Ecological Indicators 138 (2022) 108838

Contents lists available at ScienceDirect

Ecological Indicators
journal homepage: www.elsevier.com/locate/ecolind

Original Articles

Development of an integrated hydrochemical index for delineating

livestock manure-derived groundwater plumes in agro-livestock
farming areas
Jeong-Ho Lee a, b, Seong-Taek Yun a, *, Soonyoung Yu c, Chang-Hoon Yoo d, Yong-Seok Jeong d,
Kyoung-Ho Kim e, Ho-Rim Kim b, Hyunkoo Kim f
a
Department of Earth and Environmental Sciences, Korea University, Seoul 02841, Republic of Korea
b
Korea Institute of Geoscience and Mineral Resources, Daejeon 34132, Republic of Korea
c
Smart Subsurface Environment Management Research Center, Korea University, Seoul 02841, Republic of Korea
d
Department of Biology, Kyung Hee University, Seoul 02453, Republic of Korea
e
Korea Environment Institute, Sejong 30147, Republic of Korea
f
National Institute of Environmental Research, Incheon 22689, Republic of Korea

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

Keywords: To delineate a livestock manure-derived groundwater plume (LDGP) in agro-livestock farming areas with the
Agro-livestock farming area extensive use of chemical fertilizers, multilevel monitoring wells (MLWs) were installed for depth-specific
Livestock manure-derived groundwater plume sampling of groundwater, and then hydrochemical, nitrate N-O isotopic and microbiological data of shallow
(LDGP)
groundwater collected from the MLWs were evaluated with the help of multivariate statistical tools. The LDGP
Pervasive agricultural contamination
was distinguished from the pervasive agricultural contamination based on δ15Nnitrate (10‰). Fecal coliforms and
Fecal microorganism
Integrated hydrochemical index Escherichia coli were not observed in the LDGP, whereas bovine enterovirus type 2 was detected at a sample
collected at a depth of 18 m below ground level (bgl) downgradient manure piles, indicating a potential virus
risk. Among hydrochemical parameters, Ca2+, Mg2+, Na+, Cl- and SiO2(aq) were shown to be effective indicators
to trace the LDGP. On the other hand, nitrate was not effective to discriminate the LDGP because of denitrifi­
cation in deep parts (>6 m bgl) within the LDGP, although nitrate contamination was serious at shallow parts
(≤6 m bgl; 62.4 to 119.1 mg/L, median 97.9 mg/L). Thus, an integrated hydrochemical index consisting of Ca2+,
Mg2+, Na+, Cl- and SiO2(aq) was suggested based on the result of principal component analysis with isometric log-
ratio transformed data to delineate the LDGP in shallow unconsolidated aquifers overlying silicate bedrocks, and
a threshold of the index was determined to be 1.17. The application of the index to three other agro-livestock
farming areas with prevalent agricultural contamination successfully distinguished the groundwater samples
within the LDGP (11 of 40 samples) that showed the index values exceeding the threshold and a median
δ15Nnitrate of 12.5 ‰. The study results show that the hydrochemical index can be used for the quick evaluation of
a LDGP in agro-livestock farming areas given its cost efficiency and easily accessible analysis devices for water
chemistry compared to isotopes, although the broad use of the index to determine LDGPs needs to be further
verified by application to various geology and land uses. Livestock manures should be carefully managed to
protect groundwater resources, including impermeable covers above or beneath the manure piles, given the
potential risk of virus at deep parts and nitrate at shallow parts within the LDGP. The methodology to select
hydrochemical indicators for distinguishing a LDGP and the combination of the indicators to develop an index
suggested in this study will be useful to build new indices adapted to local conditions.

1. Introduction and agricultural purposes particularly in rural areas. Meanwhile, the

rural areas are faced with groundwater contamination due to various
Groundwater is an important water source for drinking, domestic contamination sources including agricultural fertilizers, livestock

* Corresponding author.
E-mail address: [email protected] (S.-T. Yun).

https://doi.org/10.1016/j.ecolind.2022.108838
Received 22 December 2021; Received in revised form 27 March 2022; Accepted 1 April 2022
Available online 4 April 2022
1470-160X/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
J.-H. Lee et al. Ecological Indicators 138 (2022) 108838

manures, and domestic wastes (e.g., Aravena et al., 1993; Baily et al., geochemical processes occurring within the LDGP, and to suggest a
2011; Power and Schepers, 1989). Recently, the livestock manures are hydrochemical index (i.e., a combination of indicators) effectively
paid attention to as a critical source of groundwater contamination (e.g., differentiating the LDGP from the pervasive agricultural contamination
Zirkle et al., 2016; Kim et al., 2021) as livestock farming has been in shallow unconsolidated aquifers in agro-livestock farming areas. For
expanded and intensified to meet the increased demand for meat (e.g., this purpose, multilevel monitoring wells (MLWs) were installed at both
Nierenberg and Mastny, 2005; Thornton, 2010). Livestock manures are upgradient and downgradient of feedlots and manure piles, and then the
spread on agricultural fields as fertilizers without composting (Ma et al., hydrochemistry, dual isotopic composition of nitrate, and fecal micro­
2018) or temporarily piled up in livestock farming fields without organisms in groundwater with depth were examined. Based on the
appropriate management (e.g., impermeable covers) until they are study result, hydrochemical parameters that differentiated the LDGP
transported for treatment (Rosov et al., 2020), which creates a livestock were chosen and combined to suggest an integrated hydrochemical
manure-derived groundwater plume (LDGP) in aquifers persistently index to trace the LDGP in a shallow unconfined aquifer. The applica­
affected by fertilizers. For instance, the amount of livestock manures bility of the index was validated using hydrochemical and isotopic
increased from 71,530 tons a day in 1992 to 177,110 tons a day in 2012 compositions from three other agro-livestock farming areas in South
in South Korea, and was estimated to be 185,069 tons a day in 2018 Korea. The methodology to select hydrochemical parameters for dis­
(KMOE, 2020). Improper disposal of livestock manures has become a tinguishing a LDGP and to combine them to develop a hydrochemical
major cause of groundwater contamination in agro-livestock farming index and the biogeochemical processes within the LDGP found in this
areas (Cho et al., 2000; Kim et al., 2019b; Kim et al., 2021). study will be useful for managing groundwater pollution by livestock
Livestock manures contain various kinds of organic and inorganic manures in agro-livestock farming areas.
contaminants (Kumar et al., 2013; Regan et al., 2017). Among the
inorganic contaminants, nitrogen compounds are a major contaminant 2. Materials and methods
found in the LDGP at three forms (nitrate, nitrite and ammonium), of
which nitrate is the final product of nitrification and shows the wide­ 2.1. Study area
spread distribution in groundwater due to high mobility and low ab­
sorptivity (Goldberg, 1989). High levels of nitrate in groundwater are This study was conducted in an agro-livestock farming area in the
reported to cause diseases in humans such as methemoglobinemia and Chungnam Province of South Korea (Fig. 1), in which the lowland began
gastric and colorectal cancer (Johnson, 2019; Temkin et al., 2019) in to be converted to livestock farms from agricultural fields approximately
addition to ecological risks such as eutrophication and algal blooming in in 1995, and the areal extent of livestock farming (feedlots) was pro­
surface water (Heisler et al., 2008). In addition, livestock manures are a gressively expanded at the downgradient area (Supplementary Fig. S1).
source of pathogens such as bacteria, parasites and viruses (Bicudo and As of 2010, the main land uses included livestock farming (about 1.55
Goyal, 2003; Hutchison et al., 2000); many of those pathogens have ha) and agricultural fields (about 0.08 ha) as in Fig. 1a, and livestock
been found to survive in the subsurface environment (e.g., Jansons et al., farms were located around agricultural fields. The geology consisted of
1989; Blanford et al., 2005; Bradbury et al., 2013). Thus, careless Jurassic biotite granite, which was locally covered with coarse- to
management of livestock manures can induce pathogen-induced dis­ medium-grained sand (KIGAM, 1995). Colluvium materials such as rock
eases including waterborne diseases on livestock and humans. debris, gravel, silt and sand were also observed when MLWs (M1 to M4)
In order to protect the rural groundwater quality from the livestock were installed in 2013 (Fig. 1b). The land surface had elevations in a
manure, it is essential to characterize the spatial and vertical extent of a range of 30 to 50 m above sea level (asl) and tended to slope gradually
LDGP. However, the evaluation of impacts by an individual pollution towards the east (Fig. 1b).
source in groundwater is challenging because various point and/or non- In the agricultural fields, which were developed in an upland at el­
point sources complicatedly coexist (e.g., Kim et al., 2015; Oh et al., evations of 44 ~ 50 m asl, crops such as peppers, sweet potatoes and
2021). For instance, major hydrochemical compositions including ni­ beans were cultivated with the application of fertilizers (100–150 kg
trate and SO42- come from both fertilizers and livestock manures (e.g., ha− 1 year− 1 as of 2013 from the communication with local farmers)
Bateman and Kelly, 2007; Park et al., 2019; Vitòria et al., 2004) and including silicate fertilizers. It should be noted that the application of
hydrochemically evolve in aquifers depending on redox conditions (e.g., fertilizers must have occurred all over the study area before 1995
Choi et al., 2011). Thus, recently the dual isotopes of nitrate are widely (Supplementary Fig. S1) and probably more intensively given that the
used to distinguish contamination sources (e.g., Gibrilla et al., 2020; Jin overuse of fertilizers resulted in the acidification and nitrate contami­
et al., 2015; Kim et al., 2021; Spalding et al., 2019), while they are nation of shallow groundwater in South Korea (Kim et al., 2019b). The
expensive and time-consuming, and occasionally not applicable because use of chemical fertilizers has been regulated since 2004, decreasing the
of the isotopic overlap of nitrate sources (e.g., Xue et al., 2009). use of chemical fertilizers down to 46% between 1994 and 2014,
An integrated hydrochemical index can be used as an alternative tool whereas the use of livestock manures increased by 30% during the same
to distinguish contamination sources as the river water quality indices period in South Korea (OECD, 2019). In the livestock farming area
(Sutadian et al., 2016; Yotova et al., 2021) and may reduce the un­ located in a lowland at elevations of 30 ~ 43 m asl, livestock manures
certainties caused by the isotopic compositions. For the development of were estimated to be produced at a rate of 5,090 kg day− 1 as of 2010
hydrochemical indices, hydrochemical indicators should be selected as based on a total of 135 dairy cows and average N (37.7 kg head-1 day− 1)
in Solovey et al. (2021) who used Cl-/Br- to determine peatlands affected produced by a dairy cow (NIER, 2014). The livestock manure was piled
by anthropopressure and Ca2+/Mg2+ to determine the dominance of up without any management (e.g., impermeable covers) at D1 and D2 in
rainwater in a fen. Then the hydrochemical indicators should be coupled Fig. 1a until it was transported for treatment.
to provide a single index. Principal component analysis (PCA) has been Three agro-livestock farming areas were additionally chosen in
widely applied to assess major geochemical processes in aquifers and to South Korea to validate the suggested hydrochemical index in a similar
choose hydrochemical indicators to address each geochemical process condition with pervasive agricultural contamination and feedlots
(e.g., Kikuchi et al., 2020; Kim et al., 2019a; Kim et al., 2021; Oh et al., (Fig. 2) and livestock mainly consisting of cows. The bedrock units in the
2021; Solovey et al., 2021). When PCA is conducted using the isometric sites A (Fig. 2a), B (Fig. 2b) and C (Fig. 2c) are Precambrian biotite
log-ratio (ilr) transformed hydrochemical parameters, the ilr co­ gneiss, Jurassic biotite granite, and Cretaceous conglomerate and shale,
ordinates of a subcomposition can be recommended as a method to respectively (KIGAM, 1995), which were covered with colluvium (sites
integrate the selected ions (Batsaikhan et al., 2021). A and B) or alluvium (site C) (Supplementary Fig. S2) similar to the
This study was conducted to select hydrochemical indicators through study area with weathering products overlying silicate bedrock
the understanding of the hydrochemistry of a LDGP and (bio-) (Fig. 1b).

2
J.-H. Lee et al. Ecological Indicators 138 (2022) 108838

Fig. 1. Location and maps of the study area in South Korea. a) Plan view of land use, showing the layout of four multilevel monitoring wells (MLWs; M1 to M4) and a
soil water sampler (Ly) and b) vertical geological profile with topography and water table, showing the sampling depths of groundwater at MLWs. Closed and open
symbols represent the sampling depths for Group I (n = 12) and II (n = 8) defined in Fig. 3, respectively, while the four black closed symbols address the sampling
depths for fecal microorganisms. In (a), the gray dotted arrow indicates the groundwater flow direction estimated based on the topography and water levels in MLWs.

The study area and the three agro-livestock farming areas A to C have in the study (see Supplementary Fig. S2).
a temperate climate with four distinct seasons and are influenced by the
East Asian monsoon. The average annual rainfall for the past 30 years 2.2.1. Sampling
(1981 to 2010) in the study area and the sites A to C were 1,286 mm, A total of twenty groundwater samples were obtained from the
1,223 mm, 1,371 mm and 1,349 mm, of which 67%, 66%, 64% and 67% MLWs at depths between 2.7 and 20 m bgl (Fig. 1b) on August 2013 for
occurred from June to September, respectively (KMA, 2011). In the the analysis of hydrochemical and isotopic compositions and dissolved
study year of 2013, the annual rainfall of the study area and the sites A to organic carbon (DOC). Groundwater samples were also taken twice
C were 1,019 mm, 1,092 mm, 1,062 mm and 1,236 mm with 57%, 51%, between August and November 2013 at four sampling points (M2-6, M3-
62% and 56% of the annual rainfall occurring from June to September, 9, M3-15 and M3-18 where M− d indicates the well name (M) and
respectively (KMA, 2014). sampling depth (‘d’ m bgl); see black closed symbols in Fig. 1b) based on
the isotopic and hydrochemical results to detect fecal microorganisms
2.2. Sample collection and analysis including fecal coliform, Escherichia coli (E. coli) and bovine enterovirus
within the LDGP. In addition, an effluent and a soil water sample were
Four MLWs (M1 to M4) were installed in 2013 in the shallow un­ collected near the livestock manure heaps and analyzed for physi­
confined aquifer consisting of the weathered granite (Fig. 1b) by core ochemical parameters and/or microorganisms, while a livestock manure
drilling down to depths between 18 and 20 m below ground level (bgl) sample was obtained for fecal microorganisms for comparison. From the
to characterize the spatial and vertical extent of groundwater contami­ three agro-livestock farming areas in Fig. 2, groundwater samples (n =
nation by livestock manures. A soil water sampler was also installed at a 40) were collected in September to October in 2013 from A1 to A4 (n =
depth of 0.4 m bgl near D1 (Ly in Fig. 1a) for comparison. Drill cores 12), B1 to B4 (n = 21), and C1 and C2 (n = 7) as in Supplementary
were used for geologic logging in Fig. 1b. The locations of MLWs were Table S1 for physicochemical parameters and dual nitrate isotopes. Note
selected based on land use, topography and groundwater flow di­ that the sampling depths in the MLWs in the sites A to C were located in
rections: M1 was located upgradient adjacent to vegetable fields, silty sand, gravel sand, mixture zone of rock debris, gravel, sand and silt,
whereas M2 to M4 were downgradient in the livestock farming area with and slightly weathered silicate bedrock, similar to those in the study
two livestock manure piles at D1 and D2. Water tables measured in four area in Fig. 1b.
MLWs supported the groundwater flow from M1 (the elevation of sur­ Prior to collection of the groundwater samples, groundwater was
face of 48 m asl) through M2 (37 m asl) and M3 (38 m asl) to M4 (33 m sufficiently and slowly purged at around 0.3 to 0.5 L min− 1 using a
asl) in the aquifer along the topography (Fig. 1b). The hydraulic con­ peristaltic pump (7523–30 Masterflex; Cole Palmer Vernon Hills, IL,
ductivity of the aquifer was estimated to be around 0.14 m d-1 at M2 USA), until pH, electrical conductivity (EC) and dissolved oxygen (DO)
based on the borehole measurements (i.e., flow direction and velocity) were stabilized to obtain representative groundwater samples at each
by using a portable water flow meter (Colloidal Borescope, Geotech). In depth. The water samples for hydrochemical parameters, nitrate N-O
a similar way, ten MLWs (A1 to A4, B1 to B4, and C1and C2) were isotopic compositions and DOC were immediately filtered using 0.45 μm
installed in shallow unconfined aquifers in 2013 in the three agro- membrane filters (Cellulose Nitrate Membrane Filters; Whatman) in the
farming areas (Fig. 2) to validate the hydrochemical index developed field, and were collected in 60 ml polyethylene bottles for

3
J.-H. Lee et al. Ecological Indicators 138 (2022) 108838

Fig. 2. Maps of three agro-livestock farming areas in South Korea: (a) site A, (b) site B, (c) site C, for which the suggested hydrochemical index in Eq. (6) was applied
to validate its applicability for delineating the livestock manure-derived groundwater plume (LDGP) in shallow unconsolidated aquifers.

hydrochemical and isotopic compositions and in 40 ml brown glass vials Hach), and alkalinity was also measured in the field by titration with
for DOC, respectively. The samples for cation analysis were acidified to 0.05 N HNO3 (or 0.5 N HNO3 for effluents) and converted to HCO3–.
pH < 2 by adding a few drops of ultra-pure nitric acid. Meanwhile, the Cations (Ca2+, Mg2+, Na+, K+) and SiO2(aq) were analyzed by ICP-
water samples for fecal coliform and E. coli were not filtered and AES (Optima 3000XL; Perkin Elmer), while anions (Cl-, SO42-, NO3–,
collected in 1 L sterile polypropylene bottles, while the samples for F-) were analyzed by IC (DX-120; Dionex) at the Center for Mineral
bovine enterovirus were filtered using NanoCeram cartridge filters Resources Research (CMR) in Korea University. The charge balances
(Argonide). In addition, a livestock manure sample for fecal microor­ were generally within ± 5%. The δ15N and δ18O values of nitrate in
ganisms was obtained using a sterilized 50 ml conical tube and spoon. groundwater were determined by a gas-isotope ratio mass spectrometer
All water samples were kept at 4 ℃ in a cooler or refrigerator until at the Isotope Science Laboratory in the University of Calgary, following
analysis. the denitrification method (Casciotti et al., 2002).
Fecal coliform and E. coli were analyzed using the Colilert Quanti-
2.2.2. Measurement and analysis tray system (IDEXX Laboratories, ME), while bovine enterovirus was
Unstable parameters such as pH, DO and EC were measured on site analyzed using reverse transcription-polymerase chain reaction (PCR).
using a portable meter (5-Star Plus multi-parameter meter; Thermo The PCR products were identified using the BLASTN program at the
Scientific Orion) and chamber to minimize the contact with air. In Virology Laboratory in Kyung Hee University. Concentrations of DOC
addition, the Fe2+ and NH3-N concentrations of water were measured in was analyzed by using a total organic carbon analyzer at the National
the field using a portable colorimeter (DR/890 Portable Colorimeter; Instrumentation Center for Environmental Management (NICEM) at

4
J.-H. Lee et al. Ecological Indicators 138 (2022) 108838

Seoul National University.

2.3. Multivariate statistical analysis

The relationships among hydrochemical parameters (i.e., Ca2+,

Mg , Na+, K+, SiO2(aq), Cl-, SO42-, NO3–, HCO3–, F-, Fe2+ and NH3-N)
2+

were interpreted using discriminant analysis (DA) and principal

component analysis (PCA) to understand the hydrochemical properties
of the LDGP and to select hydrochemical indicators for tracing the LDGP.
DA is a multivariate statistical technique to separate a set of observa­
tions (n = 20 in this study) into predefined groups (Group I and II in this
study) and to assess the statistical difference between the groups
(Johnson and Wichern, 2002). In this study, DA was used to confirm the
hydrochemical difference between two groups classified based on
δ15Nnitrate. PCA is an effective statistical procedure to reduce the
dimensionality of a multivariate dataset (p = 12 in this study) to two or
three dimensions called principal components (PCs) and helps identify
predominant patterns in a large dataset (Abdi and Williams, 2010). In
this study, PCA was performed to assess the major hydrochemical pro­
cesses in the studied aquifer and to choose hydrochemical indicators to
address each hydrogeochemical process.
Prior to the DA and PCA analyses, the raw data were transformed to
the isometric log-ratio (ilr) data by a sequential binary partition to avoid
erroneous statistical effects by applying a wrong geometry or simplex
(Egozcue et al., 2003; Egozcue and Pawlowsky-Glahn, 2005). The PCA
result on the ilr transformed data was then back-transformed to the
centered log-ratio (clr) space by following Filzmoser et al. (2009) to
interpret the PCA result in terms of the original variables. In addition,
the non-parametric Mann-Whitney U test was conducted using the clr
transformed data to examine statistically meaningful hydrochemical
parameters (i.e., indicators) to differentiate the predefined groups
because the clr transformation overcomes the statistical problems of
compositional data and moreover shows the results of statistical analysis
with the original variable names (Buccianti et al., 2006; Pawlowsky-
Glahn et al., 2007). The data transformations and multivariate statisti­
Fig. 3. Plots of isotopic compositions of nitrate in groundwater from the
cal analyses were performed by CoDaPack version 2.02.21, R version
multilevel monitoring wells (M1 to M4). Group I (n = 12) and II (n = 8) were
3.1.1 and SPSS version 25. defined based on the δ15Nnitrate (10 ‰), and then further divided depending on
Furthermore, a subcomposition representing the LDGP was chosen the sampling depths shallower (triangle) and deeper (circle) than 6 m below
based on the PCA result and Mann-Whitney U test, and then the sub­ ground level (bgl). a) Dual isotopic compositions of nitrate. The boxes (denoted
composition was used to derive an integrated hydrochemical index as a to e) and solid arrows (f and g) are from Xue et al. (2009), and the dotted
following the method described in Batsaikhan et al. (2020). Specifically, line is regression for the Group I samples collected at deep parts (8 circles). b)
the simplex of D-part subcomposition is statistically equivalent to a D-1 The nitrogen isotope composition versus nitrate concentration. DWS indicates
dimensional Euclidean space composed of ilr coordinates where D is the the drinking water standard of South Korea and US EPA.
number of compositions. For example, the ilr coordinates of the three-
part subcomposition, ilr [x1 , x2 , x3 ], are defined as: nitrate from chemical fertilizers (<5 ‰ in δ15Nnitrate) if soil organic
[√̅̅̅̅̅ √̅̅̅̅̅ ] matter is excluded from potential sources due to low nitrate levels
ilr[x1 , x2 , x3 ] = (ilr1 , ilr2 ) =
1 x1 ⋅x2
ln ,
1 x1
ln , (1) (mostly < 5 ppm; Pastén-Zapata et al., 2014; Meghdadi and Javar,
6 x3 2 2 x2 2018). Therefore, this study was focused on the identification and
characterization of the LDGP from the pervasive agricultural
Note that each coordinate has the form of ionic ratios widely used to
contamination.
interpret sources and hydrochemical processes (e.g., El Moujabber et al.,
2006; Ghabayen et al., 2006; Alcalá and Custodio, 2008; Solovey et al.,
2021). Thus, the ilr coordinates were evaluated as an integrated 3.1. Identification of the LDGP
hydrochemical index to differentiate the LDGP from the pervasive
agricultural contamination in this study. Groundwater samples were divided into two groups (Group I and II)
based on the δ15Nnitrate of 10 ‰ (Fig. 3). Group I groundwater (n = 12)
3. Results was collected from all depths in M3 (n = 5) and M4 (n = 6) located in the
downstream of D1 and D2 and the shallowest depth (≤6 m bgl) in M2 (n
Based on the spatiotemporal variation of land use from 1991 to 2010 = 1) (closed symbols in Fig. 1b and 4), and its dual isotopic compositions
(Supplementary Fig. S1) and the excessive applications of fertilizers in suggest the origin of nitrate from livestock manures (Fig. 3; δ15Nnitrate ≥
agricultural lands (see section 2.1), two dominant sources of ground­ 10.5 ‰ in Table 1). In contrast, Group II groundwater (n = 8) was
water contamination were expected in the study area: pervasive agri­ collected from all depths in M1 (n = 4) and deeper depths (>6 m bgl) in
cultural contamination by chemical fertilizers and the restricted M2 (n = 4) (open symbols in Fig. 1b and 4), and its dual isotopic com­
contamination from livestock manures. The dual isotopic compositions positions indicate the origin of nitrate from ammonium-rich chemical
(δ15N and δ18O) of nitrate in Fig. 3 also suggest the coexistence of two fertilizers (Fig. 3; δ15Nnitrate ≤ 6.5 ‰ in Table 1). DA confirmed that
nitrate sources: manure-derived nitrate (>10 ‰ in δ15Nnitrate) and Group I samples were also hydrochemically distinct from.

5
J.-H. Lee et al. Ecological Indicators 138 (2022) 108838

Table 1
Statistical summary of physicochemical parameters, nitrate isotopic compositions and hydrochemical index values for Group I and II samples.
f
Unit Group I (n = 12) Group II (n = 8) p-value
Mean Median Min. Max. Mean Median Min. Max.

Depth m bgl 11.2 10.5 2.7 20.0 13.6 14.5 6.0 20.0
pH – 6.5 6.5 5.7 7.3 6.1 6.3 5.5 6.6
EC a μS/cm 511 507 385 626 195 203 125 253
DO b mg/L 2.62 2.14 0.71 5.51 4.08 3.98 3.05 5.23
TDS c mg/L 336.9 332.0 266.9 422.0 139.2 143.4 111.8 158.4
Ca2+ mg/L 21.1 21.1 15.2 27.3 7.4 6.2 2.4 14.7 0.02
Mg2+ mg/L 10.2 10.2 6.7 13.2 2.3 2.6 1.2 3.3 0.00
Na+ mg/L 50.8 51.5 34.0 60.9 14.9 15.2 7.3 22.1 0.01
K+ mg/L 15.9 9.5 3.1 50.6 11.3 2.3 1.2 31.7 0.52
Cl- mg/L 46.5 43.6 29.5 65.5 10.4 10.1 6.3 14.0 0.00
NO3– mg/L 45.4 28.5 4.2 119.1 20.2 21.4 10.8 31.2 0.73
SO42- mg/L 27.4 23.7 2.3 93.8 15.3 9.1 1.2 41.3 0.85
HCO3– mg/L 102.7 88.2 54.9 169.3 30.3 30.5 9.2 50.3 0.06
F- mg/L 0.1 0.1 0.0 0.4 0.2 0.2 0.1 0.3 0.00
Fe2+ mg/L 0.04 0.04 0.01 0.16 0.05 0.06 0.01 0.11 0.03
NH3-N mg/L 0.12 0.07 0.01 0.54 0.04 0.02 0.01 0.12 0.68
SiO2(aq) mg/L 16.5 13.1 9.2 36.4 26.7 29.8 12.3 37.7 0.00
DOC d mg/L 5.1 2.7 1.0 27.5 1.2 0.7 0.2 5.0
δ15Nnitrate ‰ 12.9 12.5 10.5 18.2 3.1 2.3 1.3 6.5
δ18Onitrate ‰ 3.8 4.0 2.1 4.9 6.0 5.8 4.3 7.6
e
Index value 2.03 2.16 1.17 2.43 0.73 0.76 0.53 0.86
a
Electrical conductivity; b Dissolved oxygen; c Total dissolved solids; d Dissolved organic carbon; e using Eq. (6); f the Mann-Whitney U test for hydrochemical
parameters (centered log-ratio-transformed) between Group I and II. Bold denotes the significant difference (p < 0.05).

Group II samples (p = 0.00; Supplementary Fig. S3). The Mann- and Cl- were probably due to their high levels in the effluent from
Whitney U test showed that the two groups were statistically different livestock manures (Table 2), whereas the relatively low SiO2(aq) seemed
(p < 0.05) in Ca2+, Mg2+, Na+, Cl-, F-, Fe2+ and SiO2(aq) (Table 1). The to be caused by the high SiO2(aq) levels in the background (Group II)
levels of Ca2+, Mg2+, Na+ and Cl- were high in Group I samples, while groundwater since silicate fertilizers have been widely used as soil
the levels of F-, Fe2+ and SiO2(aq) were low in Group I compared to those amendments and for crop growth (Fu et al., 2009; Ma and Takahashi,
in Group II (Fig. 4 and Table 1). 2002; Samaddar et al., 2019; Van Oosten et al., 2017; Zhang et al., 2017)
Based on the statistical difference between the two groups, the LDGP including the study area. DOC was also higher in Group I (average of 5.1
was distinguished from the contamination by chemical fertilizers in the mg/L in Table 1), in particular in M3-18 (i.e., obtained 18 m bgl in M3; >
studied shallow unconfined aquifer using the spatial and vertical dis­ 20 mg/L in Supplementary Fig. S4), implying the effect of anthropogenic
tributions of δ15Nnitrate (Fig. 5c). Total dissolved solids (TDS) and con­ influences (>4 mg/L; Regan et al., 2017).
servative Cl- showed the similar trends to the δ15Nnitrate, while NO3– did However, nitrate concentrations as well as NH3-N, K+ and SO42- in
not as in the Mann-Whitney U test (p > 0.05 in Table 1). Spatially the Group I were not statistically different from those in Group II (Table 1;
LDGP was negligible at all depths in M1, while the LDGP began to Fig. 4), probably due to biogeochemical evolution. In particular, nitrate
appear at the shallowest depth in M2 and then became distinct down­ showed large variations with depth within the LDGP (Figs. 4 and 5). The
gradient (M3 and M4) at all depths, with high TDS (≥266.9 mg/L) and nitrate levels were extremely high (62.4 to 119.1 mg/L, median 97.9
Cl- (≥29.5 mg/L) as well as δ15Nnitrate (≥10.5 ‰) as in Table 1. The mg/L) in shallow parts (≤6 m bgl) exceeding a maximum allowable level
δ15Nnitrate increased by passing below the livestock manure dumping designated for drinking water (44.3 mg/L as NO3–), but abruptly
sites (D1 and D2). decreased (4.2 to 32.5 mg/L, median 20.6 mg/L) in deeper parts (>6 m
bgl) in Table 2. The concentration of K+ and NH3-N also sharply
decreased with depth (Table 1; Fig. 4). Similarly, DO decreased with
3.2. Characteristics of the LDGP depth from the median of 4.59 to 1.71 mg/L, whereas pH increased from
a slightly acidic condition (5.7 to 6.2, median 5.9) in the shallow part to
The LDGP was characterized using the hydrochemical compositions the median of 6.8 (6.4 to 7.3) in the deep part (Table 2; Fig. 4). DOC was
and fecal microorganisms of Group I groundwater (n = 12), while the observed in higher levels in deep parts than in shallow parts within the
Group II groundwater was used to define the background (surrounding) LDGP (Table 2 and Supplementary Fig. S4), which will be discussed with
groundwater for comparison. The Group I groundwater was further denitrification (section 4.2.2) and the detection of virus at deep parts
divided depending on the sampling depth shallower and deeper than 6 (section 3.2.2).
m bgl (Table 2), because the nitrate showed different biogeochemical In contrast, the chloride concentration, which is well known to be
evolution at the depth of 6 m bgl (Fig. 4). Median values were used to conservative (Cook and Herczeg, 2012) and widely used as a tracer of
describe the hydrochemical and isotopic data instead of mean values wastewater (Loizidou and Kapetanios, 1993), was consistent between
although both are provided in Tables 1 and 2 since the median is more the shallow (29.5 ~ 65.5 mg/L; median 39.8 mg/L) and deep saturated
robust against the effect of outliers (Yang et al., 2019) and thus efficient zone (32.2 ~ 62.6 mg/L; median 49.9 mg/L) within the LDGP (Fig. 4,
to show the data distribution. Table 2), although it decreased from the effluent (452.8 mg/L) through
the soil water (237.6 mg/L) in the unsaturated zone to the shallow
3.2.1. Levels and behaviors of hydrochemical compounds and DOC in the saturated zone. The abruptly decrease in nitrate but consistent chloride
LDGP with depth yielded the higher molar ratio of NO3–/Cl− in the shallow
The LDGP was characterized by increased TDS (266.9 to 422.0 mg/L, part (1.0 to 1.6, median 1.2) than in the deep part (0.1 to 0.3, median
median 332.0 mg/L) and Cl- (29.5 to 65.5 mg/L, median 43.6 mg/L) 0.3). Similarly, Ca2+, Mg2+ and Na+ were consistent regardless of depth
compared to Group II (Fig. 5; Table 1). In addition, Ca2+, Mg2+ and Na+ within the LDGP (Fig. 4).
were significantly higher, while SiO2(aq), F- and Fe2+ were lower in
Group I (Table 1; Fig. 4). The high concentrations of Ca2+, Mg2+, Na+

6
J.-H. Lee et al. Ecological Indicators 138 (2022) 108838

Fig. 4. Vertical profiles of physicochemical parameters observed in M1 to M4. The dotted blue (3.93 m bgl), green (2.43 m bgl), orange (4.96 m bgl), and brown
(1.67 m bgl) line in the first figure denotes the measured water table in M1 to M4, respectively. The closed and open symbols denote Group I and II, respectively.

3.2.2. Occurrence of fecal microorganisms in the LDGP g., bovine enterovirus, bovine adenovirus, bovine parvovirus, bovine
Large numbers of fecal coliform (8.3 × 104 MPN/g) and E. coli (1.3 × papillomavirus; Pell, 1997; Ley et al., 2002) and was detected in the
4
10 MPN/g) were counted, and bovine enterovirus type 2 (BEV2) was livestock manure obtained from the study area (Supplementary
detected in the livestock manure sample (Supplementary Table S2). Table S2).
However, fecal coliform and E. coli were not detected either in the soil The non-detection of fecal coliform or E. coli within the LDGP was
water sample or the selected Group I samples (n = 8) within the LDGP probably because they were physically filtered out due to sizes bigger
(black closed symbols in Fig. 1b). In the case of bovine enterovirus, BEV2 than pores or sorbed in soils (Balkhair, 2017; McGechan and Lewis,
was detected once at M3-18 from which the highest DOC (>20 mg/L) 2002; Stevik et al., 2004). Viruses also can be sorbed in soils and thus not
was found (Supplementary Fig. S4), while not in the soil water. The detected in groundwater. However, small-size viruses may reach to the
detection of BEV2 at M3-18 indicates that the LDGP was affected by the deep subsurface (e.g., Jansons et al., 1989; Blanford et al., 2005; Brad­
livestock manure because BEV2 is one of the viruses excreted by cows (e. bury et al., 2013) particularly through preferential flow paths (e.g.,

7
J.-H. Lee et al. Ecological Indicators 138 (2022) 108838

Fig. 5. Contour diagrams of total dissolved solids (TDS), chloride, δ15Nnitrate and NO3– for groundwater samples collected from the four multilevel monitoring wells
along A-A’ in Fig. 1a. The contours were drawn by using Surfer software (version 10).

Corapcioglu et al., 2006). Moreover, the viruses can transport in porous 3.3. Principal component analysis
media without sorption depending on physicochemical conditions,
including pH and ionic strength of solution (e.g., Zhuang and Jin, 2003; PCA was conducted to assess major hydrochemical processes in the
Michen and Graule, 2010), DOC contents (e.g., Sobsey et al., 1980; studied aquifer and to select the hydrochemical components for char­
Regan et al., 2017), and anoxic (e.g., Frohnert et al., 2014) and saturated acterizing the LDGP. The first two PCs accounted for 53% of the total
flow conditions (e.g., Lance and Gerba, 1984). The physicochemical variance in Fig. 6, in which the loadings and scores were back-
condition of high pH and low DO (Fig. 4) and high DOC (Supplementary transformed to the clr space. PC1 was positively related with Ca2+,
Fig. S4) at M3-18 seemed to cause the detection of BEV2 at this depth. Mg2+, Na+ and Cl-, while negatively with Fe2+, F- and SiO2(aq), consis­
Based on the result, the LDGP is poorly identified by the microor­ tent with the result of the Mann-Whitney U test (Table 1). The Group I
ganisms since they are influenced through natural attenuation (e.g., soil samples were well separated from the surrounding groundwater (Group
filtration). Besides, the LDGP is not risky for fecal coliform and E. coli in II) by PC1 and had positive PC1 scores (Fig. 6b), implying that PC1
this aquifer, whereas there is a potential risk of virus in the deep envi­ addressed the LDGP.
ronment since the bovine enterovirus infection is associated with The positive relation of PC1 with Ca2+, Mg2+, Na+ and Cl- was
abortion, stillbirth, infertility and neonatal death in livestock (Knowles probably due to their high levels in livestock manures (Table 2) and
and Mann, 1990). In addition, the non-detection of fecal coliform or consistency within the LDGP (Fig. 4), whereas the negative with Fe2+, F-
E. coli but the detection of BEV2 at M3-18 indicates that fecal coliform or and SiO2(aq) was caused by their depletion within the LDGP (Group I),
E. coli cannot be an indicator of the bovine enterovirus in aquifers unlike compared to the surrounding groundwater affected by agricultural ac­
the expectation that the fecal bacteria are an indicator for fecal virus in tivities (Group II in Table 1) and the rapid decrease of solubility of SiO2
2+
surface water and groundwater (US EPA, 2000). (aq) at pH < 8 (Faure, 1997). In fact, the effluent contained little Fe and
F- but high SiO2(aq) at pH of 8.1 (Table 2), while pH decreased down to <

8
J.-H. Lee et al. Ecological Indicators 138 (2022) 108838

Table 2
Statistical summary of physicochemical parameters, nitrate isotopic compositions and hydrochemical index values for Group I samples (n = 12) divided depending on
the sampling depth shallower and deeper than 6 m blow the ground level (bgl). Physicochemical data of soil water and an effluent released from livestock manures are
also shown for comparison.
Unit Shallow parts (≤6 m bgl) (n = 4) Deep parts (>6 m bgl) (n = 8) Soil water Effluent
Mean Median Min. Max. Mean Median Min. Max. (n = 1) (n = 1)

Depth m bgl 4.7 5.0 2.7 6.0 14.5 15.0 9.0 20.0 0.4 0
pH – 5.9 5.9 5.7 6.2 6.8 6.8 6.4 7.3 8.0 8.1
EC a μS/cm 532 559 385 626 500 496 456 584 2125 8750
DO b mg/L 4.53 4.59 3.42 5.51 1.67 1.71 0.71 2.27 3.01 0.09
TDS c mg/L 362.8 381.2 266.9 422.0 324.0 322.4 292.5 371.1 1614.1 7919.8
Ca2+ mg/L 20.6 21.1 18.2 21.9 21.3 22.0 15.2 27.3 118.8 83.1
Mg2+ mg/L 11.5 11.4 10.1 13.2 9.5 8.9 6.7 13.1 55.2 35.2
Na+ mg/L 44.3 45.7 34.0 51.9 54.0 57.1 39.8 60.9 137.0 248.7
K+ mg/L 32.2 36.0 6.2 50.6 7.8 8.0 3.1 14.0 136.9 774.8
Cl- mg/L 43.7 39.8 29.5 65.5 48.0 49.9 32.2 62.6 237.6 452.8
NO3– mg/L 94.3 97.9 62.4 119.1 20.9 20.6 4.2f 32.5 77.1 131.0
SO42- mg/L 20.7 22.4 2.3 36.0 30.8 23.7 6.7 93.8f 317.9 121.2
HCO3– mg/L 75.3 75.8 60.3 89.2 116.5 121.3 54.9 169.3 450.0 4998.8
F- mg/L 0.2 0.2 0.1 0.4 0.1 0.1 0.0 0.3 0.0 0.0
Fe2+ mg/L 0.03 0.04 0.01 0.04 0.05 0.03 0.01 0.16f 0.01 0.00
NH3-N mg/L 0.24 0.18 0.07 0.54 0.06 0.02 0.01 0.17 60.00 1010.00
SiO2(aq) mg/L 19.6 14.9 12.4 36.4 15.0 11.6 9.2 24.3 23.6 64.2
NO3–/Cl- molar ratio – 1.3 1.2 1.0 1.6 0.2 0.3 0.1 0.3
DOC d mg/L 1.8 1.2 1.0 3.7 6.8 4.4 1.0 27.5
δ15Nnitrate ‰ 12.6 13.1 10.5 13.7 13.1 12.0 11.1 18.2
δ18Onitrate ‰ 4.2 4.4 3.1 4.9 3.6 4.0 2.1 4.9
Index value e 1.86 2.05 1.17 2.17 2.11 2.24 1.64 2.43
a
Electrical conductivity; b Dissolved oxygen; c Total dissolved solids; d Dissolved organic carbon; e using Eq. (6); f Concentrations of NO3–, SO42- and Fe2+ at M3-9
(collected 9 m bgl in M3).

Fig. 6. The results of principal component analysis (PCA) for the isometric log-ratio (ilr) transformed data (n = 20) obtained from the four multilevel monitoring
wells (M1 to M4). a) The centered log-ratio (clr) transformed loadings of hydrochemical parameters, and b) the clr transformed scores of samples for the first two
principal components (PCs).

8 in groundwater. Consistently, the high TDS values were significantly the negative PC2, which will be discussed in section 4.2.
and positively related with Ca2+, Mg2+, Na+ and Cl- (R2 > 0.6 in Sup­
plementary Fig. S5), whereas little with Fe+ and F- (R2 = 0.0) and 4. Discussion
slightly and negatively with SiO2(aq) (R2 = 0.4) in the studied aquifer.
Moreover, the concentration of SiO2(aq) was higher in M1 and M2 Based on the study results, the contamination scenario was made
upgradient and close to the agricultural fields than in M3 and M4 (section 4.1), major hydrochemical processes within the LDGP were
(Fig. 4). discussed (section 4.2), and then an integrated hydrochemical index to
Meanwhile PC2 was positively correlated with K+, SO42-, NH3-N and distinguish the LDGP from the prevailing agricultural contamination
NO3– but negatively with HCO3–, and groundwater samples from deep was suggested and validated (section 4.3).
parts (circles in Fig. 6b) showed negative PC2 scores. Given that the
levels of K+, SO42-, NH3-N and NO3– changed with depth (Fig. 4), PC2
seemed to reflect the biogeochemical controls occurring in the aquifer, 4.1. Contamination scenario
including nitrification and adsorption onto soil particles at shallow
depths along the positive PC2 and denitrification at deep depths along Livestock manure piles at D1 and D2 seemed to be a source of the
LDGP and increase EC, TDS, Ca2+, Mg2+, Na+, Cl- and δ15Nnitrate within

9
J.-H. Lee et al. Ecological Indicators 138 (2022) 108838

the LDGP based on the facts: The Group I groundwater samples (n = 12) increasing the major cations (e.g., Ca2+, Mg2+, Na+ and K+). In fact, the
were located downstream of D1 and D2 (Fig. 1a), the effluent and soil Ca2+ + Mg2+ versus HCO3– plot (Fig. 7a) shows that the hydrochemistry
water contained high levels of these major components (Table 2), and of groundwater samples from MLWs was controlled by both natural
these components were consistent with depth within the LDGP (Figs. 4 weathering (by carbonic acid) and enhanced weathering (by nitrifica­
and 5). Specifically, the LDGP seemed to be originated from the manure tion). Note that the major cations are common weathering products of
pile at D1 and transported downgradient toward M3 and M4 and silicate minerals as well as anthropogenic contaminants (e.g., fertilizers,
deepened down to 18 m bgl at M4, and the M4 was additionally influ­ livestock manures) and can be exchanged with each other in the aquifer.
enced by the manure pile at D2. Given the high concentrations of major
elements in soil water and effluent (Table 2), livestock manures affected 4.2.2. Denitrification
groundwater through soil water. The input of nitrogen compounds from livestock manures and their
In order to properly manage livestock manures, guidelines or regu­ nitrification caused the high levels of nitrate and NH3-N at shallow parts
lations for livestock manures have been suggested (e.g., Sahoo et al., but they rapidly decreased with depth (Fig. 4). Together with increase in
2016). Livestock farms should be equipped with isolated livestock yards pH values, the abrupt decrease in nitrate concentrations with depth
with impermeable surface pavement, livestock manure storages with within the LDGP can be explained by denitrification based on the pos­
covers to prevent rainwater from seeping, and diversions to prevent itive relation between δ18O and δ15N values of nitrate in the deep parts
runoff from reaching livestock manures. If such management systems (8 circles in Group I in Fig. 3a). The dilution effect was excluded because
are not well equipped in livestock farms, the aquifers are easily of the consistent chloride concentrations regardless of depth in Figs. 4
contaminated by leachates from livestock manures. and 5.
In the present study area (Fig. 1), three kinds of causes can be The negative relationship between nitrate and bicarbonate concen­
thought for the LDGP based on the field survey. The first and the most trations in the deep parts (R2 = 0.75 except M3-9; Fig. 7b) suggests the
critical cause was the improper dumping of livestock manures at D1 and heterotrophic denitrification since heterotrophic bacteria use organic
D2. The dumping sites were not equipped with impermeable covers carbon to convert nitrate in groundwater to nitrogen gas, producing
either above or beneath the manure piles and therefore the manure bicarbonate as follows:
could react with rainwater, producing effluents into the soil and aquifer.
The second cause occurred in a gutter near the manure dumps. The 5CH2O + 4NO3- = 2 N2 + CO2 + 3H2O + 4HCO3- (3).
bottom of the gutter was not isolated by pavement and thus could be The high levels of DOC at deep parts support the condition for
affected by runoff. The third cause was observed in a canal connected to denitrification at deep parts, although the transfer pathway of DOC to
the feedlots; the canal had leaking parts. the deep environment is not clear and remains future work.
It was noticeable that M3-9 showed an extremely low concentration
of nitrate (4.2 mg/L) (Fig. 7b) as well as the increased concentration of
4.2. Major hydrochemical processes
sulfate (93.8 mg/L) and Fe2+ (0.16 mg/L) (Table 2; Fig. 4) and low
HCO3– despite high Ca2+ + Mg2+ (Fig. 7a). The high sulfate and Fe2+
Based on the study result, the hydrochemistry in the studied aquifer
and low nitrate and HCO3– can be resulted from a bacterially mediated
seemed to be affected by (1) the input of anthropogenic contaminants
denitrification reaction in the presence of pyrite as follows:
(chemical fertilizers and livestock manures), (2) their interactions with
soils and aquifer materials (e.g., enhanced weathering, sorption), (3) 5FeS2 + 14NO3- + 4H+ = 7 N2 + 10SO42- + 5Fe2+ + 2H2O (4).
infiltrating rainwater and (4) its interactions with subsurface materials.
The latter two factors seemed to evenly affect the relatively small study Such autotrophic denitrification has been well understood in previ­
area with the homogenous geology and climate conditions, implying the ous studies (Postma et al., 1991; Robertson and Cherry, 1992) and found
similar degree of natural processes in the study area. Meanwhile the locally in agro-livestock farming areas in South Korea (Kim et al.,
former two factors were different within and out of the LDGP due to the 2019a).
different major source. Within the LDGP, livestock manure-derived
compounds seemed to be controlled by nitrification and adsorption
4.3. Derivation of an integrated hydrochemical index for the LDGP
onto soil particles preferentially in shallow parts and denitrification in
deep parts according to the PC2 (Fig. 6), which will be discussed in this
4.3.1. An integrated hydrochemical index
section.
Based on the hydrochemical components that distinguished the
Group I groundwater from Group II (Fig. 6a and Table 1), the sub­
4.2.1. Nitrification and adsorption
composition of Ca2+ + Mg2+ + Na+ + Cl-, SiO2(aq) and Fe2+ + F- were
Once the effluent is released from livestock manures and reaches the
chosen as x1 , x2 and x3 in Eq. (1) respectively to derive an hydro­
aquifer, the levels of major compositions increase at shallow parts (e.g.,
chemical index for delineating the LDGP. Then the ilr coordinates are as
K+, NO3– and NH3-N in Fig. 4) because of their high levels in the effluent,
follows:
although they might decrease by passing the unsaturated zone (e.g., soil
[ 2+ ][ ]
water in Table 2). For instance, the high level of K+ in the effluent caused 1 Ca + Mg2+ + Na+ + Cl− SiO2(aq)
ilr1 = √̅̅̅ ln [ (5)
the high concentration of K+ at shallow parts. However, the concen­ 6
]2
F− + Fe2+
tration of K+ abruptly decreased with depth unlike other cations (Fig. 4),
probably due to the sorption of K+ onto soil particles. The noticeable
decrease in the concentration of K+ at M4 (Fig. 4) and in the unsaturated 1 [Ca2+ + Mg2+ + Na+ + Cl− ]
ilr2 = √̅̅̅ ln (6)
zone from the effluent (774.8 mg/L) to the soil water (136.9 mg/L) 2 [SiO2(aq) ]
(Table 1) can be explained by the adsorption onto soil particles (Chen where the unit is molar concentrations.
and Mackenzie, 1992; Liao et al., 2013; Shaviv et al., 1985). The ternary plot for the subcomposition in Fig. 8a shows that the
In addition, the nitrification reaction by bacteria (e.g., Nitrosomonas Group I samples are separated from Group II by the subcomposition.
and Nitrobacter) for the input of anthropogenic N from livestock manures According to the ilr plot in Fig. 8b, the Group I samples are differentiated
produces protons as follows: from Group II in particularly by ilr2 without overlapping, probably
NH4+ + 2O2 = NO3– + H2O + 2H+ (2). because of the negative relation of SiO2 with PC1 (Fig. 6a) and the minor
effect of F- + Fe2+ (<10 % in Fig. 8a). Thus, ilr2 in Eq. (6) was chosen as
which can enhance the rock weathering (Kim et al., 2019a, 2021), an integrated hydrochemical index to define the LDGP in the aquifer.

10
J.-H. Lee et al. Ecological Indicators 138 (2022) 108838

Fig. 7. Plots to explain the hydrochemical processes occurring in groundwater samples collected from the four multilevel monitoring wells (M1 to M4). a) Ca2+ +
Mg2+ versus HCO3–; b) HCO3– versus NO3– concentrations. In (b) the dotted line is regression for the Group I samples collected at deep parts (7 circles) except M3-9
obtained at 9 m below ground level (bgl) at M3.

Fig. 8. a) Ternary plot for the subcomposition (Ca2+ + Mg2+ + Na+ + Cl-, F- + Fe2+ and SiO2), and b) isometric log-ratio (ilr) plot for Group I and II samples. In (a)
all axes have different scales to plot the data in the center. In (b) the threshold for ilr2 was chosen using the minimum value of Group I (1.17).

The index values for Group I samples ranged from 1.17 to 2.43, while the L, median 403.8 mg/L) compared to the surrounding groundwater
values for the Group II samples from 0.53 to 0.86 (Table 1). The lowest (Table 3), while nitrate rapidly decreased with depth below 6 m bgl
value in Group I (1.17) was chosen as a threshold to define a LGDP, within the LDGP (Supplementary Fig. S6) similar to nitrate in Fig. 4.
which was validated in Section 4.3.2 through the application of the Their locations (A3-8, A3-15, B3-8, B4-4, B4-6, B4-8, B4-20, C1-9, C1-
index to three other agro-livestock farming areas. 12, C1-20 and C2-5 in Fig. 2) were quite in accord with the potential
sources of the LDGP, i.e., downstream of feedlots, livestock manure
4.3.2. Validation and usefulness heaps and gutters along the groundwater flow direction. These samples
The high correlations of Ca2+, Mg2+, Na+ and Cl- with TDS (R2 > 0.6; also had the δ15N and δ18O values of nitrate ranging from 6.7 to 20.2 ‰
Supplementary Fig. S5) and their consistent concentrations with depth (median 12.5 ‰) and from 1.1 to 8.2 ‰ (median 4.4 ‰), respectively,
in the LDGP (Figs. 4 and 5), the wide application of silicate fertilizers implying the origin of manure (Fig. 9a; Table 3) and validating the
(see section 2.1), and the negative relation of SiO2(aq) with TDS (R2 = threshold in the studied areas. It should be noted that the two samples
0.4; Supplementary Fig. S5) indicated that the components chosen by (A3-8 and A3-15) showed the δ15Nnitrate < 10‰ and were plotted in the
the statistical methods in Eq. (6) hydrochemically characterize the LDGP overlapping area with the soil organic matter (Fig. 9), while they had the
in the aquifer pervasively affected by chemical fertilizers. Thus, ilr2 was index values exceeding the threshold. Therefore, it is now concluded
applied to three agro-livestock farming areas (Fig. 2) to evaluate the that the hydrochemical index can be used to complement the dual ni­
usefulness of the integrated hydrochemical index (ilr2) to distinguish trate isotopes for source identification.
the LDGP from surrounding groundwater affected by agricultural fer­
tilizers in shallow unconfined aquifers. 5. Conclusions
The application results supported the general applicability of the
index to delineate the LDGP in shallow unconsolidated aquifers with In an agro-livestock farming area where livestock manures were
silicate bedrock in agro-livestock farming areas as in Fig. 9. Among 40 piled up in open fields without impermeable covers and fertilizers have
samples, 11 samples had the hydrochemical indices over the threshold been used for>20 years, the groundwater plume originated from live­
(1.17) for the LDGP (Fig. 9b). These samples also showed high levels of stock manures (called LDGP) were distinguished based on δ15Nnitrate,
EC (397 to 738 μS/cm, median 527 μS/cm) and TDS (314.2 to 495.8 mg/ and then a hydrochemical index was developed to delineate the LDGP.

11
J.-H. Lee et al. Ecological Indicators 138 (2022) 108838

The combined use of discriminant analysis, principal component anal­

ysis and Mann-Whitney U test for hydrochemical data showed that Ca2+,
Mg2+, Na+, Cl- and SiO2(aq) were effective indicators to trace the LDGP
because Ca2+, Mg2+, Na+ and Cl- were enriched in livestock manures
and non-reactive within the LDGP, while SiO2(aq) was relatively depleted
within the LDGP due to the widespread application of silicate fertilizers
in the study area and its solubility rapidly decreasing at pH < 8. In
contrast, the levels of nitrate and fecal microorganisms were not useful
to determine the LDGP because of biogeochemical evolution with depth.
For instance, heterotrophic denitrification involving organic matter
seemed to occur in deep (>6 m bgl) parts of the LDGP, while autotrophic
denitrification appeared locally, which resulted in the abruptly decrease
in nitrate levels in deep parts despite the extremely high nitrate levels in
shallow parts (≤6 m bgl) exceeding the drinking water standard (44.3
mg/L as NO3–). Fecal coliform and E. coli were not detected in ground­
water, whereas the bovine enterovirus was once detected with high DOC
(>20 mg/L) at a relatively deep depth, implying potential risks by virus
at deep parts as well as by nitrate at shallow parts within the LDGP.
Based on the statistically significant parameters for characterizing
LDGP, the molar ratio of Ca2+ + Mg2+ + Na+ + Cl- to SiO2(aq) was
suggested as an integrated hydrochemical index to delineate the LDGP
in the shallow unconfined aquifer where agricultural activities and
livestock feeding operations (mainly cows) coexist. The usefulness of the
index was validated by applying the index to shallow unconfined aqui­
fers in three agro-livestock farming areas in South Korea. It is expected
that the suggested hydrochemical index is beneficial as an index for the
quick assessment of LDGPs since it is based on the major ions that are
easily and cost-effectively analyzed. In addition, the index can suggest
potential locations where fecal coliform and E. coli cannot reach but the
bovine enterovirus may be detected since the index delineates the
overall area affected by livestock manures in the agro-livestock farming
area aquifer. Moreover, the index can be used to supplement the
application of dual nitrate isotopes for source identification by differ­
entiating samples in overlapping isotopic ranges of sources. However,
the applicability of the index and its threshold may be restricted to
Fig. 9. The application result of the suggested hydrochemical index in Eq. (6) shallow unconsolidated aquifers overlying silicate bedrock in agro-
to the hydrochemistry of groundwater samples (n = 40) from three agro-
livestock farming areas, and the broad use of the index needs to be
livestock farming areas in South Korea. Yellow dots denote the groundwater
further verified by application to areas with different geology (e.g.,
samples determined within the livestock manure-derived groundwater plume
carbonates or volcanic rocks), aquifer materials, land use and breeding
(LDGP; ilr2 > 1.17). a) The results plotted with the dual isotopic compositions
of nitrate and b) δ15Nnitrate versus index values. Group I and II samples from the livestock.
study area were also shown for comparison. The study results of hydrochemical and microbial characteristics of

Table 3
Statistical summary of physiochemical parameters, nitrate isotopic compositions and hydrochemical index values of groundwater samples determined in (n = 11) and
out of (n = 29) the livestock manure-derived groundwater plume (LDGP) at three agro-livestock farming areas in South Korea.
Groundwater samples in the LDGP (n = 11) Groundwater samples out of the LDGP (n = 29)
Mean Median Min. Max. Mean Median Min. Max.

Sample depth m bgl 10.5 8.0 4.0 20.0 12.6 12.0 4.0 20.0
pH 6.9 7.0 5.4 8.2 6.5 6.5 5.6 7.3
EC a μS/cm 571 527 397 738 280 254 189 434
DO b mg/L 2.4 1.7 0.3 5.9 2.9 3.1 1.3 4.7
TDS c mg/L 402.4 403.8 314.2 495.8 199.6 192.7 149.3 276.3
Ca2+ mg/L 41.6 39.1 23.7 79.7 16.1 15.6 10.0 25.8
Mg2+ mg/L 19.1 18.3 7.6 32.1 7.0 5.7 2.4 15.7
Na+ mg/L 32.4 28.0 18.0 63.2 19.7 19.3 15.5 27.6
K+ mg/L 10.9 4.5 3.9 25.8 2.6 2.4 1.5 6.6
Cl- mg/L 46.1 37.4 15.0 83.6 17.8 18.4 13.5 21.9
NO3– mg/L 43.5 21.0 1.4 101.5 15.6 8.8 2.4 56.7
SO42- mg/L 63.7 51.1 6.3 125.8 8.0 3.6 0.7 40.4
HCO3– mg/L 125.2 139.6 22.1 221.2 80.4 72.5 45.0 140.3
F- mg/L 0.2 0.2 0.0 0.6 0.1 0.1 0.0 0.4
Fe2+ mg/L 0.1 0.1 0.0 0.1 0.1 0.1 0.0 0.7
NH3-N mg/L 0.2 0.2 0.0 0.7 0.1 0.0 0.0 0.3
SiO2(aq) mg/L 19.4 19.5 14.0 26.0 32.0 33.4 20.7 40.1
δ15Nnitrate ‰ 13.5 12.5 6.7 20.2 7.2 5.6 3.8 20.6
δ18Onitrate ‰ 4.9 4.4 1.1 8.2 4.6 3.8 2.5 13.2
Index value d – 1.8 1.8 1.6 2.2 1.0 1.0 0.6 1.1
a b
Electrical conductivity; Dissolved oxygen; c Total dissolved solids; d using Eq. (6).

12
J.-H. Lee et al. Ecological Indicators 138 (2022) 108838

the LDGP with the help of dual nitrate isotopes and multivariate sta­ Cho, J.-C., Cho, H.B., Kim, S.-J., 2000. Heavy contamination of a subsurface aquifer and
a stream by livestock wastewater in a stock farming area, Wonju. Korea. Environ.
tistical analyses enable to better understand the impacts of livestock
Pollut. 109 (1), 137–146.
manures in the aquifer and the biogeochemical processes within the Choi, B.-Y., Yun, S.-T., Mayer, B., Kim, K.-H., 2011. Sources and biogeochemical behavior
LDGP for the management of shallow unconfined aquifers underneath of nitrate and sulfate in an alluvial aquifer: hydrochemical and stable isotope
agro-livestock farming areas. In particular, this study suggests to pay approaches. Appl. Geochem. 26 (7), 1249–1260.
Cook, P.G., Herczeg, A.L., 2012. Environmental tracers in subsurface hydrology. Springer
attention to the potential risk of virus and nitrate in agro-livestock Science & Business Media.
farming areas depending on depths and the redox processes consisting Corapcioglu, M.Y., Vogel, J.R., Munster, C.L., Pillai, S.D., Dowd, S., Wang, S., 2006. Virus
of nitrification and denitrification. Lastly, the methodology to select and transport experiments in a sandy aquifer. Water Air Soil Pollut. 169 (1-4), 47–65.
Egozcue, J.J., Pawlowsky-Glahn, V., 2005. Groups of parts and their balances in
combine hydrochemical indicators is widely applicable to make a compositional data analysis. Math. Geol. 37 (7), 795–828.
hydrochemical index for groundwater management in South Korea and Egozcue, J.J., Pawlowsky-Glahn, V., Mateu-Figueras, G., Barcelo-Vidal, C., 2003.
elsewhere. Isometric logratio transformations for compositional data analysis. Math. Geol. 35,
279–300.
Moujabber, M.E., Samra, B.B., Darwish, T., Atallah, T., 2006. Comparison of different
indicators for groundwater contamination by seawater intrusion on the Lebanese
Declaration of Competing Interest coast. Water Resour. Manag. 20 (2), 161–180.
Faure, G., 1997. Principles and applications of geochemistry. Prentice Hall Upper Saddle
River, NJ.
The authors declare that they have no known competing financial
Filzmoser, P., Hron, K., Reimann, C., 2009. Principal component analysis for
interests or personal relationships that could have appeared to influence compositional data with outliers. Environmetrics Off. J. Int. Environmetrics Soc. 20,
the work reported in this paper. 621–632.
Frohnert, A., Apelt, S., Klitzke, S., Chorus, I., Szewzyk, R., Selinka, H.-C., 2014. Transport
and removal of viruses in saturated sand columns under oxic and anoxic
Acknowledgements conditions–Potential implications for groundwater protection. Int. J. Hyg. Environ.
Health 217 (8), 861–870.
Fu, X.Q., Long, Z.E., Wei, S.J., Huang, W.X., 2009. Study of the effect of silicate
This research was initially performed with supports from the 2013-
bacterium fertilizer on rice soil microbe and fertility. Soil Fertil. Sci. China 5, 69–71.
2014 research fund from the National Institute of Environmental Ghabayen, S.M.S., McKee, M., Kemblowski, M., 2006. Ionic and isotopic ratios for
Research (NIER) of South Korea (Project title: Survey of Groundwater identification of salinity sources and missing data in the Gaza aquifer. J. Hydrol. 318
(1-4), 360–373.
Contamination Status and Baseline in Livestock Farming Area) and was
Gibrilla, A., Fianko, J.R., Ganyaglo, S., Adomako, D., Anornu, G., Zakaria, N., 2020.
prepared for a doctoral thesis by the first author in February 2020. The Nitrate contamination and source apportionment in surface and groundwater in
completion of this work was supported by the Korea Environment In­ Ghana using dual isotopes (15N and 18O-NO3) and a Bayesian isotope mixing model.
dustry & Technology Institute (KEITI) through the Subsurface Envi­ J. Contam. Hydrol. 233, 103658.
Goldberg, V.M., 1989. Groundwater pollution by nitrates from livestock wastes. Environ.
ronment Management Research Project (No. 2021002440003) and by Health Perspect. 83, 25–29.
the Basic Research Projects (No. 21-3412-1 and No. 21-3415) of the Heisler, J., Glibert, P.M., Burkholder, J.M., Anderson, D.M., Cochlan, W., Dennison, W.
Korea Institute of Geoscience and Mineral Resources (KIGAM) funded by C., Dortch, Q., Gobler, C.J., Heil, C.A., Humphries, E., Lewitus, A., Magnien, R.,
Marshall, H.G., Sellner, K., Stockwell, D.A., Stoecker, D.K., Suddleson, M., 2008.
the Ministry of Science and ICT of Korea. Eutrophication and harmful algal blooms: a scientific consensus. Harmful Algae 8
(1), 3–13.
Hutchison, M. L., Nicholson, F. A., Smith, K. A., Keevil, C. W., Chambers, B. J., Moore, A.,
Appendix A. Supplementary data
2000. A study on farm manure applications to agricultural land and an assessment of
the risks of pathogen transfer into the food chain. A report to the Ministry of
Supplementary data to this article can be found online at https://doi. Agricultural Fisheries and Food, January 2000.
org/10.1016/j.ecolind.2022.108838. Jansons, J., Edmonds, L., Speight, B., Bucens, M., 1989. Movement of viruses after
artificial recharge. Water Res. 23 (3), 293–299.
Jin, Z., Qin, X., Chen, L., Jin, M., Li, F., 2015. Using dual isotopes to evaluate sources and
References transformations of nitrate in the West Lake watershed, eastern China. J. Contam.
Hydrol. 177, 64–75.
Johnson, R.A., Wichern, D.W., 2002. Applied multivariate statistical analysis. Prentice
Abdi, H., Williams, L.J., 2010. Principal component analysis. Wiley Interdiscip. Rev.
hall Upper Saddle River, NJ.
Comput. Stat. 2 (4), 433–459.
Johnson, S.F., 2019. Methemoglobinemia: Infants at risk. Curr. Probl. Pediatr. Adolesc.
Alcalá, F.J., Custodio, E., 2008. Using the Cl/Br ratio as a tracer to identify the origin of
Health Care 49 (3), 57–67.
salinity in aquifers in Spain and Portugal. J. Hydrol. 359 (1-2), 189–207.
KIGAM (Korea Institute of Geoscience and Mineral Resources), 1995. Geological Maps of
Aravena, R., Evans, M.L., Cherry, J.A., 1993. Stable isotopes of oxygen and nitrogen in
Korea (in Korean).
source identification of nitrate from septic systems. Ground Water 31 (2), 180–186.
Kikuchi, T., Kohzu, A., Ouchi, T., Fukushima, T., 2020. Quantifying the sources and
Baily, A., Rock, L., Watson, C.J., Fenton, O., 2011. Spatial and temporal variations in
removal of nitrate in riparian and lotic environments based on land use and
groundwater nitrate at an intensive dairy farm in south-east Ireland: Insights from
topographic parameters of the watershed. Ecological Indicators 116, 106535.
stable isotope data. Agric. Ecosyst. Environ. 144 (1), 308–318.
Kim, H.R., Yu, S., Oh, J., Kim, K.H., Lee, J.H., Moniruzzaman, M., Kim, H.K., Yun, S.T.,
Balkhair, K.S., 2017. Modeling fecal bacteria transport and retention in agricultural and
2019a. Nitrate contamination and subsequent hydrogeochemical processes of
urban soils under saturated and unsaturated flow conditions. Water Res. 110,
shallow groundwater in agro-livestock farming districts in South Korea. Agric.
313–320.
Ecosyst. Environ. 273, 50–61.
Bateman, A.S., Kelly, S.D., 2007. Fertilizer nitrogen isotope signatures. Isotopes Environ.
Kim, H.-R., Yu, S., Oh, J., Kim, K.-H., Oh, Y.-Y., Kim, H.K., Park, S., Yun, S.-T., 2019b.
Health Stud. 43 (3), 237–247.
Assessment of nitrogen application limits in agro-livestock farming areas using
Batsaikhan, B., Yun, S.-T., Kim, K.-H., Yu, S., Lee, K.-J., Lee, Y.-J., Namjil, J., 2021.
quantile regression between nitrogen loadings and groundwater nitrate levels. Agric.
Groundwater contamination assessment in Ulaanbaatar City, Mongolia with
Ecosyst. Environ. 286, 106660.
combined use of hydrochemical, environmental isotopic, and statistical approaches.
Kim, K.H., Yun, S.T., Mayer, B., Lee, J.H., Kim, T.S., Kim, H.K., 2015. Quantification of
Sci. Total Environ. 765, 142790.
nitrate sources in groundwater using hydrochemical and dual isotopic data
Bicudo, J.R., Goyal, S.M., 2003. Pathogens and manure management systems: a review.
combined with a Bayesian mixing model. Agric. Ecosyst. Environ. 199, 369–381.
Environ. Technol. 24 (1), 115–130.
Kim, S.H., Kim, H.R., Yu, S., Kang, H.J., Hyun, I.H., Song, Y.C., Kim, H., Yun, S.T., 2021.
Blanford, W.J., Brusseau, M.L., Jim Yeh, T.C., Gerba, C.P., Harvey, R., 2005. Influence of
Shift of nitrate sources in groundwater due to intensive livestock farming on Jeju
water chemistry and travel distance on bacteriophage PRD-1 transport in a sandy
Island, South Korea: with emphasis on legacy effects. Water Res. 191, 116814.
aquifer. Water Res. 39 (11), 2345–2357.
KMA (Korea Meteorological Administration), 2011. Climatological NORMALS of Korea.
Bradbury, K.R., Borchardt, M.A., Gotkowitz, M., Spencer, S.K., Zhu, J., Hunt, R.J., 2013.
Report No. 11-1360000-000077-14 (in Korean).
Source and transport of human enteric viruses in deep municipal water supply wells.
KMA (Korea Meteorological Administration), 2014. Annual Climatological Report 2013.
Environ. Sci. Technol. 47 (9), 4096–4103.
Report No. 11-1360000-000016-10 (in Korean).
Buccianti, A., Mateu-Figueras, G., Pawlowsky-Glahn, V., 2006. Compositional data
KMOE (Korea Ministry of Environment), 2020. Statistics on the production and disposal
analysis in the geosciences: from theory to practice. Geological Society of London.
of livestock manures (in Korean). accessed on 30 December 2020. http://www.me.
Casciotti, K.L., Sigman, D.M., Hastings, M.G., Böhlke, J.K., Hilkert, A., 2002.
go.kr/.
Measurement of the oxygen isotopic composition of nitrate in seawater and
Knowles, N.J., Mann, J.A., 1990. Bovine enteroviruses, in: Virus Infections of Ruminants.
freshwater using the denitrifier method. Anal. Chem. 74 (19), 4905–4912.
Elsevier 513–516.
Chen, J.S., Mackenzie, A.F., 1992. Fixed ammonium and potassium as affected by added
Kumar, R.R., Park, B.J., Cho, J.Y., 2013. Application and environmental risks of livestock
nitrogen and potassium in three Quebec soils. Commun. Soil Sci. Plant Anal. 23 (11-
manure. J. Korean Soc. Appl. Biol. Chem. 56 (5), 497–503.
12), 1145–1159.

13
J.-H. Lee et al. Ecological Indicators 138 (2022) 108838

Lance, J.C., Gerba, C.P., 1984. Virus movement in soil during saturated and unsaturated Sahoo, P.K., Kim, K., Powell, M.A., 2016. Managing groundwater nitrate contamination
flow. Appl. Environ. Microbiol. 47 (2), 335–337. from livestock farms: implication for nitrate management guidelines. Curr. Pollut.
Ley, V., Higgins, J., Fayer, R., 2002. Bovine Enteroviruses as Indicators of Fecal Reports 2 (3), 178–187.
Contamination. Appl. Environ. Microbiol. 68 (7), 3455–3461. Samaddar, S., Truu, J., Chatterjee, P., Truu, M., Kim, K., Kim, S., Seshadri, S., Sa, T.,
Liao, Y.-L., Zheng, S.-X., Nie, J., Xie, J., Lu, Y.-H., Qin, X.-b., 2013. Long-term effect of 2019. Long-term silicate fertilization increases the abundance of Actinobacterial
fertilizer and rice straw on mineral composition and potassium adsorption in a population in paddy soils. Biol. Fertil. Soils 55 (2), 109–120.
reddish paddy soil. J. Integr. Agric. 12 (4), 694–710. Shaviv, A., Mohsin, M., Pratt, P.F., Mattigod, S.V., 1985. Potassium fixation
Loizidou, M., Kapetanios, E., 1993. Effect of leachate from landfills on underground characteristics of five southern California soils. Soil Sci. Soc. Am. J. 49 (5),
water quality. Sci. Total Environ. 128 (1), 69–81. 1105–1109.
Ma, J.F., Takahashi, E., 2002. Soil, fertilizer, and plant silicon research in Japan. Sobsey, M.D., Dean, C.H., Knuckles, M.E., Wagner, R.A., 1980. Interactions and survival
Elsevier. of enteric viruses in soil materials. Appl. Environ. Microbiol. 40 (1), 92–101.
Ma, Y., Liu, Z.-H., Xi, B.-D., He, X.-S., Li, Q.-L., Qi, Y.-J., Jin, M.-Y., Guo, Y., 2018. Solovey, T., Wojewódka-Przybył, M., Janica, R., 2021. Hydrochemical indicators of
Characteristics of groundwater pollution in a vegetable cultivation area of typical water source and contamination in fen peatlands of varying hydrogeomorphic
facility agriculture in a developed city. Ecol. Indicat. 105, 709–716. settings in northern and central Poland. Ecol. Indicat. 129, 107944.
McGechan, M.B., Lewis, D.R., 2002. SW – soil and water: transport of particulate and Spalding, R.F., Hirsh, A.J., Exner, M.E., Little, N.A., Kloppenborg, K.L., 2019.
colloid-sorbed contaminants through soil, part 1: general principles. Biosyst. Eng. 83 Applicability of the dual isotopes δ15N and δ18O to identify nitrate in groundwater
(3), 255–273. beneath irrigated cropland. J. Contam. Hydrol. 220, 128–135.
Meghdadi, A., Javar, N., 2018. Quantification of spatial and seasonal variations in the Stevik, T.K., Aa, K., Ausland, G., Hanssen, J.F., 2004. Retention and removal of
proportional contribution of nitrate sources using a multi-isotope approach and pathogenic bacteria in wastewater percolating through porous media: a review.
Bayesian isotope mixing model. Environ. Poll. 235, 207–222. Water Res. 38, 1355–1367.
Michen, B., Graule, T., 2010. Isoelectric points of viruses. J. Appl. Microbiol. 109 (2), Sutadian, A.D., Muttil, N., Yilmaz, A., Perera, B., 2016. Development of river water
388–397. quality indices—a review. Environ. Monit. Assess. 188, 58. https://doi.org/10.1007/
NIER (National Institute of Environmental Research, Korea), 2014. Guidelines for s10661-015-5050-0.
Management of Water Pollution (Report No. 11-1480523-001918-01). Temkin, A., Evans, S., Manidis, T., Campbell, C., Naidenko, O.V., 2019. Exposure-based
Nierenberg, D., Mastny, L., 2005. Happier meals: rethinking the global meat industry. assessment and economic valuation of adverse birth outcomes and cancer risk due to
Worldwatch Institute. nitrate in United States drinking water. Environ. Res. 176, 108442.
Oecd, 2019. Nutrient Balance (indicator). accessed on 30 December 2020. https://data. Thornton, P.K., 2010. Livestock production: recent trends, future prospects. Philos.
oecd.org/agrland/nutrient-balance.htm. Trans. R. Soc. B Biol. Sci. 365 (1554), 2853–2867.
Oh, J., Kim, H.-R., Yu, S., Kim, K.-H., Yun, S.-T., 2021. Delineating the impacts of poultry US EPA, 2000. Improved enumeration methods for the recreational water quality
burial leachate on shallow groundwater in a reclaimed agro-livestock farming area, indicators: Enterococci and Escherichia coli.
using multivariate statistical analysis of hydrochemical data. Environ. Sci. Poll. Res. Van Oosten, M.J., Pepe, O., De Pascale, S., Silletti, S., Maggio, A., 2017. The role of
28 (7), 7742–7755. biostimulants and bioeffectors as alleviators of abiotic stress in crop plants. Chem.
Park, J., Cho, K.H., Ligaray, M., Choi, M.-J., 2019. Organic matter composition of Biol. Technol. Agr. 4, 1–12.
manure and its potential impact on plant growth. Sustainability 11, 2346. Vitòria, L., Otero, N., Soler, A., Canals, À., 2004. Fertilizer characterization: isotopic data
Pastén-Zapata, E., Ledesma-Ruiz, R., Harter, T., Ramírez, A.I., Mahlknecht, J., 2014. (N, S, O, C, and Sr). Environ. Sci. Technol. 38, 3254–3262.
Assessment of sources and fate of nitrate in shallow groundwater of an agricultural Xue, D., Botte, J., De Baets, B., Accoe, F., Nestler, A., Taylor, P., Van Cleemput, O.,
area by using a multi-tracer approach. Sci. Total Environ. 470, 855–864. Berglund, M., Boeckx, P., 2009. Present limitations and future prospects of stable
Pawlowsky-Glahn, V., Egozcue, J.J., Tolosana Delgado, R., 2007. Lecture notes on isotope methods for nitrate source identification in surface-and groundwater. Water
compositional data analysis. http://dugi-doc.udg.edu/bitstream/handle/10256/ Res. 43 (5), 1159–1170.
297/CoDabook.pdf?sequence=1 (accessed on 15 March 2022). Yang, J., Rahardja, S., Fränti, P., 2019. Outlier detection: how to threshold outlier
Pell, A.N., 1997. Manure and microbes: public and animal health problem? J. Dairy Sci. scores?. In: In: Proceedings of the International Conference on Artificial Intelligence,
80 (10), 2673–2681. Information Processing and Cloud Computing, pp. 1–6.
Postma, D., Boesen, C., Kristiansen, H., Larsen, F., 1991. Nitrate reduction in an Yotova, G., Varbanov, M., Tcherkezova, E., Tsakovski, S., 2021. Water quality assessment
unconfined sandy aquifer: water chemistry, reduction processes, and geochemical of a river catchment by the composite water quality index and self-organizing maps.
modeling. Water Resour. Res. 27 (8), 2027–2045. Ecol. Ind. 120, 106872.
Power, J.F., Schepers, J.S., 1989. Nitrate contamination of groundwater in North Zhang, M., Liang, Y., Chu, G., 2017. Applying silicate fertilizer increases both yield and
America. Agric. Ecosyst. Environ. 26 (3-4), 165–187. quality of table grape (Vitis vinifera L.) grown on calcareous grey desert soil. Sci.
Regan, S., Hynds, P., Flynn, R., 2017. An overview of dissolved organic carbon in Hort. 225, 757–763.
groundwater and implications for drinking water safety. Hydrogeol. J. 25, 959–967. Zhuang, J., Jin, Y., 2003. Virus retention and transport through Al-oxide coated sand
Robertson, W.D., Cherry, J.A., 1992. Hydrogeology of an unconfined sand aquifer and its columns: effects of ionic strength and composition. J. Contam. Hydrol. 60 (3-4),
effect on the behavior of nitrogen from a large-flux septic system. Appl. Hydrogeol. 193–209.
0 (0), 32–44. Zirkle, K.W., Nolan, B.T., Jones, R.R., Weyer, P.J., Ward, M.H., Wheeler, D.C., 2016.
Rosov, K.A., Mallin, M.A., Cahoon, L.B., 2020. Waste nutrients from US animal feeding Assessing the relationship between groundwater nitrate and animal feeding
operations: Regulations are inconsistent across states and inadequately assess operations in Iowa (USA). Sci. Total Environ. 566, 1062–1068.
nutrient export risk. J. Environ. Manage. 269, 110738.

14