Abstract

This paper used the trophic level index (TLI) method combined with the relevant data from 2014 to 2017 to evaluate the water quality of Gaoyou Lake. Meanwhile, based on principal component analysis (PCA) and multiple linear regression (MLR) models on chlorophyll a (Chla), this research developed predictions and an early warning scheme for eutrophication in Gaoyou Lake. The results showed the following: 1. The TLI of Gaoyou Lake showed a significant increasing trend, and the lake was in the state of light to moderate eutrophy. 2. According to the PCA eigenvalues that were greater than 1, principal components (PCs) with a cumulative contribution rate of 76.04% were obtained, and a linear model was further obtained: CChla = 6.146 + 1.209 (Score 1) + 0.583 (Score 2) + 1.095 (Score 3). 3. The credibility of the early warning system reached 75%, which met the requirements of this study. This study provides a scientific basis for the control of eutrophication and improvement of water quality.

INTRODUCTION

In recent years, with the continuous expansion of cities and the rapid development of the economy, large amounts of nutrient-rich domestic and industrial sewage have been generated (Yang et al. 2017). This sewage has worsened the eutrophication of water bodies, threatening the utilization of water resources and the ecological environment (WHO 2006; Smith et al. 2016). Nevertheless, improvements in water eutrophication remain a challenge, especially in developing countries (WHO 2004; John et al. 2014). Therefore, it is necessary to evaluate and predict water eutrophication.

Information on the trophic status of a lake provides an indication of an ecosystem's current structure and function (Schaeffer et al. 1988; Jacobson et al. 2017; Nevalainen & Luoto 2017). For this reason, many studies have established a variety of methods for assessing water quality, including the fuzzy decision method (Bin & Jianshe 1991), set pair analysis method (Fanxiu et al. 2000), and comprehensive index method (Yang et al. 2008). Furthermore, the method based on the trophic level index (TLI) for nutrient assessments is currently widely used to determine lake water quality (China Environmental Monitoring Station 2001). In general, the trophic levels of lakes are assessed using five main indicators of responses to nutrients: Chla (chlorophyll a), TP (total phosphorus), TN (total nitrogen), SD (Secchi disc) and CODMn (permanganate index) (Xiang et al. 2014). The TLI is based on these five indicators. Accordingly, the TLI method has been widely used for the assessment of surface water quality. For example, Wang et al. (2019) used a combination of the water quality index (WQI) and TLI to analyse eutrophication of the estuary of the Wuli River in Taihu, China. Based on this method, Liu et al. (2012) evaluated the eutrophication of the lakes in Yunnan, China, and Trolle et al. (2014) used the TLI to analyse lakes such as Lake Benmore in the South Island of New Zealand. Obviously, these studies have a significant and positive impact on the evaluation of water quality eutrophication. However, simply evaluating eutrophication can provide only the current nutritional status of a lake and cannot be used to analyse future water quality.

Water quality predictions can be used to predict future trends in water quality in a changing environment and minimize microbial risks (Hamilton & Schladow 1997; Ge & Frick 2009). At present, the methods for the prediction of eutrophication of freshwater bodies such as rivers, reservoirs and lakes are relatively mature. Prediction methods include regression analysis, neural network models, the decision tree method, and support vector machine models (Maier & Dandy 1996; Biggs 2000; Karul et al. 2000; Atkins et al. 2007; Rene & Saidutta 2008; Faruk 2010). An approach that has received growing interest and is becoming popular in ecological and environmental modelling is principal component analysis (PCA) (Barbieri et al. 1999; Perkins & Underwood 2000; Flink et al. 2001; Wannaz et al. 2003; Zhang & Li 2016; Pérez-Arribas et al. 2017). Unlike other methods, PCA offers an objective method for handling a large set of biotic and abiotic data and aids in reducing the complexity of multidimensional systems by maximizing the variance of the component loads and eliminating invalid components (Jackson 1993; Petersen et al. 2001; Bengraïne & Marhaba 2003; Loska & Wiechuła 2003). Moreover, PCA has been employed either alone or in combination with other environmental factors to model biological and ecological processes (Vink & Van der Zee 1997; Zimmerman & Canuel 2001; Bengraïne & Marhaba 2003). In this study, the PCA method was combined with Chla to obtain the relationship between principal components (PCs) and Chla to effectively predict the eutrophication of a water body.

The basis of establishing an early warning system is to evaluate the status quo and predict the future situation. In addition, the process from evaluation and prediction to early warning is also a process that can be used to deepen the understanding of the system. Thus, it is necessary to establish an early warning system using a forecasting scheme. In recent years, great progress has been made in the study of methods and models for early warning systems (Park et al. 2015; Wang et al. 2018); however, the accuracies of these systems are still not ideal. Chla is a nutrient source for most aquatic plants and can reflect the comprehensive status of water eutrophication (Çamdevýren et al. 2005). Therefore, it is of great importance to use Chla as an early warning indicator.

Gaoyou Lake is one of the ten largest freshwater lakes in China. This lake is an important source of drinking water and home to a variety of fish and aquatic plants. Gaoyou Lake has many functions, such as fish farming, industrial and agricultural water, irrigation, tourism, shipping and flood control (Wang & Cao 2005). Gaoyou Lake was once known to have the highest water quality in Jiangsu Province, but in recent years, due to the continuous development of the fishery and tourism industries, the pollution level of the water has been accelerated, resulting in eutrophication (Ji et al. 2009; Wei et al. 2010). In this regard, although some scholars have evaluated the water quality of this lake in the past few years (Zhao et al. 2015; Jiang et al. 2017), simple evaluations and analyses cannot determine the future development trends of water quality. This study combines research on eutrophication, prediction and early warning to conduct comprehensive and in-depth research on the water in Gaoyou Lake.

The purpose of this study is to (1) use the TLI method to assess the changes in the eutrophication of water in Gaoyou Lake in recent years; (2) use the multiple linear regression analysis method to establish a prediction model, predict the Chla concentration and the overall trend, and compare the fitting degrees and variation trends of the measured and predicted Chla concentrations; and (3) provide early warning of water quality eutrophication.

MATERIALS AND METHODS

Study area

Gaoyou Lake (119°06′-119°25′E, 32°42′-33°41′N) is located in the city of Tianchang in Anhui Province and Gaoyou city in Jiangsu Province (Figure 1). The lake is 48 kilometres long and has a maximum width of 28 kilometres. The total water area is 760.67 square kilometres, and this lake is in the Huaihe River Basin. Gaoyou Lake is one of the large lakes in the basin and is known to have the best water quality in Jiangsu Province, China. At the same time, Gaoyou Lake is the main position of the Huaihe River waterway improvement project and is closely related to the South-to-North Water Transfer Project. In recent years, due to the development of industries in the counties and cities around Gaoyou Lake, the water quality has been polluted to varying degrees, and the phenomenon of eutrophication of water bodies has emerged (Tang & Fan 2009).

Figure 1

Map of the sampling sites in Gaoyou Lake, China.

Figure 1

Map of the sampling sites in Gaoyou Lake, China.

Sample collection and water quality monitoring

Given the characteristics of the environment surrounding Gaoyou Lake, the Gaoyou Lake water intake point in Tianchang city, Anhui Province, was selected as the monitoring point (119.1859E, 32.8065N), and monitoring was conducted at three measuring points to the left, in the middle and to the right of the intake point. The average was taken as the final data. The monitoring time was from January 2014 to December 2017, and the monitoring frequency was once a month for a total of 48 times. The TN, ammonium (NH4-N), TP and potassium permanganate index (CODMn) were measured in the laboratory according to standard methods (Jin & Tu 1990). Water temperature (T) and dissolved oxygen (DO) were measured on-site with a multiparameter water quality sonde (YSI, 6600V2, USA). Transparency was measured by a SD, and Chla was determined using a chlorophyll a fluorescence detector (FuloroQuik, AMI, USA).

Trophic level index (TLI) calculations

Taking Chla as the benchmark parameter, parameters such as CODMn, TN, TP, SD and other significant correlations with Chla were selected as evaluation factors, and weight analysis was carried out to obtain the water quality evaluation results. The method is used to establish a TLI model as follows (Zhang et al. 2011): 
formula
where TLI (j) represents the composite index of j with the correlative weight Wj. With Chla as the reference parameter, the normalized correlation weight of the jth parameter is calculated as: 
formula
rij is the correlation coefficient between the jth parameter and the reference parameter Chla; m is the number of evaluation parameters. The correlations between Chla and other parameters of Chinese lakes, rij, and rij2, are shown in Table S1. The nutritional status index is calculated as follows:
  • (1)

    TLI (Chla) = 10 (2.5 + 1.086 ln Chla)

  • (2)

    TLI (TP) = 10 (9.436 + 1.624 ln TP)

  • (3)

    TLI (TN) = 10 (5.453 + 1.694 ln TN)

  • (4)

    TLI (SD) = 10 (5.118 − 1.94 ln SD)

  • (5)

    TLI (CODMn) = 10 (0.109 + 2.661 ln CODMn)

where the units for Chla are mg/m3, and the units for SD are m; the units for the other indexes are mg/L.

To illustrate the eutrophication status of the lake, a series of consecutive numbers from 0 to 100 were used to classify the nutritional status of the lake (Table 1) (Wang et al. 2019).

Table 1

Classification of nutritional status

Parameter Evaluation level Qualitative evaluation Water quality level 
TLI() <30 Oligotropher Excellent 
30 <TLI() <50 Mesotropher Good II 
50 <TLI() <60 Light eutropher Mild pollution III 
60 <TLI() <70 Middle eutropher Moderate pollution IV 
TLI() >70 Hyper eutropher Severe pollution 
Parameter Evaluation level Qualitative evaluation Water quality level 
TLI() <30 Oligotropher Excellent 
30 <TLI() <50 Mesotropher Good II 
50 <TLI() <60 Light eutropher Mild pollution III 
60 <TLI() <70 Middle eutropher Moderate pollution IV 
TLI() >70 Hyper eutropher Severe pollution 

Note: Under the same nutritional status, the higher the comprehensive nutritional index, the more nutritious it is.

Construction of eutrophication prediction model

Factor analysis linearly combines multivariate data, uses dimensionality reduction, and integrates many original variables into fewer PCs, which can reveal the structural characteristics of the data itself and is widely used to reduce the loss of initial information (Moore 1981). The basic mathematical model is as follows: 
formula
In the formula, F1, F2, F3…FK represent the common PC factors, &p is a special factor, and wpk represents the structural load of the pth variable on the kth factor.
According to this mathematical model, the principal component method is chosen to extract the common factors between the variables. The basic mathematical model is as follows: 
formula
where F is the principal component, ZX1…ZXn is the value of the original normalized variable matrix X, a1…amn is the eigenvector corresponding to the eigenvalue of the covariance matrix of the original variable matrix X, n is the number of variables, and m is the number of samples (Tipping & Bishop 1999).
In this study, the regression equation was calibrated, and SPSS data analysis software was selected. A linear model of PCs and Chla is obtained by multivariate linear regression of PCs with Chla: 
formula

In the formula, a0 is a constant coefficient, kn is the regression coefficient of the scores of the nth PC, Sn is the score of the nth PC, and CChla is the content of Chla.

Construction of the water quality warning model

Early warnings generally consist of five processes: identifying the warning, determining the source of the warning, analysing the warning, pre-alarming, and eliminating the warning. Among these steps, identifying the warning is particularly important. Warnings are usually divided into four levels: no warning, light warning, moderate warning and strong warning. The alarm levels can be displayed as different colours in a four-colour display system, such as blue, green, yellow, and red.

The key to predicting a warning is determining the boundaries of the alert. The alarm can be called the threshold value, which is used to determine a reasonable measure that is suitable for the indicator of an early warning. This value is used as a criterion for proposing the normal operation of the forecasted object and determining whether the alarm is generated and its severity during the operation of the forecasted object. When the indicator value exceeds the warning limit, the alarm will be triggered, and the warning can be predicted. According to the severity of the forecast, corresponding measures can be proposed to improve the water quality of the lake.

RESULTS AND DISCUSSION

Eutrophication assessment based on the TLI

The TLI values in this study ranged from 49.56 to 60.02. The results of the water eutrophication evaluation were generally light eutrophy, and the average of the qualitative evaluation was mild pollution (Table 1 and Table 2). The results showed that the trophic level of Gaoyou Lake was generally good. Although the water quality still meets the Chinese drinking water standard after treatment, the TLI fluctuates greatly (Figure 2). Therefore, it is still necessary to evaluate eutrophication in the future.

Figure 2

Monthly variation of nutritional status index in Gaoyou Lake from 2014 to 2017.

Figure 2

Monthly variation of nutritional status index in Gaoyou Lake from 2014 to 2017.

From the temporal perspective, during the study period from 2014 to 2017, the highest TLI value occurred from July to October, and the lowest value generally occurred from December to January, which most likely occurred because temperature influences algae production. This temporal trend confirms that temperature is one of the main factors affecting the eutrophication of water bodies (Zhang 1993). Some researchers used the same method to study eutrophication changes in Gaoyou Lake from 2010 to 2014 (Çamdevýren et al. 2005). The results showed that the TLI of the whole lake and various functional zones ranged between 51.6 and 59.9, all of which exhibited light eutrophy, which is similar to the results of our findings. Therefore, the water quality of Gaoyou Lake is stable and not easily altered.

From a spatial perspective, since Gaoyou Lake is a shallow water lake, the water from Lake Hongzehu to the northwest of Gaoyou Lake flows into the Yangtze River after flowing into Gaoyou Lake. Therefore, the water in the eastern part of Gaoyou Lake is affected by the upstream water supply, and the degree of eutrophication in this area cannot represent the status of the whole lake. The western part of Gaoyou Lake was selected as the sampling point, and the water environment was relatively stable in this area. Gaoyou Lake is the main area of aquatic product production in Jiangsu Province. Simultaneously, this lake is a very large aquaculture base. Some of the remaining feed and animal manure are decomposed, allowing a large amount of nitrogen to enter the water body. Although Gaoyou Lake can be categorized as a water-based lake, the number of species and quantity of plankton have increased significantly compared with those recorded during the 1993 survey (Zhang 1993). Consequently, the possibility that Gaoyou Lake will be further vegetated is not excluded.

Table 2

Water quality assessment of Gaoyou Lake in 2014–2017

Time TIL TIL TIL TIL TIL TLI (Evaluation result Qualitative evaluation 
Chla TN TP SD CODMn 
First half of 2014 44.73 48.54 57.82 68.47 55.35 54.17 Light eutropher Mild pollution 
Second half of 2014 46.83 57.67 63.65 69.32 54.66 57.49 Light eutropher Mild pollution 
First half of 2015 39.91 48.13 38.95 71.51 53.89 49.56 Mesotropher Good 
Second half of 2015 42.61 48.10 42.68 70.14 53.22 50.60 Light eutropher Mild pollution 
First half of 2016 40.84 53.12 39.92 71.30 53.39 51.31 Light eutropher Mild pollution 
Second half of 2016 43.25 53.05 41.04 72.87 54.48 52.08 Light eutropher Mild pollution 
First half of 2017 47.00 68.07 56.98 75.50 53.09 58.99 Light eutropher Mild pollution 
Second half of 2017 47.61 59.58 64.50 76.37 57.33 60.02 Middle eutropher Moderate pollution 
Time TIL TIL TIL TIL TIL TLI (Evaluation result Qualitative evaluation 
Chla TN TP SD CODMn 
First half of 2014 44.73 48.54 57.82 68.47 55.35 54.17 Light eutropher Mild pollution 
Second half of 2014 46.83 57.67 63.65 69.32 54.66 57.49 Light eutropher Mild pollution 
First half of 2015 39.91 48.13 38.95 71.51 53.89 49.56 Mesotropher Good 
Second half of 2015 42.61 48.10 42.68 70.14 53.22 50.60 Light eutropher Mild pollution 
First half of 2016 40.84 53.12 39.92 71.30 53.39 51.31 Light eutropher Mild pollution 
Second half of 2016 43.25 53.05 41.04 72.87 54.48 52.08 Light eutropher Mild pollution 
First half of 2017 47.00 68.07 56.98 75.50 53.09 58.99 Light eutropher Mild pollution 
Second half of 2017 47.61 59.58 64.50 76.37 57.33 60.02 Middle eutropher Moderate pollution 

Chla prediction based on the PCA

According to PCA, out of the seven PCs, only three PCs with eigenvalues higher than 1 were selected for multiple linear regression analysis. The selected PCs explained 76.04% of the total variation in the variables in PCA (Table 3). Furthermore, the communality values of the variables were found to be high in the selected PCs, for example, 69% in TN, 79% in TP and 77% in CODMN (Table 4). These results further confirm the appropriateness of the selected number of PCs (Pituch & Stevens 2015) used in modelling.

Table 3

Total variances for explanation

Observations Initial eigen value
 
Eigen value Contribution rate of variance Contribution rate of cumulative 
2.288 32.692% 32.692% 
1.653 23.614% 56.307% 
1.381 19.731% 76.038% 
0.625 8.925% 84.962% 
0.415 5.931% 90.893% 
0.347 4.963% 95.856% 
0.290 4.144% 100% 
Observations Initial eigen value
 
Eigen value Contribution rate of variance Contribution rate of cumulative 
2.288 32.692% 32.692% 
1.653 23.614% 56.307% 
1.381 19.731% 76.038% 
0.625 8.925% 84.962% 
0.415 5.931% 90.893% 
0.347 4.963% 95.856% 
0.290 4.144% 100% 
Table 4

Results of principal component analysis

Variables Loading of variables
 
Communalities 
PC1 PC2 PC3 
CODMn 0.839 0.134 −0.213 0.771 
TP 0.163 0.882 0.049 0.792 
NH4-N −0.187 0.158 0.824 0.737 
TN 0.24 −0.229 0.765 0.692 
DO 0.129 0.683 −0.224 0.806 
0.326 0.753 −0.122 0.811 
SD 0.453 −0.201 0.482 0.723 
Variables Loading of variables
 
Communalities 
PC1 PC2 PC3 
CODMn 0.839 0.134 −0.213 0.771 
TP 0.163 0.882 0.049 0.792 
NH4-N −0.187 0.158 0.824 0.737 
TN 0.24 −0.229 0.765 0.692 
DO 0.129 0.683 −0.224 0.806 
0.326 0.753 −0.122 0.811 
SD 0.453 −0.201 0.482 0.723 

The high values of communalities indicate that the variances in the variables were efficiently reflected in the regression analysis. All seven variables were included in the three selected PCs. However, only certain variables possessed significant loads within each PC. For example, the only meaningful load in PC1 was found to be CODMn, while TP, DO and T possessed the most significant loads in PC2. However, NH4-N and TN were significant in PC3.

The weights of the variables and the values of the standardized variables were then multiplied to obtain the scores of the PCs. The scores obtained from PCA were used as independent variables in the stepwise multiple linear regression analysis to determine the most significant PCs for the Chla level. Accordingly, score 1, score 2 and score 3 were found to have significant linear relationships with Chla.

All scores possessed a positive impact. In other words, the Chla level would be expected to increase as the scores increased. Consequently, a total increase in significant variables with a score of 1, namely, CODMN, would lead to an increase in the Chla level. On the other hand, an increase in significant variables with a score of 2, e.g., TP, DO and T, would result in an increase in Chla abundance. Similarly, an increase in significant variables with a score of 2, e.g., NH4-N and TN, would result in an increase in Chla abundance.

The predicted (modelled) values of Chla were determined according to a regression model. The significance test results (Table 5) indicate that the regression equation and the regression coefficient passed the significance test at the defined confidence level, indicating that the equation and the data were well fitted. The equation was valid. After calculation and verification, the correlation model between Chla and the scores was obtained: 
formula
Table 5

Multiple linear regression coefficients

Included independent variables Regression coefficient P 
Constant 6.146 0.000** 
Score 1 1.209 0.000** 
Score 2 0.583 0.000** 
Score 3 1.095 0.000** 
Included independent variables Regression coefficient P 
Constant 6.146 0.000** 
Score 1 1.209 0.000** 
Score 2 0.583 0.000** 
Score 3 1.095 0.000** 

*p < 0.05.

**P < 0.01.

The comparison between the measured values and the fitted values and the error are used to reflect the fitting accuracy of the model. Figures 3 and 4 show the comparison and error of the measured and fitted values of the Chla in Gaoyou Lake from 2014 to 2017. The measured value is consistent with the trend of the fitted value, and the error is approximately less than 1.0 μg/L, which is acceptable. From the trend of Chla from 2014–2017, the overall Chla was highest in 2014. Although the Chla from 2014–2015 exhibited an overall declining trend, the overall Chla increased significantly from 2015–2017. This prediction is consistent with the results of the eutrophication assessment. At the same time, from 2014–2017, the Chla contents in summer and autumn were much higher than those in spring and winter. This phenomenon mainly occurred because the water temperature is higher in summer and autumn than in spring and winter, which is more conducive to the growth and absorption of nitrogen and phosphorus by algae (Vollenweider et al. 1974; Raven & Geider 1988; Dai et al. 2016).

Figure 3

Comparison of monthly predicted and actual values of chlorophyll a in Gaoyou Lake from 2014 to 2017.

Figure 3

Comparison of monthly predicted and actual values of chlorophyll a in Gaoyou Lake from 2014 to 2017.

Figure 4

Fitting error values of chlorophyll a in Gaoyou Lake from 2014 to 2017.

Figure 4

Fitting error values of chlorophyll a in Gaoyou Lake from 2014 to 2017.

Under normal conditions, the phosphorus concentrations were highly correlated with the Chla concentrations in lakes (Jones & Bachmann 1976; Van Nieuwenhuyse & Jones 1996). Nevertheless, the nitrogen concentrations and transparency were highly correlated with the Chla concentrations in this study. Furthermore, the concentrations of phytoplankton and Chla in aquatic systems also depend on several variables, including nutrient availability, quality and quantity of light, temperature, physico-chemical properties of the water mass and grazing pressure (Håkanson et al. 2003). Therefore, the concentrations of Chla are largely affected by the physical and chemical characteristics of water.

Various models and approaches have been employed by different researchers to model primary production in aquatic environments (Maier & Dandy 1996; Wehde et al. 2001; An & Park 2002; Faruk 2010; Kauer et al. 2015; Zan et al. 2018). Some researchers have defined the relations of Chla with biotic or abiotic factors as univariate, while others have adapted a multivariate linear or nonlinear approach (Scardi 2001; An & Park 2002). Regardless of the scope of the model, the R2-value is generally accepted as a standard criterion for determining the predictive success of the models (Håkanson et al. 2003). In the present study, a linear regression equation combined with PCA was used to establish an algal biomass response model in Gaoyou Lake with Chla as the dependent variable and the PCs as the independent variables. The R2-value of the above multiple linear regression equation is 0.886, suggesting that the model can accurately describe the relationship between PCs and Chla.

Early warning scheme for eutrophication classification

Gaoyou Lake, which is a source of drinking water, should implement Class II water standards. According to the standard values of lakes and reservoirs, the Chla concentration in Class II water should be below 4 μg/L. However, analysis in combination with specific water environments can more accurately define nutrient levels because this accumulation of algae is closely related to the species and external conditions required for growth (Guven & Howard 2006). Gaoyou Lake is a shallow lake, and shallower water depths are more conducive to algae growth. The water quality of Gaoyou Lake has been monitored by the state and the province for 20 years. The Chla concentration of Gaoyou Lake in the algae development period in summer and autumn is between 8 and 18 μg/L, and the concentration is less than 15 μg/L most of the year. Based on the above factors, we believe that the critical value of the Chla is 8 μg/L, which is in line with the water quality of Gaoyou Lake; that is, the concentration of the Chla exceeds 8 μg/L, which may pose a threat to safe water supply.

The above analysis combined with the early warning alert system described above was used to formulate a plan for an early warning system in Gaoyou Lake (Table S2). The algae biomass response prediction model for the entire lake was established using the monitoring data from Gaoyou Lake from 2014 to 2017, and the monitoring data from Gaoyou Lake in 2017 triggered a warning. The warning results are shown in Table S3. The forecast level was between the green and red levels. Among the 12 data points, 9 of the prediction are accurate, representing 75% of the total, and the early warning results meet the requirements.

CONCLUSIONS

In summary, the TN and TP concentrations in Gaoyou Lake have exceeded the standard in recent years, indicating light and moderate eutrophy and gradually showing a deteriorating trend. Therefore, to ensure ecological protection of the water body, it is necessary to pay attention to the water quality changes and nutritional status of the lake. On this basis, the establishment of a four-colour system for a water quality warning scheme is intuitive and has been shown to be accurate in practical applications. The warning scheme reflects the eutrophication level of Gaoyou Lake and issues a timely alarm to trigger the implementation of targeted pollution prevention measures, which will effectively improve the water quality. Based on the needs to further improve the accuracy of prediction and early warning, the early warning systems should be made more practical and better able to meet the needs of water quality prediction and early warning.

ACKNOWLEDGEMENTS

This work was supported by the National Science Foundation of China (41601573), the Key University Science Research Project of Anhui Province (KJ2019A0641), the Linkage Project of Anhui Public Welfare Technology Application Research (1704f0804053) and the Science and the Technology Innovation Strategy and Soft Science Research Special Project of Anhui Province (1706a02020048). Anonymous reviewers are acknowledged for their constructive comments and helpful suggestions.

REFERENCES

REFERENCES
Atkins
J. P.
,
Burdon
D.
&
Allen
J. H.
2007
An application of contingent valuation and decision tree analysis to water quality improvements
.
Marine Pollution Bulletin
55
(
10–12
),
591
602
.
Barbieri
P.
,
Andersson
C.
,
Massart
D.
,
Predonzani
S.
,
Adami
G.
&
Reisenhofer
E.
1999
Modeling bio-geochemical interactions in the surface waters of the Gulf of Trieste by three-way principal component analysis (PCA)
.
Analytica Chimica Acta
398
(
2–3
),
227
235
.
Bengraïne
K.
&
Marhaba
T. F.
2003
Using principal component analysis to monitor spatial and temporal changes in water quality
.
Journal of Hazardous Materials
100
(
1–3
),
179
195
.
Biggs
B. J.
2000
Eutrophication of streams and rivers: dissolved nutrient-chlorophyll relationships for benthic algae
.
Journal of the North American Benthological Society
19
(
1
),
17
31
.
Bin
C.
&
Jianshe
S.
1991
The fuzzy decision method of water quality nutrition evaluation of lakes
.
Environmental Science
5
,
88
91
.
Çamdevýren
H.
,
Demýr
N.
,
Kanik
A.
&
Keskýn
S.
2005
Use of principal component scores in multiple linear regression models for prediction of Chlorophyll-a in reservoirs
.
Ecological Modelling
181
(
4
),
581
589
.
China Environmental Monitoring Station
2001
Lakes (Reservoirs) Eutrophication Assessment Methods and Classification Technology Requirements
.
China Environmental Monitoring Station
,
Beijing
.
Dai
R.
,
Wang
P.
,
Jia
P.
,
Zhang
Y.
,
Chu
X.
&
Wang
Y.
2016
A review on factors affecting microcystins production by algae in aquatic environments
.
World Journal of Microbiology and Biotechnology
32
(
3
),
51
.
Fanxiu
L.
,
Yan
X.
&
Wu
C.
2000
Set pair analysis method of water quality eutrophication assessment of lakes
.
Chongqing Environmental Science
6
,
10
11
.
Faruk
D. Ö.
2010
A hybrid neural network and ARIMA model for water quality time series prediction
.
Engineering Applications of Artificial Intelligence
23
(
4
),
586
594
.
Flink
P.
,
Lindell
L. T.
&
Östlund
C.
2001
Statistical analysis of hyperspectral data from two Swedish lakes
.
Science of the Total Environment
268
(
1–3
),
155
169
.
Guven
B.
&
Howard
A.
2006
Modelling the growth and movement of cyanobacteria in river systems
.
Science of the Total Environment
368
(
2–3
),
898
908
.
Hamilton
D. P.
&
Schladow
S. G.
1997
Prediction of water quality in lakes and reservoirs. Part I – model description
.
Ecological Modelling
96
(
1–3
),
91
110
.
Ji
J. G.
,
Zhong
K. W.
,
Chen
B.
&
Zou
Y.
2009
Analysis on the influence of aquaculture on water environment in Gaoyou Lake and countermeasures
.
Jiangsu Water Resources
2009
(
11
),
34
35
.
Jiang
L. N.
,
Wu
J. F.
&
Ren
X. M.
2017
Analysis of water quality and eutrophication in Gaoyou Lake in recent years
.
Pollution Control Technology
30
,
23
25
.
Jin
X. C.
&
Tu
Q. Y.
1990
The Standard Methods for Observation and Analysis in Lake Eutrophication
.
Chinese Environmental Science Press
,
Beijing
, pp.
240
.
John
V.
,
Jain
P.
,
Rahate
M.
&
Labhasetwar
P.
2014
Assessment of deterioration in water quality from source to household storage in semi-urban settings of developing countries
.
Environmental Monitoring and Assessment
186
(
2
),
725
734
.
Jones
J. R.
&
Bachmann
R. W.
1976
Prediction of phosphorus and chlorophyll levels in lakes
.
Water Pollution Control Federation
48
,
2176
2182
.
Karul
C.
,
Soyupak
S.
,
Çilesiz
A. F.
,
Akbay
N.
&
Germen
E.
2000
Case studies on the use of neural networks in eutrophication modeling
.
Ecological Modelling
134
(
2–3
),
145
152
.
Kauer
T.
,
Kutser
T.
,
Arst
H.
,
Danckaert
T.
&
Nõges
T.
2015
Modelling primary production in shallow well mixed lakes based on MERIS satellite data
.
Remote Sensing of Environment
163
,
253
261
.
Liu
W. Z.
,
Li
S.
,
Bu
H.
,
Zhang
Q.
&
Liu
G.
2012
Eutrophication in the Yunnan Plateau lakes: the influence of lake morphology, watershed land use, and socioeconomic factors
.
Environmental Science and Pollution Research
19
(
3
),
858
870
.
Maier
H. R.
&
Dandy
G. C.
1996
The use of artificial neural networks for the prediction of water quality parameters
.
Water Resources Research
32
(
4
),
1013
1022
.
Pérez-Arribas
L. V.
,
León-González
M. E.
&
Rosales-Conrado
N.
2017
Learning principal component analysis by using data from air quality networks
.
Journal of Chemical Education
94
(
4
),
458
464
.
Petersen
W.
,
Bertino
L.
,
Callies
U.
&
Zorita
E.
2001
Process identification by principal component analysis of river water-quality data
.
Ecological Modelling
138
(
1–3
),
193
213
.
Pituch
K. A.
&
Stevens
J. P.
2015
Applied Multivariate Statistics for the Social Sciences: Analyses with SAS and IBM's SPSS
.
Routledge
,
New York, NY
.
Raven
J. A.
&
Geider
R. J.
1988
Temperature and algal growth
.
New Phytologist
110
(
4
),
441
461
.
Rene
E.
&
Saidutta
M.
2008
Prediction of water quality indices by regression analysis and artificial neural networks international
.
Journal of Environmental Research
2
(
2
),
183
188
.
Schaeffer
D. J.
,
Herricks
E. E.
&
Kerster
H. W.
1988
Ecosystem health: I. Measuring ecosystem health
.
Environmental Management
12
(
4
),
445
455
.
Tang
C.
&
Fan
X.
2009
Present situation of water environment in Gaoyou Lake and Shaobo Lake and countermeasures
.
China Water Resources
2009
(
11
),
48
49
.
Tipping
M. E.
&
Bishop
C. M.
1999
Probabilistic principal component analysis
.
Journal of the Royal Statistical Society: Series B (Statistical Methodology)
61
(
3
),
611
622
.
Trolle
D.
,
Spigel
B.
,
Hamilton
D. P.
,
Norton
N.
,
Sutherland
D.
,
Plew
D.
&
Allan
M. G.
2014
Application of a three-dimensional water quality model as a decision support tool for the management of land-use changes in the catchment of an oligotrophic lake
.
Environmental Management
54
(
3
),
479
493
.
Van Nieuwenhuyse
E. E.
&
Jones
J. R.
1996
Phosphorus chlorophyll relationship in temperate streams and its variation with stream catchment area
.
Canadian Journal of Fisheries and Aquatic Sciences
53
(
1
),
99
105
.
Vollenweider
R. A.
,
Talling
J. F.
&
Westlake
D. F.
1974
A Manual on Methods for Measuring Primary Production in Aquatic Environments
.
Blackwell Scientific Publications
,
London, UK
.
Wang
M.
&
Cao
Z.
2005
Discussion on the protection and development of wetland resources in Gaoyou Lake
.
Journal of Jiangsu Forestry Science Technology
32
(
1
),
55
56
.
Wang
J.
,
Fu
Z.
,
Qiao
H.
&
Liu
F.
2019
Assessment of eutrophication and water quality in the estuarine area of Lake Wuli, Lake Taihu, China
.
Science of The Total Environment
650
,
1392
1402
.
Wannaz
E.
,
Zygadlo
J.
&
Pignata
M.
2003
Air pollutants effect on monoterpenes composition and foliar chemical parameters in Schinus areira L
.
Science of the Total Environment
305
(
1–3
),
177
193
.
Wehde
H.
,
Backhaus
J. O.
&
Hegseth
E. N.
2001
The influence of oceanic convection in primary production
.
Ecological Modelling
138
(
1–3
),
115
126
.
Wei
W. J.
,
Fu
L. X.
,
Chen
R. M.
&
Sun
W. X.
2010
Water quality and phytoplankton survey and nutritional status evaluation in Gaoyou
.
Lake Resources and Environment in The Yangtze Basin
19
(
S1
),
106
110
.
World Health Organisation
.
2004
Meeting the MDG Drinking Water and Sanitation Target: A mid-Term Assessment of Progress
.
World Health Organization
,
Geneva, Switzerland
.
World Health Organisation
.
2006
Meeting the MDG Drinking Water and Sanitation Target: the Urban and Rural Challenge of the Decade
.
World Health Organization
,
Geneva, Switzerland
.
Xiang
B.
,
Song
J. W.
,
Wang
X. Y.
,
Zhen
J.
&
Gao
R.
2014
Improving accuracy of Eutrophication State Index estimation in Chaohu Lake by moderate resolution remote sensing data driven method
. In:
Remote Sensing and Modeling of the Atmosphere, Oceans, and Interactions V. International Society for Optics and Photonics
, Vol.
9265
, pp.
926509
.
Yang
X.
,
Wu
X.
,
Hao
H.
&
He
Z.
2008
Mechanisms and assessment of water eutrophication
.
Journal of Zhejiang University Science B
9
(
3
),
197
209
.
Zhang
T. Q.
1993
An investigating report of phytoplankton's resources of Gao. Bao. Shaobo Lake
.
Journal of Aquaculture
6
,
12
14
.
Zhang
L.
&
Li
H.
2016
Evaluation system of urban environmental pollution based on the principal component analysis
. In:
2016 International Conference on Smart City and Systems Engineering (ICSCSE)
.
IEEE
, Vol.
2016
, pp.
47
49
.
Zhang
J.
,
Ni
W.
,
Luo
Y.
,
Stevenson
R. J.
&
Qi
J.
2011
Response of freshwater algae to water quality in Qinshan Lake within Taihu Watershed
.
China Physics and Chemistry of the Earth, Parts A/B/C
36
(
9–11
),
360
365
.
Zhao
Z.
,
Wang
Y. B.
&
Wang
L. L.
2015
Study on the trend of water quality and eutrophication in the eco-functional regions of Gaoyou Lake in recent years
.
Zhi Huai
2016
(
1
),
21
22
.

Supplementary data