Water Research 223 (2022) 118998

Contents lists available at ScienceDirect

Water Research
journal homepage: www.elsevier.com/locate/watres

Groundwater metabolome responds to recharge in fractured

sedimentary strata
Christian Zerfaß a, Robert Lehmann b, Nico Ueberschaar c, Carlos Sanchez-Arcos a, 1, Kai
Uwe Totsche b, Georg Pohnert a, *
a
Department of Bioorganic Analytics, Institute of Inorganic and Analytical Chemistry, Friedrich Schiller University, Jena, Germany
b
Department of Hydrogeology, Institute of Geosciences, Friedrich Schiller University, Jena, Germany
c
Mass Spectrometry Platform, Faculty for Chemistry and Earth Sciences, Friedrich Schiller University, Jena, Germany

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

Understanding the sources, structure and fate of dissolved organic matter (DOM) in groundwater is paramount
Original content: MTBLS3450 (Original data)
for the protection and sustainable use of this vital resource. On its passage through the Critical Zone, DOM is
subject to biogeochemical conversions. Therefore, it carries valuable cross-habitat information for monitoring
Keywords:
and predicting the stability of groundwater ecosystem services and assessing these ecosystems’ response to
Groundwater
fluctuations caused by external impacts such as climatic extremes. Challenges arise from insufficient knowledge
Dissolved Organic Matter
Recharge on groundwater metabolite composition and dynamics due to a lack of consistent analytical approaches for long-
Discharge term monitoring. Our study establishes groundwater metabolomics to decipher the complex biogeochemical
Metabolomics transport and conversion of DOM. We explore fractured sedimentary bedrock along a hillslope recharge area by a
Aquifer 5-year untargeted metabolomics monitoring of oxic perched and anoxic phreatic groundwater. A summer with
extremely high temperatures and low precipitation was included in the monitoring. Water was accessed by a
monitoring well-transect and regularly collected for liquid chromatography-mass spectrometry (LC-MS) inves­
tigation. Dimension reduction of the resulting complex data set by principal component analysis revealed that
metabolome dissimilarities between distant wells coincide with transient cross-stratal flow indicated by
groundwater levels. Time series of the groundwater metabolome data provides detailed insights into subsurface
responses to recharge dynamics. We demonstrate that dissimilarity variability between groundwater bodies with
contrasting aquifer properties coincides with recharge dynamics. This includes groundwater high- and lowstands
as well as recharge and recession phases. Our monitoring approach allows to survey groundwater ecosystems
even under extreme conditions. Notably, the metabolome was highly variable lacking seasonal patterns and did
not segregate by geographical location of sampling wells, thus ruling out vegetation or (agricultural) land use as
a primary driving factor. Patterns that emerge from metabolomics monitoring give insight into subsurface
ecosystem functioning and water quality evolution, essential for sustainable groundwater use and climate
change-adapted management.

1. Introduction
The recent IPCC report highlights the potential impact of climate
change on groundwater quality and quantity. It notes a lack of quanti­
tative data on groundwater development that prevents the reliable
prediction of potential impacts (Caretta et al., 2022). Particularly the
composition of dissolved organic matter (DOM) and its fate in
groundwater reservoirs are rather obscure. DOM signatures are essential
for assessing groundwater ecosystem functioning and vulnerability to
land use, water resource exploitation, and climate change-driven threats
(Drake et al., 2020; Humphrey et al., 2016; Phillips et al., 2012; Riedel
and Weber, 2020; Rodell et al., 2018; Rodríguez-Cardona et al., 2022;
United Nations Office for Disaster Risk Reduction (UNDRR), 2021;
Wada et al., 2012). Groundwater DOM sources, composition,

Abbreviations: CZ, Critical zone; DOC, dissolved organic carbon; DOM, dissolved organic matter; LC-MS, liquid chromatography coupled to mass spectrometry;
PCA, principal component analysis.
* Corresponding author.
E-mail address: georg.pohnert@uni-jena.de (G. Pohnert).
1
Present Address: Institute of Zoology, Faculty of Mathematics and Sciences, University Cologne, Cologne, Germany

https://doi.org/10.1016/j.watres.2022.118998
Received 23 March 2022; Received in revised form 1 August 2022; Accepted 16 August 2022
Available online 20 August 2022
0043-1354/© 2022 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
C. Zerfaß et al.
persistence, and transformation reflect biogeochemical processes
below-ground (Kaiser and Kalbitz, 2012; Lehmann and Kleber, 2015; Li
et al., 2017; Roth et al., 2019; Schmidt et al., 2011). DOM also plays a
crucial role in the transport of organics and in the supply of the
below-ground microbial communities. These multiple functions attract a
high interest for DOM in Critical Zone (CZ) research worldwide (Grant
and Dietrich, 2017; Küsel et al., 2016; Li et al., 2017; Richardson, 2017;
Riebe et al., 2017).
For sustainable resource management and groundwater protection, a
sound understanding of recharge areas’ hydrogeological and ecological
functioning is required (Lehmann and Totsche, 2020).
Multi-disciplinary, temporally highly-resolved and long-term moni­
toring efforts can provide valuable insight and support concepts for
management. However, such approaches are still rare, except for tar­
geted investigations of contaminants (Fernandes et al., 2021). To date,
we have no information about spatial and temporal patterns and dy­
namics of the entirety of metabolites and their transformation products
(here termed as metabolome) in the groundwater.
Such information is especially required for vulnerable recharge areas
where cross-compartment matter exchange (Lehmann and Totsche,
2020) and dynamics of microbial activity contribute to the evolution of
groundwater quality (Benk et al., 2019). Fractured sedimentary bedrock
with mixed lithologies hosts important groundwater reservoirs world­
wide. These reservoirs are challenging to investigate due to the interplay
of transient flow patterns and various sedimentary or surface-fed sources
of organic matter. In addition, the dynamics of microbial communities,
their metabolic activity, and their long-term persistence influence the
groundwater metabolome. Assessing the temporal dynamics of the
groundwater metabolome may provide cross-habitat information in
multiple ways: Metabolic composition reflects the input from
above-ground and contributions from mobile and rock-associated mi­
crobial populations in high-permeability (fractures) and
low-permeability (rock matrices) environments. In addition, abiotic
transformations contribute to metabolic diversity below-ground (Cho­
rover and Amistadi, 2001).
Untargeted metabolomics techniques are emerging tools in envi­
ronmental monitoring of compound transportation and transformation
(Crimmins and Holsen, 2019). They have also been successfully used to
characterize the fraction of metabolites in DOM in ground- (Aukes and
Schiff, 2021; Benk et al., 2019; Hofmann et al., 2020), soil- (Danczak
et al., 2020; Lu et al., 2020; Seifert et al., 2016; Ueberschaar et al., 2021)
and surface-waters (Aukes and Schiff, 2021; Danczak et al., 2020; Drake
et al., 2020; Garayburu-Caruso et al., 2020; Hawkes et al., 2020; Lu
et al., 2020; Lynch et al., 2019; Patriarca and Hawkes, 2021; Pemberton
et al., 2020; Raeke et al., 2017; Seifert et al., 2016; Tanentzap et al.,
Upper Buntsandstein). (C) Examples of drill cores from the screen sections of the monitoring wells, representing contrasting compartments with similar lithology.
Left: Weathered mudrock and stained fractures indicate oxic conditions in diverse aeration zone habitats. Right: Rock colors and brittle mudstone beds indicate
oxygen-deficient conditions in low-weathered strata. c: calcimudstone, f: fracture, m: mudrock/-stone, r: rudstone (bioclastic limestone).
2
Water Research 223 (2022) 118998
2019; Wologo et al., 2021). As sub-surface metabolites are not well
characterized, exact compound annotation is challenging. Therefore,
previous studies focused on selected marker molecules and compound
class categorization. However, fingerprint analyzes by multivariate ap­
proaches such as principal component analysis (PCA) also proved a
powerful tool for the analysis of aggregated monitoring data (Benk et al.,
2019, 2018; Schwab et al., 2017; Ueberschaar et al., 2021). These ap­
proaches allowed for the categorization of similarities across compart­
ments that can be linked to subsurface connectivity (Küsel et al., 2016;
Lehmann and Totsche, 2020) and the composition of groundwater
microbiomes (Benk et al., 2019; Yan et al., 2020).
Here we present a long-term, high-resolution analysis of ground­
water metabolome patterns in Triassic limestone-mudstone alternations
of a topographic recharge area (Fig. 1). We profiled both perched and
phreatic groundwater for over 5 years at multiple locations. The
assessment is based on untargeted metabolomics using data from liquid
chromatography coupled to mass spectrometry (LC-MS). A two-step data
analysis approach is chosen. Metabolome data per individual moni­
toring well is analyzed over time giving seasonal and intra-annual data
at specific locations. In addition, data are analyzed by generating a
dissimilarity score at each timepoint, by which evolving metabolomes
are investigated over time. While a detailed analysis of the metabolomes
at a single time point was conducted previously (Sanchez-Arcos et al.,
2022), the data analysis presented here aimed at exploring the metab­
olome structure and dynamics in distinct zones of the groundwater
recharge area shown in Fig. 1.
The metabolomic dataset is embedded in an extensive assessment of
environmental data. This includes water levels, inorganic ions, DOC
(sum parameter) and other basic hydrochemical parameters such as pH,
temperature, dissolved oxygen and electric conductivity. These large
sets of basic parameters allow for the reconstruction of flow regimes and
ecosystem compartmentalization (Lehmann and Totsche, 2020). In this
context, untargeted metabolomics enables the understanding of sub­
surface ecosystem functioning by untangling the specific information
contained in DOM. The metabolome reflects the signal resulting from
the surface and subsurface sources (e.g. sedimentary carbon turnover)
and all abiotic and biotic transformations of organic metabolites within
the Critical Zone until the time of sampling. The groundwater metab­
olome thus contains cross-habitat information, potentially revealing
whole-ecosystem functioning and stability. Notably, the metabolic
output will not associate with individual organisms but with microbial
communities. Thus, metabolomic assessment represents a proxy for
metabolic activity in cross-habitat microbial populations.
Fig. 1. Configuration and aquifer rock
properties of the study site. (A) Location of
the well transect in central Germany. (B)
Schematic cross section of the hillslope
groundwater recharge area of the Hainich
Critical Zone Exploratory (CZE). Arrow roses
highlight transient and prevailing flow di­
rections in two contrasting compartments, the
deep aeration zone (downward/stratiform >
upward) and the phreatic zone (upward >
downward), in discharge direction. Dashed blue
lines show presumed groundwater flow pat­
terns. Bedrock strata belong to lithostrati­
graphic subgroups Lower Muschelkalk (mu),
Middle Muschelkalk (mm), Upper Muschelkalk
(mo) with Trochitenkalk formation (moTK),
Warburg Fm. (moW), Meissner Fm. (moM),
above the regional flow barrier (Röt formation,
C. Zerfaß et al.
2. Materials and methods
2.1. Site description
The exemplary studied groundwater recharge area is part of the
Hainich Critical Zone Exploratory (or observatory) in NW Thuringia,
central Germany. It was established by the Collaborative Research
center AquaDiva focusing on subsurface biodiversity, groundwater
quality, and its surface- and subsurface controls. The exploratory and the
selection criteria for the wells are introduced in (Küsel et al., 2016). The
site represents fractured sedimentary bedrock environments, or more
specifically, widely distributed mixed carbonate-/siliciclastic rock al­
ternations, as well as groundwater reservoirs in agricultural catchments
with forested recharge areas in low-mountain terrain. The monitoring
well transect spans ca. 5 km along the eastern hillslope of the Hainich
mountain range, comprising different relief positions, aquifers, and
depths (see Fig. 1). The wells access perched groundwater bodies in the
upper slope of a thick aeration zone, and phreatic groundwater in
downslope positions with depths between 7 and 88 m below ground
level (Kohlhepp et al., 2017). In this study, analysis includes data from
each three wells in mid- to footslope positions (H41-H43 and H51-H53,
phreatic zone). Only one upper midslope well (H32, perched ground­
water) was included in the analysis since it was the only one at that site
that accessed groundwater throughout the period of observation.
We demonstrate our metabolome analysis for two representative
compartments within the studied groundwater recharge area: perched
groundwater in the shallow aeration zone (well H32), and deeper
phreatic groundwater (H52) that locates in between two other phreatic
zone observer wells (H51 and H53).
2.2. Groundwater sampling and monitoring
Groundwater samples were provided during joint, usually monthly
sampling campaigns as described previously (Küsel et al., 2016; Leh­
mann and Totsche, 2020). Each 5 l of groundwater per well were
sampled in triplicate borosilicate bottles by employing submersible
pumps (MP1 or SQE 5–70, Grundfos, Denmark) and PE-HD tubing, after
representative conditions were achieved by monitoring physicochem­
ical parameters until stabilization. Groundwater levels, temperatures,
specific electrical conductivity and inorganic ion contents were
measured as described previously (Lehmann and Totsche, 2020).
2.3. Untargeted metabolomics by LC-MS
Samples collected during a three-day campaign were stored dark and
chilled until subsequent transfer and immediate processing. Metabolite
extraction was carried out by GF/C filtering (1.5 µm, VWR #516–0875)
followed by solid phase extraction (SPE) on Strata-X 33 µm polymeric
reversed phase cartridges (500 mg, Phenomenex #8B-S100-HCH). Prior
to extraction, the SPE cartridges were placed on a Supelco Visiprep DL
24 manifold (Merck/Sigma-Aldrich #57,265) and conditioned with 5 ml
LC-MS grade methanol. Subsequently, GF/C filters placed in a Swinnex
47 mm diameter filter holder (Carl Roth, #XE88.1) were assembled onto
the cartridges, and connected via Teflon tubes to the sample bottles.
Flow of sample was maintained by applying vacuum (up to –500 mbar as
measured by a manifold-attached manometer). After 5 l groundwater
samples had run through, 5 ml of LC-MS grade water were applied for
washing the SPE cartridges (GF/C filter already detached), and vacuum
was applied for 5 min for cartridge drying. For elution, 4 ml of a 1:1
mixture methanol/acetonitrile (LC-MS grade) was applied and extracts
were collected in 4 ml glass vials. The solvent was evaporated by
applying a nitrogen stream or vacuum and subsequently dissolved in
100 µl of a 1:1 mixture of methanol/tetrahydrofuran (LC-MS grade).
Samples were transferred to LC glass vials (with insets) and if sediments
formed, clear supernatants were transferred to new vials by pipetting.
From these, 1 µl was injected into the LC-MS in technical triplicates.
3
Water Research 223 (2022) 118998
LC-MS was performed on a Dionex UltiMate 3000 chromatography
system coupled to a Q-Exactive Plus orbitrap mass spectrometer
(Thermo Fisher Scientific, Germany). A heated electrospray ionization
(HESI) source was used with the following ionization and mass spec­
trometry settings: mass range 100–1500 m/z, resolution 70,000 m/Δm,
AGC target 3 × 106, polarity positive and negative (only positive used in
later analysis), maximum inject time 200 ms, spray voltage 3.0 kV
(positive mode) and 2.5 kV (negative mode), capillary temperature
360 ◦ C. LC separation was conducted at 25 ◦ C column temperature, with
slightly different methods as follows: LC Method 1 (samples until June
2017): LC-column: Kinetex Core Shell (Phenomenex; 1.7 u, C18, 100A,
100 × 2.1 mm). Solvent A: H2O, 2% acetonitrile, 0.1% formic acid.
Solvent B: acetonitrile. LC-gradient: 0–0.2 min 100% A; 0.02–8 min
linear gradient to 0% A; 8–9 min 0% A; 9.1–10 min 100% A. LC Method 2
(samples March 2018 till July 2019): LC-column: Accucore TM C18
(Thermo Fisher Scientific, 2.1 × 100 mm, 2.6 mm). Solvent A: H2O, 2%
acetonitrile, 0.1% formic acid. Solvent B: acetonitrile, 0.1% formic acid.
LC-gradient: 0–0.2 min 100% A; 0.02–8 min linear gradient to 0% A;
8–12 min 0% A; 12.1–14 min 100% A. LC Method 3 (samples from March
2020; equivalent gradient to LC Method 1 with longer recording towards the
end of method): LC-column: Kinetex Core Shell (Phenomenex; 1.7 u, C18,
100A, 100 × 2.1 mm). Solvent A: H2O, 2% acetonitrile, 0.1% formic
acid. Solvent B: acetonitrile. LC-gradient: 0–0.2 min 100% A; 0.2–8 min
linear gradient to 0% A; 8–11 min 0% A; 11.1–12 min 100% A.
2.4. Data processing and statistical analysis
Data acquired by liquid chromatography coupled to mass spec­
trometry is accessible via the Metabolights repository (Haug et al., 2019)
with study identifier MTBLS3450 (Zerfaß et al., 2021). Thermo Fisher
Xcalibur raw LC-MS output files were converted to mzXML files with
ProteoWizard (Chambers et al., 2012) and subsequently analyzed in
R-CoreTeam (2017) with the Bioconductor (Huber et al., 2015) package
XCMS (Benton et al., 2010; Smith et al., 2006; Tautenhahn et al., 2008).
Datasets were analyzed with standardized XCMS parameters, and
crucial steps are described in the following. During data conversion to
mzXML by ProteoWizard, positive-mode LC-MS traces were extracted in
the time range 50–570 s, in which recordings were overlapping for
LC-MS methods 1 to 3 as described above. The converted files were read
into, and analyzed by, XCMS with function readMSdata, peak identifi­
cation and integration with parameters defined in centWaveParam,
retention time adjustment was carried out with the obiwarp method,
and grouping and peak filling from baseline data was executed with
groupChromPeaks and fillChromPeaks, respectively. The table of inte­
grated peaks was extracted with the featureValues function.
For hierarchical clustering of metabolite feature traces, peak in­
tensities were z-normalized (Cooksey, 2020; Miller et al., 2015), and
averages and standard deviations were calculated for the technical
replicates (i.e. triplicate LCMS injections) and groundwater replicates
(triplicates, standard deviation via pooled variance). Clustering was
carried out by the R-function cor with Pearson correlation for complete
observations. Variability was assessed by ANOVA with Benjamini
Hochberg correction (Antonelli et al., 2019) (R-functions aov and p.
adjust).
For principal component analyzes, zero peak intensities were
changed to "not applicable" (NA), and the peak intensity table was log2-
transformed. Again, averages and standard deviations were calculated
for the technical replicates (i.e. triplicate LC-MS injections) and
groundwater replicates (triplicates, standard deviation via pooled vari­
ance). Principal component analysis (Ivanisevic and Want, 2019;
Nguyen and Holmes, 2019; Worley and Powers, 2012) (PCA) was car­
ried out with R’s prcomp function on centered peak intensity data,
omitting NAs, on the groundwater-replicate averages dataset. With the
resulting rotation data, sample averages and standard deviations were
transformed into the PC1-PC2 space.
For long-term analysis of consistent analysis periods, data for one
C. Zerfaß et al.
sampling point at different time points was fed into the PCA. Notably,
measurement parameters may drift over time, such as for (a) changes in
chromatography columns and LC methods and (b) due to the long time
period, drifts in e.g. LC-column capacities and separation efficiency that
may complicate integrated data-analysis involving retention time
alignments of chromatograms. We therefore conducted an additional
analysis where PCAs were carried out for each sampling campaign (i.e.
time points) with a complete set of seven sampling locations from the
Hainich CZE wells H32, H41–3, and H51–3 (compare Fig. 1). Distances
between each two datapoints (i.e. between the measurement averages)
in the PC1-PC2 space were calculated, and the standard deviations of the
distances were calculated via the pooled (weighted) variance in the di­
rection of the connecting line as follows. Knowing the differences be­
tween datapoints in the direction of the PC1 and PC2 axis, respectively,
the distance was inferred from Pythagoras’ law as
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
( )̅
Δwell1,well2 = Δ2well1,well2 (PC1) + Δ2well1,well2 (PC2)
where Δwell1,well2 indicates the distance between two data points in the
PC1-PC2-space, and Δ2well1,well2(PC1) and Δ2well1,well2(PC2) the squares of
the distances in the principal components 1 and 2, respectively.
The angle α of the connecting line (measured from PC1 axis) was
inferred by
( 2 )
Δ (PC1)
α = arccos well1,well2 (2)
Δwell1,well2
where arccos is the arccosine (also sometimes expressed as cos− 1).
The standard deviation in the PC1-PC2 space was considered as
elliptical space (Wang et al., 2015; Yuill, 1971) with limits defined by
the respective standard deviations in the first two principal components
(δ(PC1) and δ(PC2)). From these limits, the coordinates of the standard
deviation ellipse for all angles α were associated with an angle ϕ on the
limit circles, so that
〈 〉 〈 〉 collected at three geographical locations from sampling wells at one
coordinatePC1 δ(PC1) ∗ cos(ϕ)
= (3)
coordinatePC2 δ(PC2) ∗ sin(ϕ)
where “coordinates” define the standard deviation ellipse as vectors
with respective values in the direction of PC1 and PC2 for each angle ϕ,
and δ signifies standard deviations.
The length of the standard deviation vector at the angle α (that de­
termines the orientation of the connecting line between the data points
of well 1 and 2) can therefore be calculated from the individual standard
deviations δ(PC1) and δ(PC2) by determining the angle ϕ (defining
angle that determines the value on the standard deviational ellipse in the
orientation of α) with the following equation
⎛√ ̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
√⎛
√ half-year (Fig. S1). Periods of low specific electrical conductivity during
⎜√
⎜ √⎜ cos(α)2 ∗ δy
2
ϕ = arccos⎜ √⎜ δx2
⎝ √⎝ 1 − cos(α)2 + cos(α)2 ∗ δy22 ⎠ ⎠
δx
and using Pythagoras’ law as
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
( )
δ(Δwell1 (α)) = (δ(PC1, well1) ∗ cos(ϕ))2 + (δ(PC2, well2) ∗ sin(ϕ))2
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
( )
δ(Δwell2 (α)) = (δ(PC1, well2) ∗ cos(ϕ))2 + (δ(PC2, well2) ∗ sin(ϕ))2
This allows to calculate a standard deviation for each of the sampling
wells in the PC1-PC2 space in the direction of the connecting line be­
tween the data points. The standard deviation of the well-distance was
calculated via the pooled (weighted) variance from the individual
standard deviations of the datapoints, with equal sample sizes (leaving
Water Research 223 (2022) 118998
the weighting term obsolete). The difference between the wells was
thereafter divided by the standard deviation of the difference, which
expresses the distance as multiples of its standard deviation:
Δwell1,well2 Δwell1,well2
differences in multiples of σ = ( ) = √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
( 2 )
δ Δwell1,well2 (α)
δ (α) + δ2 (α) well1 well2
(7)
where σ (sigma) refers to the standard deviation of the distance between
the data points for well 1 and well 2 in the PC1-PC2-space. Note for
statistical assessment that for a normal-distributed random sample, the
1σ and 2σ ranges mark the intervals expectedly containing 68% and 95%
of observations, respectively (Wang et al., 2015). To analyze the
dissimilarity of wells over time, the well differences in multiples of σ
were compared by plotting them against the date of sampling
campaigns.
(1)
3. Results and discussion
3.1. Dimension reduction of metabolome data reveals responses to high-
flow conditions in the aeration and phreatic zone
The long-term monitoring of groundwater bodies with contrasting
aquifer properties along a topographic recharge area provided a
metabolomic data set suitable for determining spatiotemporal patterns.
Groundwater sampling, filtration, and extraction on reversed-phase
material were optimized for reproducibility, recovery, and robustness.
Extracts were measured with high resolution LC-MS using reversed-
phase ultra high-pressure liquid chromatography (UHPLC) for separa­
tion. Recently, a systematic analysis of the metabolomes from different
wells at a single time point was reported (Sanchez-Arcos et al., 2022).
Here, we focus on pattern recognition in the long-term data sets without
the aim to elucidate the structure of the detected metabolites. The
five-year monitoring dataset comprised 35 time points when data was
(H32) or three (H41-H43, H51-H53) depths. This data set was system­
atically mined to identify patterns according to location, depth, season
and recharge events. In addition, basic hydrochemical and conventional
monitoring parameters, including groundwater levels and temperature,
were collected along with the metabolome.
The thick hillslope aeration zone, containing perched groundwater
bodies (Fig. 1), is accessed by well H32. Multiple pronounced periods of
recharge in the winter half-year interspersed by low-stand phases were
observed there during the monitoring period from 2014 to 2020
(Fig. 2A). During recharge, the Ca/Mg-carbonate rich water is oxygen­
ated (up to 3.8 mg/L dissolved oxygen, not shown) and the temperature
dips (Fig. 2A). Further, nitrate is supplied via percolation in the winter
⎞̅ ⎞
⎟ ⎟ early recharge phases indicate short residence times (Figure S1). Dis­
⎟ ⎟ (4)
⎟ solved organic carbon (DOC, sum parameter) and potassium did not
show significant responses to recharge, whereas during recession pha­
ses, warming, increasing sulfur/sulfate, sodium, and calcium point to
longer residence (Fig. S1) (Bakalowicz, 1994; Hudak, 2004).
The metabolome of well H32 from 09/2014 to 03/2017 (consistent
monitoring with mostly monthly sampling campaigns) revealed a highly
variable composition, with more than 75% of the metabolite features
(5)
significantly changing in intensity over time (Fig. 2B, D). Notably, the
metabolomes at different time points were characterized by singular
emergences of metabolite features (high z-normalised intensities indi­
(6) cated in red), rather than recurrent, or seasonally reoccuring, profiles
(Fig. 2B). The lack of recurrent signals characterizing the studied setting
of thin-bedded alternations of carbonate-/siliciclastic-rock may also
apply for other groundwater bodies in widely distributed mixed-
lithology sedimentary reservoirs. Lacking large fractures or karst con­
duits, thin/-bedded alternations represent retarded flow systems
(White, 2012), thus generally exhibiting less pronounced DOM
4
C. Zerfaß et al. Water Research 223 (2022) 118998

Fig. 2. Single-well (H32) metabolome time-series analysis. (A) Groundwater level and temperature traces. Recharge phases (gray-shaded) are inferred from the
multi-level head data and the environmental tracer temperature (red). Seasonal cooling is observed during the winter half-year highstands. (B) Hierarchical cluster
analysis of z-normalised (Cooksey, 2020; Miller et al., 2015) metabolite feature intensities. Red color in the heatmap indicates high positive values (i.e. high
measured intensities of underlying LC-MS signals), blue color negative z-normalised values (i.e. low intensity or absence of underlying signal. (C) Principal
component analysis of time-series data. Arrows are drawn as guides for chronological order of datapoints; the first datapoint of each year is labelled in blue font. (D)
False discovery rate (FDR; from Benjamini-Hochberg corrected ANOVA) distribution as indicator of significance of metabolite feature intensity variability in the
time series.

5
C. Zerfaß et al.
dynamics by input from the subsurface. DOM dynamics might result
from multidirectional flow dynamics within groundwater recharge areas
(Lehmann and Totsche, 2020), supplying water and DOM from different
sources, including sedimentary carbon. Such sediments are common in
freshwater aquifers worldwide, and the observed patterns might be
prototypic (Franco et al., 2017).
Interannual variability may generally depend on weather and cli­
matic conditions (droughts, precipitation, temperature) (Humphrey
et al., 2016; Phillips et al., 2012; Riedel and Weber, 2020; Rodell et al.,
2018; United Nations Office for Disaster Risk Reduction (UNDRR),
2021). Elsewhere, geochemical tracer investigations revealed that re­
sponses to intra-annual periodicity in precipitation and associated
recharge is reflected in the groundwater composition (Bonneau et al.,
2018; Bowen et al., 2019; Jasechko et al., 2014; Tweed et al., 2005).
Emergence and dissipation of metabolite signals could be linked to these
factors within the course of a year.
The first two (determining) principle components revealed consid­
erable patterns when metabolic signatures throughout the sampling
period were analyzed (Fig. 2C). The most striking was an observed shift
coinciding with the high-flow period 12/2015–05/2016, causing a
disturbance in the PCA (Fig. 2C, lower half). Similarly, the high-flow
periods 11/2014–06/2015 (upper-left of PCA) and 11/2016–04/2017
(Fig. 2C, upper right) caused separate clusters, interspersed by distinct
phases such as the low-stand phase 06/2015–11/2016 (Fig. 2C, center-
left).
The much deeper phreatic zone well H52 (Fig. 1) was characterized
by damped, though high-magnitude water table fluctuation of more
than 18 m due to seasonal recharge dynamics (Figs. 3A; Supplementary
Fig. S2). Compared to the perched groundwater, the oxygen-free Ca/Mg-
carbonate water is low in nitrate and high in sulfate. As indicated by less
surface-fed input, this aquifer zone is characterized by more isolated
conditions (Lehmann and Totsche, 2020). During recharge phases,
increased calcium sulfate (Fig. S2) and hydraulic heads indicate
ascending flow from deeper strata (Lehmann and Totsche, 2020).
Increased EC25 associated with increasing sodium and potassium and
higher hydraulic heads in the shallower observer (H53, Fig. 3A) indicate
descending flow during recessions. These observations highlight
considerable cross-stratal exchange (Lehmann and Totsche, 2020) in the
stratified bedrock. DOC, again, showed insignificant fluctuation, point­
ing to the limited merit of the sum parameter in such a sedimentary
setting.
The metabolome in the phreatic groundwater (11/2014 to 03/2017)
was also highly variable and exhibited singular emergences of metabo­
lites at high intensities (Fig. 3B, D). Like in the well H32 in the uphill
aeration zone and the other wells of the phreatic zone, recurrent fluc­
tuations of metabolite signals were missing. Nevertheless, we again
observed an association of the PCA with the groundwater fluctuation
(Fig. 3A, C). The recharge-phase 11/2015–05/2016 contained a clear
dispersion in the bottom-left of the PCA, while the subsequent discharge
phase caused a cluster transition (08–09/2016, center-right) with
another dispersion during the low-stand 01–03/2017 (center-right).
While less distinct, the recharge phase 12/2014–05/2015 caused a PCA-
shift (range 04–06/2015).
3.2. General coincidence of temporal metabolome pattern changes and
recharge dynamics across sampling sites
The association of metabolome PCAs with water levels emerged as a
consistent observation across the recharge area. This was observed in
shallow as well as in deep wells spanning different habitat conditions
(Lazar et al., 2019) and microbial communities (Yan et al., 2021, 2020).
We discuss here in detail results from wells H32 and H52 exemplarily as
representatives for aeration zone and phreatic zone wells. However, in
all cases (H32, H41 to − 43, H51 to − 53), the metabolome was variable,
with pronounced shifts in the PCA-time series, particularly in response
to high-flow / recharge phases (Figs. 2 and 3 for H32 and H52, as well as
6
Water Research 223 (2022) 118998
Supplementary Figs. S3 to S8 for the other wells).
A summary is given in Supplementary Table 1, highlighting distinct
phases in the PCAs. The association of the first two principal components
with water level fluctuation, representing cross-stratal exchange in the
fractured bedrock strata, revealed the strong responses to recharge dy­
namics. It is conceivable that high-flow phases will modify the DOM
profile by co-transporting organic matter into the groundwater. In
accordance, the dissolved organic carbon (DOC) levels in H32 and H52
were often variable during recharge periods (Supplementary Figs. S1
and S2).
Transport of organic matter by water fluxes can change the
groundwater metabolome by the input of novel compounds and mi­
crobial biochemical transformations will cause additional variability. At
our study site, it has been recently reported that the microbiome
composition is impacted by water recharge phases as observed for the
metabolome (Yan et al., 2021, 2020). It is conceivable that a changing
microbiome will cause a shift in biochemical functions and therefore
impact the metabolome composition. Other impacts on the metabolome
will result from changing geochemical parameters, for instance, pH and
oxygen levels that influence abiotic transformations such as
mineral-catalysed transformations (Zhang et al., 2021).
The abovementioned processes can rationalize most of the metab­
olome profile variations over time. However, there remain unexplained
changes. For instance, a disturbance in 09/2015 was observed in H52
(Fig. 3), which was also observed in the other H5 wells (Supplementary
Figs. S7 and S8, Table S1), but not in the shallow H32 (Fig. 2; infor­
mation on location H4 not available). This disturbance could neither be
explained by the sampling procedure (identical) nor did it have a clear
representation in other measured parameters. A possible explanation
might be that the corresponding pronounced recession period, followed
by an extreme lasting groundwater highstand after 2013, might have
influenced the prevalence of the specific metabolites.
3.3. Inter-well metabolome similarity suggests cross-compartment
exchange
Having established that water level fluctuations leave an imprint in
the metabolome over time, we queried whether the metabolome also
provides evidence for cross-compartment exchange along the hillslope
recharge area (Fig. 1). In addition to the metabolome data, we recorded
quality parameters like temperature, dissolved oxygen, various solutes
and DOC that have been previously used as environmental tracers to
explore local aquifer systems (Figs. S1 and S2) (Benischke, 2021; Jiang
et al., 2021). These parameters fall short in resolution as aggregate pa­
rameters like DOC are too general, with only weak spatiotemporal
variation. Also, major inorganic ions only allow interpretation at low
resolution compared to the numerous organic analytes accessed by
metabolomics techniques. Other tracers like fluorescent dyes or trace
metals need to be deliberately released, thereby already impacting the
environment. We therefore considered that untargeted LC-MS-based
metabolomics monitors a suitable naturally-occurring organic tracer
set for investigating cross-compartment exchange. While the extraction
of individual marker signals from such datasets would carry the risk of
extracting false positive correlations, we relied again on dimensionality
reduction by PCA to determine fingerprints that are more robust against
overfitting (Nguyen and Holmes, 2019).
We considered metabolome data of the seven monitoring wells H32,
H41 to − 43, and H51 to − 53, covering the period from 11/2014 to 06/
2020. For each timepoint, PCAs were generated to determine the prin­
cipal discriminating components for inter-well dissimilarity (see Mate­
rials and Methods). From these, the PCA distances between every two
wells were inferred and expressed as multiples of the distances’ standard
deviations (Fig. 4 and Supplementary Figs. S9–S13). This allowed us to
compare metabolome profiles acquired over the extended monitoring
period.
Well H52 was most distinct (i.e., had high scores for the
C. Zerfaß et al. Water Research 223 (2022) 118998

Fig. 3. Single-well (H52) metabolome time-series analysis. (A) Groundwater level and temperature traces for H52 (blue; in comparison with H51 (black) and
H53 (gray/yellow)). Recharge phases (gray-shaded) are inferred from the environmental tracer temperature (red) and multi-level head data. (B) Hierarchical cluster
analysis of z-normalised metabolite feature intensities. (C) Principal component analysis of time-series. Arrows are drawn as guides for the chronological order of
datapoints; the first datapoint of each year is labelled in blue font. (D) False discovery rate (FDR; from Benjamini-Hochberg corrected ANOVA) distribution as
indicator of significance of metabolite feature intensity variability in the time series.

7
C. Zerfaß et al.
dissimilarity) from the other wells (Fig. 4); however, phases of similarity
to well H53 emerged in 06–07/2015, 11/2016, 04/2018, 07/2019 and
04/2020. At all these timepoints, the H5 wells were at or near a peak of
groundwater highstand (Fig. 3 and Supplementary Fig. S3), suggesting
that ascending flow caused cross-stratal exchange and mixing. There
were also phases of similarity between H52 and the well H51 at times
around the highstands in 06/2015 and 09/2016, and an additional
phase in 03/2016 where H51 and H52 were approaching their
highstand.
Notably, in several phases when H52 and H53 were similar, the
metabolome of the deeper well H51 diverged, and had instead higher
similarity to the aeration zone well H32 (04/2018; 2.5 km further up­
slope, see Fig. 1) or the phreatic wells H41 to H43 (07/2019 and 04/
2020) than to the wells H52 and H53 (Supplementary Fig. S13). These
metabolome observations support earlier findings (Benk et al., 2019) of
transient local flow patterns that connect even deep and isolated com­
partments with the surface or shallow upslope zones via ascending flow
(Lehmann and Totsche, 2020).
In accordance, multivariate analyzes from aggregated, long-term
Hainich monitoring data on the microbiome (Yan et al., 2020), physi­
cochemical parameters (Kohlhepp et al., 2017; Schwab et al., 2017), and
phospholipid-derived fatty acids (PLFA) (Schwab et al., 2017) classed
the wells H52 and H53 as similar and distinct from the others, while a
previous metabolomics study (Benk et al., 2019) placed H52 in prox­
imity to H4 wells (1 km uphill, see Fig. 1), but separate from H53 (same
location, but 15 m above). The present analysis synergizes these findings
by clarifying the changes in well-dissimilarities over longer monitoring
periods.
The aeration zone well (H32), accessing a perched, oxic groundwater
body, showed variable similarities with all wells but H53, which was
consistently different. Phases of similarities to other wells were consis­
tently observed following to all high-flow phases in H32, as follows: 11/
2014–01/2015 (to H41–43, H51–52), 06/2015 (H41, H43, H51–52),
03–04/2016 (H42), 07/2017 (H41–43, H52), 04/2018 (H41, H51), 07/
2019 (H41, H43, H51) and 04–05/2020 (H41, H43, H51).
This demonstrates that PCA of metabolomics data from long term
surveys can be used to derive time-traces of dissimilarities by inferring
the sample distances in the PC1-PC2 space. Acknowledging the obstacles
8
Water Research 223 (2022) 118998
Fig. 4. Metabolome dissimilarities of wells H32
(left) and H52 (right) compared against other
wells included in the analysis (as shown in the
legend). Well distances are expressed as multi­
ples of the standard deviation σ in the PC1-PC2
space from principal component analysis (PCA)
at individual sampling campaigns (i.e. time
points). The distances were log2-transformed
for presentation, and the 1σ and 2σ threshold
are indicated (log2(1)=0 and log2(2)=1,
respectively). Note for statistical assessment
that for a normal-distributed random sample,
the 1σ and 2σ ranges mark the intervals
expectedly containing 68% and 95% of obser­
vations, respectively (Wang et al., 2015). Sam­
pling campaign dates are detailed in
Supplementary Table S2.
of long-term data acquisition, where metabolomics measurement pa­
rameters, such as operator, device calibration, separation column, and
method may have changed, the extraction of PCA-distances per time
point allows integration of long-term monitoring data. We can conclude
that the sedimentary-rock groundwaters vary in dissimilarity over time
across the hillslope recharge area. Phases of low dissimilarity emerge
coincident with groundwater recharge dynamics, covering surface-fed
as well as subsurface matter exchange.
Previously, it has been queried to which extent the surface vegeta­
tion or (agricultural) land use in a recharge area impacts underlying
groundwater systems (Küsel et al., 2016). Metabolome investigations of
soil waters sampled in lysimeters revealed discrete profiles for the input
metabolome in forest, pasture and agricultural areas (Ueberschaar et al.,
2021). For the groundwater wells in Fig. 1, we have demonstrated
elsewhere that the metabolomes of the wells studied on a single sam­
pling day were discernible in relation to the factors land-use type,
aquifer system and sampling depth (Sanchez-Arcos et al., 2022). We can
now show that for the long-term groundwater metabolomes (Fig. 4),
vertical transport, or the locations’ respective land use types on the
surface are not the major determinants. Instead, periods of similarity
occur between H32 and wells at the H4 and H5 sites with different land
use and different aquifer properties (Lazar et al., 2019; Lehmann and
Totsche, 2020). No continuous segregation by land-use type was
observed, implying that its impact on the well metabolomes is either
transient or secondary. Changing similarities are meanwhile mainly
driven by variable flow patterns and water stands. Notably, these may
become more fluctuating with weather and climatic extremes, both
which are expected to become more frequent under climate change
predictions (IPCC, 2022). In effect, it can be expected that groundwater
metabolic diversity may rise further thus posing unprecedented chal­
lenges to groundwater microbes and ecosystems.
4. Conclusion
We show that groundwater metabolomics can resolve the sum
parameter DOC with much detail and disclose hidden information on
groundwater ecosystems’ functioning and dynamics. We further
demonstrate that groundwater flows exert a major impact on the
C. Zerfaß et al.
groundwater metabolome variability, which will thus affect the carbon
and energy provision to belowground microbial communities. Such in­
formation is essential to discover ecosystem patterns and responses, such
as groundwater recharge events or climatic extremes that are expected
to increase in severity as predicted in climate change scenarios, and may
impact the discharge-recharge cycle. Generally, the groundwater
metabolome and its temporal dynamics provide cross-habitat informa­
tion from multiple sources of metabolites, covering contributions from
microbial populations, as well as abiotic transformations. This allows,
for example, the recognition of similarities caused by high-flow phases
in wells that are several kilometers apart. The metabolomics approach
has the advantage over tracer experiments that it determines natural
organic components of the groundwater without the need to introduce
markers. It closes a critical gap in the interdisciplinary investigation of
groundwater ecosystems.
Author contributions
All authors have approved the final version of the submitted article.
Data availability
Data acquired by liquid chromatography coupled to mass spec­
trometry is accessible via the Metabolights repository (Haug et al., 2019)
https://www.ebi.ac.uk/metabolights/MTBLS3450 (study identifier
MTBLS3450).
Funding
DFG SFB 1076, project number 218627073, “AquaDiva”. EXC 2051 –
Project-ID 390713860 “Balance of the Microverse”.
CRediT authorship contribution statement
Christian Zerfaß: Methodology, Investigation, Writing – review &
editing, Writing – original draft. Robert Lehmann: Methodology,
Investigation, Writing – review & editing, Writing – original draft. Nico
Ueberschaar: Methodology, Investigation, Writing – review & editing.
Carlos Sanchez-Arcos: Methodology, Investigation, Writing – review &
editing. Kai Uwe Totsche: Methodology, Investigation, Writing – re­
view & editing, Writing – original draft, Conceptualization. Georg
Pohnert: Methodology, Investigation, Writing – review & editing,
Writing – original draft, Conceptualization.
Declaration of Competing Interest
All authors declare they have no competing interests.
Data availability
We have shared a link to the data.
MTBLS3450 (Original data) (Metabolights).
Acknowledgments
This project was funded by the Collaborative Research center
AquaDiva of the Deutsche Forschungsgemeinschaft (DFG, German
Research foundation – SFB 1076 Project number 218627073 and EXC
2051 – Project-ID 390713860 “Balance of the Microverse”). We thank
Falko Guthmann and Heiko Minkmar for maintaining the Hainich CZE
infrastructure and conducting sampling, and Tim Baumeister and Ann-
Sophie Lehnert for support in sample processing.
9
Water Research 223 (2022) 118998
Supplementary materials
Supplementary material associated with this article can be found, in
the online version, at doi:10.1016/j.watres.2022.118998.
References
Antonelli, J., Claggett, B.L., Henglin, M., Kim, A., Ovsak, G., Kim, N., Deng, K., Rao, K.,
Tyagi, O., Watrous, J.D., Lagerborg, K.A., Hushcha, P.V., Demler, O.V., Mora, S.,
Niiranen, T.J., Pereira, A.C., Jain, M., Cheng, S., 2019. Statistical workflow for
feature selection in human metabolomics data. Metabolites 9. https://doi.org/
10.3390/metabo9070143.
Aukes, P.J.K., Schiff, S.L., 2021. Composition Wheels: visualizing dissolved organic
matter using common composition metrics across a variety of Canadian ecozones.
PLoS ONE 16, e0253972. https://doi.org/10.1371/journal.pone.0253972.
Bakalowicz, M., 1994. Water geochemistry: water quality and dynamics. Groundwater
Ecology. Elsevier, pp. 97–127. https://doi.org/10.1016/B978-0-08-050762-
0.50011-5.
Benischke, R., 2021. Review: advances in the methodology and application of tracing in
karst aquifers. Hydrogeol. J. 29, 67–88. https://doi.org/10.1007/s10040-020-
02278-9.
Benk, S.A., Li, Y., Roth, V.N., Gleixner, G., 2018. Lignin dimers as potential markers for
14C-young terrestrial dissolved organic matter in the critical zone. Front. Earth Sci.
6, 168. https://doi.org/10.3389/feart.2018.00168.
Benk, S.A., Yan, L., Lehmann, R., Roth, V.N., Schwab, V.F., Totsche, K.U., Küsel, K.,
Gleixner, G., 2019. Fueling diversity in the subsurface: composition and age of
dissolved organic matter in the critical zone. Front. Earth Sci. 7, 296. https://doi.
org/10.3389/feart.2019.00296.
Benton, H.P., Want, E.J., Ebbels, T.M.D., 2010. Correction of mass calibration gaps in
liquid chromatography-mass spectrometry metabolomics data. Bioinformatics 26,
2488–2489. https://doi.org/10.1093/bioinformatics/btq441.
Bonneau, J., Burns, M.J., Fletcher, T.D., Witt, R., Drysdale, R.N., Costelloe, J.F., 2018.
The impact of urbanization on subsurface flow paths – a paired-catchment isotopic
study. J. Hydrol. 561, 413–426. https://doi.org/10.1016/j.jhydrol.2018.04.022.
Bowen, G.J., Cai, Z., Fiorella, R.P., Putman, A.L., 2019. Isotopes in the water cycle:
regional- to global-scale patterns and applications. Annu. Rev. Earth Planet. Sci. 47,
453–479. https://doi.org/10.1146/annurev-earth-053018-060220.
Caretta, M., Mukherji, A., Arfanuzzaman, M., Betts, R., Gelfan, A., Hirabayashi, Y.,
Lissner, T., Liu, J., Lopez Gunn, E., Morgan, R., Mwanga, S., Supratid, S., et al., 2022.
Chapter 4 - Water. In: Pörtner, H.O., Roberts, D., Tignor, M., Poloczanska, E.,
Mintenbeck, K., Alegría, A., Craig, M., et al. (Eds.), Climate Change 2022: Impacts,
Adaptation, and Vulnerability. Contribution of Working Group II to the Sixth
Assessment Report of the Intergovernmental Panel on Climate Change. Cambridge
University Press. https://www.ipcc.ch/report/ar6/wg2/. In press.
Chambers, M.C., Maclean, B., Burke, R., Amodei, D., Ruderman, D.L., Neumann, S.,
Gatto, L., Fischer, B., Pratt, B., Egertson, J., Hoff, K., Kessner, D., Tasman, N.,
Shulman, N., Frewen, B., Baker, T.A., Brusniak, M.Y., Paulse, C., Creasy, D.,
Flashner, L., Kani, K., Moulding, C., Seymour, S.L., Nuwaysir, L.M., Lefebvre, B.,
Kuhlmann, F., Roark, J., Rainer, P., Detlev, S., Hemenway, T., Huhmer, A.,
Langridge, J., Connolly, B., Chadick, T., Holly, K., Eckels, J., Deutsch, E.W.,
Moritz, R.L., Katz, J.E., Agus, D.B., MacCoss, M., Tabb, D.L., Mallick, P., 2012.
A cross-platform toolkit for mass spectrometry and proteomics. Nat. Biotechnol. 30,
918–920. https://doi.org/10.1038/nbt.2377.
Chorover, J., Amistadi, M.K., 2001. Reaction of forest floor organic matter at goethite,
birnessite and smectite surfaces. Geochim. Cosmochim. Acta 65, 95–109. https://
doi.org/10.1016/S0016-7037(00)00511-1.
Cooksey, R.W., 2020. Descriptive statistics for summarising data. Illustrating Statistical
Procedures: Finding Meaning in Quantitative Data. Springer Singapore, Singapore,
pp. 61–139. https://doi.org/10.1007/978-981-15-2537-7_5.
Crimmins, B.S., Holsen, T.M., 2019. Non-targeted screening in environmental monitoring
programs. In: Woods, A.G., Darie, C.C. (Eds.), Advancements of Mass Spectrometry
in Biomedical Research. Springer, Cham, pp. 731–741. https://doi.org/10.1007/
978-3-030-15950-4_43.
Danczak, R.E., Chu, R.K., Fansler, S.J., Goldman, A.E., Graham, E.B., Tfaily, M.M.,
Toyoda, J., Stegen, J.C., 2020. Using metacommunity ecology to understand
environmental metabolomes. Nat. Commun. 11, 6369. https://doi.org/10.1038/
s41467-020-19989-y.
Drake, T.W., Podgorski, D.C., Dinga, B., Chanton, J.P., Six, J., Spencer, R.G.M., 2020.
Land-use controls on carbon biogeochemistry in lowland streams of the Congo Basin.
Glob. Change Biol. 26, 1374–1389. https://doi.org/10.1111/gcb.14889.
Fernandes, J.P., Almeida, C.M.R., Salgado, M.A., Carvalho, M.F., Mucha, A.P., 2021.
Pharmaceutical compounds in aquatic environments-occurrence, fate and
bioremediation prospective. Toxics 9. https://doi.org/10.3390/toxics9100257.
Franco, D., Coulibaly, K., Kunberger, T., Kinzli, K.-.D., Arbelaez, S., Missimer, T.M.,
2017. Evolution of heterogeneous mixed siliciclastic/carbonate aquifers containing
metastable sediments. Groundwater 55, 784–796. https://doi.org/10.1111/
gwat.12529.
Garayburu-Caruso, V.A., Danczak, R.E., Stegen, J.C., Renteria, L., Mccall, M.,
Goldman, A.E., Chu, R.K., Toyoda, J., Resch, C.T., Torgeson, J.M., Wells, J.,
Fansler, S., Kumar, S., Graham, E.B., 2020. Using community science to reveal the
global chemogeography of river metabolomes. Metabolites 10. https://doi.org/
10.3390/metabo10120518.
Grant, G.E., Dietrich, W.E., 2017. The frontier beneath our feet. Water Resour. Res. 53,
2605–2609. https://doi.org/10.1002/2017WR020835.
C. Zerfaß et al.
Haug, K., Cochrane, K., Nainala, V.C., Williams, M., Chang, J., Jayaseelan, K.V.,
O’Donovan, C., 2019. MetaboLights: a resource evolving in response to the needs of
its scientific community. Nucleic Acids Res. 48, D440–D444. https://doi.org/
10.1093/nar/gkz1019.
Hawkes, J.A., D’Andrilli, J., Agar, J.N., Barrow, M.P., Berg, S.M., Catalán, N., Chen, H.,
Chu, R.K., Cole, R.B., Dittmar, T., Gavard, R., Gleixner, G., Hatcher, P.G., He, C.,
Hess, N.J., Hutchins, R.H.S., Ijaz, A., Jones, H.E., Kew, W., Khaksari, M., Palacio
Lozano, D.C., Lv, J., Mazzoleni, L.R., Noriega-Ortega, B.E., Osterholz, H.,
Radoman, N., Remucal, C.K., Schmitt, N.D., Schum, S.K., Shi, Q., Simon, C.,
Singer, G., Sleighter, R.L., Stubbins, A., Thomas, M.J., Tolic, N., Zhang, S., Zito, P.,
Podgorski, D.C., 2020. An international laboratory comparison of dissolved organic
matter composition by high resolution mass spectrometry: are we getting the same
answer? Limnol. Oceanogr. Methods 18, 235–258. https://doi.org/10.1002/
lom3.10364.
Hofmann, R., Uhl, J., Hertkorn, N., Griebler, C., 2020. Linkage between dissolved organic
matter transformation, bacterial carbon production, and diversity in a shallow
oligotrophic aquifer: results from flow-through sediment microcosm experiments.
Front. Microbiol. 11, 543567 https://doi.org/10.3389/fmicb.2020.543567.
Huber, W., Carey, V.J., Gentleman, R., Anders, S., Carlson, M., Carvalho, B.S., Bravo, H.
C., Davis, S., Gatto, L., Girke, T., Gottardo, R., Hahne, F., Hansen, K.D., Irizarry, R.A.,
Lawrence, M., Love, M.I., MacDonald, J., Obenchain, V., Oleś, A.K., Pagès, H.,
Reyes, A., Shannon, P., Smyth, G.K., Tenenbaum, D., Waldron, L., Morgan, M., 2015.
Orchestrating high-throughput genomic analysis with Bioconductor. Nat. Methods
12, 115–121. https://doi.org/10.1038/nmeth.3252.
Hudak, P.F., 2004. Principles of Hydrogeology. CRC Press. https://doi.org/10.1201/
9781420057911.
Humphrey, V., Gudmundsson, L., Seneviratne, S.I., 2016. Assessing global water storage
variability from GRACE: trends, seasonal cycle, subseasonal anomalies and extremes.
Surv. Geophys. 37, 357–395. https://doi.org/10.1007/s10712-016-9367-1.
IPCC Report. In: Pörtner, H.O., Roberts, D., Tignor, M., Poloczanska, E., Mintenbeck, K.,
Alegría, A., et al. (Eds.), 2022. Climate Change 2022: Impacts, Adaptation, and
Vulnerability. Contribution of Working Group II to the Sixth Assessment Report of
the Intergovernmental Panel On Climate Change. Cambridge University Press
https://www.ipcc.ch/report/ar6/wg2/. In press.
Ivanisevic, J., Want, E.J., 2019. From samples to insights into metabolism: uncovering
biologically relevant information in LC-HRMS metabolomics data. Metabolites 9,
308. https://doi.org/10.3390/metabo9120308.
Jasechko, S., Birks, S.J., Gleeson, T., Wada, Y., Fawcett, P.J., Sharp, Z.D., McDonnell, J.
J., Welker, J.M., 2014. The pronounced seasonality of global groundwater recharge.
Water Resour. Res. 50, 8845–8867. https://doi.org/10.1002/2014WR015809.
Jiang, S., Ibánhez, J.S.P., Wu, Y., Zhang, J., 2021. Geochemical tracers in submarine
groundwater discharge research: practice and challenges from a view of climate
changes. Environ. Rev. 29, 242–259. https://doi.org/10.1139/er-2020-0093.
Kaiser, K., Kalbitz, K., 2012. Cycling downwards – dissolved organic matter in soils. Soil
Biol. Biochem. 52, 29–32. https://doi.org/10.1016/j.soilbio.2012.04.002.
Kohlhepp, B., Lehmann, R., Seeber, P., Küsel, K., Trumbore, S.E., Totsche, K.U., 2017.
Aquifer configuration and geostructural links control the groundwater quality in
thin-bedded carbonate–siliciclastic alternations of the Hainich CZE, central
Germany. Hydrol. Earth Syst. Sci. 21, 6091–6116. https://doi.org/10.5194/hess-21-
6091-2017.
Küsel, K., Totsche, K.U., Trumbore, S.E., Lehmann, R., Steinhäuser, C., Herrmann, M.,
2016. How deep can surface signals be traced in the critical zone? Merging
biodiversity with biogeochemistry research in a central german muschelkalk
landscape. Front. Earth Sci. 4 https://doi.org/10.3389/feart.2016.00032.
Lazar, C.S., Lehmann, R., Stoll, W., Rosenberger, J., Totsche, K.U., Küsel, K., 2019. The
endolithic bacterial diversity of shallow bedrock ecosystems. Sci. Total Environ. 679,
35–44. https://doi.org/10.1016/j.scitotenv.2019.04.281.
Lehmann, J., Kleber, M., 2015. The contentious nature of soil organic matter. Nature
528, 60–68. https://doi.org/10.1038/nature16069.
Lehmann, R., Totsche, K.U., 2020. Multi-directional flow dynamics shape groundwater
quality in sloping bedrock strata. J. Hydrol. 580, 124291 https://doi.org/10.1016/j.
jhydrol.2019.124291.
Li, L., Maher, K., Navarre-Sitchler, A., Druhan, J., Meile, C., Lawrence, C., Moore, J.,
Perdrial, J., Sullivan, P., Thompson, A., Jin, L., Bolton, E.W., Brantley, S.L.,
Dietrich, W.E., Mayer, K.U., Steefel, C.I., Valocchi, A., Zachara, J., Kocar, B.,
Mcintosh, J., Tutolo, B.M., Kumar, M., Sonnenthal, E., Bao, C., Beisman, J., 2017.
Expanding the role of reactive transport models in critical zone processes. Earth Sci.
Rev. 165, 280–301. https://doi.org/10.1016/j.earscirev.2016.09.001.
Lu, Q., He, D., Pang, Y., Zhang, Y., He, C., Wang, Y., Zhang, H., Shi, Q., Sun, Y., 2020.
Processing of dissolved organic matter from surface waters to sediment pore waters
in a temperate coastal wetland. Sci. Total Environ. 742, 140491 https://doi.org/
10.1016/j.scitotenv.2020.140491.
Lynch, L.M., Sutfin, N.A., Fegel, T.S., Boot, C.M., Covino, T.P., Wallenstein, M.D., 2019.
River channel connectivity shifts metabolite composition and dissolved organic
matter chemistry. Nat. Commun. 10, 459. https://doi.org/10.1038/s41467-019-
08406-8.
Miller, M.J., Kennedy, A.D., Eckhart, A.D., Burrage, L.C., Wulff, J.E., Miller, L.A.D.,
Milburn, M.V., Ryals, J.A., Beaudet, A.L., Sun, Q., Sutton, V.R., Elsea, S.H., 2015.
Untargeted metabolomic analysis for the clinical screening of inborn errors of
metabolism. J. Inherit. Metab. Dis. 38, 1029–1039. https://doi.org/10.1007/
s10545-015-9843-7.
Nguyen, L.H., Holmes, S., 2019. Ten quick tips for effective dimensionality reduction.
PLoS Comput. Biol. 15, e1006907 https://doi.org/10.1371/journal.pcbi.1006907.
Patriarca, C., Hawkes, J.A., 2021. High molecular weight spectral interferences in mass
spectra of dissolved organic matter. J. Am. Soc. Mass Spectrom. 32, 394–397.
https://doi.org/10.1021/jasms.0c00353.
10
Water Research 223 (2022) 118998
Pemberton, J.A., Lloyd, C.E.M., Arthur, C.J., Johnes, P.J., Dickinson, M., Charlton, A.J.,
Evershed, R.P., 2020. Untargeted characterisation of dissolved organic matter
contributions to rivers from anthropogenic point sources using direct-infusion and
high-performance liquid chromatography/Orbitrap mass spectrometry. Rap.
Commun. Mass Spectrom. 34 (Suppl 4), e8618. https://doi.org/10.1002/rcm.8618.
Phillips, T., Nerem, R.S., Fox-Kemper, B., Famiglietti, J.S., Rajagopalan, B., 2012. The
influence of ENSO on global terrestrial water storage using GRACE. Geophys. Res.
Lett. 39, L16705. https://doi.org/10.1029/2012GL052495.
R-CoreTeam, 2017. R: a Language and Environment for Statistical Computing. https://
www.r-project.org/.
Raeke, J., Lechtenfeld, O.J., Tittel, J., Oosterwoud, M.R., Bornmann, K., Reemtsma, T.,
2017. Linking the mobilization of dissolved organic matter in catchments and its
removal in drinking water treatment to its molecular characteristics. Water Res. 113,
149–159. https://doi.org/10.1016/j.watres.2017.01.066.
Richardson, J.B., 2017. Critical Zone (Ed.). In: White, W.M. (Ed.), Encyclopedia of
Geochemistry. Encyclopedia of Earth Sciences Series. Springer, Cham. https://doi.
org/10.1007/978-3-319-39193-9_355-1.
Riebe, C.S., Hahm, W.J., Brantley, S.L., 2017. Controls on deep critical zone architecture:
a historical review and four testable hypotheses. Earth Surf. Process. Landforms 42,
128–156. https://doi.org/10.1002/esp.4052.
Riedel, T., Weber, T.K.D., 2020. Review: the influence of global change on Europe’s
water cycle and groundwater recharge. Hydrogeol. J. 28, 1939–1959. https://doi.
org/10.1007/s10040-020-02165-3.
Rodell, M., Famiglietti, J.S., Wiese, D.N., Reager, J.T., Beaudoing, H.K., Landerer, F.W.,
Lo, M.H., 2018. Emerging trends in global freshwater availability. Nature 557,
651–659. https://doi.org/10.1038/s41586-018-0123-1.
Rodríguez-Cardona, B.M., Wymore, A.S., Argerich, A., Barnes, R.T., Bernal, S.,
Brookshire, E.N.J., Coble, A.A., Dodds, W.K., Fazekas, H.M., Helton, A.M., Johnes, P.
J., Johnson, S.L., Jones, J.B., Kaushal, S.S., Kortelainen, P., López-Lloreda, C.,
Spencer, R.G.M., McDowell, W.H., 2022. Shifting stoichiometry: long-term trends in
stream-dissolved organic matter reveal altered C:N ratios due to history of
atmospheric acid deposition. Glob. Chang. Biol. 28, 98–114. https://doi.org/
10.1111/gcb.15965.
Roth, V.N., Lange, M., Simon, C., Hertkorn, N., Bucher, S., Goodall, T., Griffiths, R.I.,
Mellado-Vázquez, P.G., Mommer, L., Oram, N.J., Weigelt, A., Dittmar, T.,
Gleixner, G., Griffiths, I.R., Mellado-Vazquez, P.G., Mommer, L., Oram, N.J.,
Weigelt, A., Dittmar, T., Gleixner, G., 2019. Persistence of dissolved organic matter
explained by molecular changes during its passage through soil. Nat. Geosci. 12,
755–761. https://doi.org/10.1038/s41561-019-0417-4.
Sanchez-Arcos, C., Ueberschaar, N., Pohnert, G., 2022. Aquifer system and depth specific
chemical patterns in fractured-rock groundwater from the Critical Zone revealed by
untargeted LC-MS-based metabolomics. Water Res. 219, 118566 https://doi.org/
10.1016/j.watres.2022.118566.
Schmidt, M.W.I., Torn, M.S., Abiven, S., Dittmar, T., Guggenberger, G., Janssens, I.A.,
Kleber, M., Kögel-Knabner, I., Lehmann, J., Manning, D.A.C., Nannipieri, P.,
Rasse, D.P., Weiner, S., Trumbore, S.E., 2011. Persistence of soil organic matter as an
ecosystem property. Nature 478, 49–56. https://doi.org/10.1038/nature10386.
Schwab, V.F., Herrmann, M., Roth, V.-.N., Gleixner, G., Lehmann, R., Pohnert, G.,
Trumbore, S., Küsel, K., Totsche, K.U., Kuesel, K., Totsche, K.U., 2017. Functional
diversity of microbial communities in pristine aquifers inferred by PLFA- and
sequencing-based approaches. Biogeosciences 14, 2697–2714. https://doi.org/
10.5194/bg-14-2697-2017.
Seifert, A.G., Roth, V.N., Dittmar, T., Gleixner, G., Breuer, L., Houska, T., Marxsen, J.J.,
2016. Comparing molecular composition of dissolved organic matter in soil and
stream water: influence of land use and chemical characteristics. Sci. Total Environ.
571, 142–152. https://doi.org/10.1016/j.scitotenv.2016.07.033.
Smith, C.A., Want, E.J., O’Maille, G., Abagyan, R., Siuzdak, G., 2006. XCMS: processing
mass spectrometry data for metabolite profiling using nonlinear peak alignment,
matching, and identification. Anal. Chem. 78, 779–787. https://doi.org/10.1021/
ac051437y.
Tanentzap, A.J., Fitch, A., Orland, C., Emilson, E.J.S., Yakimovich, K.M., Osterholz, H.,
Dittmar, T., 2019. Chemical and microbial diversity covary in fresh water to
influence ecosystem functioning. Proc. Natl. Acad. Sci. U. S. A. 116, 24689–24695.
https://doi.org/10.1073/pnas.1904896116.
Tautenhahn, R., Böttcher, C., Neumann, S., 2008. Highly sensitive feature detection for
high resolution LC/MS. BMC Bioinform. 9, 504. https://doi.org/10.1186/1471-
2105-9-504.
Tweed, S.O., Weaver, T.R., Cartwright, I., 2005. Distinguishing groundwater flow paths
in different fractured-rock aquifers using groundwater chemistry: dandenong
Ranges, southeast Australia. Hydrogeol. J. 13, 771–786. https://doi.org/10.1007/
s10040-004-0348-y.
Ueberschaar, N., Lehmann, K., Meyer, S., Zerfass, C., Michalzik, B., Totsche, K.U.,
Pohnert, G., 2021. Soil solution analysis with untargeted GC–MS—a case study with
different lysimeter types. Front. Earth Sci. 8 https://doi.org/10.3389/
feart.2020.563379.
United Nations Office for Disaster Risk Reduction (UNDRR), 2021. GAR Special Report
on Drought 2021. Geneva. https://www.undrr.org/publication/gar-special-report-
drought-2021.
Wada, Y., van Beek, L.P.H., Bierkens, M.F.P., 2012. Nonsustainable groundwater
sustaining irrigation: a global assessment. Water Resour. Res. 48 https://doi.org/
10.1029/2011WR010562.
Wang, B., Shi, W., Miao, Z., 2015. Confidence analysis of standard deviational ellipse and
its extension into higher dimensional euclidean space. PLoS ONE 10, e0118537.
https://doi.org/10.1371/journal.pone.0118537.
White, W.B., 2012. Conceptual models for carbonate aquifers (1969, reprint). Ground
Water 50, 180–186. https://doi.org/10.1111/j.1745-6584.2012.00923.x.
C. Zerfaß et al.
Wologo, E., Shakil, S., Zolkos, S., Textor, S., Ewing, S., Klassen, J., Spencer, R.G.M.,
Podgorski, D.C., Tank, S.E., Baker, M.A., O’Donnell, J.A., Wickland, K.P., Foks, S.S.
W., Zarnetske, J.P., Lee-Cullin, J., Liu, F., Yang, Y., Kortelainen, P., Kolehmainen, J.,
Dean, J.F., Vonk, J.E., Holmes, R.M., Pinay, G., Powell, M.M., Howe, J., Frei, R.J.,
Bratsman, S.P., Abbott, B.W., 2021. Stream dissolved organic matter in permafrost
regions shows surprising compositional similarities but negative priming and
nutrient effects. Global Biogeochem. Cycles 35, e2020GB006719. https://doi.org/
10.1029/2020GB006719.
Worley, B., Powers, R., 2012. Multivariate analysis in metabolomics. Curr. Metabolomics
1, 92–107. https://doi.org/10.2174/2213235X11301010092.
Yan, L., Hermans, S.M., Totsche, K.U., Lehmann, R., Herrmann, M., Küsel, K., 2021.
Groundwater bacterial communities evolve over time in response to recharge. Water
Res. 117290 https://doi.org/10.1016/j.watres.2021.117290.
11
Water Research 223 (2022) 118998
Yan, L., Herrmann, M., Kampe, B., Lehmann, R., Totsche, K.U., Küsel, K., 2020.
Environmental selection shapes the formation of near-surface groundwater
microbiomes. Water Res. 170, 115341 https://doi.org/10.1016/j.
watres.2019.115341.
Yuill, R.S., 1971. The standard deviational ellipse; an updated tool for spatial description.
Geogr. Ann. Ser. B 53, 28. https://doi.org/10.2307/490885.
Zerfaß, C., Lehmann, R., Ueberschaar, N., Sanchez-Arcos, C., Totsche, K.U., Pohnert, G.,
2021. MTBLS3450. MetaboLights Repository https://www.ebi.ac.uk/metabolights/
MTBLS3450.
Zhang, J., McKenna, A.M., Zhu, M., 2021. Macromolecular characterization of compound
selectivity for oxidation and oxidative alterations of dissolved organic matter by
manganese oxide. Environ. Sci. Technol. 55, 7741–7751. https://doi.org/10.1021/
acs.est.1c01283.