Abstract

This study was conducted to investigate the groundwater quality features of the Yongding River in Beijing, China, and its relationship with urban development and ecological restoration projects. The Yongding River has been cut off all year around and the ecological environment has continued to deteriorate. Therefore, a series of river ecological restoration projects of ‘Five Lakes on One Route’ have been implemented. In order to characterize the physico-chemical properties of groundwater and evaluate the effects of these projects on groundwater quality, by using principal component analysis, this study analyzed spatial and temporal variation on the basis of 11 water quality parameters at 10 monitoring sites of ‘Five Lakes on One Route’ for Yongding River during April and September of 2011 and 2016. Principal component analysis demonstrated that relatively poor groundwater is mainly distributed in Fengtai District residential and industrial land, and the groundwater in Mentougou District woods is generally better. The groundwater quality at eight monitoring sites kept the same level or became better, and the construction of the river ecological restoration projects of ‘Five Lakes on One Route’ is important for protecting the groundwater resource.

INTRODUCTION

Groundwater plays key ecological and social roles within urban systems. It is very useful for interpreting groundwater quality data to understand specific hydrogeochemical processes. The deterioration of groundwater quality is influenced by natural processes including climatic conditions, weathering processes, and sediment transport and also by anthropogenic inputs including urban development and expansion, and industrial and agricultural practices (Muangthong & Shrestha 2015). Although the situation seems to have improved with a series of ecological restoration projects in a general increase since the 1990s, the quality of groundwater in large cities such as Beijing is still far from satisfactory. A large number of recent urban ecology studies have focused on management and restoration of urban groundwater, however, knowledge regarding the basic mechanism and factors influencing groundwater quality in urban areas is still limited (Conte et al. 2012). Therefore, regular monitoring and evaluating of the groundwater quality around rivers are equally important for managing water resources and the water environment.

Since the 1980s, the flow of the Yongding River, which is the largest river in Beijing, has been continually decreasing because of reduced outflow from the Guanting Reservoir. The river channel in the plain has been continuously interrupted, thus it has completely dried up and the surrounding groundwater level has continued to decline. In order to create a beautiful water environment, the Beijing Water Authority (BWA) approved the construction of the Yongding River Green Ecological Corridor in 2009. In February 2010, the construction of the river ecological restoration project from the Sanjiadian to Lugouqiao section of Yongding Diversion Channel was started. In 2012, the project of the 18.4 km river channel ecological restoration was completed.

‘Five Lakes on One Route’ (from Sanjiadian to Lugouqiao) for Yongding River is an economically and ecologically important area situated in Beijing, where anthropogenic activities are very intensive. Using management methods which can reveal the critical issues seems necessary due to some limitations such as labor and funding. Techniques of multivariate statistical analysis (such as cluster analysis, principal component analysis (PCA), factor analysis and discriminants analysis) have been used widely to gain understanding of water quality with respect to spatial and temporal variability, hydrochemical facies, flow paths, and other factors that influence groundwater quality (Pacheco et al. 2017). PCA is a multivariate statistical technique used to reduce the dimensionality of the data set by explaining the correlation among a large group of variables in terms of a small number of underlying factors or principal components without losing much information (Leigh 1993). Most of the previous uses have been to find out the major contributing parameters involved in deciding the geochemistry of groundwater samples. In this study, we applied PCA as a management and decision support system tool for analyzing spatial and temporal variations of groundwater quality from 2011 to 2016. We studied the correlation between groundwater quality of the natural and human processes these two major factors, and then evaluated the influence of river ecological restoration projects on groundwater quality in ‘Five Lakes on One Route’ for Yongding River.

Based on this, we can achieve a better understanding of the behavior of the groundwater system over temporal and spatial scales. At the same time, this study can provide reference and experience for the optimization of the technical scheme for the eco-restoration and the allocation of the water resources concerned in the Yongding River in the future.

MATERIALS AND METHODS

Study area

The Yongding River is located in the southwest part of Beijing (Figure 1(b)) and is known as the mother river of Beijing. The Yongding River is 747 km long with a watershed area of 47,016 km2. Since the 1980s, due to the years of drought, the lower reach of the Yongding River basin plain has not been able to get enough water supplies, which has caused river dry-up, aquatic biological extinction and water ecological environment degradation.

Figure 1

(a) Location map of the study site in China; (b) locations of the study area in Beijing and locations of cross-section A–A′; (c) the sketch map of the study area and the distribution of the 10 monitoring sites; (d) the hydrogeological cross-section along A–A′.

Figure 1

(a) Location map of the study site in China; (b) locations of the study area in Beijing and locations of cross-section A–A′; (c) the sketch map of the study area and the distribution of the 10 monitoring sites; (d) the hydrogeological cross-section along A–A′.

In 2009, ‘Beijing Yongding River Green ecological corridor construction plan’ was approved by BWA. ‘Five Lakes on One Route’ for Yongding River has carried out the research on ecological restoration and ecological water requirement. ‘Five Lakes on One Route’ (from Sanjiadian to Lugouqiao), with a length of about 17 km, gradually falls from the mountains into the low mountains and plains (Yu et al. 2017). From upstream to downstream, the area is divided into Mencheng Lake, Lianshi Lake, Yuanbo Lake, Xiaoyue Lake and Wanping Lake. The total area is 800 hm2, and the length of the circulating pipeline is 22 km. In ecological restoration, plant revetment technology is used to cover the original concrete hard revetment with soil. The implemented flood control projects, ecological restoration works in dikes, infiltration reduction projects, embankment ecological restoration projects, and water quality improvement projects have been put into action (Qiang et al. 2018). According to local conditions, the greening area is increased to realize the functions of soil conservation, water conservation, flood control and scour prevention (Figure 1(c)). At the same time, it also provides a safe and stable habitat for amphibians, thus properly handling the contradiction between the flood control of river channels and greening space of riverbeds. The Beijing Plain is mainly composed of the Yongding River alluvial fan and the Chaobai River alluvial fan. Rivers are the main recharge sources of aquifers. The vertical hydrogeologic conditions in these two sub river basins are as follows (Figure 1(d)).

Groundwater sampling and analysis

To understand the variability of the spatial and temporal processes affecting groundwater quality, along the 17 km stretch of river, sampling of groundwater was carried out in 10 monitoring sites over a period of 6 years (2009–2016). Water samples were collected twice in the year, namely, in April and September which are months in the dry and wet seasons. This is because in these 2 months, the impacts on the groundwater from precipitation are generally the smallest and the largest of the entire year, respectively, because of the climatic and aquifer characteristics in this area.

To obtain representative samples from the monitoring wells, each well was purged using a stainless-steel bailer before the samples were collected (Zhai et al. 2015). Wells were pumped at rates less than 2 L per minute to minimize vertical mixing of groundwater during sampling. The samples were collected at a depth of 0.5–2 m below the groundwater level, placed into plastic bottles (2.5 L), transported to the laboratory and stored at 0–4 °C for subsequent chemical analysis (Wang et al. 2015). These water quality parameters were CODMn, THard, total dissolved solids (TDS), Cl, SO42−, Ca2+, Mg2+, Na+, K+, HCO3 and NO3 (Ogwueleka 2015). Dichromate reflux method was used to determine CODMn. THard was determined by complexometric titration and TDS was determined by an automatic TDS meter. Cl, SO42−, Ca2+, Mg2+, Na+, K+, HCO3 and NO3 were determined by ion chromatography using ion chromatographer (Metrohm 782 Basic IC) (Khan et al. 2017).

PCA algorithm

This section first explains briefly PCA. Principal component analysis is a kind of statistics technique that people use to determine patterns of data in high dimension. The PCA decomposes the covariance matrix of the variables into its principal components and their magnitudes, thereby enabling grouping of like variables based on the magnitude of their covariance. In this case, a component is considered significant if the eigenvalue is greater than one. Mathematically, PCA normally involve the following five major steps: (1) collect data with their corresponding features, and then build a matrix; (2) normalize the data; (3) build the correlation matrix; (4) rearrange the data by their eigenvalues; (5) develop the factor loading matrix and perform a varimax rotation on the factor loading matrix to infer the principal components (Ouyang 2005). The way PCA works is to reduce the number of dimensions of the data. Consequently, it is used to process a large amount of data into a smaller range.

RESULTS AND DISCUSSION

Groundwater quality summary statistics

Box plot analysis and piper diagram analysis of the water quality data were performed using the SPSS 23 and RockWare-AqQA. The data are presented graphically in time series to allow visual evaluation alongside the summary text (Zhao et al 2010). Basic statistics of the 6-years data set on river water quality such as minimum, maximum, average was summarized in Figure S1 (see online Appendix), which showed some descriptive statistics for the 11 physico-chemical parameters monitored in 10 groundwater samples (de Andrade et al. 2008). Ca2+ and HCO3 are the major cation and anion in the study area, so the groundwater composition is dominated mainly by limestone dissolution. The water quality standard for groundwater III (GB/T 14848-2017), TH 450 mg/L, TDS 1,000 mg/L, SO42− 250 mg/L, Cl 250 mg/L, Na+ 200 mg/L, NO3 20 mg/L and CODMn 3 mg/L. In most of the time, the SO42−, Cl, Na+ and CODMn in groundwater met the criterion, but TH, TDS and NO3 were higher than the criterion. The excessive nitrate concentration suggested hydrological systems of the study area can be influenced by anthropogenic pollution.

Principal component analysis

In order to further understand temporal and spatial processes of groundwater quality, this study is to first apply PCA sequentially to data of each sampling period, and then utilize PCA to different sampling periods of each site. In this study, PCA is done using the 11 chemical variables. The variables for PCA were THard, TDS, SO42−, K+, Na+, Ca2+, Mg2+, HCO3, Cl, NO3 and CODMn.

For example in 2016, the correlation matrix of the 11 water quality analysis variables (Table 1) was calculated from data normalized as described in the subsection ‘PCA algorithm’. According to the Kaiser criterion, our discussion should focus only on the first two components, which have factors higher than 1. Table 2 presents the results of the PCA for groundwater quality. The eigenvalues for the first two components are 6.089 and 3.186, which explain 84.32% of the total variance in the data set. The first component (PC1) accounts for 55.36% of the variance and is composed of positive loadings for TH, TDS, SO42−, Na+, Ca2+, Mg2+, Cl and NO3. The second component (PC2) accounts for 28.96% of the variance and is composed of positive loadings for K and CODMn. These two new variables (principal components) represent linear combinations of the 11 initial variables, reducing the complexity of the analysis and the amount of variables to consider.

Table 1

Correlation matrix of the 11 groundwater quality parameters

THTDSSO42−K+Na+Ca2+Mg2+HCO3ClNO3CODMn
TH 1.000           
TDS 0.941 1.000          
SO42− 0.838 0.933 1.000         
K+ 0.017 0.262 0.319 1.000        
Na+ 0.475 0.608 0.589 0.589 1.000       
Ca2+ 0.697 0.605 0.683 −0.183 0.201 1.000      
Mg2+ 0.711 0.610 0.680 −0.118 0.300 0.951 1.000     
HCO3 0.736 0.568 0.551 −0.317 0.170 0.935 0.927 1.000    
Cl 0.478 0.636 0.736 0.590 0.733 0.532 0.536 0.383 1.000   
NO3 0.453 0.611 0.683 0.752 0.559 0.411 0.473 0.267 0.752 1.000  
CODMn −0.048 0.191 0.252 0.946 0.429 −0.170 −0.152 −0.316 0.531 0.729 1.000 
THTDSSO42−K+Na+Ca2+Mg2+HCO3ClNO3CODMn
TH 1.000           
TDS 0.941 1.000          
SO42− 0.838 0.933 1.000         
K+ 0.017 0.262 0.319 1.000        
Na+ 0.475 0.608 0.589 0.589 1.000       
Ca2+ 0.697 0.605 0.683 −0.183 0.201 1.000      
Mg2+ 0.711 0.610 0.680 −0.118 0.300 0.951 1.000     
HCO3 0.736 0.568 0.551 −0.317 0.170 0.935 0.927 1.000    
Cl 0.478 0.636 0.736 0.590 0.733 0.532 0.536 0.383 1.000   
NO3 0.453 0.611 0.683 0.752 0.559 0.411 0.473 0.267 0.752 1.000  
CODMn −0.048 0.191 0.252 0.946 0.429 −0.170 −0.152 −0.316 0.531 0.729 1.000 
Table 2

Loadings of 11 variables on two significant principal components

VariablePC1PC2
TH 0.847 −0.303 
TDS 0.900 −0.038 
SO42− 0.931 0.006 
K+ 0.361 0.916 
Na+ 0.667 0.420 
Ca2+ 0.779 −0.526 
Mg2+ 0.805 −0.478 
HCO3 0.698 −0.650 
Cl 0.829 0.337 
NO3 0.775 0.484 
CODMn 0.299 0.895 
Eigenvalue 6.089 3.186 
% of variance 55.36 28.96 
% cumulative variance 55.36 84.32 
VariablePC1PC2
TH 0.847 −0.303 
TDS 0.900 −0.038 
SO42− 0.931 0.006 
K+ 0.361 0.916 
Na+ 0.667 0.420 
Ca2+ 0.779 −0.526 
Mg2+ 0.805 −0.478 
HCO3 0.698 −0.650 
Cl 0.829 0.337 
NO3 0.775 0.484 
CODMn 0.299 0.895 
Eigenvalue 6.089 3.186 
% of variance 55.36 28.96 
% cumulative variance 55.36 84.32 
From the eigenvectors obtained in the PCA, the principal component synthesis model, F, can be given as: 
formula
(1)

The notation used to denote samples is as follows (Mahapatra et al. 2012): Z is the monitoring well, the number following X denotes the well number, and the coefficients are the eigenvectors. The spatial distribution of the component scores, which are the linear combinations for each variable at each well, is shown in Figure 2(a). The colours white and gray respectively denote the dry and rainy seasons. The numbers from 2011 to 2016 respectively denote the years of sampling.

Finally, the comprehensive scores of 10 monitoring sites are calculated, and the quantitative description of the water pollution degree is given to each site. The higher the score, the more serious the pollution degree.

Spatial variation in groundwater quality

The comprehensive scores identified above are plotted in Figure 2 to analyze the spatial variability of the groundwater quality governing processes. From Figure 2(a), the relatively poor groundwater is mainly distributed in Fengtai District and the groundwater in Mentougou District is generally better. This was likely due to rapid economic growth; land use has become one of the most important factors responsible for groundwater quality. According to the survey of land use types in Mentougou District and Fengtai District in Figure 3(b), it can be seen that the land use types in Mentougou District are mainly woodland and grassland, accounting for 87.79% of the total area; urban village and industrial and mining land in Fengtai District account for the largest proportion, up to 63.79%, and transportation land also accounts for 9.03%. In addition, the Yingshanzui monitoring site is the most polluted of the 10 monitoring sites and the Liyuanzhuang monitoring site is of better groundwater quality relative to the other sites in ‘Five Lakes on One Route’. This was likely due to the machinery factories and concretion plants around the Yingshanzui monitoring site. These factors have hampered ecological restoration of the downstream in ‘Five Lakes on One Route’ for Yongding River. In fact, the distribution patterns of groundwater quality have been affected by the various sources of contamination, but a more in-depth statistical analysis is required to understand other sources of contamination. These findings reflect that the groundwater quality differences among monitoring sites have been caused by different types of land use and human activities around the monitoring sites. Therefore, we need to strengthen management and supervision of human activities in Fengtai District.

Figure 2

Spatial distribution and temporal distribution of principal component synthesis scores.

Figure 2

Spatial distribution and temporal distribution of principal component synthesis scores.

Figure 3

Groundwater level vs time of some monitoring wells and land use types in the study area.

Figure 3

Groundwater level vs time of some monitoring wells and land use types in the study area.

Temporal variation in groundwater quality

Similarly, the temporal variation trend in the groundwater system also can be identified by the representation of the principal component synthesis model. We found that the ecological restoration project has had the greatest impact on ‘Five Lakes on One Route’ for Yongding River in recent years. In order to investigate whether the groundwater quality changed before and after the ecological restoration project, this study used the groundwater quality data of 2009, 2013 and 2016 for PCA. The reason for choosing 2016 is that the impact of the river restoration project on groundwater quality is a long-term process and often lags behind, and it is still necessary to continuously sample and monitor the groundwater in the study area for a long time. The analysis steps are the same as those described above. From Figure 2(b), it may be noted that the groundwater quality only at the Meishikou and Liyuanzhuang monitoring sites tended to decrease and the eight others improved significantly in 2016, but the ecological projects were just completed in 2013. At that time, the groundwater quality had not obviously improved compared with that before the projects. In 2016, the groundwater quality in the eight monitoring points had improved significantly. Our study results suggest that the construction of the river ecological restoration project of ‘Five Lakes on One Route’ is important for protecting the groundwater resource and there is also a delayed effect.

The decline of groundwater quality at the Liyuanzhuang monitoring site only occurred during the flood season. The impact on the groundwater level from precipitation is generally the largest of the entire year during this time. Figure 3(a) shows that the groundwater level has a downward trend from upstream to downstream. Due to the close distance between monitoring sites, the groundwater in the upstream will be supplied to the downstream, which may be an important natural reason for the deterioration of the water quality at the Liyuanzhuang monitoring site. From 2014 to 2015, we found that the average concentration of nitrate nitrogen at the Meishikou monitoring site increased from 1.97 mg/L to 35.6 mg/L. Therefore, there were nitrogen-containing pollution sources near Meishikou. Nitrate nitrogen pollution in groundwater might be caused by surface sewage discharge, and groundwater was polluted by river leakage: urban septic tank, leakage of polluted pipelines, rainwater leaching from garbage dumps and so on. In addition, after the South-to-North Water Diversion Project went to Beijing, the groundwater level increased and there was a garbage dump called Laoshanxi near the Meishikou monitoring site, which might dissolve the nitrogen in the soil. We still need to do further investigation on which factor played the leading role.

CONCLUSIONS

Investigation of the groundwater quality from Sanjiadian to Lugouqiao showed the evolution of groundwater chemistry in the study area. The study results showed that both natural and human processes were the two major factors for the chemical compositions of groundwater.

PCA further analyzed spatial and temporal variations which are important for the design of groundwater protection schemes and which were used to identify various underlying natural and anthropogenic processes. The groundwater quality data from 10 monitoring sites showed that the groundwater quality differences among monitoring sites have been caused by different types of land use and human activities around the monitoring sites. The groundwater quality in most of these monitoring sites kept at the same level or became better. The construction of the river ecological restoration projects of ‘Five Lakes on One Route’ is important for protecting the groundwater resource. However, as the impact of the River Restoration Project on groundwater quality is a long-term process and often lags behind, it is still necessary to continuously sample and monitor the groundwater in the study area for a long time. Once problems are found, remedial measures shall be taken in a timely manner. The conclusion of this study will not only help in the study of groundwater chemical characteristics but also shed light on the methodologies that can understand the cause of groundwater contamination, and therefore provide a good reference for groundwater resource management in urban areas.

ACKNOWLEDGEMENTS

This research is supported by Natural Science Foundation of Beijing, China (Nos. Z160001, Z170004). We also thank the anonymous reviewers for their useful suggestions.

SUPPLEMENTARY DATA

The Supplementary Data for this paper are available online at http://dx.doi.org/10.2166/ws.2019.119.

REFERENCES

REFERENCES
Conte
G.
Bolognesi
A.
Bragalli
C.
Branchini
S.
De Carli
A.
Lenzi
C.
Masi
F.
Massarutto
A.
Pollastri
M.
Principi
I.
2012
Innovative urban water management as a climate change adaptation strategy: results from the implementation of the project ‘Water Against Climate Change (WATACLIC)’
.
Water
4
(
4
),
1025
1038
.
de Andrade
E. M.
Palácio
H. A. Q.
Souza
I. H.
de Oliveira Leão
R. A.
Guerreiro
M. J.
2008
Land use effects in groundwater composition of an alluvial aquifer (Trussu River, Brazil) by multivariate techniques
.
Environmental Research
106
(
2
),
170
177
.
Leigh
S.
1993
Book Reviews: A user's guide to principal components
.
Technometrics
35
(
1
),
83
85
.
Mahapatra
S. S.
Sahu
M.
Patel
R. K.
Panda
B. N.
2012
Prediction of water quality using principal component analysis
.
Water Quality, Exposure and Health
4
,
93
104
.
Pacheco Castro
R.
Pacheco Ávila
J.
Ye
M.
Cabrera Sansores
A.
2017
Groundwater quality: analysis of its temporal and spatial variability in a karst aquifer
.
Groundwater
56
,
62
72
.
Qiang
Y.
Yu
Z.
Cuiping
Y.
Dong
H.
2018
Review and suggestions on river ecological management in urban section of Yongding River in Beijing
.
China Flood & Drought Management
28
(
3
),
28
32
,
40
.

Supplementary data