Table 5. For the CVW dataset, five VFs were found; these VFs accounted for 81.55 % of the total variations. VF1 accounted for 28.95 % of the total variations and had significantly positive loadings on Sal and TDS. This result indicates the effect of internal impurity on the wetland. VF2 accounted for 20.17 % of the total variations, with positive loadings on Cl−, ORP, and TN as well as moderate positive relations with NO3 −-N. This result demonstrates the nutrition salt pollution source of N and P from the RW. VF3 accounted for 13.76 % of the total variations and had a positive correlation with pH as well as a negative correlation with DO. The results indicate the effect of acid and alkali pollutants. VF4 accounted for 10.87 % of the total variations, and its characteristic factors were Chl-a and PCY. This result can be expressed as the effect of algae breeding. VF4 accounted for 7.81 % of the total variations, thus verifying significant relations with CODMn and TP. This result indicates organic pollution and phosphorus pollution as point pollution caused by the RW supplement. For the POP dataset, five VFs were found to account for 90.57 % of the total variations. VF1 accounted for 25.61 % of the total variations and had significant positive relations with Sal and TDS. This result indicates the effect of internal impurity. VF2 accounted for 22.38 % of the total variations, thus demonstrating relations with TN, TP, ORP, and PCY. The result also highlights nutrient salt pollution. VF3 accounted for 16.01 % of the total variations, thus demonstrating relations with DO, TN, NO3 −-N, and pH. This result can be illustrated by the effect of denitrification on water quality. VF4 accounted for 12.87 % of the total variations and was correlated with DO and NH4 +-N; VF4 therefore represents ammonia nitrogen pollutant in point pollution. VF5 accounted for 8.70 % of the total variations, and its characteristic factors were Chl-a, Cl−, CODMn, and pH. This result indicates the common influence of algae activity and organic pollutants. For the MOP and ML dataset, only three VFs were found, possibly because of the relatively less sampling points in these two functional areas. In the MOP, VF1 (variation 56.01 %; characterization factors Chl-a, CODMn, DO, pH, PCY, ORP, NH4 +-N, Sal, and TDS), VF2 (variation 30.15 %; characterization factors CODMn, NO3 −-N, PCY, TN, and TP), and VF3 (variation 13.84 %; characterization factors Cl−, NH4 +-N, PCY, and pH) indicated that the major pollutants successively originated from algal blooms and wetland water impurities, nitrogen, and phosphorus nutrients, as well as atmospheric sedimentation (non-point pollution) and ammonia nitrogen (point pollution). In ML, VF1 (variation 50.24 %; characterization factors Chl-a, CODMn, PCY, ORP, and TDS), VF2 (variation 29.3 %; characterization factors Cl−, DO, ORP, pH, TDS, and TP), and VF3 (variation 20.46 %; characterization factors Cl−, NH4 +-N, NO3 −-N, and TN) indicated that water pollution successively resulted from eutrophication, organic pollution, and phosphorus and nitrogen pollution. The comprehensive analysis of the main factors in each functional area reveals that the major pollution was the effect of internal impurity on the wetland and that only in the CVW and POP did nutrition salt (N and P; VF2) serve as a key influential indicator. It suggests that throughout the CVW and POP, N and P had been removed effectively. Especially in POP, the cascade landscape design enhanced water flow force and shortened hydraulic retention time, both of which sufficiently decreased eutrophication. The ML was easily influenced by eutrophication and atmospheric sedimentation (non-point pollution) because it received the largest volume of water and the longest hydraulic retention time.

Table 5 Factor loading values and explained variance of water quality parameters in functional zones
Full size table

Conclusions

The total water quality in Olympic Park was relatively satisfactory. The pollution indicators were all under surface water class IV. The result verifies the relatively high water quality of the CW in Olympic Park, which is suitable for landscape reuse. The spatial and temporal distributions of NO3 −-N, NH4 +-N, TN, and PCY were uneven. Water quality in summer was the best (belonging to level III) because of strong nutrient assimilation caused by vigorous growth and frequent metabolism. The water quality in autumn was the worst (class V) because of plant decay and liquid water variation. In different functional areas, water quality in the POP belonged to class III because of high vegetation coverage and rapid water circulation. Water quality in the CVW was not as good as that in the POP, possibly because RW is the starting functional area of the CW and because the removal of N and P was incomplete. By contrast, water quality in the MOP belonged to class V because of its small vegetation area and hydrophilic landscape, which can easily be disrupted by human activities. Class V in the ML could possibly be the result of its wide receiving area for water and long retention time.

The water environment had a unique pattern of temporal and spatial variations that resulted from both natural conditions and human activities. The use of CODMn, NO3 --N, ORP, and TN should be sufficient to illustrate temporal variations. A close monitoring of these four parameters should therefore be conducted in future works. In particular, water quality in spring was complex, thus warranting frequent monitoring. According to the spatial analysis of water quality, DO and pH demonstrated a significant relation with spatial variations. Cl, Chl-a, and TN were significantly affected by human activities. The significant variation in PCY occurred in the ML.

The primary source of pollution in the spatial and temporal areas differed. In a seasonal scale, internal impurity was the main indicator that influenced water quality in each season. Organic pollution in spring was the severest. Nutrient pollution (N and P) affected water quality in summer. Water quality in fall was easily threatened by eutrophication. In a spatial scale, the main influential indicator in the CVW and POP was nitrogen salt, including N and P. As for the other functional areas, internal impurity remained the main influential indicator.

The weak effect of non-point pollution in the CW resulted from complete rain flood facilities. Pollution mainly came from supplement water. The combination of plant and treatment processes influenced nitrogen, phosphorus, and organic pollutant removal efficiency. In future water quality regulations, we should promptly open the gates between functional areas, build cascade landscapes to enhance water circulation and avoid eutrophication, and promptly clear withered plants in autumn to avoid the backflow of nutrient elements.