Journal of Environmental Management 273 (2020) 111185

Contents lists available at ScienceDirect

Journal of Environmental Management

journal homepage: http://www.elsevier.com/locate/jenvman

Research article

Assessment of the hydrodynamics role for groundwater quality using an

integration of GIS, water quality index and multivariate
statistical techniques
Yunhai Fang a, Tianyuan Zheng b, *, Xilai Zheng a, **, Hui Peng a, Huan Wang a, Jia Xin a,
Bo Zhang a
a
Key Laboratory of Marine Environment Science and Ecology, Ministry of Education, Shandong Provincial Key Laboratory of Marine Environment and Geological
Engineering (MEGE), College of Environmental Science and Engineering, Ocean University of China, Qingdao, 266100, China
b
College of Engineering, Ocean University of China, Qingdao, 266100, China

A R T I C L E I N F O A B S T R A C T

Keywords: To explore the impact of groundwater hydrodynamics on water quality, a cost-effective geospatial model was
Water quality index developed using geographic information system (GIS) technology and the Dupuit assumption. Meanwhile, the
Groundwater hydrodynamics groundwater quality in the Dagu River Basin was evaluated based on the water quality index (WQI) and
GIS technology
multivariate statistical analyses. In April (dry season) and September (rainy season) 2017, the groundwater level
Dagu river
was automatically monitored from 115 wells, and the water quality including 21 hydrochemical parameters was
sampled from 37 wells. Results reveal that the WQI values varied from 35.01 to 64.74, with mean values of 51.89
and 47.87 in the rainy and dry seasons. Approximately 80% of the samples exhibited moderate water quality,
with no significant difference between the rainy and dry seasons. Nitrate pollution and the integrated water
quality in the central and northern regions were generally worse than that in the southern region. The Darcy
velocity in the central and northern regions was relatively high with a maximum rate of 0.56 m/d, compared
with the southern region. This correlation illustrates the effect of groundwater hydrodynamics on quality. The
sowing of greater chemical fertilizers combined with faster groundwater movement is likely responsible for the
large–scale nitrate pollution in the central and northern regions. Results also proved the accuracy of the geo­
spatial model with a valid uncertainty. The geospatial model provides a valuable alternative for the spatial
analysis of the effect of groundwater hydrodynamics on water quality.

1. Introduction determining the water quality in aquatic bodies, and it plays an

increasingly significant role in water resources management (Abbasnia
The deterioration of water quality as a result of direct and indirect et al., 2018, 2019; Machiwal et al., 2019; Saleh et al., 2019; Wang et al.,
anthropogenic activities reduces available water resources; this decline 2017; Wu et al., 2018; Yousefi et al., 2017). Compared with traditional
further threatens human health, ecosystem services, and national secu­ water quality evaluation, this method converts multiple environmental
rity (Sotomayor et al., 2018). Extensive investigations have been parameters into a single number expressing the status of water quality
attempted to conduct water physicochemical monitoring and quality and effectively provides integrated information regarding the overall
assessment with the objective of providing guidance for water resources quality. Lumb et al. (2011) have reviewed the evolution of WQI, and
management. To date, several water quality assessment methods (e.g., readers may get more details about this method.
water quality index (WQI), neural networks, analytic hierarchy process, Based on the identification of integrated water quality status, a few
and variable fuzzy sets theory (Fang et al., 2019)) have been designed to studies have explored the factors that influence water quality (Ferreira
determine the status and key contributors to quality variations. The WQI et al., 2019; Wang et al., 2017; Wu et al., 2017). Wu et al. (2017)
method, commonly used worldwide, is a simple but powerful tool for assessed the water quality of Lake Poyang by WQI and analyzed the

* Corresponding author.
** Corresponding author. College of Environmental Science and Engineering Ocean University of China, Qingdao, 266100, China.
E-mail addresses: [email protected] (T. Zheng), [email protected] (X. Zheng).

https://doi.org/10.1016/j.jenvman.2020.111185
Received 13 February 2020; Received in revised form 4 June 2020; Accepted 1 August 2020
Available online 5 August 2020
0301-4797/© 2020 Elsevier Ltd. All rights reserved.
Y. Fang et al. Journal of Environmental Management 273 (2020) 111185

influence of water level on quality using correlation analysis. Ground­ that the boundary conditions can be accurately described. Apart from
water is an important source of freshwater, and groundwater movement water quality monitoring data, this numerical model requires time­
and solute transport directly affect water quality (Majumder and Eldho, –consuming training and calibration procedures, as well as detailed
2019). Traditionally, the numerical simulation of groundwater con­ datasets including hydrology, geology, meteorology, and hydrogeology
tributes to understanding these processes (Majumder and Eldho, 2019; (Kavetski et al., 2006). Skilled simulators and associated disciplinary
Xiao et al., 2019). However, previous studies usually evaluated knowledge are also the key requisites to accomplish such a goal. These
groundwater quality using a single environmental parameter and preconditions may be a challenge for policy-makers and practitioners in
analyzed the causal relationship with groundwater hydrodynamics (Lal water resources management. Therefore, the development of a fast,
and Datta, 2019; Mohammadi et al., 2018; Xiao et al., 2019). Koh et al. simple, and inexpensive method to explore the impact of groundwater
(2016) numerically simulated the degradation of groundwater quality hydrodynamics on water quality is essential, particularly in developing
by the index of nitrate in a layered aquifer system. For the evaluation by countries where basic data and technology are limited.
a single environmental parameter, different parameters may lead to This study was conducted as a preliminary investigation on
various results of water quality assessment (Wu et al., 2018). The rela­ groundwater quality by 21 hydrochemical parameters and groundwater
tionship between groundwater hydrodynamics and quality may be level in the Dagu River Basin, a typical groundwater source in China.
confusing due to these changes. Therefore, the influence of groundwater The primary objectives of this study were (1) to identify the integrated
hydrodynamics on integrated water quality regarding the overall in­ water quality status based on the WQI and multivariate statistical ana­
formation needs further investigation. lyses; (2) to develop a geospatial model for simple and cost-effective
Establishing the relationship between groundwater hydrodynamics analysis of groundwater hydrodynamics; and (3) to determine whether
and quality in actual aquifer system based on numerical simulation is a the regional variation in integrated groundwater quality may be asso­
complex course (Xiao et al., 2019). Before the simulation, the processes ciated with the characteristics of groundwater movement. The geo­
of groundwater dynamics in the aquifer should be well understood so spatial model and findings from this study will be further applied to

Fig. 1. (a) Study area and location of sampling points; Hydrogeological cross-section from point A to B in (a). The base line of elevation is the mean sea level. (black
and white in print).

2
Y. Fang et al. Journal of Environmental Management 273 (2020) 111185

enhance groundwater management system and to better protect fresh­ include parameters selection, normalization, and aggregation. In this
water resources. study, principal component analysis (PCA) and factor analysis (FA) were
employed to determine the principal water quality parameters and its
2. Data and methods weights. This process was implemented in SPSS 22.0. The study by Qian
et al. (2007) provided further details on the application of PCA and FA.
2.1. Study area Next, the measured values of selected parameters were used for
normalization. This step was conducted based on the national standard
The Dagu River Groundwater Source (DRGS) (36◦ 18′ 00′′ –36◦ 45′ for groundwater quality (China Gaqsiq, 2017). The completed criteria
45 N, 120◦ 04′ 48′′ –120◦ 21′ 00′′ E) is the most stable and crucial
′′
are listed in Table S3. Finally, the aggregation formula (eq. (1), proposed
freshwater resource in Qingdao, with an area of 456 km2 (Fig. 1). The by Wu et al. (2018)), combining the score of various parameters, was
groundwater is mainly stored in a phreatic aquifer (usually 4–8 m invoked to obtain the WQI.
thickness) composed of sand and gravel. The bottom and edge of DRGS ⎛ ⎞
are mostly enclosed by the cretaceous siltstone and clay rock, which ∑n ⎜ ⎟
form a system of impervious base and boundary that facilitates WQI = ⎜ Pi × C i ⎟ (1)
⎝∑n ⎠
groundwater accumulation. The overlying cohesive soil layer is i=1 Pi
composed of a thin layer of clayey sand. The burial conditions are
i=1

conducive to the infiltration of atmospheric precipitation, which is the where n is the total number of major environmental parameters, Pi is the
primary source of groundwater recharge in the DRGS. The average weight of parameter i, and Ci is the score of parameter i after normali­
rainfall is 667.95 mm/a, with significant seasonal difference between zation. The WQI ranges from 0 to 100, with the higher value repre­
the rainy period (May–September) and the dry period (October–April). senting better water quality. The classification method proposed by
Fulazzaky et al. (2010) was used to divide groundwater quality into five
2.2. Sample collection and laboratory analysis categories according to the WQI value: excellent (80–100), good
(60–80), moderate (40–60), low (20–40), and bad (0–20).
To manage groundwater resources in the area, the Dagu River Au­
thority has performed considerable work, including regular monitoring 2.4. Calculation of Darcy velocity
of groundwater quality and installation wells to monitor the ground­
water level automatically. To investigate seasonal effects on ground­ A new geospatial model was established to calculate Darcy velocity
water quality, samples including 21 hydrochemical parameters were and to analyze the groundwater hydrodynamics across the considered
monitored in both the rainy (September) and dry (April) seasons. These aquifer using the integration of Geographic information system (GIS),
include pH, total hardness (TH), total dissolved solids (TDS), chloride the Dupuit assumption, and hydrogeological data. The Dupuit assump­
(Cl), fluoride (F), sulfate (SO4), ammonium (NH4–N), nitrate (NO3–N), tion is a well-known theory that lays a foundation for many hydrological
nitrite (NO2–N), permanganate index (CODMn), volatile phenol, cya­ problems (Bear, 1972). GIS technology can be employed to build pre­
nide, arsenic, mercury, cadmium, chromium (hexavalent), lead, iron, diction surfaces based on limited data, along with the reduction in
copper, manganese, and zinc. pH was measured in situ via a portable monitoring costs. Previous studies have demonstrated the usefulness of
instrument (Hach HQ40D) with a pH electrode. Other chemical pa­ GIS technology for water quality assessment (Yousefi et al., 2019a,
rameters were measured at the laboratory of the Water Environment 2019b).
Monitoring Center (Qingdao, China) according to a system of national The Dupuit assumption (Bear, 1972) was proposed based on the fact
and professional standards (China Epa, 2004; China Gaqsiq, 2017). that the slope of phreatic surface is small for the steady-state flow of
Table S1 in the supplementary materials provides the detection limits. groundwater (Fig. 2). Thus, the vertical component of velocity can be
For this study, the groundwater quality datasets were collected in April ignored. The Darcy velocity at each point on the vertical section (B) can
(dry season) and September (rainy season) 2017 from 37 monitoring be expressed as
wells (Fig. 1). Certain changes in monitoring strategies, however, have
limited the availability of data from only 26 wells during the dry season. vx ≈ vs = − K tan θ = − K
dH
=− K
dz
(2)
The groundwater quality datasets constructed in this study include 37 × dx dx
21 (rainy season) and 26 × 21 (dry season).
where θ is the slope angle, H is the hydraulic head, and K is the hydraulic
The monitoring frequency of groundwater level was twice a day, at
conductivity. It has been previously demonstrated that the error caused
8:00 and 20:00, respectively. The groundwater level was automatically
by this algorithm is slight when dHdx << 1.
measured through a multi–parameter equipment (YSI–6920) with a
Before the calculation of Darcy velocity based on eq. (2), three items
depth sensor. The monitoring process was guided by professional stan­
need to be completed: 1) groundwater level data; 2) prediction surface
dards (China Epa, 2004). Previous evidence suggested that the impact of
groundwater hydrodynamics on quality was not an instantaneous pro­ of the water table; and 3) hydraulic gradient dH dx . The latter two can be
cess. Thus, in this study, it was explored on a monthly scale. The completed with GIS technology.
groundwater level datasets were collected from 115 monitoring wells
(Fig. S1) on April 1–30th and September 1–29th, 2017. 2.4.1. Determination of potentiometric surface
Hydraulic conductivity has a critical influence on groundwater flow The prediction surface of the water table could be established by the
and is an indispensable parameter for calculating the seepage velocity. A geostatistical interpolation (kriging) methods in ArcGIS (version 9.2),
hydrogeological survey indicates that the study area can be divided into based on the dataset of 115 auto–monitoring wells. These methods are
different subregions according to the permeability (Fig. S2). Table S2 capable of providing a measure of prediction accuracy. The optimal
lists the hydraulic conductivity values used in this study; these data have predictions were identified by cross–validation and validation tools,
been verified by Zeng et al. (2017) using numerical models. based on the prediction error indices, including mean, mean standard­
ized, and root mean squared standardized prediction errors. For unbi­
2.3. WQI calculation ased predictions with valid uncertainty, the optimal predictions must
satisfy the following conditions: the mean prediction error (M) and
The WQI is defined as a simple value that reflects the status of water mean standardized prediction error (S) should approach 0, and the root
quality and potential trends by a summary of different environmental mean squared standardized error (R) should approach 1.
parameters (Lumb et al., 2011). The primary steps of WQI calculation

3
Y. Fang et al. Journal of Environmental Management 273 (2020) 111185

Fig. 2. Flowchart of WQI method and the geospatial model that was established to analyze groundwater hydrodynamics across the considered aquifer. This protocol
was applied to identify the influence of groundwater hydrodynamics on integrated water quality. (black and white in print).

2.4.2. Determination of hydraulic gradient groundwater quality parameters and the double–sample test of Kolmo­
The hydrological analysis module of ArcGIS provides a means to gorov–Smirnov to check whether the samples in the two periods were
calculate the direction and hydraulic gradient of groundwater flow. The derived from the same population. The parameters pH, Cl, SO4, TH, TDS,
flow direction function considered a surface of continuous groundwater F, CODMn, NH4–N, NO2–N, and NO3–N were not significantly different
level as input and two rasters as outputs. The flow direction raster dis­ during the rainy and dry periods (all p > 0.05). For other parameters
played the direction of flow out of each cell. The direction was deter­ (not listed in the table, such as manganese, iron), the measured values
mined by the steepest descent or maximum drop. The hydraulic gradient were lower than the detection limit. The average TH value of the rainy
raster exhibited the ratio of the maximum change in elevation from each season (475 mg/L) was slightly lower than that of the dry season (488
cell along the direction of flow to the length between the centers of cells. mg/L). The TH values of all water samples ranged from 225 to 844 mg/
L, and 46% samples had low water quality (450–650 mg/L). Of all the
3. Results and discussion water samples, 49% had TDS concentrations in the status of low quality
(1000–2000 mg/L), and other 51% were of moderate water quality
3.1. Water quality characteristics (500–1000 mg/L). The fluoride (F) concentration in 92% of the water
samples was lower than the excellent quality threshold (1 mg/L), and
3.1.1. Environmental parameters features 8% of the water samples (from W10, W12, W22, and W28 during the
Table 1 provides a statistical summary of the measured values of the rainy season and W22 during the dry season) had low water quality (1–2

Table 1
Statistical summary of groundwater hydrochemistry.
a
Parameters Rainy period (Sep. n = 37) Dry period (Apr. n = 26) P

Min Max Mean Standard-deviation Min Max Mean Standard-deviation

pH 6.5 7.9 7.1 0.3 6.6 8.2 7.2 0.4 0.83

Cl 66 214 141 35 96 231 148 35 0.96
SO4 87 410 190 61 75 395 204 75 0.40
TH 226 750 475 135 277 844 489 126 0.59
TDS 683 1550 1029 228 706 1420 1008 196 0.99
F 0.1 1.6 0.5 0.4 0.1 1.2 0.4 0.3 0.52
CODMn 0.7 2.5 1.2 0.5 0.6 1.8 1.1 0.2 0.39
NH4–N – 0.23 0.03 0.05 – 0.76 0.08 0.20 1.00
NO2–N – 0.09 0.01 0.02 – 0.04 0.01 0.01 0.77
NO3–N 0.50 128.00 37.34 31.89 0.44 134.00 37.46 30.78 0.91

Note: all the units are in mg/L except for pH; – means lower than the detection limit; a Kolmogorov–Smirnov test.

4
Y. Fang et al. Journal of Environmental Management 273 (2020) 111185

mg/L). The NH4–N concentration of 85% of the water samples was lower total variance; this indicates the impact of organic pollution.
than the detection limit, and a maximum concentration of 0.76 mg/L The primary parameters during the dry season are similar to the plot
was observed in wells W32 and W37 during the dry season. The con­ during the rainy season, and the analysis results of principal factors and
centration of NO3–N varied from 0.44 to 134.00 mg/L, and the mean Pearson’s correlations test are summarized in Tables S5 and S7,
concentrations during the rainy and dry periods were 37.34 and 37.46 respectively.
mg/L, respectively. In the rainy season, 40% of the water samples
exceeded the minimum limit of bad water quality (30 mg/L), and this 3.1.3. Groundwater quality assessment based on WQI
proportion was 46% in the dry season. The maximum content of pH, Cl, Most of the samples that accounted for 75.67% of the 37 sites during
SO4, CODMn, and NO2–N was all lower than the upper limit of moderate the rainy season exhibited a moderate water quality status (Fig. 3a). A
water quality. total of 6 and 3 samples had good and low water quality, respectively.
WQI values varied from 35.01 to 64.74 (Fig. 4), with mean and median
3.1.2. Key environment variables based on PCA and FA values of 51.89 and 51.59, respectively. During the dry season, 80.76%
The results of Kaiser–Meyer–Olkin and Bartlett’s sphericity tests water samples exhibited moderate water quality (Fig. 3b). A total of 2
(KMO＞0.5, p＜0.001) show that the water quality datasets were suit­ and 3 samples had good and low water quality, respectively. WQI values
able for variable reduction by PCA and FA (Wang et al., 2017). The varied from 37.62 to 62.89, with a mean of 47.87 and a median of 47.11
analysis results of the principal factors, eigenvalues, variance contri­ (Fig. 4).
butions, and varimax–rotated loads are summarized in Tables S4–S5. The WQI values during the rainy and dry periods were not signifi­
The rotated loading value can be used to identify important environ­ cantly different (p > 0.05), and the Kolmogorov–Smirnov test indicates
mental parameters as long as the loading value was “strong” (absolute that they are part of the same population. The Gibbs diagram results of
value greater than 0.75) (Matiatos et al., 2018). In the rainy season, Liu et al. (2019), Xiong et al. (2020), and Yin et al. (2018) indicate that
significant environment parameters included TH (0.90), TDS (0.89), F precipitation had little effect on the groundwater chemical composition
(0.76), CODMn (0.86), NH4–N (0.82), NO2–N (0.84), and NO3–N (0.78). in this area. In addition, the annual precipitation in recent years has
Significant parameters were TH (0.86), TDS (0.93), F (0.88), NH4–N been lower than the average annual precipitation. Thus, the precipita­
(0.95), NO2–N (− 0.89), and NO3–N (0.83) in the dry season. tion may slightly affect the groundwater chemical composition. The
The results of PCA can be used to analyze the principal factors cor­ sampling results of Liu et al. (2017, 2019) also show that the seasonal
responding to different processes that control water chemistry and differences in groundwater chemical composition were not obvious in
source of variation in the water quality data. In the rainy season, the first this area. These explains the reason that WQI values were not signifi­
four principal factors accounted for 75.05% of the total variance. The cantly different between the rainy season and the dry season.
first factor explained 30.24% of the total variance. The closely related
parameters were TH and TDS. This factor accounts for elevated 3.2. Description of groundwater hydrodynamics
groundwater salinity in the study area. The Pearson’s correlations test
for the environmental parameters show that TDS was significantly 3.2.1. Prediction of groundwater table
correlated with TH, Cl, SO4, CODMn, and NO3–N (Table S6), which The groundwater level datasets in 115 sites during the rainy and dry
indicate that these chemical components are the main source of TDS. Liu periods are summarized in Table S8. Analysis of variance (one–way
et al. (2017) analyzed the groundwater hydrochemical types in this ANOVA) was performed to investigate the effect of time variation on the
study area of the same year. Their results show that the groundwater groundwater level at each site. The results show that there was signifi­
hydrochemical type in the central and northern regions was dominated cant seasonal variation (p < 0.05) in the groundwater level at most sites
by NO3, and that in the southern region was mainly HCO3, followed by (101, accounting for 88%). Table S9 shows in detail the prediction ac­
HCO3/Cl and HCO3/SO4. The cations in the study area were dominated curacies of the continuous surface based on the optimal Kriging method.
by Ca2+/Mg2+/Na+. Their results support the correlation between A statistical summary of Table S9 is listed in Table 2. The prediction
TDS and these chemical components. The Gibbs diagram has been errors (M, S, and R) of the cross–validation and validation approached
widely applied to recognize the hydrological effects of the evolution of their corresponding ideal values (0, 0, and 1), which indicate that the
groundwater chemical composition. The Gibbs diagram results of Liu continuous surfaces of the groundwater level are unbiased with valid
et al. (2019), Xiong et al. (2020), and Yin et al. (2018) indicate that the uncertainty. The prediction results of continuous groundwater level are
groundwater with high salinity was primarily controlled by rock shown in Fig. S1.
weathering–solubilization and evaporation–concentration in the study
area. In addition to the factors that can be reflected in the Gibbs dia­ 3.2.2. Temporal− spatial distribution of Darcy velocity
gram, cation exchange is also an important influence factor. The nega­ Fig. 5 exhibits the spatial distribution of average Darcy velocity
tive correlation (R2 = 0.76) between (Na+ – Cl-) and (Ca2+ + Mg2+) – during the rainy and dry periods. The distribution characteristics of flow
(HCO3- + SO42-) conducted by Liu et al. (2019) confirmed the exchange velocity and direction were similar in the two periods. Groundwater
process between Ca2+/Mg2+ and Na+. Xiong et al. (2019, 2020) and initially flowed down from the north and the flow direction was trans­
Yin et al. (2018) also found this exchange process in the study area. formed, proceeding southwest in the south–central region, while the
Factor 2 contributed 17.40% to the total variance with high loadings flow direction in the southernmost region was westward (Fig. S1). These
of F and NO3–N. Nitrate contamination in the study area has been re­ features were consistent with the previous report of Zhang et al. (2019).
ported in previous studies. Liu et al. (2017) and Yu et al. (2020) applied The velocity varied from 0.10 to 0.56 m/d, which agrees with the
stable isotope to identify the source of nitrate in groundwater. Their consensus of natural groundwater flow (Onderka et al., 2013). Overall,
results reveal that chemical fertilizers, soil nitrogen, and manure and the flow velocity near the boundary region was less than that of the
sewage were the dominant source of nitrate. The concentration of central part. The flow velocity in the central and northern regions was
fluoride in a small amount of water samples was high, which may be higher than that of the southern region. Note that the maximum hy­
related to the hydrolysis of fluorite (Banerjee, 2015; Yousefi et al., draulic gradient in all datasets was 0.004, much less than 1; this in­
2018). dicates that there is a valid uncertainty in the calculation of Darcy
Factor 3 explained 14.99% of the total variance. The closely related velocity based on the Dupuit assumption.
parameters were NH4–N and NO2–N; these reflect the effects of bio­
logical and microbial processes on nitrogen cycle in the soil (Rivett et al., 3.3. Effect of groundwater hydrodynamics on quality
2008). Yu et al. (2020) also confirmed this biological process in the
study area. With loading of CODMn, factor 4 contributed 12.41% to the The Darcy velocity in the central and northern regions was relatively

5
Y. Fang et al. Journal of Environmental Management 273 (2020) 111185

Fig. 3. Spatial distribution of groundwater quality based on WQI values (a: rainy period; b: dry period). (black and white in print).

hydrodynamics on quality. The land use type in the region is dominated

by agricultural cultivation areas (70.95%) (Yu et al., 2020). The in­
vestigations by Liu et al. (2017) and Yu et al. (2020) indicate that the
central and northern regions were mainly planted with vegetables, while
the southern region was dominated by grain crops. Vegetable cultivation
requires more fertilizer than grain crop (Min et al., 2011). The sowing of
greater chemical fertilizers combined with faster groundwater move­
ment may be responsible for the large–scale nitrate pollution in the
central and northern regions, which further deteriorates the integrated
water quality. Yin et al. (2018) also reveal that the increased sulfate
content in the northern region was likely attributed to the coupling ef­
fect of groundwater movement and sulfide mineral dissolution. There­
fore, the impact of groundwater hydrodynamics on quality may not be
ignored in the central and northern regions.
Specifically, the concentrations of environmental parameters were
considerably close in the northern W1–W3 wells, and the WQI values
indicated moderate water quality regardless of the seasons. In this area,
the subsurface environment is relatively analogous in terms of ground­
water depth and geological structure, whereas the surface environment
Fig. 4. WQI values during the rainy and dry periods. (black and white in print).
is different with respect to soil utilization (the pattern of land use in the
study area have been described by Yu et al. (2020)). According to field
high, compared with the southern region (Fig. 5). The spatial distribu­
surveys, the average fertilizer application amounts in the W1–W3 re­
tion of WQI shows that the water quality in the central and northern
gions were 2212, 3000, and 3862 kg/ha, respectively. However, the
regions was generally worse than that of the southern region. This
NO3–N contents were close with values of 23.21, 26.30, and 31.14
negative correlation illustrates the effect of groundwater
mg/L, during the rainy season; this may come from the hydrodynamic

Table 2
Statistical description of the prediction errors during the rainy and the dry seasons.
Validation Type Prediction error Rainy period (Sep.) Dry period (Apr.)

Min Max Mean Min Max Mean

Cross− validation Mean (M) − 0.004 − 0.001 − 0.002 − 0.015 0.002 − 0.001
Mean standardized (S) − 0.002 − 0.001 − 0.001 − 0.010 0.001 − 0.001
Root mean square standardized (R) 1.090 1.093 1.091 1.056 1.081 1.073
Validation Mean (M) − 1.133 − 0.888 − 0.997 − 0.873 − 0.538 − 0.781
Mean standardized (S) − 0.621 − 0.508 − 0.561 − 0.555 − 0.372 − 0.506
Root mean square standardized (R) 1.611 1.711 1.662 1.498 1.775 1.643

6
Y. Fang et al. Journal of Environmental Management 273 (2020) 111185

Fig. 5. Spatial distribution of average Darcy velocity (a: rainy period; b: dry period). (black and white in print).

mixing at a high Darcy velocity of 0.56 m/d. Other parameters were also TH, TDS, F, CODMn, NH4–N, NO2–N, and NO3–N were identified as
considerably close, such as CODMn with values of 1.9, 1.6, and 1.2 significant water quality parameters. The WQI values varied from 35.01
mg/L; the mixing effect can be verified. A similar situation also occurred to 64.74, with mean values of 51.89 and 47.87 in the rainy and dry
in the W5–W6 region, and the values of both WQI and environmental seasons. Approximately 80% of the samples exhibited moderate water
parameters were comparatively alike. quality, with no significant difference between the rainy and dry sea­
sons. The average concentrations of TH, TDS, and NO3–N exceeded the
3.4. Groundwater quality management for sustainable development limit of drinking water standard for China, especially NO3–N that was
1.87 times the limit. Nitrate pollution and the integrated water quality in
There are two challenges for groundwater quality management in the central and northern regions were generally worse than that in the
the study area, nitrate pollution and elevated groundwater salinity. The southern region. In the central and northern regions, the Darcy velocity
average concentrations of TH, TDS, and NO3–N exceeded the limit of calculated by a new geospatial model was relatively high with a
drinking water standard for China (GB/T 14,848–2017). Especially maximum rate of 0.56 m/d, compared with the southern region. This
NO3–N was 1.87 times larger than the limit. Groundwater in this area correlation illustrates the effect of groundwater hydrodynamics on
has to be treated to remove and reduce the contents of these chemical quality. The sowing of greater chemical fertilizers combined with faster
components for drinking. groundwater movement is likely responsible for the large–scale nitrate
Chemical fertilizers, soil nitrogen, and manure and sewage are the pollution in the central and northern regions. Reducing the use of
dominant source of nitrate in this region (Liu et al., 2017; Yu et al., chemical fertilizers is a decisive measure for sustainable groundwater
2020). The residential areas in this region are mainly scattered villages quality management, especially in the central and northern regions.
and towns with low urbanization and unqualified sewage treatment The study illustrated the effective performance of the new geospatial
facilities. Therefore, it is necessary to upgrade and construct sewage model for the spatial analysis of the effect of groundwater hydrody­
treatment facilities in the study area to prevent groundwater pollution. namics on water quality. The prediction errors (M, S, and R) were
Especially in the southern region, villages and towns in this area are significantly close to the corresponding ideal values (0, 0, and 1),
relatively dense and concentrated. The nitrate concentration in the proving the accuracy of the new model with a valid uncertainty. For
central and northern regions was greater than that in the southern re­ some cases being lack of the hydrogeological information, this approach
gion. This phenomenon is likely attributed to greater fertilizer for can alternatively be employed to model the dynamic water quality and
vegetable cultivation and faster groundwater movement in the central to enhance groundwater management efficiency with a relatively
and northern regions. Yu et al. (2020) also found that chemical fertilizer accessible requirement for the basic data and technology.
was the main source of nitrate with a contribution of 54.32% in the
central and northern regions. So reducing the use of chemical fertilizers CRediT authorship contribution statement
is an important measure, especially in the central and northern regions.
The reduction in nitrate content helps to relieve the elevated ground­ Yunhai Fang: Investigation, Methodology, Validation, Conceptual­
water salinity, because nitrate is one of the main source of TH and TDS. ization, Writing - original draft, Writing - review & editing. Tianyuan
In addition, the uncontaminated freshwater can be recharged to reduce Zheng: Resources, Writing - review & editing, Visualization, Supervi­
the groundwater salinity. sion. Xilai Zheng: Supervision, Funding acquisition, Resources. Hui
Peng: Investigation, Validation. Huan Wang: Validation, Formal anal­
4. Conclusions ysis. Jia Xin: Supervision. Bo Zhang: Formal analysis.

Based on GIS, WQI, and multivariate statistical analyses, the

groundwater quality in the Dagu River Basin was empirically analyzed.

7
Y. Fang et al. Journal of Environmental Management 273 (2020) 111185

Declaration of competing interest Matiatos, I., Paraskevopoulou, V., Lazogiannis, K., Botsou, F., Dassenakis, M.,
Ghionis, G., Alexopoulos, J.D., Poulos, S.E., 2018. Surface–ground water interactions
and hydrogeochemical evolution in a fluvio-deltaic setting: the case study of the
The authors declare that they have no known competing financial Pinios River delta. J. Hydrol. 561, 236–249. https://doi.org/10.1016/j.
interests or personal relationships that could have appeared to influence jhydrol.2018.03.067.
the work reported in this paper. Min, J., Zhao, X., Shi, W.M., Xing, G.X., Zhu, Z.L., 2011. Nitrogen balance and boss in a
greenhouse vegetable system in southeastern China. Pedosphere 21 (4), 464–472.
https://doi.org/10.1016/s1002-0160(11)60148-3.
Acknowledgments Mohammadi, A.A., Yousefi, M., Soltani, J., Ahangar, A.G., Javan, S., 2018. Using the
combined model of gamma test and neuro-fuzzy system for modeling and estimating
lead bonds in reservoir sediments. Environ. Sci. Pollut. Res. 25, 30315–30324.
We thank the Dagu River Authority for providing the groundwater https://doi.org/10.1007/s11356-018-3026-7.
quality and level data. We are also grateful for the anonymous reviewers Onderka, M., Banzhaf, S., Scheytt, T., Krein, A., 2013. Seepage velocities derived from
thermal records using wavelet analysis. J. Hydrol. 479, 64–74. https://doi.org/
for their useful comments to improve the content. The National Natural 10.1016/j.jhydrol.2012.11.022.
Science Foundation of China (grant number 41731280) supported this Qian, Y., Migliaccio, K.W., Wan, Y., Li, Y., 2007. Surface water quality evaluation using
work and the Joint Funds of National Science Foundation of China–­ multivariate methods and a new water quality index in the Indian River Lagoon,
Florida. Water Resour. Res. 43 (8), 199–212. https://doi.org/10.1029/
Shandong Province (grant number U1806210). 2006WR005716.
Rivett, M.O., Buss, S.R., Morgan, P., Smith, J.W.N., Bemment, C.D., 2008. Nitrate
attenuation in groundwater: a review of biogeochemical controlling processes.
Appendix A. Supplementary data
Water Res. 42 (16), 4215–4232. https://doi.org/10.1016/j.watres.2008.07.020.
Saleh, H.N., Valipoor, S., Zarei, A., Yousefi, M., Asghari, F.B., Mohammadi, A.A.,
Supplementary data to this article can be found online at https://doi. Amiri, F., Ghalehaskar, S., Khaneghah, A.M., 2019. Assessment of groundwater
quality around municipal solid waste landfill by using Water Quality Index for
org/10.1016/j.jenvman.2020.111185.
groundwater resources and multivariate statistical technique: a case study of the
landfill site, Qaem Shahr City, Iran. Environ. Geochem. Health. https://doi.org/
References 10.1007/s10653-019-00417-0.
Sotomayor, G., Hampel, H., Vázquez, R.F., 2018. Water quality assessment with
emphasis in parameter optimisation using pattern recognition methods and genetic
Abbasnia, A., Alimohammadi, M., Mahvi, A.H., Nabizadeh, R., Yousefi, M., algorithm. Water Res. 130, 353–362. https://doi.org/10.1016/j.
Mohammadi, A.A., Pasalari, H., Mirzabeigi, M., 2018. Assessment of groundwater watres.2017.12.010.
quality and evaluation of scaling and corrosiveness potential of drinking water Wang, J., Liu, G., Liu, H., Lam, P.K.S., 2017. Multivariate statistical evaluation of
samples in villages of Chabahr city, Sistan and Baluchistan province in Iran. Data In dissolved trace elements and a water quality assessment in the middle reaches of
Brief 16, 182–192. https://doi.org/10.1016/j.dib.2017.11.003. Huaihe River, Anhui, China. Sci. Total Environ. 583, 421–431. https://doi.org/
Abbasnia, A., Yousefi, N., Mahvi, A.H., Nabizadeh, R., Radfard, M., Yousefi, M., 10.1016/j.scitotenv.2017.01.088.
Alimohammadi, M., 2019. Evaluation of groundwater quality using water quality Wu, Z., Wang, X., Chen, Y., Cai, Y., Deng, J., 2018. Assessing river water quality using
index and its suitability for assessing water for drinking and irrigation purposes: case water quality index in Lake Taihu Basin, China. Sci. Total Environ. 612, 914–922.
study of Sistan and Baluchistan province (Iran). Hum. Ecol. Risk Assess. 25 (4), https://doi.org/10.1016/j.scitotenv.2017.08.293.
988–1005. https://doi.org/10.1080/10807039.2018.1458596. Wu, Z., Zhang, D., Cai, Y., Wang, X., Zhang, L., Chen, Y., 2017. Water quality assessment
Banerjee, A., 2015. Groundwater fluoride contamination: a reappraisal. Geosci. Front. 6 based on the water quality index method in Lake Poyang: the largest freshwater lake
(2), 277–284. https://doi.org/10.1016/j.gsf.2014.03.003. in China. Sci. Rep. 7 (1), 17999. https://doi.org/10.1038/s41598-017-18285-y.
Bear, J., 1972. Dynamics of Fluids in Porous Media. American Elsevier Pub. Co. https:// Xiao, K., Li, H., Shananan, M., Zhang, X., Wang, X., Zhang, Y., Zhang, X., Liu, H., 2019.
doi.org/10.1016/0013-7952(73)90047-1. Coastal water quality assessment and groundwater transport in a subtropical
Epa, PR China, 2004. Technical Specification for Environmental Monitoring of mangrove swamp in Daya Bay, China. Sci. Total Environ. 646, 1419–1432. https://
Groundwater. HJ/T 164-2004). doi.org/10.1016/j.scitotenv.2018.07.394.
Fang, Y., Zheng, X., Peng, H., Wang, H., Xin, J., 2019. A new method of the relative Xiong, G., An, Q., Fu, T., Chen, G., Xu, X., 2020. Evolution analysis and environmental
membership degree calculation in variable fuzzy sets for water quality assessment. management of intruded aquifers of the Dagu River Basin of China. Sci. Total
Ecol. Indicat. 98, 515–522. https://doi.org/10.1016/j.ecolind.2018.11.032. Environ. 719, 137260. https://doi.org/10.1016/j.scitotenv.2020.137260.
Ferreira, D.M., Fernandes, C.V.S., Kaviski, E., Fontane, D., 2019. Water quality modelling Xiong, G., Fu, T., Han, J., Chen, G., Xu, X., Xu, X., Liu, G., Liu, W., 2019.
under unsteady state analysis: strategies for planning and management. J. Environ. Hydrogeochemical and isotopic characteristics of groundwater in Dagu River basin.
Manag. 239, 150–158. https://doi.org/10.1016/j.jenvman.2019.03.047. Adv. Mar. Sci. 37 (4), 626–637. https://doi.org/10.3969/j.issn.1671-
Fulazzaky, M.A., Teng, W.S., Masirin, M.I.M., 2010. Assessment of water quality status 6647.2019.04.009 (In Chinese).
for the Selangor River in Malaysia. Water Air Soil Pollut. 205 (1–4), 63–77. https:// Yin, Z., Lin, Q., Xu, S., 2018. Spatial—temporal variations and controlling factors of
doi.org/10.1007/s11270-009-0056-2. groundwater hydrochemical characteristics in the Dagu River basin. Geol. Rev. 64
Gaqsiq, PR China, 2017. Standard for Groundwater Quality (GB/T 14848-2017). (4), 1030–1043. https://doi.org/10.16509/j.georeview.2018.04.017 (In Chinese).
Kavetski, D., Kuczera, G., Franks, S.W., 2006. Bayesian analysis of input uncertainty in Yousefi, M., Asghari, F.B., Zuccarello, P., Conti, G.O., Ejlali, A., Mohammadi, A.A.,
hydrological modeling: 1. Theory. Water Resour. Res. 42 (3), W03407. https://doi. Ferrante, M., 2019a. Spatial distribution variation and probabilistic risk assessment
org/10.1029/2005WR004376. of exposure to fluoride in ground water supplies: a case study in an endemic fluorosis
Koh, E.H., Lee, E., Lee, K.K., 2016. Impact of leaky wells on nitrate cross-contamination region of northwest Iran. Int. J. Environ. Res. Publ. Health 16, 564. https://doi.org/
in a layered aquifer system: Methodology for and demonstration of quantitative 10.3390/ijerph16040564.
assessment and prediction. J. Hydrol. 541, 1133–1144. https://doi.org/10.1016/j. Yousefi, M., Ghalehaskar, S., Asghari, F.B., Ghaderpoury, A., Dehghani, M.,
jhydrol.2016.08.019. Ghaderpoori, M., Mohammadi, A.A., 2019b. Distribution of fluoride contamination
Lal, A., Datta, B., 2019. Multi-objective groundwater management strategy under in drinking water resources and health risk assessment using geographic information
uncertainties for sustainable control of saltwater intrusion: solution for an island system, northwest Iran. Regul. Toxicol. Pharmacol. 107, 104408. https://doi.org/
country in the South Pacific. J. Environ. Manag. 234, 115–130. https://doi.org/ 10.1016/j.yrtph.2019.104408.
10.1016/j.jenvman.2018.12.054. Yousefi, M., Ghoochani, M., Mahvi, A.H., 2018. Health risk assessment to fluoride in
Liu, G.Q., Zhou, S.Y., Huang, X.D., Liu, T., Xu, D., Yue, C.D., 2017. Multiple methods to drinking water of rural residents living in the Poldasht city, Northwest of Iran.
recognize sources of underground water nitrate contamination in plain area of Dagu Ecotoxicol. Environ. Saf. 148, 426–430. https://doi.org/10.1016/j.
River, Qingdao, China. Acta Scientiac Circumstantiac 37 (1), 347–356. https://doi. ecoenv.2017.10.057.
org/10.13671/j.hjkxxb.2016.0163 (In Chinese). Yousefi, M., Saleh, H.N., Mohammadi, A.A., Mahvi, A.H., Ghadrpoori, M., Suleimani, H.,
Liu, G.Q., Zhu, L.W., Sun, Y.X., 2019. Hydrochemical characteristics and origins of salt 2017. Data on water quality index for the groundwater in rural area Neyshabur
groundwater in the lower reaches of Dagu River. Periodical of Ocean University of County, Razavi province, Iran. Data In Brief 15, 901–907. https://doi.org/10.1016/j.
China 49 (5), 84–92. https://doi.org/10.16441/j.cnki.hdxb.20180089 (In Chinese). dib.2017.10.052.
Lumb, A., Sharma, T., Bibeault, J., 2011. A review of genesis and evolution of water Yu, L., Zheng, T., Zheng, X., Hao, Y., Yuan, R., 2020. Nitrate source apportionment in
quality index (WQI) and some future directions. Water Qual. Expos. Health 3 (1), groundwater using Bayesian isotope mixing model based on nitrogen isotope
11–24. https://doi.org/10.1007/s12403-011-0040-0. fractionation. Sci. Total Environ. 137242 https://doi.org/10.1016/j.
Machiwal, D., Islam, A., Kamble, T., 2019. Trends and probabilistic stability index for scitotenv.2020.137242.
evaluating groundwater quality: the case of quaternary alluvial and quartzite aquifer Zeng, X., Dong, J., Wang, D., Wu, J., Zhu, X., Xu, S., Zheng, X., Xin, J., 2017. Identifying
system of India. J. Environ. Manag. 237, 457–475. https://doi.org/10.1016/j. key factors of the seawater intrusion model of Dagu river basin, Jiaozhou Bay.
jenvman.2019.02.071. Environ. Res. 165, 425–430. https://doi.org/10.1016/j.envres.2017.10.039.
Majumder, P., Eldho, T.I., 2019. Reactive contaminant transport simulation using the Zhang, X., Lin, Q., Huang, X., Xu, S., 2019. Numerical simulation coupling soil water/
analytic element method, random walk particle tracking and kernel density groundwater and estimation of groundwater recharge in Dagu river basin. Acta
estimator. J. Contam. Hydrol. 222, 76–88. https://doi.org/10.1016/j. Pedol. Sin. 1, 1–13. https://doi.org/10.11766/trxb201802120016 (in Chinese).
jconhyd.2019.01.006.