You are on page 1of 9

Journal of Environmental Management 314 (2022) 115101

Contents lists available at ScienceDirect

Journal of Environmental Management


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

Research article

APCS-MLR model: A convenient and fast method for quantitative


identification of nitrate pollution sources in groundwater
Lu Yu a, c, Tianyuan Zheng a, b, *, Ruyu Yuan a, b, Xilai Zheng a, b
a
College of Environmental Science and Engineering, Ocean University of China, Qingdao, 266100, China
b
Key Lab of Marine Environmental Science and Ecology, Ministry of Education, Ocean University of China, Qingdao, 266100, China
c
Ecological Environment Research and Development Center, Weihai Innovation Institute, Qingdao University, Weihai, 264200, China

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

Keywords: Nitrate (NO−3 ) contamination in groundwater has diverse sources and complicated transformation processes. To
Nitrate source effectively control NO−3 pollution in groundwater systems, quantitative and accurate identification of NO−3
Hydrochemical indicators sources is critical. In this work, we applied hydrochemical characteristics and isotope analysis to determine NO−3
Multivariate statistical model
source apportionment. For the first time, the NO−3 source contributions were calculated using hydrochemical
Bayesian isotope mixing model
indicators combined with multivariate statistical model (PCA-APCS-MLR). The results interpret that chemical
Method comparison
fertilizers (58.11%) and natural sources (22.69%) were the primary NO−3 sources in the vegetable cultivation
area (VCA) which were rather close to the estimation by Bayesian isotope mixing model (SIAR). In particular, the
contributions of chemical fertilizers in the VCA differed by only 3.79% between the two methods. Compared with
previous approaches e.g. SIAR, the key advantage of the proposed PCA-APCS-MLR model is that it only requires
the hydrochemical indicators which can be easily measured. A series of complicated experiments including
measurement of isotope data of NO−3 in groundwater, monitoring of in-situ pollution source information and
calculation of isotopic enrichment factor can be simply avoided. The PCA-APCS-MLR model offers a much more
convenient and faster method to determine the contribution rates of NO−3 pollution sources in groundwater.

1. Introduction intensive agriculture being a major source due to the use of chemical
fertilizers and animal manure, along with the discharge of sewage (Xue
Groundwater nitrate (NO−3 ) pollution has become a significant et al., 2012; Gutierrez et al., 2018; Han et al., 2018; Lentz and Lehrsch,
environmental problem worldwide, particularly in agricultural areas 2019). Accurate identification of NO−3 pollution sources is of great sig­
(Rivett et al., 2008; Choi et al., 2017; Jia et al., 2018; Xin et al., 2019; Yu nificance for effective groundwater management.
et al., 2019). Excessively high NO−3 -N concentrations in groundwater The methods commonly used to determine the NO−3 sources mainly
can cause serious health problems (when the groundwater is used as include water quality analysis and stable isotope analysis (Cui et al.,
drinking water) and negative environmental effect (such as eutrophi­ 2018; Torres-Martinez et al., 2021). Principal component analysis (PCA)
cation produced by the transport of NO−3 to the surface water) (Bastani is an important method for water quality analysis, having been widely
and Harter, 2019; Xin et al., 2021). As a result, the World Health Or­ used for pollution source determination in aquatic systems (Wang et al.,
ganization (WHO) has stipulated that the maximum NO−3 -N concentra­ 2017; Taufiq et al., 2019; Zhu et al., 2019). PCA extracts principal
tion threshold in groundwater for drinking water purposes is 11.3 mg/L. components from water quality index parameters by reducing the
In China, the groundwater quality standard (GB/T14848-2017) stipu­ number of input variables, allowing potential pollution sources to be
lates that the concentration of NO−3 -N for drinking water shall not exceed identified qualitatively according to the major pollution factors extrac­
20 mg/L. However, in many agricultural regions and areas with inten­ ted (Kim et al., 2015a; Matiatos, 2016; Meghdadi and Javar, 2018;
sive human activity, the NO−3 concentrations in groundwater tend to Torres-Martinez et al., 2021). However, PCA cannot estimate the pro­
exceed drinking water standards (Burow et al., 2010; Kringel et al., portional contributions of pollution sources. Therefore, some studies
2016; Motevalli et al., 2019; Yuan et al., 2020). Groundwater NO−3 have utilized absolute principal component score (APCS) analysis
sources are generally attributed to anthropogenic activities, with combined with multivariate linear regression (MLR), to further quantify

* Corresponding author. College of Environmental Science and Engineering, Ocean University of China, Qingdao, 266100, China.
E-mail address: zhengtianyuan@ouc.edu.cn (T. Zheng).

https://doi.org/10.1016/j.jenvman.2022.115101
Received 12 February 2022; Received in revised form 8 April 2022; Accepted 16 April 2022
Available online 23 April 2022
0301-4797/© 2022 Elsevier Ltd. All rights reserved.
L. Yu et al. Journal of Environmental Management 314 (2022) 115101

the impact of different pollution sources (Zhou et al., 2007; Su et al., NO−3 in groundwater and in-situ pollutant sources, as well as to inves­
2011; Gholizadeh et al., 2016; Jin et al., 2019; Rashid et al., 2020). tigate the ϵ value of nitrogen transformation process. Therefore, it is
Currently, the APCS-MLR method has mainly been applied to air particularly critical to find a convenient and reliable quantitative anal­
pollution, heavy metals and organic pollution in water, but it has not yet ysis method of NO−3 pollution sources. In contrast, measurement of
been used for NO−3 source apportionment in groundwater (Thurston and conventional hydrochemical indicators is much more simple.
Spengler, 1985; Meng et al., 2018; Chen et al., 2019a,b; Yan et al., To address the above issues, in this work, we proposed the APCS-
2020). Therefore, the applicability and accuracy of the APCS-MLR MLR model which can accurately capture the NO−3 source apportion­
method for NO−3 source apportionment still requires further assess­ ment, while avoid all the complicated experiments. The objectives were
ment and verification. This will further expand and improve the appli­ as follows: (1) to provide a convenient and fast method for the quanti­
cation of water quality analysis in quantitative source apportionment of tative identification of NO−3 pollution sources in groundwater (PCA-
NO−3 . APCS-MLR model); (2) to verify the feasibility and applicability of the
As NO−3 pollution sources exhibit distinctive isotope signatures, ni­ PCA-APCS-MLR model in quantitative identification of NO−3 pollution
trogen and oxygen isotope analysis and isotope mixing models have sources by comparing with SIAR model. Through quantitative identifi­
been applied to identify NO−3 sources (Moore and Semmens, 2008; cation of NO−3 sources from multiple perspectives, this study will provide
Fenech et al., 2012; Denk et al., 2017; Guo et al., 2020). In particular, the a useful tool for effective prevention and control of NO−3 pollution in
Bayesian isotope mixing model (SIAR) is increasingly being used for groundwater.
NO−3 source apportionment, due to its consideration of the isotopic
fractionation effect (Xia et al., 2017; Liu et al., 2018; Li et al., 2019; Sui 2. Materials and methods
et al., 2020; Zhang et al., 2020). However, the results of NO−3 source
apportionment are often uncertain due to the overlap in source isotope 2.1. Study area
compositions and the uncertainty of isotopic fractionation factors
(Nakagawa et al., 2017; Nikolenko et al., 2018; Biddau et al., 2019; Kruk The Dagu River groundwater source area (120◦ 04′ 48′′ – 120◦ 21′ 00′′
et al., 2020). Previous researchers have reviewed that the accuracy of E, 36◦ 18′ 00′′ – 36◦ 45′ 45′′ N) is located in the middle and lower reaches
NO−3 source apportionment can be improved by using the isotopic values of the Dagu River (Qingdao, China), with an area of 460 km2 (Fig. 1a).
of in-situ pollution sources and seeking suitable isotopic enrichment The groundwater in the study area is the main water source for agri­
factor (ϵ) (Zhang et al., 2018; Yu et al., 2020). However, a series of cultural and industrial activities in the surrounding area. The ground­
complicated experiments are required to obtain sufficient isotope data of water of Dagu River is mainly distributed in the sand and gravel layer

Fig. 1. Location of the Dagu River groundwater reservoir and distribution of groundwater sampling sites with the land-use (a); typical geological profile of the
middle (a-a’) (b).

2
L. Yu et al. Journal of Environmental Management 314 (2022) 115101

from the Quaternary alluvial diluvium (Ke et al., 2021). The typical groundwater quality standard for drinking water (GB/T14848-2017). If
vertical geological profile of Dagu River groundwater source area is the Pi > 1, then the water quality index i value exceeds the standard limit
shown in Fig. 1b. The overlying layer of the aquifer is mainly clay sandy and the water is considered polluted, with larger Pi values indicating a
soil and sandy clay with a thickness of 2–5 m. The phreatic aquifer is more serious degree of pollution (Wang et al., 2020).
mainly composed of coarse sand, medium coarse sand, and a little
gravel, with an average thickness of 5 m. The bottom layer is mainly 2.4. Multivariate statistical model
composed of Cretaceous clay rock and sandstone with poor perme­
ability, forms a good waterproof layer and is conducive to the storage of 2.4.1. Principal component analysis (PCA)
groundwater (Fang et al., 2020). The groundwater of Dagu River is Principal component analysis (PCA) can extract principal compo­
recharged by precipitation, river infiltration, and irrigation, and nents (PCs) from multiple index parameters by reducing the dimension
discharge occurs through artificial exploitation, runoff, and evaporation of multivariate datasets (Singh et al., 2005). The main steps in PCA
(Nai et al., 2020). identification of pollution sources include: (1) assessing the feasibility of
The study area is an alluvial plain with a flat terrain. The region is a original indicators for factor analysis; (2) reducing the dimension of
typical agricultural planting area in Shandong Province (composed of variables and determining the number of PCs; (3) obtaining a rotation
71% agricultural land). Based on field investigations, the study area can factor loading matrix to analyze the rationality of the extracted results;
be classified as vegetable cultivation area (VCA) and grain cultivation (4) identifying the pollution sources based on the PCs (Chen et al.,
area (GCA) (Fig. 1a) (Yu et al., 2020). Intensive agricultural activities, 2019a,b; Yuan et al., 2020).
especially in the VCA, have led to the use of large amounts of chemical The number of PCs is determined by analysis of their eigenvalue (>1)
fertilizers. According to the Qingdao Statistical Yearbook in 2019, the (Liu et al., 2019; Yan et al., 2020). Larger variance contribution rates
average amount of chemical fertilizer applied in Qingdao was 517 indicate that PCs have a stronger ability to explain the original data.
kg/hm2, which is about 4-fold higher than the global average. However, Factor loadings reflect the relative importance of each variable in PCs,
some studies have reported that the agricultural utilization of nitrogen with factor loading values of >0.75, 0.75–0.5 and 0.5–0.3 defined as
fertilizers in China remains below 50% overall (Cai et al., 2018; Hu strong, medium and weak, respectively (Xiao et al., 2016; Liu et al.,
et al., 2019). The N not absorbed by plants gradually accumulates in the 2019).
soil and is transformed into NO−3 by microorganisms, with NO−3 then
leaching into groundwater through unsaturated zones, resulting in the 2.4.2. Absolute principal component score (APCS) - multivariate linear
accumulation of NO−3 in groundwater. regression (MLR)
The absolute principal component score (APCS) was determined on
2.2. Sample collection and analysis the basis of PCA results, with multivariate linear regression (MLR)
performed using measured pollutant concentrations as the dependent
A total of 34 groundwater samples were collected from wells in variable and APCS values as the independent variable (Su et al., 2011;
March 2019, including 24 VCA and 10 GCA groundwater samples Gholizadeh et al., 2016). The contribution of different pollution sources
(Fig. 1a). 200 mL groundwater samples were collected from each sam­ was determined according to regression coefficient values. The
pling site using a water sampler, with all samples filtered through 0.45 APCS-MLR model can be expressed using Equations (2)–(5) as follow
μm filters and stored at 4 ◦ C for hydrochemical analyses. About 50 mL (Huang et al., 2018; Yan et al., 2020):
groundwater were filtered through 0.22 μm filters and stored in poly­ (0 − xi ) − xi
ethylene bottles at − 20 ◦ C for isotope analyses. More details about the (Z0 )i = = (2)
σi σi
groundwater and pollution sources sampling and isotope testing were
provided by Yu et al. (2020). ∑
I

Groundwater temperature, pH, electrical conductivity (EC), and (A0 )p = Spi (Z0 )i (3)
dissolved oxygen (DO) content were measured on site using a portable
i=1

multi-parameter water quality meter (HQ40d, Hach, US). The nitrogen APCSjp = (Az )jp − (A0 )p (4)
concentrations in groundwater samples were determined by spectro­
photometry (Unico 2800 UV/VIS). The detection method included ul­ ∑
P
traviolet spectrophotometry for NO−3 -N detection, the Nessler’s reagent Ci = bi + bpi × APCSjp (5)
colorimetric method for ammonium (NH+ 4 -N) detection and ethyl­
p=1

enediamine colorimetry for nitrite (NO−2 -N) detection. The metal cations
potassium (K+), sodium (Na+), calcium (Ca2+), magnesium (Mg2+), and where, (Z0)i indicates the normalized concentration of indicator i in an
anions chloride (Cl− ) and sulfate (SO2− anthropogenic sample, where its concentration is 0; xi indicates the
4 ) were measured using ion
chromatography (ICS-3000, Dionex, US). Bicarbonate (HCO−3 ) concen­ average concentration of indicator i in all samples; σi indicates the
trations were determined by acid-base titration. The total hardness of standard deviation of all samples for indicator i; (A0)p indicates the
water was calculated based on the Ca2+ and Mg2+ concentrations. The principal component score in a “zero” polluted sample; Spi indicates the
total dissolved solids (TDS) concentration was calculated as the com­ scoring coefficient of indicator i in the principal component p; APCSjp is
bined content of ions minus half of the concentration of HCO−3 . the APCS of sample j in principal component p; (Az)jp indicates the
principal component score of sample j in principal component p; Ci in­
dicates the concentration of indicator i; bi indicates a constant multiple
2.3. Single pollution index method regression term; bpi indicates the regression coefficient of principal
component p for indicator i.
The single pollution index method was used to evaluate water In the APCS-MLR model, bpi × APCSjp is used to assess the contri­
quality, which is expressed according to Equation (1) as follows (Men bution of principal component p to indicator i in sample j. In this study,
et al., 2018): the average value of bpi × APCSjp was established for all samples as the
PI = Ci /Ci0 (1) contribution of each source to NO−3 pollution (Ma et al., 2018; Yan et al.,
2020).
where, PI is the pollution index; Ci is the measured value of the water
quality indicator i; Ci0 is the evaluation standard value for water quality
indicator i, based on the standard defined by level III of the Chinese

3
L. Yu et al. Journal of Environmental Management 314 (2022) 115101

2.5. Bayesian isotope mixing model (SIAR) Cl− , NH+


4 -N, pH, and Na exceeded the standard limits in 14.7%, 11.8%,
+

5.9%, and 2.9% of samples, respectively. Results indicate that most


In this study, the Bayesian isotope mixing model (SIAR) was also groundwater in the study area was seriously polluted by NO−3 and that
used to determine the NO−3 sources apportionment, expressed as follows the level of groundwater hardness was high.
(Parnell et al., 2010): From the perspective of cropping system, the groundwater quality in
∑K ( ) the study area exhibited distinct spatial differences, especially in terms
Xij = k=1
Pk Sjk + εjk + νij of NO−3 pollution, which was more serious in the VCA than in the GCA
( ) (Fig. 2b). In the VCA, the average NO−3 PI value was 3.8 and the
Sjk ∼ N μjk , ω2jk (6) maximum PI value was 8.63. In the GCA, the average NO−3 PI value was
1.3 and the maximum PI value was 3.2. The average PI values for TDS,
( ) SO2−
4 , Cl , Na , and NH4 -N did not exceed 1, suggesting that these in­
− + +
εjk ∼ N λjk , τ2jk dicators did not cause significant pollution within the study area, while a
( ) low level of local pollution was present. Results indicate that NO−3 was
vjk ∼ N 0, σ 2j the dominant contaminant in the groundwater sources of the Dagu
River.
where Xij is the isotopic value j of sample i; Sjk is the source value k of
isotope j; Pk is the proportion of source k, which needs to be estimated by 3.2. Hydrochemical characteristics
the SIAR model; ϵjk is the enrichment factor of isotope j on source k; and
νij is the residual error. In this study, the SIAR model was performed with The hydrochemical characteristic of groundwater in the study area
the SIAR package in the R Programming Language (R version ×64 are shown in the Piper’s diagram (Fig. 3). The cation in groundwater
3.4.2). samples was mainly Ca2+. The percentages of Ca2+ milligram equivalent
in most sampling sites were more than 50%, with no significant spatial
2.6. Statistical analysis divergence observed between different cropping systems. Combined
cation and anion analysis showed that two hydrochemical types were
Correlation analysis, PCA and multiple linear regression were carried identified, including Ca–HCO3 (Cl + NO3) and Ca–SO4 (Cl + NO3). The
out using IBM SPSS version 25.0. The correlation between different in­ Ca– HCO3 (Cl + NO3) type was mainly distributed in the GCA, while the
dicators were analyzed by Pearson’s correlation coefficients, with sta­ Ca–SO4 (Cl + NO3) type was mostly distributed in the VCA, indicating
tistical significance at the p < 0.05 level. Kaiser-Meyer-Olkin (KMO) and that the cropping system significantly influenced the groundwater
Bartlett spherical tests were used to further analyze the correlation be­ chemical composition.
tween indicators (Zhang et al., 2010; Meghdadi and Javar, 2018). KMO Studies have reported that Ca–HCO3 type ions are mainly derived
values closer to 1 represent stronger correlations between indicators and from the dissolution of carbonate rocks, while Ca− (Cl + NO3) and
therefore higher accuracy factor analysis outcomes. Only KMO value > Ca–SO4 (Cl + NO3) ions mainly originate from anthropogenic activities,
0.5 and Bartlett spherical tests at the p < 0.05 level is suitable for factor such as chemical fertilizer utilization and wastewater discharge (Kim
analysis and principal component analysis (Gholizadeh et al., 2016). et al., 2015b; Wang et al., 2017). Therefore, the hydrochemical
composition of groundwater in the GCA was affected by both natural
and anthropogenic activities, while the in the VCA it was mainly influ­
3. Results and discussion
enced by anthropogenic activities, especially the extensive use of fer­
tilizers containing potassium sulfate, resulting in the increase in SO2−
4
3.1. Groundwater pollution assessment
concentrations.
The hydrochemical parameter statistics for all groundwater samples
are summarized in Table S1. The water quality in the study area was 3.3. Source apportionment using the PCA-APCS-MLR model
evaluated according to level III of the Chinese groundwater quality
standard (GB/T14848-2017) for drinking water (Table S2). As shown in 3.3.1. Identification of potential pollution sources using PCA
Fig. 2a, the most serious pollutants were NO−3 -N and total hardness, with Based on the results of correlation analysis, the indicators DO, K+,
88.2% (30 samples) of groundwater samples exceeding the groundwater Na+, Ca2+, Mg2+, Cl− , SO2− 4 , and NO3 were used for PCA. In the VCA,

quality standard limits. Concentrations of SO2− 4 and TDS exceeded the the KMO (0.71) and Bartlett’s test (p < 0.001) results confirmed the
standard limits in 32.4% (11 samples) of groundwater samples, while suitability of the dataset for PCA (Gholizadeh et al., 2016; Meghdadi and

Fig. 2. Rate of standard limit exceedance in groundwater samples (a) and average pollution index values (b) for the assessed pollutants.

4
L. Yu et al. Journal of Environmental Management 314 (2022) 115101

Fig. 3. Piper’s diagram showing the hydrochemical types of groundwater samples.

Javar, 2018). However, in the GCA, the KMO value of 0.43 indicated factor loading, allowing pollution sources to be inferred. For the VCA
that data was not suitable for PCA analysis due to the small number of (Fig. 4), in PC1, Na+ and Mg2+ exhibited strong loads, while Cl− , K+ and
sampling sites. Therefore, this study only carried out source appor­ SO2−
4 exhibited medium loads. The correlation coefficients of Na with
+

tionment for the VCA. Mg2+ and Cl− were all >0.75 (p < 0.01), indicating that these ions
The extraction results of PCs in the VCA were listed in Table S3. In originate from the same pollution source. For the conventional
the VCA, two PCs were extracted with eigenvalues of >1 and a cumu­ groundwater ions, they were mainly derived from the dissolution of
lative variance percentage of 76.98%, indicating that these PCs could natural minerals (Wang et al., 2017). K+ is easily adsorbed by clay and
adequately explain most of the information in the original dataset (Liu utilized by plants, resulting in low groundwater concentrations and
et al., 2019; Yan et al., 2020). After the factor loading matrix was rotated therefore, K+ was also considered to originate from natural minerals.
using the maximum variance method, PC1 and PC2 explained 39.70% SO2−
4 is usually attributed to the dissolution of sulfate sediments and the
and 37.28% of the total variance, respectively. oxidation of sulfur minerals under natural conditions (Matiatos, 2016).
The dominance of each indicator on the PCs was obtained by rotating However, SO2− 4 concentrations in some sampling sites exceeded the

Fig. 4. Rotation factor loading for the vegetable cultivation area.

5
L. Yu et al. Journal of Environmental Management 314 (2022) 115101

groundwater standard limit, which was likely to be caused by the use of Table 1
K2SO4 compound fertilizer. In summary, PC1 was mainly formed from δ15N and δ18O values and isotopic enrichment factor (ϵ) of various pollution
natural sources, with an additional contribution from chemical fertil­ sources.
izers in areas with excessive SO2−4 levels. Source δ15N (‰) δ18O (‰)
In PC2, Ca2+, NO−3 -N and DO exhibited strong loadings (Fig. 4). The Mean SD ϵ Mean SD ϵ
concentration of NO−3 -N in the natural state usually does not exceed 3
AD − 4.56 1.19 0 64.75 5.99 0
mg/L, which is considered the threshold for contribution from anthro­
NS 4.48 0.7 − 8.7 ± 2.7 2.85 0.16 0
pogenic activity (Ogrinc et al., 2019). However, the evaluation results CF − 0.29 1.96 − 8.7 ± 2.7 2.85 0.16 0
(section 3.1) indicate that the degree of NO−3 pollution in the study area M&S 2.52 1.92 14.7 ± 2.8 2.85 0.16 0
was serious and therefore, it can be concluded that NO−3 mainly derived
from anthropogenic activities. Generally, Ca2+ in groundwater origi­
nates mainly from the weathering and dissolution of sediments. How­
ever, a significant positive correlation was observed between Ca2+ and
NO−3 -N with a regression coefficient of 0.79 (p < 0.01), indicating that
Ca2+ may also be derived from anthropogenic activities (Rashid et al.,
2018; Jehan et al., 2019). According to the field investigations, vege­
table planting usually requires a large amount of nitrogen fertilizers,
such as urea and ammonium nitrogen fertilizer. After NH+ 4 enters the
2+ 2+
soil, ion exchange occurs between NH+ 4 and Ca , with Ca extracted
from the soil leaching into groundwater. H+ produced by the trans­
formation of NH+ 4 to NO3 could enhance the leaching of soluble calcium,

leading to an increase in Ca2+ concentrations in groundwater. In addi­


tion, as vegetable crops commonly require calcium for growth, calcium
fertilizers are widely used, further increasing the Ca2+ content in
groundwater. The average concentration of DO in the VCA was 6.67
mg/L, suggesting that there was no obvious organic pollution input in Fig. 5. Contributions of NO−3 sources established by SIAR model in the vege­
table cultivation area (VCA) and grain cultivation area (GCA).
this area (Gholizadeh et al., 2016). Therefore, PC2 was attributed to
chemical fertilizer pollution sources.
the NO−3 pollution mainly came from CF (54.32%) and NS (37.75%),
3.3.2. Determination of pollution source contributions followed by M&S (6.63%) and AD (1.3%). In the GCA, NS (33.67%), CF
Based on PCA, natural sources (NS) and chemical fertilizers (CF) (33.27%), and M&S (30.16%) were the main contributors to NO−3
were determined to be the dominant NO−3 sources in the VCA. The pollution, while AD contributed only 2.90%. The contribution pro­
regression equation described in Equation (7) was obtained using APCS- portions of CF and M&S showed significant spatial variation under
MLR model: different cropping systems, while the difference of NS and AD was not
obvious. This was because intensive vegetable cultivation required more
C = 11.85 × APCS1 + 36.26 × APCS2 − 39.20 (P < 0.05) (7) fertilizer than grain cultivation (Min et al., 2011). The survey found that
2
The adjusted regression coefficient (A-R ) was equal to 0.808, indi­ CF was the most widely used fertilizer in vegetable planting due to its
cating that the model fitted well and could explain 80.8% of the NO−3 quick and efficient effect. Land use type (Fig. 1a) indicated that resi­
source contributions, while the remaining 19.2% was attributed to dential areas in the GCA were more concentrated than in the VCA. The
multiple mixed sources (MMS) (Gholizadeh et al., 2016; Yan et al., residential areas were dominated by rural and town with imperfect
2020). sewage treatment facilities, resulting in the increase of M&S contribu­
The contributions of NO−3 sources established by the APCS-MLR tions in the GCA.
model showed that the dominant NO−3 source was CF in the VCA, ac­ The application of ϵ significantly changed the contributions of NO−3
counting for 58.11%. This finding was consistent with the monitoring pollution sources, especially in the VCA (Fig. 5). Due to isotopic frac­
result from field investigation, finding that intensive vegetable cultiva­ tionation effect, in the VCA, the contribution of NS increased by 23.70%,
tion results in the use of large amounts of chemical fertilizers. The sec­ while the contribution of CF and M&S decreased by 6.74% and 17.03%,
ond major source was NS (mineralization of soil organic nitrogen), respectively. In the GCA, the contribution of NS increased by 9.10%,
which accounted for 22.69% of NO−3 pollution sources. Since NS also while the contribution of CF and M&S decreased by 7.24% and 1.65%,
includes a small CF contribution (in areas with excessive SO2− respectively. The accuracy of NO−3 source apportionment could be
4 ), the
overall contribution of CF may be underestimated in the results of this improved by determining and applying the ϵ according to the trans­
study. formation processes of different nitrogen sources (Yu et al., 2020).
However, to investigate the ϵ of nitrogen transformation processes will
3.4. Source apportionment by SIAR model increase the workload and consume more time and money, which makes
NO−3 source apportionment more complicated.
In this study, δ15N and δ18O were used to calculate the contribution
of four potential pollution sources (CF, NS, atmospheric deposition 3.5. Comparison of source apportionment methods
(AD), and manure and sewage (M&S)). The δ15N-NO−3 and δ18O-NO−3 of
groundwater samples were shown in Table S4. The isotopic values of the To verify the validity and feasibility of the PCA-APCS-MLR model,
in-situ pollution sources and the ϵ in the Dagu River groundwater source this study analyzed and compared NO−3 source apportionment results
area were obtained from the literature of Yu et al. (2020, 2021) from the perspectives of hydrochemical characteristics and isotopes. The
(Table 1). contributions of NO−3 sources in the VCA were determined by APCS-MLR
According to the operation results of SIAR model, the source con­ and SIAR models respectively, as shown in Fig. 6. In the VCA, the results
tributions under different cropping systems were shown in Fig. 5. Pre­ of both source apportionment methods showed that CF and NS were the
vious studies (Yu et al., 2020) have shown that nitrogen transformation main sources of NO−3 pollution. However, the APCS-MLR model failed to
and isotopic fractionation occurred during NO−3 pollution in the study identify M&S. The results of SIAR model indicated that M&S contributed
area. Considering the nitrogen isotopic fractionation effect, in the VCA, was only 6.63%. The low contribution was the main reason for the

6
L. Yu et al. Journal of Environmental Management 314 (2022) 115101

(1) According to the PCA-APCS-MLR model, CF and NS were the


primary contributors to NO−3 contamination in the VCA, with
contributions of 58.11% and 22.69%, respectively. Therefore,
optimizing the fertilization mode and reducing the loss of nitro­
gen fertilizer are the key to control the NO−3 pollution in the
intensive vegetable cultivation area.
(2) The results of NO−3 source apportionment determined by the PCA-
APCS-MLR model were close to those estimated by the SIAR
model in the VCA. In particular, the contributions of CF differed
by only 3.79% between the two methods. The main disagreement
between the two methods was the effective identification of M&S.
Future research can examine representative indicators of M&S to
improve the PCA-APCS-MLR model applications.
(3) The PCA-APCS-MLR model offers a convenient and fast method
for the quantitative identification of NO−3 pollution sources in
groundwater. The major advantage of this model is that it avoids
obtaining isotope data and isotopic enrichment factor, which
greatly reduces the experimental time and cost, and improves the
Fig. 6. Comparative analysis of nitrate pollution sources in the vegetable simplicity of NO−3 source apportionment. The hydrochemical in­
cultivation area using different source apportionment methods. dicators are the only required information which can be easily
measured.
difficulty in identification, leading to the inclusion of M&S in MMS in
the APCS-MLR model. Another reason was that few indicators were Author statement
selected for principal component analysis, which could not fully reflect
the water quality characteristics, leading to difficulty in identifying Lu Yu: Conceptualization, Methodology, Software, Validation,
some pollution sources. Therefore, when the PCA-APCS-MLR model is Formal analysis, Writing - original draft, Writing - review & editing,
used for NO−3 source apportionment, more hydrochemical indicators Visualization. Tianyuan Zheng: Resources, Supervision, Project admin­
should be determined, such as chemical oxygen demand (COD), phos­ istration, Writing - review & editing. Ruyu Yuan: Investigation, Data
phate (PO3− 4 ), total coliform group and bacterial count. This will help curation. Xilai Zheng: Funding acquisition.
extract more representative principal components and improve the
comprehensiveness of pollution source identification.
Declaration of competing interest
Although the results generated using the two methods differed to
some degree, they were largely consistent. In particular, the identified
The authors declare that they have no known competing financial
contribution of CF pollution sources in the VCA differed by only 3.79%
interests or personal relationships that could have appeared to influence
between the two methods. The PCA-APCS-MLR model has the advantage
the work reported in this paper.
of not needing detailed information on pollution sources, requiring only
hydrochemical indicators to extract the PCs and infer pollution sources,
Acknowledgements
which can save cost and time. The SIAR model can estimate the
contribution of pollution sources using the isotopic characteristics of
This work was supported by the National Natural Science Foundation
groundwater and pollution sources. However, overlap in the isotopic
of China (grant number 41731280), the Joint Funds of National Natural
characteristics of different sources and the complexity of isotopic frac­
Science Foundation of China - Shandong Province (grant number
tionation, can reduce the accuracy of source apportionment (Nikolenko
U1806210), the National Natural Science Foundation of China (grant
et al., 2018; Biddau et al., 2019; Yu et al., 2020). In conclusion,
number 51978348).
compared with SIAR model, the PCA-APCS-MLR model is a simpler and
faster method for quantitative identification NO−3 pollution sources in
Appendix A. Supplementary data
groundwater. The PCA-APCS-MLR model can be extended to determine
NO−3 pollution sources in different areas. However, it should be noted
Supplementary data to this article can be found online at https://doi.
that water quality indicators related to possible pollution sources need
org/10.1016/j.jenvman.2022.115101.
to be monitored in different areas. For example, heavy metals (plum­
bum, cadmium, chrome, etc.), phenols and petroleum hydrocarbons
pollutants should also be monitored in industrial areas, while COD, References
3−
NH+ 4 , PO4 , surfactant, total coliform group and bacterial count should Bastani, M., Harter, T., 2019. Source area management practices as remediation tool to
be monitored in livestock farming areas or urban areas. address groundwater nitrate pollution in drinking supply wells. J. Contam. Hydrol.
226, 103521. https://doi.org/10.1016/j.jconhyd.2019.103521.
Biddau, R., Cidu, R., Da Pelo, S., Carletti, A., Ghiglieri, G., Pittalis, D., 2019. Source and
4. Conclusions
fate of nitrate in contaminated groundwater systems: assessing spatial and temporal
variations by hydrogeochemistry and multiple stable isotope tools. Sci. Total
Nitrate has become the dominant contaminant in the Dagu River Environ. 647, 1121–1136. https://doi.org/10.1016/j.scitotenv.2018.08.007.
groundwater source area. In this study, NO−3 sources and their propor­ Burow, K.R., Nolan, B.T., Rupert, M.G., Dubrovsky, N.M., 2010. Nitrate in groundwater
of the United States, 1991-2003. Environ. Sci. Technol. 44, 4988–4997. https://doi.
tional contributions were determined using hydrochemical indicators org/10.1021/es100546y.
and isotopic characteristics. This is the first study to compare the results Cai, J., Xia, X.L., Chen, H.B., Wang, T., Zhang, H.L., 2018. Decomposition of fertilizer use
generated using a multivariate statistical model and a Bayesian isotope intensity and its environmental risk in China’s grain production process.
Sustainability 10 (2), 498. https://doi.org/10.3390/su10020498.
mixing model for NO−3 source apportionment. These findings will benefit Chen, R.H., Chen, H.Y., Song, L.T., Yao, Z.P., Meng, F.S., Teng, Y.G., 2019a.
the development of NO−3 source apportionment models, while also Characterization and source apportionment of heavy metals in the sediments of Lake
providing useful information for the reduction of NO−3 contamination in Tai (China) and its surrounding soils. Sci. Total Environ. 694, 133819. https://doi.
org/10.1016/j.scitotenv.2019.133819.
groundwater. The main conclusions are as follows: Chen, R.H., Teng, Y.G., Chen, H.Y., Hu, B., Yue, W.F., 2019b. Groundwater pollution and
risk assessment based on source apportionment in a typical cold agricultural region

7
L. Yu et al. Journal of Environmental Management 314 (2022) 115101

in Northeastern China. Sci. Total Environ. 696, 133972. https://doi.org/10.1016/j. Liu, J., Shen, Z.Y., Yan, T.Z., Yang, Y.C., 2018. Source identification and impact of
scitotenv.2019.133972. landscape pattern on riverine nitrogen pollution in a typical urbanized watershed,
Choi, W.J., Kwak, J.H., Lim, S.S., Park, H.J., Chang, S.X., Lee, S.M., Arshad, M.A., Yun, S. Beijing, China. Sci. Total Environ. 628–629, 1296–1307. https://doi.org/10.1016/j.
I., Kim, H.Y., 2017. Synthetic fertilizer and livestock manure differently affect δ15N scitotenv.2018.02.161.
in the agricultural landscape: a review. Agric. Ecosyst. Environ. 237, 1–15. https:// Liu, L.L., Tang, Z., Kong, M., Chen, X., Zhou, C.C., Huang, K., Wang, Z.P., 2019. Tracing
doi.org/10.1016/j.agee.2016.12.020. the potential pollution sources of the coastal water in Hong Kong with statistical
Cui, J., Zhou, F.W., Gao, M., Zhang, L.Y., Zhang, L.M., Du, K., Leng, Q.M., Zhang, Y.Z., models combining APCS-MLR. J. Environ. Manag. 245, 143–150. https://doi.org/
He, D.Y., Yang, F.M., Chan, A., 2018. A comparison of various approaches used in 10.1016/j.jenvman.2019.05.066.
source apportionments for precipitation nitrogen in a mountain region of southwest Ma, W.C., Tai, L.Y., Qiao, Z., Zhong, L., Wang, Z., Fu, K.X., Chen, G.Y., 2018.
China. Environ. Pollut. 241, 810–820. https://doi.org/10.1016/j. Contamination source apportionment and health risk assessment of heavy metals in
envpol.2018.06.028. soil around municipal solid waste incinerator: a case study in North China. Sci. Total
Denk, T.R.A., Mohn, J., Decock, C., Lewicka-Szczebak, D., Harris, E., Butterbach-Bahl, K., Environ. 631–632, 348–357. https://doi.org/10.1016/j.scitotenv.2018.03.011.
Kiese, R., Wolf, B., 2017. The nitrogen cycle: a review of isotope effects and isotope Matiatos, I., 2016. Nitrate source identification in groundwater of multiple land-use
modeling approaches. Soil Biol. Biochem. 105, 121–137. https://doi.org/10.1016/j. areas by combining isotopes and multivariate statistical analysis: a case study of
soilbio.2016.11.015. Asopos basin (Central Greece). Sci. Total Environ. 541, 802–814. https://doi.org/
Fang, Y.H., Zheng, T.Y., Zheng, X.L., Peng, H., Wang, H., Xin, J., Zhang, B., 2020. 10.1016/j.scitotenv.2015.09.134.
Assessment of the hydrodynamics role for groundwater quality using an integration Meghdadi, A., Javar, N., 2018. Quantification of spatial and seasonal variations in the
of GIS, water quality index and multivariate statistical techniques. J. Environ. proportional contribution of nitrate sources using a multi-isotope approach and
Manag. 273, 111185. https://doi.org/10.1016/j.jenvman.2020.111185. Bayesian isotope mixing model. Environ. Pollut. 235, 207–222. https://doi.org/
Fenech, C., Rock, L., Nolan, K., Tobin, J., Morrissey, A., 2012. The potential for a suite of 10.1016/j.envpol.2017.12.078.
isotope and chemical markers to differentiate sources of nitrate contamination: a Men, C., Liu, R.M., Wang, Q.R., Guo, L.J., Shen, Z.Y., 2018. The impact of seasonal varied
review. Water Res. 46 (7), 2023–2041. https://doi.org/10.1016/j. human activity on characteristics and sources of heavy metals in metropolitan road
watres.2012.01.044. dusts. Sci. Total Environ. 637–638, 844–854. https://doi.org/10.1016/j.
Gholizadeh, H.M., Melesse, A.M., Reddi, L., 2016. Water quality assessment and scitotenv.2018.05.059.
apportionment of pollution sources using APCS-MLR and PMF receptor modeling Meng, L., Zuo, R., Wang, J.S., Yang, J., Teng, Y.G., Shi, R.T., Zhai, Y.Z., 2018.
techniques in three major rivers of South Florida. Sci. Total Environ. 566–567, Apportionment and evolution of pollution sources in a typical riverside groundwater
1552–1567. https://doi.org/10.1016/j.scitotenv.2016.06.046. resource area using PCA-APCS-MLR model. J. Contam. Hydrol. 218, 70–83. https://
Guo, Z.F., Yan, C.Z., Wang, Z.S., Xu, F.F., Yang, F., 2020. Quantitative identification of doi.org/10.1016/j.jconhyd.2018.10.005.
nitrate sources in a coastal peri-urban watershed using hydrogeochemical indicators Min, J., Zhao, X., Shi, W.M., Xing, G.X., Zhu, Z.L., 2011. Nitrogen balance and loss in a
and dual isotopes together with the statistical approaches. Chemosphere 243, greenhouse vegetable system in southeastern China. Pedosphere 21 (4), 464–472.
125364. https://doi.org/10.1016/j.chemosphere.2019.125364. https://doi.org/10.1016/S1002-0160(11)60148-3.
Gutierrez, M., Biagioni, R.N., Alarcon-Herrera, M.T., Rivas-Lucero, B.A., 2018. An Moore, J.W., Semmens, B.X., 2008. Incorporating uncertainty and prior information into
overview of nitrate sources and operating processes in arid and semiarid aquifer stable isotope mixing models. Ecol. Lett. 11 (5), 470–480. https://doi.org/10.1111/
systems. Sci. Total Environ. 624, 1513–1522. https://doi.org/10.1016/j. j.1461-0248.2008.01163.x.
scitotenv.2017.12.252. Motevalli, A., Naghibi, S.A., Hashemi, H., Berndtsson, R., Pradhan, B., Gholami, V., 2019.
Han, L., Huang, M.S., Ma, M.H., Wei, J.B., Hu, W., Chouhan, S., 2018. Evaluating sources Inverse method using boosted regression tree and k-nearest neighbor to quantify
and processing of nonpoint source nitrate in a small suburban watershed in China. effects of point and non-point source nitrate pollution in groundwater. J. Clean.
J. Hydrol. 559, 661–668. https://doi.org/10.1016/j.jhydrol.2017.04.057. Prod. 228, 1248–1263. https://doi.org/10.1016/j.jclepro.2019.04.293.
Hu, L.X., Zhang, X.H., Zhou, Y.H., 2019. Farm size and fertilizer sustainable use: an Nai, H., Xin, J., Liu, Y., Zheng, X.L., Lin, Z.W., 2020. Distribution and molecular
empirical study in Jiangsu, China. J. Integr. Agric. 18 (12), 2898–2909. https://doi. chemodiversity of dissolved organic nitrogen in the vadose zone-groundwater
org/10.1016/s2095-3119(19)62732-2. system of a fluvial plain, northern China: implications for understanding its loss
Huang, Y., Deng, M.H., Wu, S.F., Japenga, J., Li, T.Q., Yang, X., He, Z.L., 2018. pathway to groundwater. Sci. Total Environ. 723, 137928. https://doi.org/10.1016/
A modified receptor model for source apportionment of heavy metal pollution in j.scitotenv.2020.137928.
soil. J. Hazard Mater. 354, 161–169. https://doi.org/10.1016/j. Nakagawa, K., Amano, H., Takao, Y., Hosono, T., Berndtsson, R., 2017. On the use of
jhazmat.2018.05.006. coprostanol to identify source of nitrate pollution in groundwater. J. Hydrol. 550,
Jehan, S., Khan, S., Khattak, S.A., Muhammad, S., Rashid, A., Muhammad, N., 2019. 663–668. https://doi.org/10.1016/j.jhydrol.2017.05.038.
Hydrochemical properties of drinking water and their sources apportionment of Nikolenko, O., Jurado, A., Borges, A.V., Knller, K., Brouyre, S., 2018. Isotopic
pollution in Bajaur agency, Pakistan. Measurement 139, 249–257. https://doi.org/ composition of nitrogen species in groundwater under agricultural areas: a review.
10.1016/j.measurement.2019.02.090. Sci. Total Environ. 621, 1415–1432. https://doi.org/10.1016/j.
Jia, Y.F., Xi, B.D., Jiang, Y.H., Guo, H.M., Yang, Y., Lian, X.Y., Han, S.B., 2018. scitotenv.2017.10.086.
Distribution, formation and human-induced evolution of geogenic contaminated Ogrinc, N., Tamse, S., Zavadlav, S., Vrzel, J., Jin, L.X., 2019. Evaluation of geochemical
groundwater in China: a review. Sci. Total Environ. 643, 967–993. https://doi.org/ processes and nitrate pollution sources at the Ljubljansko polje aquifer (Slovenia): a
10.1016/j.scitotenv.2018.06.201. stable isotope perspective. Sci. Total Environ. 646, 1588–1600. https://doi.org/
Jin, G.Q., Fang, W., Shafi, M., Wu, D.T., Li, Y.Q., Zhong, B., Ma, J.W., Liu, D., 2019. 10.1016/j.scitotenv.2018.07.245.
Source apportionment of heavy metals in farmland soil with application of APCS- Parnell, A.C., Inger, R., Bearhop, S., Jackson, A.L., 2010. Source partitioning using stable
MLR model: a pilot study for restoration of farmland in Shaoxing City Zhejiang, isotopes: coping with too much variation. PLoS One 5 (3), 1–5. https://doi.org/
China. Ecotoxicol. Environ. Saf. 184, 109495. https://doi.org/10.1016/j. 10.1371/journal.pone.0009672.
ecoenv.2019.109495. Rashid, A., Farooqi, A., Gao, X.B., Zahir, S., Noor, S., Khattak, J.A., 2020. Geochemical
Ke, S.N., Chen, J.J., Zheng, X.L., Sun, X.W., 2021. Reference ion method: a simple and modeling, source apportionment, health risk exposure and control of higher fluoride
fast method for quantitatively identifying the source of nitrate and denitrification in groundwater of sub-district Dargai, Pakistan. Chemosphere 243, 125409. https://
rate in groundwater. Sci. Total Environ. 769, 144555. https://doi.org/10.1016/j. doi.org/10.1016/j.chemosphere.2019.125409.
scitotenv.2020.144555. Rashid, A., Guan, D.X., Farooqi, A., Khan, S., Zahir, S., Jehan, S., Khattak, S.A., Khan, M.
Kim, H., Kaown, D., Mayer, B., Lee, J.Y., Hyun, Y., Lee, K.K., 2015b. Identifying the S., Khan, R., 2018. Fluoride prevalence in groundwater around a fluorite mining area
sources of nitrate contamination of groundwater in an agricultural area (Haean in the flood plain of the River Swat, Pakistan. Sci. Total Environ. 635, 203–215.
basin, Korea) using isotope and microbial community analyses. Sci. Total Environ. https://doi.org/10.1016/j.scitotenv.2018.04.064.
533, 566–575. https://doi.org/10.1016/j.scitotenv.2015.06.080. Rivett, M.O., Buss, S.R., Morgan, P., Smith, J.W., Bemment, C.D., 2008. Nitrate
Kim, K.H., Yun, S.T., Mayer, B., Lee, J.H., Kim, T.S., Kim, H.K., 2015a. Quantification of attenuation in groundwater: a review of biogeochemical controlling processes.
nitrate sources in groundwater using hydrochemical and dual isotopic data Water Res. 42 (16), 4215–4232. https://doi.org/10.1016/j.watres.2008.07.020.
combined with a Bayesian mixing model. Agric. Ecosyst. Environ. 199, 369–381. Singh, K.P., Malik, A., Sinha, S., 2005. Water quality assessment and apportionment of
https://doi.org/10.1016/j.agee.2014.10.014. pollution sources of Gomti river (India) using multivariate statistical techniques—a
Kringel, R., Rechenburg, A., Kuitcha, D., Fouepe, A., Bellenberg, S., Kengne, I.M., case study. Anal. Chim. Acta 538, 355–374. https://doi.org/10.1016/j.
Fomo, M.A., 2016. Mass balance of nitrogen and potassium in urban groundwater in aca.2005.02.006.
Central Africa, Yaounde/Cameroon. Sci. Total Environ. 547, 382–395. https://doi. Su, S.L., Li, D., Zhang, Q., Xiao, R., Huang, F., Wu, J.P., 2011. Temporal trend and source
org/10.1016/j.scitotenv.2015.12.090. apportionment of water pollution in different functional zones of Qiantang River,
Kruk, M.K., Mayer, B., Nightingale, M., Laceby, J.P., 2020. Tracing nitrate sources with a China. Water Res. 45 (4), 1781–1795. https://doi.org/10.1016/j.
combined isotope approach (δ15NNO3, δ18ONO3 and δ11B) in a large mixed-use watres.2010.11.030.
watershed in southern Alberta, Canada. Sci. Total Environ. 703, 135043. https://doi. Sui, Y.Y., Ou, Y., Yan, B.X., Rousseau, A.N., Fang, Y.T., Geng, R.Z., Wang, L.X., Ye, N.,
org/10.1016/j.scitotenv.2019.135043. 2020. A dual isotopic framework for identifying nitrate sources in surface runoff in a
Lentz, R.D., Lehrsch, G.A., 2019. Temporal changes in δ18O and δ15N of nitrate nitrogen small agricultural watershed, northeast China. J. Clean. Prod. 246, 119074. https://
and H2O in shallow groundwater: transit time and nitrate-source implications for an doi.org/10.1016/j.jclepro.2019.119074.
irrigated tract in southern Idaho. Agric. Water Manag. 212, 126–135. https://doi. Taufiq, A., Effendi, A.J., Iskandar, I., Hosono, T., Hutasoit, L.M., 2019. Controlling
org/10.1016/j.agwat.2018.08.043. factors and driving mechanisms of nitrate contamination in groundwater system of
Li, C., Li, S.L., Yue, F.J., Liu, J., Zhong, J., Yan, Z.F., Zhang, R.C., Wang, Z.J., Xu, S., 2019. Bandung Basin, Indonesia, deduced by combined use of stable isotope ratios, CFC
Identification of sources and transformations of nitrate in the Xijiang River using age dating, and socioeconomic parameters. Water Res. 148, 292–305. https://doi.
nitrate isotopes and Bayesian model. Sci. Total Environ. 646, 801–810. https://doi. org/10.1016/j.watres.2018.10.049.
org/10.1016/j.scitotenv.2018.07.345.

8
L. Yu et al. Journal of Environmental Management 314 (2022) 115101

Thurston, G.D., Spengler, J.D., 1985. A quantitative assessment of source contributions mining-impacted sediments. Chemosphere 239, 124748. https://doi.org/10.1016/j.
to inhalable particulate matter pollution in metropolitan Boston. Atmos. Environ. 19 chemosphere.2019.124748.
(1), 9–25. https://doi.org/10.1016/0004-6981(85)90132-5. Yu, C.Q., Huang, X., Chen, H., Godfray, H.C.J., Wright, J.S., Hall, J.W., Gong, P., Ni, S.Q.,
Torres-Martinez, J.A., Mora, A., Mahlknecht, J., Daessle, L.W., Cervantes-Aviles, P.A., Qiao, S.C., Huang, G.R., Xiao, Y.C., Zhang, J.B., Feng, Z., Ju, X.T., Ciais, P.,
Ledesma-Ruiz, R., 2021. Estimation of nitrate pollution sources and transformations Stenseth, N.C., Hessen, D.O., Sun, Z.L., Yu, L., Cai, W.J., Fu, H.H., Huang, X.M.,
in groundwater of an intensive livestock-agricultural area (Comarca Lagunera), Zhang, C., Liu, H.B., Taylor, J., 2019. Managing nitrogen to restore water quality in
combining major ions, stable isotopes and MixSIAR model. Environ. Pollut. 269, China. Nature 567 (7749), 516–520. https://doi.org/10.1038/s41586-019-1001-1.
115445. https://doi.org/10.1016/j.envpol.2020.115445. Yu, L., Zheng, T.Y., Zheng, X.L., Hao, Y.J., Yuan, R.Y., 2020. Nitrate source
Wang, F.F., Guan, Q.Y., Tian, J., Lin, J.K., Yang, Y.Y., Yang, L.Q., Pan, N.H., 2020. apportionment in groundwater using Bayesian isotope mixing model based on
Contamination characteristics, source apportionment, and health risk assessment of nitrogen isotope fractionation. Sci. Total Environ. 718, 137242. https://doi.org/
heavy metals in agricultural soil in the Hexi Corridor. Catena 191, 104573. https:// 10.1016/j.scitotenv.2020.137242.
doi.org/10.1016/j.catena.2020.104573. Yu, L., Zheng, T.Y., Hao, Y.J., Zheng, X.L., 2021. Determination of the nitrogen isotope
Wang, S.Q., Zheng, W.B., Currell, M., Yang, Y.H., Zhao, H., Lv, M.Y., 2017. Relationship enrichment factor associated with ammonification and nitrification in unsaturated
between land-use and sources and fate of nitrate in groundwater in a typical soil at different temperatures. Environ. Res. 202, 111670. https://doi.org/10.1016/
recharge area of the North China Plain. Sci. Total Environ. 609, 607–620. https:// j.envres.2021.111670.
doi.org/10.1016/j.scitotenv.2017.07.176. Yuan, R.Y., Zheng, T.Y., Zheng, X.L., Liu, D.S., Xin, J., Yu, L., Liu, G.Q., 2020.
Xia, Y.Q., Li, Y.F., Zhang, X.Y., Yan, X.Y., 2017. Nitrate source apportionment using a Identification of groundwater nitrate pollution sources in agricultural area using
combined dual isotope, chemical and bacterial property, and Bayesian model PCA and SIAR methods. Episodes 43 (2), 739–749. https://doi.org/10.18814/
approach in river systems. J. Geophys. Res.: Biogeosciences 122, 2–14. https://doi. epiiugs/2020/020047.
org/10.11821/dlxb201601001. Zhang, H., Xu, Y., Cheng, S.Q., Li, Q.L., Yu, H.R., 2020. Application of the dual-isotope
Xiao, M.S., Bao, F.Y., Wang, S., Cui, F., 2016. Water quality assessment of the Huaihe approach and Bayesian isotope mixing model to identify nitrate in groundwater of a
River segment of Bengbu (China) using multivariate statistical techniques. Water multiple land-use area in Chengdu Plain, China. Sci. Total Environ. 717, 137134.
Resour. 43 (1), 166–176. https://doi.org/10.1134/s0097807816010115. https://doi.org/10.1016/j.scitotenv.2020.137134.
Xin, J., Liu, Y., Chen, F., Duan, Y.J., Wei, G.L., Zheng, X.L., Li, M., 2019. The missing Zhang, M., Zhi, Y.Y., Shi, J.C., Wu, L.S., 2018. Apportionment and uncertainty analysis of
nitrogen pieces: a critical review on the distribution, transformation, and budget of nitrate sources based on the dual isotope approach and a Bayesian isotope mixing
nitrogen in the vadose zone-groundwater system. Water Res. 165, 114977. https:// model at the watershed scale. Sci. Total Environ. 639, 1175–1187. https://doi.org/
doi.org/10.1016/j.watres.2019.114977. 10.1016/j.scitotenv.2018.05.239.
Xin, J., Wang, Y., Shen, Z.L., Liu, Y., Wang, H.T., Zheng, X.L., 2021. Critical review of Zhang, Z., Tao, F.L., Du, J., Shi, P.J., Yu, D.Y., Meng, Y.B., Sun, Y., 2010. Surface water
measures and decision support tools for groundwater nitrate management: a surface- quality and its control in a river with intensive human impacts–a case study of the
to-groundwater profile perspective. J. Hydrol. 598, 126386. https://doi.org/ Xiangjiang River, China. J. Environ. Manag. 91 (12), 2483–2490. https://doi.org/
10.1016/j.jhydrol.2021.126386. 10.1016/j.jenvman.2010.07.002.
Xue, D.M., Baets, B.D., Cleemput, O.V., Hennessy, C., Berglund, M., Boeckx, P., 2012. Use Zhou, F., Huang, G.H., Guo, H.C., Zhang, W., Hao, Z.J., 2007. Spatio-temporal patterns
of a Bayesian isotope mixing model to estimate proportional contributions of and source apportionment of coastal water pollution in eastern Hong Kong. Water
multiple nitrate sources in surface water. Environ. Pollut. 161, 43–49. https://doi. Res. 41 (15), 3429–3439. https://doi.org/10.1016/j.watres.2007.04.022.
org/10.1016/j.envpol.2011.09.033. Zhu, A.P., Chen, J.Y., Gao, L., Shimizu, Y., Liang, D.M., Yi, M., Cao, L.X., 2019. Combined
Yan, B., Xu, D.M., Chen, T., Yan, Z.A., Li, L.L., Wang, M.H., 2020. Leachability microbial and isotopic signature approach to identify nitrate sources and
characteristic of heavy metals and associated health risk study in typical copper transformation processes in groundwater. Chemosphere 228, 721–734. https://doi.
org/10.1016/j.chemosphere.2019.04.163.

You might also like