See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/348110575

Triple oxygen isotopes in the water cycle

Article in Chemical Geology · December 2020

DOI: 10.1016/j.chemgeo.2020.120026

CITATIONS READS

93 1,174

9 authors, including:

Emily J Beverly Tyler Huth

University of Houston Washington University in St. Louis
69 PUBLICATIONS 1,375 CITATIONS 23 PUBLICATIONS 358 CITATIONS

SEE PROFILE SEE PROFILE

Christopher J Poulsen Ian Winkelstern

University of Oregon Grand Valley State University
221 PUBLICATIONS 11,555 CITATIONS 39 PUBLICATIONS 797 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Emily J Beverly on 03 March 2021.

The user has requested enhancement of the downloaded file.

 Chemical Geology 565 (2021) 120026

Contents lists available at ScienceDirect

Chemical Geology
journal homepage: www.elsevier.com/locate/chemgeo

Invited review article

Triple oxygen isotopes in the water cycle

Phoebe G. Aron a, *, Naomi E. Levin a, Emily J. Beverly b, Tyler E. Huth a, Benjamin H. Passey a,
Elise M. Pelletier a, Christopher J. Poulsen a, Ian Z. Winkelstern c, Drake A. Yarian a
a
Department of Earth and Environmental Sciences, University of Michigan, Ann Arbor, MI 48109, United States
b
Department of Earth and Atmospheric Sciences, University of Houston, Houston, TX, 77204, United States
c
Department of Geology, Grand Valley State University, Allendale, MI, 49401, United States

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

Editor: Michael E. Boettcher The past decade has seen a remarkable expansion of studies that use mass-dependent variations of triple oxygen
isotopes (16O, 17O, 18O) in isotope hydrology and isotope geochemistry. Recent technological and analytical
Keywords: advances demonstrate that small deviations of δ′ 18O and δ′ 17O from a mass-dependent reference relationship are
Triple oxygen isotopes systematic and are explained by well-known equilibrium and kinetic fractionations. Measurements of δ′ 18O and
Meteoric water
δ′ 17O complement traditional metrics like deuterium-excess, constrain isotope effects of kinetic fractionation that
Mass-dependent fractionation
are impossible to discern with δ18O alone, and help reconstruct past environmental conditions from geologic
records. In this review, we synthesize published meteoric (derived from precipitation) water triple oxygen
isotope data with a new, near-global surface water dataset of δ′ 18O, δ′ 17O, δ2H, deuterium-excess, and ∆′ 17O,
where ∆′ 17O is defined as δ′ 17O – λref δ′ 18O, δ′ notation is a logarithmic definition of the common δ value (δ′ =ln(δ
+ 1), and λref is equal to 0.528. The expanded dataset shows that meteoric water δ′ 18O and δ′ 17O fit multiple
regression lines and indicates that one global meteoric water line does not adequately describe all triple oxygen
isotope data. Instead, this isotope system may be sensitive to processes such as moisture transport, rainout, and
evaporation that do not affect the water cycle equally across the globe. This review provides a practical guide to
understand ∆′ 17O variation in waters, explains the utility of this isotope system in hydrologic and paleoclimate
studies, and outlines directions of future work that will expand the use of ∆′ 17O.

1. Introduction synthesizes new and published meteoric (derived from precipitation)

water isotope data and explains the processes that drive triple oxygen
Ratios of 18O to 16O are among the most common isotopic mea­ isotope variation in the water cycle. This review is intended as an
surements in Earth science and play a critical role tracing biogeo­ introduction to triple oxygen isotope hydrology and as a primer to how
chemical cycles and reconstructing past climate conditions (Dansgaard, this emerging field may contribute to hydrologic and paleoclimate
1964; Joussaume et al., 1984; Zachos, 2001). Studies of 17O, the rarest research.
stable oxygen isotope (Table 1), have lagged because 17O/16O ratios Oxygen isotopes fractionate due to non-mass-dependent and mass-
were long considered invariant, too difficult to measure, or redundant to dependent effects (Bao et al., 2016; Thiemens et al., 2012). Non-mass-
18
O/16O ratios (Gat, 1996). However, recent technological and analyt­ dependent fractionation arises from chemical effects, including nu­
ical advances show that small, mass-dependent deviations between clear spin, transition state chemistry, molecular symmetry, and photo­
17
O/16O and 18O/16O contain new information about water cycling and chemical reactions (Criss and Farquhar, 2008; Thiemens and
past environmental conditions (Barkan and Luz, 2005). This review Heidenreich, 1983). These effects can result in large variations between
captures the emerging field of triple oxygen isotope (16O, 17O, 18O) 17
O/16O and 18O/16O and have a range of applications in atmospheric
hydrology at an important moment: many laboratories are now able to chemistry, planetary science, and biological productivity that are
make 17O measurements and large datasets are rapidly emerging, but already well reviewed (Bao et al., 2009; Bhattacharya et al., 2000;
triple oxygen isotope variability is not yet fully understood and there are Blunier et al., 2012; Blunier et al., 2002; Luz et al., 2009; Thiemens,
important inconsistencies between studies. Therefore, this review 2006; Thiemens et al., 1995). In contrast, mass-dependent effects arise

* Corresponding author.
E-mail address: [email protected] (P.G. Aron).

https://doi.org/10.1016/j.chemgeo.2020.120026
Received 27 June 2020; Received in revised form 27 October 2020; Accepted 10 December 2020
Available online 31 December 2020
0009-2541/© 2020 Elsevier B.V. All rights reserved.
P.G. Aron et al. Chemical Geology 565 (2021) 120026

Table 1 Clearly the potential applications of triple oxygen isotopes in hy­

The three stable oxygen isotopes. drology and paleoclimate are immense. However, this field is still rela­
Isotope Symbol Protons Neutrons Mass (u) Natural terrestrial tively young, global variations are not yet well characterized, and
abundance (%) explanations of triple oxygen isotope variability are scattered among the
Oxygen-16 16
O 8 8 15.995 99.757 literature (Table 2) or are not well understood. The seminal description
Oxygen-17 17
O 8 9 16.999 0.038 of modern meteoric water triple oxygen isotopes (Luz and Barkan, 2010)
Oxygen-18 18
O 8 10 17.999 0.205 laid the foundation for a decade of research (Fig. 2 and Table 2), but
there is now far more variability than initially realized (Fig. 3a and b)
and triple oxygen isotope ratios do not neatly fit a global meteoric water
from differences in bond energy, reaction rate, and diffusivity (Matsu­
line (Sharp et al., 2018). Therefore, building upon more than a decade of
hisa et al., 1978; Young et al., 2002) that result in very small variations
work, this review re-examines meteoric water δ′ 18O, δ′ 17O, δ2H, ∆′ 17O,
between 17O/16O and 18O/16O. These mass-dependent variations are
and d-excess data. We synthesize published and new meteoric water
sensitive to fractionation during equilibrium isotope exchange and the
data to evaluate the δ′ 18O–δ′ 17O relationship (Section 5), explain the
diffusion of water vapor through air, and have recently gained attention
hydrologic processes and mass-dependent fractionations that drive
as a new way to study modern hydrology and reconstruct past envi­
variation in ∆′ 17O and d-excess (Section 6), review analytical methods
ronmental conditions (e.g., Bao et al., 2016; Barkan and Luz, 2005; Luz
and considerations for ∆′ 17O measurements (Section 7), and present
and Barkan, 2010; Pack and Herwartz, 2014; Rumble et al., 2007).
directions of future triple oxygen isotope work (Section 9).
Decoupling fractionation effects from equilibrium and kinetic pro­
cesses is critical to interpreting isotope data and identifying processes
such as Rayleigh distillation and evaporation. In modern waters, the 2. Isotope terminology and fractionation
degree of kinetic fractionation is often quantified using deuterium-
excess (d-excess = δ2H – 8*δ18O, Dansgaard, 1964; see Section 2.1 for A summary of common symbols, explanations, and values in hy­
the definition of δ notation). However, d-excess varies with both tem­ drologic triple oxygen isotope studies is provided in Table 3.
perature and relative humidity, so interpretations of d-excess data are
not always straightforward (Gat, 1996). Mass-dependent variations be­ 2.1. Isotope notation
tween 17O/16O and 18O/16O can also quantify kinetic fractionation, but
are relatively insensitive to temperature (Barkan and Luz, 2005). Isotope partitioning between two substances (A and B) is expressed
Therefore, triple oxygen isotopes and d-excess provide complementary as an isotopic fractionation factor, α:
information to track evapotranspiration, moisture transport, and pre­
cipitation processes (e.g., Galewsky et al., 2016), identify temperature αA− B = RA /RB (1)
and relative humidity conditions at moisture sources (Landais et al.,
where R is the ratio of the rare to common isotope (e.g., 2H/1H, 18O/16O,
2008, 2012a, 2012b; Uemura et al., 2010; Winkler et al., 2012), and link
or 17O/16O). During a single mass-dependent fractionating process, α
seasonal or glacial-interglacial isotope cycles to climate conditions (Risi
values of coexisting phases (e.g., liquid and vapor), materials (e.g.,
et al., 2010; Schoenemann and Steig, 2016).
water and mineral), or components related by simple processes (e.g.,
Translating these principles to the past and differentiating equilib­
diffused gas and residual gas) are related by a power law relationship
rium and kinetic fractionation effects in paleoclimate records is chal­
derived from mass law theory (Matsuhisa et al., 1978; Young et al.,
lenging because most geologic archives (e.g., carbonates, sulfates,
2002):
phosphates, etc.) do not have both oxygen- and hydrogen-containing
minerals. Much like d-excess (Fig. 1), triple oxygen isotopes add a de­ *
αA− B = (* αA− B )θ (2)
gree of freedom (17O/16O) to paleoclimate records and can clarify some
processes and fractionations that cannot be resolved with traditional where θ, the fractionation exponent, is a constant that defines the
oxygen isotope ratios (18O/16O) alone (e.g., Rech et al., 2019). For relationship between α values and * denotes a heavy mass number (e.g.,
example, triple oxygen isotope ratios preserved in minerals add new 17 or 18 for oxygen, 2 for hydrogen).
information about aridity and paleo-humidity (Alexandre et al., 2019; Triple oxygen isotope fractionation exponents are well characterized
Gázquez et al., 2018; Passey and Ji, 2019; Surma et al., 2018), enable for equilibrium (θeq) and kinetic (θdiff) processes. Liquid-vapor θeq is
reconstructions of the isotopic composition of ancient waters (e.g., predicted by mass law theory (Young et al., 2002, θeq = 0.529) and has
Gehler et al., 2011; Herwartz et al., 2015; Liljestrand et al., 2020; Passey been verified empirically (Barkan and Luz, 2005, θeq = 0.529 ± 0.001);
and Ji, 2019), and constrain effects of diagenesis and formation condi­ the kinetic fractionation exponent for diffusion of water vapor through
tions of sedimentary records (Levin et al., 2014). air (θdiff) is derived from the kinetic theory of gases and the ideal gas law

Fig. 1. Schematic showing the similarities between (A) d-excess and (B) Δ′ 17O. Note that Δ′ 17O is defined from δ′ 18O and δ′ 17O. See Eq. (4) for the definition of
δ′ notation

2
P.G. Aron et al. Chemical Geology 565 (2021) 120026

Table 2
Summary of published hydrologic triple oxygen water isotope studies.
Water type Timeframe Location Analysis Reference
method

Plant Water
Leaf April 2015 Europe and IRMS Landais et al.,
Israel 2006
Stem and leaf Summer 2012 Central IRMS Li et al., 2017
Kenya
Meteoric Water
Various Various, 2002-2010 global IRMS Luz and
Barkan, 2010
Precipitation Seasonal, short convective cell Niger IRMS Landais et al.,
2010
Surface water Spring 2011 Iran IRMS Surma et al.,
2015
Tap water 2008-2011 Continental IRMS Li et al., 2015
United
States
Precipitation, cave drip March 2012-July 2014 Switzerland Picarro Affolter et al.,
2015
Precipitation, surface water, cave drip 2012-2013 Spain Picarro Gázquez
et al., 2017
Surface water March 2014 Atacama IRMS Surma et al.,
Desert, 2018
Chile
Precipitation Event scale, 2014-2018 Central LGR Tian et al.,
United 2018, Tian
States et al., 2019
Precipitation Event scale, 2012-2016 Namibia LGR Kaseke et al.,
2018
Tap water Monthly, December 2014 to November 2015 China LGR Tian et al.,
2019
Precipitation Weekly, January 2011 to December 2012 Japan Picarro Uechi and
Uemura,
2019
Surface water June 2014 Western IRMS Passey and Ji,
United 2019
States
Surface water 2017-2018 Northern IRMS Bergel et al.,
Israel 2020
Surface water and ocean 2015-2016 Pacific Picarro Bershaw
Northwest, et al., 2020
United
States
Surface water Various, 2016-2019 global IRMS this study
Polar Precipitation
Snow and ice Last 150,000 years Vostok, IRMS Landais et al.,
Antarctica 2008
Snow 2000 Vostok, IRMS Landais et al.,
Antarctica 2012a
Snow and water vapor 2003-2005 Greenland IRMS Landais et al.,
2012b
Ice Glacial-interglacial cycles East IRMS Winkler et al.,
Antarctica 2012
Ice LGM to Holocene WAIS IRMS Schoenemann
Divide, et al., 2014
Antarctica
Snow December 2009-January 2010 Coast to IRMS Pang et al.,
Dome A 2015
transect,
East
Antarctica
Snow 1999-2011 East IRMS Touzeau
Antarctica et al., 2016
Snow January 2010 East IRMS Pang et al.,
Antarctica 2019
Modeling
Vapor Glacial-interglacial cycles Vostok, Single column Risi et al.,
Antarctica model 2010
Precipitation Modern and LGM global LMDZ Risi et al.,
(atmospheric 2013
transport
GCM)
Precipitation Modern seasonal cycle Antarctica Intermediate Schoenemann
complexity and Steig,
model 2016

Analysis method abbreviations: isotope ratio mass spectrometer (IRMS), Los Gatos Research (LGR), Laboratory of Dynamic Meteorology (LMDZ), general circulation
model (GCM).

3
P.G. Aron et al. Chemical Geology 565 (2021) 120026

Fig. 2. Geographical distribution of published meteoric water triple oxygen isotope data, colored by sample type. Plant water includes water extracted from stems
and leaves. Surface and subsurface includes surface water, soil water, groundwater, cave water, and tap water. New surface water samples reported in this review are
outlined in gold; published studies are listed in Table 2.

Fig. 3. Scatterplots of meteoric water isotope data. Color indicates sample type. New surface water data reported in this study are outlined in gold. The top row
shows plots of δ′ 18O versus Δ′ 17O from (A) Luz and Barkan (2010), (B) all published data as of 2020 (Table 2), and (C) published precipitation, surface, and sub­
surface waters within a normal meteoric range (–25 to 10‰) as of 2020. The dashed box in (B) outlines the limits of the data in (C). The bottom row shows plots of
(D) δ18O versus d-excess and (E) d-excess versus Δ′ 17O from all the available published meteoric water triple oxygen isotope data (Table 2). There are more points in
(B) than (D) or (E) because not all studies include δ2H data.

4
P.G. Aron et al. Chemical Geology 565 (2021) 120026

Table 3
Common symbols, explanations, and values for triple oxygen isotopes.
Symbol Value Explanation Reference

θeq 0.529 liquid-vapor equilibrium fractionation exponent Barkan and Luz, 2005
θdiff 0.518 water vapor diffusion fractionation exponent Barkan and Luz, 2007
18 18
αeq Variable (temperature dependent) O/16Ol-v equilibrium fractionation factor Majoube, 1971
17 17
αeq αeq = (18αeq)θeq 17
O/16Ol-v equilibrium fractionation factor Eq. (2)
18 18
αdiff 1 (pure turbulent transport) to 1.0285 O/16Ol-v diffusive transport fractionation factor Merlivat, 1978
(transport via pure molecular diffusion)
17 17
αdiff αdiff = (18αdiff)θdiff 17
O/16Ol-v diffusive transport fractionation factor Eq. (2)
λref 0.528 slope of the δ′ 18O–δ′ 17O reference line commonly used in hydrologic studies Luz and Barkan, 2010
δ Rsample “delta” McKinney et al., 1950
δ= − 1
Rstandard
δ′ δ = ln(δ + 1)
′
“delta prime” Hulston and Thode, 1965; Miller, 2002
∆′ 17O ∆′ 17O = δ′ 17O – λref δ′ 18O “cap 17O” Barkan and Luz, 2007

(Marrero and Mason, 1972, θdiff = 0.5184) and has also been confirmed also varies non-linearly when it is defined with δ notation, but a loga­
experimentally (Barkan and Luz, 2007, θdiff = 0.5185 ± 0.0003). The rithmic definition of d-excess is typically considered only at high lati­
small, but statistically significant, difference between θeq and θdiff values tudes or when δ18O variation is large (Dütsch et al., 2017; Schoenemann
means that triple oxygen isotopes can differentiate equilibrium and ki­ et al., 2014; Uemura et al., 2012). The δ′ notation is imperative for triple
netic fractionation in materials that contain oxygen atoms. oxygen isotopes because the non-linear ∆′ 17O calculation artifact is the
In nature, isotopic compositions rarely reflect fractionation from a same order of magnitude as analytical ∆′ 17O precision and natural ∆′ 17O
single process, but instead integrate multiple fractionating processes and variability in the water cycle (Fig. 5). The slope of the δ′ 18O–δ′ 17O
several θ values. Following convention from triple oxygen isotope reference line is discussed in Section 5, but hydrologic studies, including
literature, we use λ notation to differentiate relationships that integrate this review, typically use a value of 0.528 for λref. Values of ∆′ 17O are
multiple fractionating processes (λ) from those that result from a single generally very small and are expressed in units of per meg (1,000 per
fractionating process (θ). The most familiar λ value in isotope hydrology meg = 1‰).
is the slope (~ 8) of the oxygen-hydrogen global meteoric water line, The triple oxygen isotope literature uses multiple terms to express
δ2H = 8*δ18O + 10 (Craig, 1961), where δ notation is expressed in per δ′ 18O and δ′ 17O deviations from the reference relationship in waters. For
mil and defined as example, ∆′ 17O, ∆17O, 17O-excess, and 17Oexcess are all found in triple
oxygen isotope literature (Sharp et al., 2018; Passey and Ji, 2019; Luz
Rsample
δ= − 1 (3) and Barkan, 2010; Landais et al., 2010, respectively). These terms are
Rstandard
equivalent and are each defined as in Eq. (5) in this review, but the
Linear meteoric water isotope relationships are ubiquitous in isotope different notation can cause confusion among studies. The 17O-excess
hydrology because they provide a useful reference frame from which to and 17Oexcess terms are advantageous because they highlight the relative
assess isotopic variability and quantify non-equilibrium fractionation excess of 17O in meteoric waters as compared to ocean water and
(Gat, 1996). However, these relationships are not always truly linear emphasize similarities between the triple oxygen isotope system and d-
because mass-dependent fractionation follows a power law function (Eq. excess (Fig. 1). The “capital delta” notation is advantageous because ∆
(2)). This non-linearity is rarely observed in natural waters (e.g., Craig, notation is defined as the isotopic deviation from a reference relation­
1961; Dansgaard, 1964; Rozanski et al., 1993) because the range of ship and is used in multiple isotope systems (e.g., ∆34S and ∆25Mg; Criss
isotope values on Earth is relatively small and the scatter of data points and Farquhar, 2008; Young and Galy, 2004).
around an apparent linear relationship is too great to resolve the slight We prefer and recommend the ∆′ 17O term (Eq. (5) and Sharp et al.,
curvature (Fig. 4a and c). However, curvature appears over a sufficiently 2018) because this notation explicitly indicates that the ∆′ 17O parameter
large isotopic range (Fig. 4b and d). This curvature is concave when the is calculated using δ′ values (not δ values) and defines this parameter as
slope between isotopic compositions is greater than 1 (Fig. 4b) and the deviation from a reference relationship.
convex when the slope between isotopic compositions is less than 1
(Fig. 4d). Logarithmic δ′ (“delta prime”) notation linearizes the expo­ 3. Motivation from a decade of ∆′17O observations
nential relationship between isotopic compositions (Fig. 4b and d;
Hulston and Thode, 1965; Miller, 2002): Meteoric water isotope patterns are best observed from amount-
δ′ = ln(δ + 1) (4) weighted precipitation (Dansgaard, 1964; Rozanski et al., 1993) or
flowing surface waters (Kendall and Coplen, 2001) because these waters
This δ′ notation is used in all triple oxygen isotope studies and some integrate fractionating processes in the hydrosphere, atmosphere, and
studies of d-excess (e.g., Dütsch et al., 2017). biosphere. Efforts to understand patterns in δ18O and δ2H have culmi­
nated in the global meteoric water line (Craig, 1961), δ18O and δ2H
2.2. Definition of ∆′ 17O isoscapes (e.g., Bowen, 2010), and well-tuned isotope-enabled general
circulation models (Brady et al., 2019; Joussaume et al., 1984) that
Compilations of δ′ 18O and δ′ 17O almost always appear linear reflect δ18O and δ2H variations across almost every region on Earth.
(Fig. 4c), so the most practical way to view and interpret triple oxygen These products are a point of reference for nearly every hydrologic and
isotope data is as a deviation from a reference line (Barkan and Luz, paleoclimate study of δ18O and δ2H (e.g., Jasechko, 2019; Noone et al.,
2007): 2013; Poulsen et al., 2010; Rowley and Garzione, 2007).
Comprehensive global and continent scale water isotope studies such
∆′ 17 O = δ′17 O–λref δ′ 18 O (5)
as those by Craig (1961), Dansgaard (1964), Rozanski et al. (1993), and
In this definition, λref is the slope of a mass-dependent reference line Kendall and Coplen (2001) do not yet exist for triple oxygen isotopes.
and δ′ notation ensures that isotopic deviations are calculated from a Global variability of meteoric water ∆′ 17O and δ′ 18O was first described
δ′ 18O–δ′ 17O relationship that is exactly linear. This δ′ notation is critical in 2010 from a dataset comprised of two international standards, SLAP
because a non-linear calculation artifact biases ∆′ 17O when δ18O and (Standard Light Antarctic Precipitation) and GISP (Greenland Ice Sheet
δ17O values are used instead of δ′ 18O and δ′ 17O values (Fig. 5). d-excess Precipitation) (Barkan and Luz, 2005); 29 Antarctic snow samples

5
P.G. Aron et al. Chemical Geology 565 (2021) 120026

Fig. 4. Scatterplots of meteoric water isotope values. Published data (Table 2) are colored by sample type; new data reported in this review are outlined in gold. Over
the natural range of (A) δ18O versus δ2H and (C) δ′ 18O versus δ′ 17O, meteoric water isotope relationships are nearly linear. The same data are shown in (B) and (D),
respectively, but are overlaid with calculated δ18O, δ17O, or δ2H (dashed black line) and calculated δ′ 18O, δ′ 17O, or δ′ 2H (solid black line) over larger isotopic ranges.
There are fewer points in (A) and (B) than (C) and (D) because some triple oxygen isotope studies do not include δ2H data. The curvature between δ-δ relationships is
concave when the slope between isotope values is greater than 1 (B, δ18O–δ2H) and convex when the slope between isotope values is less than 1 (D, δ18O–δ17O).

(Landais et al., 2008); and 52 meteoric waters from locations mostly data, 2) evaluate the global δ′ 18O–δ′ 17O relationship and present
scattered throughout Europe and Asia (Luz and Barkan, 2010). From this updated triple oxygen isotope meteoric water relationships that better
compilation, Luz and Barkan (2010) defined a global meteoric water fit the available data, and 3) explain the hydrologic processes that drive
line, established λref as the slope of this line, and set the expectation that ∆′ 17O variation in the water cycle. Following the framework established
∆′ 17O values of meteoric water should be relatively invariant over a ~ by Craig (1961), Dansgaard (1964), Rozanski et al. (1993), and Kendall
70‰ range in δ′ 18O (Fig. 3a). However, observations since this seminal and Coplen (2001), we prefer a narrow definition of meteoric water that
work show that meteoric water ∆′ 17O is far more variable than initially includes only precipitation and surface water for the second objective.
recognized (compare Fig. 3a and b). Moreover, the meteoric water δ′ 18O For the third objective, we include a much wider definition of meteoric
and δ′ 17O data published since 2010 can fit multiple regression lines, water (plant water, precipitation, surface and subsurface, snow and ice,
suggesting that the global meteoric water line defined in 2010 may not and ocean water) to explain triple oxygen isotope variation in as many
actually represent global meteoric waters (Miller, 2018; Sharp et al., parts of the hydrosphere as possible.
2018).
We use this review paper to provide both a synopsis of what we know 4. New surface water data
about δ′ 18O, δ′ 17O, and ∆′ 17O in meteoric water and as a guide for how
practitioners might use triple oxygen isotope data in hydrologic and The new surface water dataset reported in this review includes δ18O,
paleoclimate studies. As part of this review, we report a new, near-global δ17O, δ2H, d-excess, and ∆′ 17O data from 104 rivers and lakes (Fig. 2).
surface water dataset that spans 6 continents and 17 Köppen climate The samples are part of a global crowdsourced dataset, and the isotope
classes and expands the published meteoric water triple oxygen isotope data are reported in Supplements 1–3 and Figs. 3, 4, and 6. These data
dataset. In the following sections we 1) introduce our new surface water are briefly summarized here because they are included in the evaluation

6
P.G. Aron et al. Chemical Geology 565 (2021) 120026

0.3‰ for δ17O, 0.9‰ for δ18O, and 10 per meg for ∆′ 17O. The analytical
error on our measurements is nearly identical to all other published
meteoric water isotope data.

4.1. Isotopic compositions of the new surface water dataset

Surface water δ18O ranges from − 20.3 to 9.6‰, d-excess ranges from
− 31.8 to 21.1‰, and ∆′ 17O ranges from − 45 to 54 per meg, where ∆′ 17O
is defined with λref equal to 0.528. Most isotopic compositions cluster
between − 14 to − 5‰, 3 to 13‰, and 14 to 33 per meg (δ18O, d-excess,
and ∆′ 17O, respectively), but have high standard deviations (6.1‰,
11.1‰, and 18 per meg, respectively) and are poorly described by
average values. Values of ∆′ 17O are moderately to strongly positively
correlated with d-excess (Pearson’s r = 0.73) and moderately to strongly
negatively correlated with δ′ 18O (r = − 0.64). Both ∆′ 17O and d-excess
are very weakly correlated or uncorrelated with latitude, longitude,
elevation, mean annual temperature, mean annual precipitation, and
mean annual relative humidity (all r < ± 0.3). Rivers tend to have lower
Fig. 5. Comparison of Δ′ 17O (solid black line) versus Δ*17O (dashed black line)
δ18O, higher d-excess, and higher ∆′ 17O values than lakes, although
across a common range of meteoric water δ′ 18O values. The gray bar shows
some rivers and lakes in arid regions are isotopically similar (Supple­
typical Δ′ 17O analytical precision (± 10 per meg). Note that Δ′ 17O is calculated
ment 1). The slope (λ = 0.5262 ± 0.0002) through δ′ 18O and δ′ 17O
with δ′ values while Δ*17O is calculated with δ values. Without δ′ notation,
Δ*17O varies non-linearly as a function of δ′ 18O, and introduces a bias in Δ*17O values was determined from a Model II linear regression. The uncer­
that is greater than analytical precision and a similar magnitude to environ­ tainty on this slope, as well as all other slopes and intercepts throughout
mentally driven variability. δ′ notation linearizes the definition of Δ′ 17O and this review, is the standard error from a Model II linear regression.
removes the non-linear calculation artifact. Climate data (mean annual precipitation, temperature, and relative
humidity) from the sampling locations were extracted from the CRU 2.0
of the triple oxygen isotope meteoric water line (Section 5) and expla­ dataset (New et al., 2002) and are included in Supplement 1.
nations of ∆′ 17O variability (Section 6). A complete description of our
sample collection and analytical methods are in Section 8. 5. Triple oxygen isotope meteoric water lines and the reference
Briefly, the δ18O and δ2H values were measured with a Picarro slope
L2130-i cavity ringdown spectrometer and δ18O and δ17O values were
analyzed with a Nu Perspective isotope ratio mass spectrometer. All 5.1. Triple oxygen isotope meteoric water lines
isotopic analyses were done at the University of Michigan. The analyt­
ical precision of the Picarro δ18O and δ2H measurements was deter­ Meteoric water lines define the most fundamental relationships in
mined from replicate injections of deionized water and was better than isotope hydrology and provide a point of reference from which to
0.1‰ and 0.3‰, respectively. The root mean square error (RMSE) of interpret isotope data (e.g., Brooks et al., 2010; Craig, 1961; Jasechko,
replicate Nu measurements of USGS reference waters was better than 2019). Here, we use the well-established δ18O–δ2H global meteoric
water line (Craig, 1961) as a model to re-evaluate and update the triple

Fig. 6. Spatial variation of published (Table 2) meteoric water Δ′ 17O (per meg). Water types are differentiated by shape and Δ′ 17O values are differentiated by color.
Note that the breaks on the color bar are spaced to highlight the Δ′ 17O variability between –10 and 40 per meg.

7
P.G. Aron et al. Chemical Geology 565 (2021) 120026

oxygen isotope meteoric water relationship. that are included in the updated meteoric water line is:
The δ18O–δ2H global meteoric water line was initially built from ~
δ′ 17 O = 0.5267*δ′18 O ( ± 0.0002) + 0.013 ( ± 0.002‰) (7)
400 precipitation, river, and lake samples (Craig, 1961). Later, the
meteoric water δ18O–δ2H relationship was re-evaluated using amount- Excluding rivers from very arid environments that may be affected
weighted and arithmetic mean monthly precipitation from a near- by evaporation (Passey and Ji, 2019; Surma et al., 2015), the regression
global distribution of IAEA/WMO sites (Dansgaard, 1964; Rozanski line is:
et al., 1993) and rivers from the United States (Kendall and Coplen,
2001). These re-evaluated global meteoric water lines have slightly δ′ 17 O = 0.5268*δ′18 O ( ± 0.0002) + 0.015 ( ± 0.002‰) (8)
higher slopes and intercepts, but are statistically indistinguishable from The samples included in Eqs. (7) and (8) range from − 20.5 to 9.4‰
the original line defined by Craig (1961), indicating that the δ18O–δ2H in δ18O, and these regression lines are nearly indistinguishable from the
global meteoric water line is well characterized and represents global best-fit line from a recent compilation of meteoric waters with δ18O
variation in δ18O and δ2H (Gat, 1996). values greater than − 20‰ (δ′ 17O = 0.52654*δ′ 18O (± 0.00036) + 0.014
The triple oxygen global meteoric water line was first defined as (Luz (± 0.003); Sharp et al., 2018). In contrast, the regression line through
and Barkan, 2010) samples with δ18O values less than − 20‰,
δ′17 O = 0.528 ( ± 0.0001)*δ′ 18 O + 0.033 ( ± 0.003) (6) δ′ 17 O = 0.5285*δ′18 O ( ± 0.00005) + 0.0461 ( ± 0.0026‰) (9)

from GISP and SLAP (Barkan and Luz, 2005), 29 Vostok snow samples has a higher slope, higher intercept, and is similar to the 2010 global
(Landais et al., 2008), and a set of 52 meteoric waters (precipitation, meteoric water line (Eq. (6)). Importantly, Eq. (9) is defined from every
surface water, cave water, and snow) mostly from Europe and Asia (Luz sample with a δ18O value less than − 20‰ because currently there are
and Barkan, 2010). The basic features of this line, an empirically only 12 samples (11 precipitation (Tian and Wang, 2019) and 1 river
determined slope and positive y-intercept, are similar to the oxygen- (this review)) that meet the criteria established by the δ18O–δ2H global
hydrogen global meteoric water line. However, the 2010 Luz and Bar­ meteoric water line. Instead, more than 97% of the triple oxygen isotope
kan δ′ 18O–δ′ 17O global meteoric water line was constructed with a large samples with δ18O values less than − 20‰ are from snow and ice from
proportion of high latitude precipitation and samples (lakes, snow, and Antarctica or Greenland.
evaporated snow) with isotopic compositions that are not representative Differences between Eqs. (7)–(9) suggests that triple oxygen isotopes
of average freshwater from the mid- and low-latitudes (Miller, 2018; do not fit a single, global meteoric water line. These differences may be
Sharp et al., 2018). Building upon data that have been published since associated with regional hydrologic processes as more than 95% of the
2010, this review evaluates and updates the triple oxygen isotope global samples included in Eq. (8) are from locations equatorward of 60◦ N and
meteoric water line. 60◦ S whereas more than 97% of the samples in Eq. (9) are from locations
Following the approaches to build and evaluate the δ18O–δ2H poleward of 60◦ N and 60◦ S. Alternatively, differences between Eqs. (8)
meteoric water relationship, the triple oxygen isotope meteoric water and (9) may be related to water type, a sampling bias because most
line should be defined from a regression through δ′ 18O and δ′ 17O of samples are clustered in small regions (Figs. 2 and 6), or insufficient
integrated monthly precipitation (Dansgaard, 1964; Rozanski et al., data. For example, the Luz and Barkan (2010) triple oxygen isotope
1993) and/or flowing surface waters (rivers) (Kendall and Coplen, global meteoric water line was defined from all of the published pre­
2001). Precipitation data are preferable because they are generally cipitation, surface water, snow, and ice data in 2010, but does not fit all
unevaporated, but sample collection requires substantial effort and to of the observations available today. Eqs. (7)–(9) were defined from a
date only eight studies have reported any precipitation δ17O data subset of the published data in 2020 and better fit the observations
(Table 2). For now, river water is a reasonable alternative because it available today, but may also become outdated with future work. For
often represents the isotopic composition of amount-weighted seasonal now, the available triple oxygen isotope data do not fit a single global
precipitation (e.g., Kendall and Coplen, 2001). However, the isotopic meteoric water line. Future studies of flowing surface water and
composition of river water can be affected by post-precipitation pro­ monthly precipitation are needed to further evaluate and properly
cesses such as evaporation or isotopic exchange with atmospheric vapor, establish this relationship. If triple oxygen isotopes do not fit a single
so ultimately it will be important to re-evaluate the triple oxygen isotope global meteoric water line, that means the triple oxygen isotope system
meteoric water line with only integrated monthly precipitation data.
This evaluation process will be especially important for δ′ 18O and δ′ 17O
because one of the main applications of triple oxygen isotopes is to Table 4
constrain evaporation. Observed λ (λobs) by water type. See Table 2 for references.
The available triple oxygen isotope data (Table 2) from which we
Sample subset λobs ± standard error
defined an updated meteoric water line include 1 river from north­
All data 0.5273 ± 0.00005
western Switzerland (Affolter et al., 2015), 9 rivers from the western
Luz and Barkan (2010) 0.5282 ± 0.0003
United States (Passey and Ji, 2019), 15 rivers from southern Spain Plant water 0.5188 ± 0.0004
(Gázquez et al., 2017), 17 rivers from the Sistan Basin in eastern Iran Precipitation 0.5273 ± 0.0001
(Surma et al., 2015), 18 rivers from locations throughout Asia and Ocean 0.5278 ± 0.001
Europe (Luz and Barkan, 2010), 57 rivers from the Pacific Northwest in Snow and ice 0.5285 ± 0.00006
Surface and subsurface 0.5261 ± 0.0001
the United States (Bershaw et al., 2020), and 84 rivers from our new
surface water dataset (Section 4.1 and Supplement 1). Available pre­
cipitation data include amount-weighted monthly (Landais et al., 2010;
Tian et al., 2018) and reported monthly (Gázquez et al., 2017; Uechi and Table 5
Uemura, 2019) values. We do not include published data from indi­ Temperature dependence of equilibrium fractionation factors and λRayleigh.
vidual precipitation samples because these data are not representative of Temperature (C) 18
αeq 17
αeq λRayleigh
monthly averages, nor do we include data from precipitation samples
Calculation Explanation Majoube, 1971 (18αeq)0.529 (17αeq – 1)/(18αeq – 1)
that that were collected without any measure to prevent evaporation or
for which the accuracy and precision of ∆′ 17O measurements were not 40 1.00823 1.00435 0.5280
25 1.00937 1.00495 0.5278
explicitly presented. 0 1.01172 1.00618 0.5275
The δ′ 18O–δ′ 17O regression through the river and precipitation data − 25 1.01483 1.00782 0.5272

8
P.G. Aron et al. Chemical Geology 565 (2021) 120026

Fig. 7. Variation of liquid δ′ 18O and Δ′ 17O during Rayleigh distillation in (A) a schematic and (B) δ′ 18O–δ′ 17O isotope space. Water starts in the ocean (1), evaporates,
and condenses, leaving airmasses with (2) 80%, (3) 60%, and (4) 40% of the initial airmass remaining. In the schematic (A), f is the percentage of the initial airmass
that remains after rainout. The isotopic composition of precipitation (steps 2-4) was calculated at 25◦ C and is assumed to be in isotopic equilibrium with the vapor
shown in the middle row of Fig. 11. Because λRayleigh is approximately equal to λref, Δ′ 17O is relatively insensitive to Rayleigh distillation.

Fig. 8. Box and whisker plot of water Δ′ 17O values. Water types are listed individually, but are colored according to broader categories to give a sense of variation
within groups. The numbers in parentheses indicate the total number of published observations for each water type and include the new surface water data reported
in this review. In each box, the bolded line is the median Δ′ 17O value, the upper and lower hinge correspond to the 1st and 3rd quartiles, respectively, and the whiskers
correspond to no more than 1.5 times the interquartile range (IQR, the variation between the 1st and 3rd quartiles). The individually plotted points fall outside the
IQR. In general, more evaporated waters have lower Δ′ 17O values and less evaporated waters have higher Δ′ 17O values.

captures different parts of the hydrologic cycle that may not dominate
Table 6 equally across the globe.
Pearson correlation coefficient between ∆′ 17O and d-excess or δ′ 18O.
Water type ∆′ 17O–d-excess Correlation ∆′ 17O–δ′ 18O Correlation
5.2. Triple oxygen isotope reference slope
All Data 0.59 − 0.17
Plant Water 0.95 − 0.93 For much of the past decade it was assumed that all unevaporated
Lake 0.64 − 0.79
Rain 0.21 − 0.19
meteoric water δ′ 18O and δ′ 17O values plotted on a single global mete­
River 0.28 − 0.09 oric water line with an observed slope (λobs) equal to the reference slope
Ocean NA* − 0.03 (λref). With new observations, it is now known that meteoric water δ′ 18O
Snow or Ice − 0.35 0.37 and δ′ 17O values do not plot on a single water line, but instead that λobs
*
There is no reported correlation between ∆′ 17O and d-excess for ocean water can be quite variable (Table 4) and that the slope of a meteoric water
because only one sample (Bershaw et al., 2020) has reported δ18O, δ17O, and δ2H line (λmwl) can vary among subsets of samples (Eqs. (7)–(9); Miller,
of ocean water. 2018; Sharp et al., 2018). Variations among these values makes it critical

9
P.G. Aron et al. Chemical Geology 565 (2021) 120026

to understand differences between λobs, λmwl, and λref. 0.001‰).

The observed slope (λobs) is the slope through any particular dataset The δ′ 18O versus ∆′ 17O isotope space is analogous to δ18O versus d-
or group of samples. This value varies among studies, regions, and excess (Fig. 1). However, plots of δ18O versus d-excess are relatively
subsets of samples, and records hydrologic information because λobs uncommon because most hydrologically relevant deviations from the
values are lower (closer to θdiff) when kinetic processes dominate iso­ δ18O–δ2H reference line are already visible in plots of δ18O versus δ2H.
topic compositions and higher (closer to θeq) when equilibrium frac­ For example, the very evaporated plant waters in Fig. 4a clearly deviate
tionation controls isotopic compositions (Table 4). The slope of a from the rest of the less evaporated or unevaporated data. These de­
meteoric water line (λmwl) defines an average or representative rela­ viations are visible in plots of δ18O versus δ2H because the difference
tionship among unevaporated waters. The δ18O–δ2H model suggests that between the reference slope (8) and kinetic slopes (typically ~4–6;
this is value is constant globally, but triple oxygen isotope λmwl values Bowen et al., 2019; Gonfiantini et al., 2018) is the same or similar order
vary among subsets of samples or water types (Eqs. (7)–(9)). The of magnitude as δ18O and δ2H variations (typically a few to 10s of ‰).
reference slope (λref) is the reference frame from which ∆′ 17O is defined Similar deviations are not visible in the δ′ 18O versus δ′ 17O plot (Fig. 4c)
(Eq. (5)) and must remain constant. Typically, hydrologic triple oxygen because the difference between the reference slope (0.528) and kinetic
isotope studies use a value of 0.528 for λref, and we recommend that slope (θdiff = 0.518) is orders of magnitude smaller than variations of
authors continue to use this value. However, a value of 0.528 for λref is δ′ 18O and δ′ 17O (a few to 10s of ‰). Therefore, mass-dependent triple
somewhat arbitrary, so it is important to understand where this value oxygen isotope variations are instead presented in δ′ 18O–∆′ 17O isotope
comes from and how it affects calculated ∆′ 17O values. space.
The value of λref was initially defined from a set of meteoric water
δ′ 18O and δ′ 17O measurements made with electrolysis and continuous
flow IRMS that fit a line with λobs of 0.528 (Meijer and Li, 1998, λobs = 6.1. Variability of ∆′ 17O in meteoric water
0.5281 ± 0.0015). This slope was later confirmed with a different set of
water samples and higher precision dual inlet IRMS measurements (Luz Average meteoric water ∆′ 17O is 20 to 30 per meg, but ranges from
and Barkan, 2010 λobs = 0.528 ± 0.0001), and defined as λref. That these less than –250 per meg to greater than 100 per meg (Fig. 3b). Generally,
λobs values are closer to 0.529 (θeq) than to 0.518 (θdiff) and are nearly more evaporated waters have lower ∆′ 17O values and less evaporated
identical to the λ value during Rayleigh distillation (λRayleigh, Table 5) waters have higher ∆′ 17O values (Fig. 8). Among all published meteoric
indicates that most meteoric waters in these initial studies were more water data, ∆′ 17O is positively correlated with d-excess (r = 0.59,
strongly affected by Rayleigh distillation than by kinetic fractionation. Fig. 3e) and uncorrelated with δ′ 18O (r = –0.17, Fig. 3b), but these
Of course, the isotopic composition of meteoric water reflects more correlations vary substantially among water types (Table 6). For
than just Rayleigh distillation, and observations over the last decade example, ∆′ 17O and d-excess of highly evaporated waters such as plant
show that λobs is not always equal to 0.528. When λobs is equal to λref (e. water and lakes are strongly positively correlated (r = 0.95 and 0.64,
g., during Rayleigh distillation), δ′ 18O and δ′ 17O fractionate along a line respectively, Table 6) whereas ∆′ 17O and δ′ 18O are strongly negatively
parallel to λref and ∆′ 17O values remain constant (Figs. 3a, 7, and 8a). correlated (r = –0.93 and –0.79, respectively, Table 6). These plant
When λobs is less than λref, δ′ 18O and ∆′ 17O are negatively correlated. The water and lake correlations are so strong because both d-excess and
magnitude of this effect varies on a case-by-case basis and authors ∆′ 17O are sensitive to relative humidity and vary as a function of kinetic
should carefully consider differences between λobs and λref as a part of fractionation. The slope between d-excess and ∆′ 17O is ~ 0.7 to 2.0 per
∆′ 17O interpretations. meg ‰− 1 (e.g., Landais et al., 2010; Li et al., 2015), but the exact value
varies as a function of relative humidity and fractionation factors (Bar­
6. Hydrologic processes that affect ∆′17O kan and Luz, 2007). Correlations from less evaporated waters such as
precipitation and rivers are weaker between ∆′ 17O and d-excess (r =
The previous section focused on important distinctions among triple 0.28 and 0.21, respectively, Table 6) or δ′ 18O (r = –0.19 and –0.09,
oxygen isotope slopes, but the magnitude of these variations is very respectively, Table 6). These weak correlations mean that processes
small and δ′ 18O–δ′ 17O compilations almost always appear exactly linear other than evaporation must account for the ∆′ 17O variation of precip­
(Fig. 4c). Instead, triple oxygen isotope variations are commonly pre­ itation and river water.
sented and interpreted in δ′ 18O–∆′ 17O isotope space (for example, Published meteoric water ∆′ 17O data are poorly suited for spatial
Fig. 3a–c). This space highlights mass-dependent deviations from the analysis and there are no clear spatial trends in Fig. 6. In part this occurs
reference relationship (Farquhar and Thiemens, 2000; McKeegan and because most ∆′ 17O observations are clustered in small regions and very
Leshin, 2001) and is a helpful way to simultaneously visualize isotopic few datasets systematically span latitudinal or elevational gradients
compositions of δ′ 18O (‰) and ∆′ 17O (per meg, where 1 per meg = (Figs. 2 and 6), but also because ∆′ 17O is primarily sensitive to kinetic
fractionation and relatively insensitive to temperature effects and

Table 7
Processes and explanations of ∆′ 17O variation.
Process ∆′ 17O Response Magnitude of ∆′ 17O Response Explanation

Evaporation from the ocean increase Typically ~ 20-30 per meg. Higher values with low humidity and/or θeq vs. θdiff and θdiff less than λref
low turbulence at the evaporating site.
Condensation increase ~ 10 per meg. Higher ∆′ 17O expected in colder conditions. θeq greater than λref
Recycling increase Typically < 20 per meg θdiff less than λref
Stratospheric intrusions increase ? Addition of stratospheric water vapor
Depends on stratospheric and tropospheric ∆′ 17O values
Post-condensation evaporation decrease Potentially > 200 per meg in plant water, typically no θdiff less than λref
more than ~ 50-60 per meg in surface water
Mixing decrease 0 to >100 per meg. Depends on the mixing fraction and Non-linear response
initial δ18O and ∆′ 17O of the mixing waters
Supersaturation decrease ~ 10 to 30 per meg θdiff less than λref
Rayleigh distillation temperature dependent < 10 per meg. Larger effect at lower temperatures. λRayleigh ≈ λref
Convection ? ? ?

10
P.G. Aron et al. Chemical Geology 565 (2021) 120026

Rayleigh distillation (see Section 6.2.4 and Luz and Barkan, 2010). In water and snow or ice (Fig. 3b) disappear among precipitation and
this way, expectations of spatial ∆′ 17O variation are more similar to d- surface waters (Fig. 3c). However, the ∆′ 17O values of precipitation and
excess than to δ18O. Spatial ∆′ 17O patterns are also complicated because surface waters vary by more than 80 per meg, significantly greater than
there is often more ∆′ 17O variation within a single study than between the precision of well-tuned ∆′ 17O measurements, and systematic trends
studies. For example, Li et al. (2017) link more than 50 per meg (~ 40 to exist within this cloud of isotope data that are related to hydrologic
–15 per meg) of ∆′ 17O variation to evapotranspiration and local relative processes and well-known mass-dependent fractionations. The next
humidity in central Kenya. However, a similar range of ∆′ 17O variability sections describe the processes that drive this variation.
is related to summer sublimation in Antarctica (Pang et al., 2019) or From a mechanistic point of view, ∆′ 17O variability arises from 1)
relative humidity above remote moisture sources in southern Japan changes in θ values (θeq versus θdiff), 2) differences between the values of
(Uechi and Uemura, 2019). Individually, each of these datasets tell a λref and θeq or θdiff, or 3) non-linear isotope responses that result from the
compelling story of local-to-regional hydrology; together, they do not logarithmic δ′ notation (Table 7). Initial ∆′ 17O variation during evapo­
paint a robust picture of large-scale spatial ∆′ 17O patterns. ration from the ocean is well explained by the Craig and Gordon (1965)
Isotope-enabled climate models can fill some gaps that are missing model, but phase changes and mixing following this initial evaporation
from observations (e.g., Risi et al., 2013), and meteoric water ∆′ 17O can combine, compound, or negate each other and complicate ∆′ 17O
values do tend to be lower in arid regions due to sub-cloud and surface data.
evaporation and higher in regions where annual precipitation is domi­ To simplify interpretations of ∆′ 17O data, we first use the theoretical
nated by cold-season rain or locations that receive a high degree of framework of the Craig and Gordon (1965) model to explain triple ox­
recycled moisture (see Section 6.2.3.1). Hydrologic mixing may also ygen isotope variation during initial evaporation and condensation
decrease ∆′ 17O values in regions where airmasses or other water bodies (Section 6.2.1, Figs. 9 and 10) and then describe how distinct hydrologic
combine (Landais et al., 2010; Li et al., 2015; Risi et al., 2013). For processes can further affect ∆′ 17O (Table 7 and Sections 6.2.2 and 6.2.3).
example, lower tap water ∆′ 17O values in the central United States may For beginning readers, these sections are an introduction to the types of
result from atmospheric mixing of moisture that originated in the Pacific questions that triple oxygen isotopes can help answer. For experienced
Ocean and the Gulf of Mexico (Li et al., 2015). Latitudinal ∆′ 17O gra­ researchers, these sections outline the theoretical framework, common
dients are observed in tap waters from the United States (Li et al., 2015) isotopic models, and hydrologic processes that may explain ∆′ 17O ob­
but not in China (Tian et al., 2019), although additional work is needed servations. This focus on process is not intended to explain global or site-
to understand this pattern (Fig. 6) because tap water ∆′ 17O values are specific observations, but rather to provide a mechanistic understanding
complicated by non-local isotope signals. of meteoric water ∆′ 17O variability. Code to calculate isotopic variation
Temporal ∆′ 17O patterns are also still relatively uncertain because during initial evaporation and condensation (Supplement 4), mixing
only a few studies have focused on this type of variability (Table 2). Still, (Supplement 5), and post-condensation evaporation (Supplement 6) is
a seasonal pattern of mid-latitude precipitation ∆′ 17O is emerging, with included to help readers understand triple oxygen isotope variation.
lower values in the summer and higher values in the winter (Affolter
et al., 2015; Li et al., 2015; Tian et al., 2018; Uechi and Uemura, 2019). 6.2.1. Why do most meteoric waters have positive ∆′ 17O values?
This seasonal pattern may be related to variations in relative humidity The Craig and Gordon (1965) model of evaporation explains why
and evaporative conditions above remote moisture sources (Tian et al., most meteoric waters have positive ∆′ 17O values. To understand this, we
2018; Uechi and Uemura, 2019). Condensation temperature may also break the theoretical Craig and Gordon (1965) framework into step-by-
play a role in seasonal ∆′ 17O variation (Table 5), but these effects are step processes and show isotopic variation schematically (Fig. 9) and
likely smaller than those related to relative humidity because ∆′ 17O is step-wise (Figs. 10 and 11) as water evaporates from the ocean and then
more sensitive to kinetic fractionation than to temperature. condenses. This section focuses on δ′ 17O, δ′ 18O, and ∆′ 17O; δ18O, δ2H,
and d-excess are included in Figs. 10 and 11 to highlight similarities
6.2. Hydrologic processes that affect ∆′ 17O between the δ′ 18O–δ′ 17O and δ18O–δ2H isotope systems. Arrows, color,
and bolding in Figs. 9 and 10 highlight step-wise isotopic variation. The
Most meteoric waters occupy a crowded region in δ′ 18O–∆′ 17O symbols, color (red versus blue), and bolding (bold versus nonbold) are
isotope space, and many of the distinct patterns that differentiate plant the same in Figs. 9 and 10. In these figures, color differentiates

Fig. 9. Generalized pathway of δ′ 18O and δ′ 17O fractionation

in the water cycle. In this δ′ 18O versus δ′ 17O isotope space,
Δ′ 17O is defined as the deviation from the reference line. The
white background indicates isotope spaces with positive Δ′ 17O
values; the gray background indicates isotope spaces with
negative Δ′ 17O values. Water at different points in the hy­
drologic cycle is differentiated with symbols and numbered 1-
5 to show stepwise variation associated with evaporation and
condensation. Beginning from the ocean (SMOW, 1), water
evaporates into a saturated equilibrium layer (2), diffuses
through the unsaturated atmosphere (3), condenses (4), and
evaporates (5). Vapor is noted with open symbols; liquid is
noted with filled symbols. The symbols do not indicate end
points, but instead show points on a trajectory (marked with
arrows) along which isotopic compositions can continue. Red
(warmer) and blue (colder) outlines show the effects of tem­
perature dependent fractionations. For clarity, values of λref,
θeq (solid line), and θdiff (dotted line) are included. Additional
details and a step-by-step description of δ′ 18O and δ′ 17O
fractionations are shown in Fig. 10.

11
P.G. Aron et al. Chemical Geology 565 (2021) 120026

Fig. 10. Step-wise variation of δ′ 17O, δ′ 18O, δ2H, Δ′ 17O, and d-excess as water evaporates and condenses in a closed system. Isotopic variation is shown between
δ′ 18O and δ′ 17O (column A), δ′ 18O and Δ′ 17O (column B), δ18O and d-excess (column C), and δ18O and δ2H (column D). The solid black lines in columns (A) and (D)
show the triple oxygen and oxygen-hydrogen reference relationship, respectively. In each column, water begins in the ocean (row 1), evaporates into a saturated
layer (row 2), diffuses through an unsaturated atmosphere (row 3), condenses to meteoric water (row 4), and evaporates (row 5). In row 4, we assume the isotopic
composition of precipitation is equal to that of surface water. The isotopic compositions of evaporated vapor and remaining liquid were calculated assuming pan
evaporation at 16◦ C and 10% evaporation. In all panels, vapor is shown with open symbols and liquid is shown with filled symbols. In columns B and C, black arrows
show the isotopic variation associated with each row; gray arrows are included to show the ‘trajectory’ of isotope variability. Fractionation associated with equi­
librium evaporation (row 2), atmospheric vapor diffusion (row 3), and condensation (row 4) are shown under different temperature and humidity scenarios with
color and bolded symbols. When multiple scenarios are included, the star (*) notes which scenario is used in subsequent calculations (rows).

12
P.G. Aron et al. Chemical Geology 565 (2021) 120026

Fig. 11. Isotope variations of δ18O (A, D, G,), Δ′ 17O (B, E, H), and d-excess (C, F, I) during evaporation from the ocean and subsequent Rayleigh distillation. Row 1
(A, B, C) shows the sensitivity of evaporated vapor to temperature and relative humidity. Rows 2 (D, E, F) and 3 (G, H, I) show the temperature sensitivity of
remaining vapor and condensed liquid, respectively. Fractionation under each temperature scenario is independent and does not represent a ‘trajectory’ from ocean
water to precipitation. The final (RH = 0.4) 25◦ C water vapor from the top row is the starting vapor in the middle row. The final (f = 0.4) 0◦ C vapor in the middle row
is in isotopic equilibrium with the liquid in the bottom row. Note the different y-axes in each row.

temperature and bolding shows the sensitivity of isotopic compositions (Fig. 10, row 3), ∆′ 17O values of atmospheric vapor (open downward-
to relative humidity. Black arrows in Fig. 10 show the isotopic variation facing triangles) increase because the value of θdiff (0.518) is less than
associated with each step, and gray arrows give a sense of the isotopic the value of λref (Figs. 9 and 10, panel 3b). The magnitude of this kinetic
‘trajectory’. effect is negatively correlated with turbulence above the saturated layer
First, beginning from the ocean (Figs. 9 and 10, row 1), equilibrium and sensitive to the relative humidity at the site of evaporation (Barkan
isotope exchange (Fig. 10, row 2) occurs between water vapor (open and Luz, 2007; Criss, 1999; Merlivat, 1978; Uemura et al., 2010). The
upward-facing triangles) and liquid water (solid black circles) in a sensitivity of ∆′ 17O to relative humidity is labeled in Fig. 9 and shown
saturated layer near the evaporating surface. Because the value of θeq with bolded (higher relative humidity) or nonbolded (lower relative
(0.529) is greater than that of λref (0.528), δ′ 18O and δ′ 17O of the vapor humidity) symbols in the 3rd row of Fig. 10. When relative humidity is
fall below the reference line and ∆′ 17O is slightly negative (–9 per meg higher, kinetic fractionation is smaller and vapor ∆′ 17O values remain
and –11 per meg at 25◦ C and 5◦ C, respectively, Figs. 9 and 10, panel 2b). closer to zero. When relative humidity is lower, vapor ∆′ 17O values are
The difference between the red (warmer) and blue (colder) open higher (Figs. 9 and 10, panel 3b).
upward-facing triangles is very small within a typical temperature range Third, equilibrium condensation (Fig. 10, row 4) proceeds along a
(~0–30◦ C) because the value of θeq is relatively insensitive to temper­ slope (θeq = 0.529) that is greater than the value of λref. This fraction­
ature (Barkan and Luz, 2005) and is similar to the value of λref. ation increases the ∆′ 17O value of the more condensed phase (precipi­
Second, during diffusion through the unsaturated atmosphere tation, solid squares in Figs. 9 and 10, panel 4b) and decreases the ∆′ 17O

13
P.G. Aron et al. Chemical Geology 565 (2021) 120026

(Fig. 3b; Cernusak et al., 2016; Landais et al., 2006; Li et al., 2017). This
fractionation is independent of initial δ18O, such that low ∆′ 17O values
do not require high δ18O and vice versa (Fig. 12).
Evaporation occurs in the hydrosphere under steady-state (constant
water levels) or non-steady-state (progressive water loss) conditions.
The isotopic composition of remaining water in both of these scenarios is
predicted by well-established models that apply to δ18O, δ17O, and δ2H
(Criss, 1999; Gázquez et al., 2018; Passey and Ji, 2019; Surma et al.,
2018). First, steady-state evaporation occurs in simple flow-through or
closed-basin systems where the volume of inflowing water is equal to the
volume of water loss via evaporation, outflow, and/or groundwater
seepage. In this scenario, the isotopic composition of the remaining
water is:
αeq αdiff (1 − h)RI hXE RA
RW = (10)
XE + αeq αdiff (1 − h)(1 − XE )

where αeq is the temperature dependent equilibrium fractionation factor

(Barkan and Luz, 2005; Horita and Wesolowski, 1994; Majoube, 1971),
αdiff is the kinetic fractionation factor, h is the relative humidity
Fig. 12. Schematic of δ′ 18O versus Δ′ 17O based on isotope mass balance models normalized to the temperature of the evaporating surface, XE is the
of steady-state evaporation. During evaporation, the Δ′ 17O response is inde­ volumetric ratio of water lost to evaporation relative to inflowing water,
pendent of initial δ′ 18O composition. and RI, RA, and RW are the isotope ratios of inflowing water, atmospheric
water vapor, and remaining evaporated water, respectively. Impor­
value of the less condensed phase (remaining vapor, Fig. 11e). Precipi­ tantly, this model assumes that the isotopic composition of the evapo­
tation and vapor ∆′ 17O values are slightly sensitive to temperature rating body is well mixed and that inflowing water is isotopically
during condensation (compare the red and blue squares in Figs. 9 and 10 uniform and unevaporated, which is not always the case (e.g., Surma
and variations in Fig. 11e) because equilibrium fractionation is larger at et al., 2018; Surma et al., 2015). Second, non-steady-state evaporation
lower temperatures than at higher temperatures (Fig. 9; 10, panel 4b; occurs in isolated water bodies that evaporate to dryness. The isotopic
and 11e) (Horita and Wesolowski, 1994; Majoube, 1971). The increase composition of the water body during this ‘pan’ evaporation is (Criss,
in precipitation ∆′ 17O associated with equilibrium condensation is 1999; Passey and Ji, 2019; Surma et al., 2018):
typically smaller than kinetic effects during diffusion because the value RW = ƒu (RWi – RWss ) + RWss (11)
of λref is closer to the value of θeq than the value of θdiff.
Fourth, post-condensation evaporation (Fig. 10, row 5) increases where f is the fraction of water remaining, RW is the isotope ratio of the
vapor (open diamonds) ∆′ 17O values and decreases the ∆′ 17O values of evaporating body, RWi is the isotope ratio of the initial water, RWss is the
the remaining liquid (solid diamonds) because the value of θdiff is less predicted final steady-state isotope ratio, and the exponent u relates
than the value of λref (Figs. 9 and 10, panel 5b). This kinetic fractionation equilibrium and kinetic fractionation factors by the relative humidity
explains why some very evaporated waters such as plant waters and (h), where u is:
lakes (Figs. 3b and 8) have negative ∆′ 17O values.
1 − αeq αdiff (1 − h)
u= (12)
6.2.2. Hydrologic processes that decrease ∆′ 17O αeq αdiff (1 − h)

and RWss is:

6.2.2.1. Evaporative effects on remaining water. Evaporation leaves the
remaining water body with higher δ18O and δ17O compositions (Gat, αeq hRA
RWss = (13)
1996; Gonfiantini et al., 2018) and a lower ∆′ 17O value because the 1 − αeq αdiff (1 − h)
value of θdiff is less than the value of λref (Figs. 9 and 10). These effects
Observations and modeling show that the isotopic effects of
are most pronounced in leaf waters, lakes, and some slow-moving rivers

Fig. 13. Variations of (A) δ18O, (B) d-excess, and (C) Δ′ 17O values when mixing VSMOW (water 1) and SLAP (water 2). δ18O and d-excess vary linearly with mixing.
Δ′ 17O responds nonlinearly to mixing because it is defined with δ′ notation.

14
P.G. Aron et al. Chemical Geology 565 (2021) 120026

evaporation are greatest when relative humidity is low and/or when 6.2.3. Hydrologic processes that increase ∆′ 17O
values of αdiff or XE are high (Criss, 1999; Gázquez et al., 2018; Passey
and Ji, 2019; Surma et al., 2018), although additional work is needed to 6.2.3.1. Moisture recycling. Moisture recycling increases ∆′ 17O of
clarify the role of turbulence at the evaporating site and to constrain the evaporated vapor and subsequent precipitation because the value of θdiff
values of αdiff and XE (Passey and Ji, 2019). is less than the value of λref (Figs. 9 and 10). Maintaining isotopic and
mass balance and following the logic that describes fractionation of
6.2.2.2. Mixing. Due to the logarithmic δ′ notation used in the defini­ remaining water during evaporation (Section 6.2.2.1), evaporated vapor
tion of ∆′ 17O, mixing water bodies with different δ18O compositions has lower δ′ 18O, lower δ′ 17O, and higher ∆′ 17O than the initial water
decreases ∆′ 17O (Fig. 13c; Luz and Barkan, 2010; Matsuhisa et al., 1978). body from which it evaporated. The ∆′ 17O of this vapor is inversely
This phenomenon is most pronounced when ∆′ 17O of the mixed water related to the relative humidity at the site of evaporation, with lower
bodies are identical and δ18O are very different. The δ18O and d-excess relative humidity resulting in higher vapor ∆′ 17O values (Figs. 9 and 10).
responses to mixing are a linear function of the fraction of each mixed Condensation of this recycled moisture further increases ∆′ 17O (Figs. 9
water (Fig. 13a and b) because δ18O and d-excess are defined with δ and 10). d-excess also increases with moisture recycling (Aemisegger
notation (not logarithmic δ′ notation). Isotopic effects of mixing also et al., 2014; Salati et al., 1979; Tian et al., 2019), so ∆′ 17O and d-excess
affect analytical systems as gases move through prep lines and isotope of evaporated vapor are positively correlated (Figs. 3e and 10, panels 5b
analyzers (see Section 7.2.6 for additional details). Code in Supplement and 5c).
5 is provided so researchers can explore the isotopic effects of mixing in
natural and analytical settings. 6.2.3.2. Stratospheric intrusions. Stratospheric water vapor undergoes
non-mass-dependent fractionation and has extremely high (greater than
6.2.2.3. Supersaturation. Kinetic effects during condensation under 1,000 per meg) ∆′ 17O values (Miller, 2018; Winkler et al., 2012) that
very cold (< ~ –20◦ C) supersaturated conditions cause low ∆′ 17O values may contribute to surface water fluxes. Stratospheric intrusions are
in snow and ice (Angert et al., 2004; Jouzel and Merlivat, 1984; Landais possible during the Antarctic winter when the tropopause is low (Franz
et al., 2012a, 2012b; Landais et al., 2008; Pang et al., 2019; Pang et al., et al., 2005; Roscoe et al., 2004), but are generally considered negligible
2015; Risi et al., 2010; Schoenemann et al., 2014; Schoenemann and in most hydrologic triple oxygen isotope studies (e.g., Landais et al.,
Steig, 2016; Winkler et al., 2012) and a pattern of downward tailing 2008; Luz and Barkan, 2010) and are not evident in Fig. 3.
∆′ 17O at very low (δ18O < –30‰) isotopic compositions (Fig. 3b). Su­
persaturation is common in high latitude and polar regions and develops 6.2.4. Hydrologic processes with little effect on ∆′ 17O
when the saturation vapor pressure of a condensing surface is less than
the vapor pressure surrounding a water droplet or ice crystal (Schoe­ 6.2.4.1. Rayleigh distillation. Rayleigh distillation explains many of the
nemann et al., 2014). Under these conditions, a strong vapor pressure spatial patterns observed in meteoric water δ18O, δ17O, and δ2H (Gat,
gradient develops between water vapor and the condensing surface. 1996; Risi et al., 2013), but has little effect on ∆′ 17O (Fig. 7) because
Water vapor must diffuse across this gradient to condense, a process that λRayleigh (Table 5) is nearly identical to λref (0.528) (Fig. 7b, Luz and
causes kinetic fractionation and lowers ∆′ 17O values of the condensate. Barkan, 2010). In other words, during Rayleigh distillation δ′ 18O and
Although equilibrium effects during condensation (Section 6.2.1) and δ′ 17O vary along a line that is nearly parallel to λref, so ∆′ 17O remains
moisture recycling in Antarctica (Pang et al., 2019) generally increase essentially constant. A slight temperature dependence of λRayleigh
∆′ 17O, low ∆′ 17O values observed in polar regions (Fig. 3b) suggest that (0.5278 at 25 ◦ C, 0.5272 at –25 ◦ C; Table 5) and can increase (decrease)
strong kinetic effects dominate under very cold supersaturated condi­ the ∆′ 17O value of vapor (condensate) by a few per meg when λRayleigh is
tions (Angert et al., 2004; Jouzel and Merlivat, 1984; Landais et al., less than λref (Fig. 11), but this variation is generally smaller than
2012a). analytical precision (approximately ± 10 per meg). The temperature

Fig. 14. Temperature dependence of the slopes in (A) δ18O–δ2H and (B) δ′ 18O–δ′ 17O isotope systems. Slopes are shown for equilibrium fractionation (θeq, solid black
line), the diffusion of water vapor through the air (θdiff, dashed black line), and the meteoric water reference relationship (solid gray line). The reference slopes are
well-established (Craig, 1961 and Luz and Barkan, 2010, respectively) and do not vary with temperature. Similarly, θdiff values are independent of temperature
(Barkan and Luz, 2007). The θdiff value associated with the δ18O–δ2H relationship is generally between ~ 2.5 and 8 (Gonfiantini et al., 2018). The triple oxygen
(δ′ 18O–δ′ 17O) θeq value is insensitive to temperature (Barkan and Luz, 2005); the oxygen-hydrogen (δ18O–δ2H) θeq value varies slightly with temperature. These
different θeq temperature sensitivities result in a slight temperature dependence in d-excess but little temperature dependent variation in Δ′ 17O.

15
P.G. Aron et al. Chemical Geology 565 (2021) 120026

sensitivity of λRayleigh arises from the equilibrium fractionation factors (α weeks of near-constant analysis; for laser-based systems, an analytical
values) for 18O and 17O (Barkan and Luz, 2005; Luz and Barkan, 2010; session is typically defined as a batch or tray of samples or a ‘calibration
Majoube, 1971) because window’ as in Schauer et al. (2016) that can span a few days to a few
( 17 ) months.
αeq − 1
(
λRayleigh = 18 ) (14)
αeq − 1 7.1.1. Dual inlet isotope ratio mass spectrometry
As a result, values of λRayleigh are slightly higher at warmer temper­ The IRMS cobalt(III) fluoride method was first described by Baker
atures and slightly lower at cooler temperatures (Table 5). At very warm et al. (2002) and was later modified by Barkan and Luz (2005) to
temperatures (> 35 ◦ C), λRayleigh is almost identical to λref, and ∆′ 17O is improve the precision of δ17O and δ18O measurements. In this process,
indeed invariant during Rayleigh distillation (Table 5, Fig. 11). water is fluorinated to convert liquid water to oxygen gas:
2H2 O(l) + 4CoF3(s) →O2(g) + 4HF(g) + 4CoF2(s) . (16)
6.2.4.2. Temperature sensitivity of ∆′ 17O. Theoretical calculations and
laboratory experiments show that ∆′ 17O is relatively insensitive to Following fluorination, O2 gas passes through a series of traps and
temperature, and far less sensitive to temperature than d-excess molecular sieves to remove reaction byproducts and capture the purified
(Figs. 10, 11, and 14b; Barkan and Luz, 2005; Cao and Liu, 2011). These sample. Triple oxygen isotope ratios are measured on O2 gas using
different temperature sensitivities occur because the triple oxygen Thermo-Finnigan Delta Plus (e.g., Luz and Barkan, 2010), Thermo-
liquid-vapor θeq value is independent of temperature (Fig. 14b; Barkan Finnigan 253 (e.g., Schoenemann et al., 2014), or Nu Perspective (this
and Luz, 2005; Cao and Liu, 2011), whereas the oxygen-hydrogen study) mass spectrometers. In total, a single measurement takes ~ 3
liquid-vapor θeq value varies slightly with temperature (Fig. 14a; hours and samples are typically analyzed twice. IRMS methods and an
Horita and Wesolowski, 1994; Majoube, 1971). For both the triple ox­ analytical workflow from our lab at the University of Michigan are
ygen and oxygen-hydrogen systems, the θeq value is defined as: described in Section 8.2.

θl− =
ln (A αl− v )
(15) 7.1.2. Laser absorption spectrometry
Because the fluorination and IRMS methods are complex and require
v
ln (B αl− v )
significant laboratory infrastructure (see Section 8.2), there is interest in
where αl-v is the temperature dependent equilibrium liquid-vapor frac­ using cavity ring-down spectroscopy (Picarro Inc.) or cavity-enhanced
tionation factor for A (2H/1H or 17O/16O) and B (18O/16O). The triple laser absorption (Los Gatos Research, LGR) to measure triple oxygen
oxygen θl-v value is nearly invariant (0.529, Fig. 14b) because 17O/16O isotope ratios. These laser absorption spectrometers can achieve similar
and 18O/16O are subject to the same temperature effects; the oxygen- ∆′ 17O precision as IRMS techniques, have a smaller laboratory footprint,
hydrogen θl-v value (~ 8, Fig. 14a) varies slightly with temperature are cheaper and more portable than mass spectrometers, and simulta­
because 2H/1H fractionation is governed by a different temperature neously measure δ18O, δ17O, and δ2H (Berman et al., 2013; Schauer
dependent relationship than 18O/16O fractionation (Horita and Weso­ et al., 2016; Steig et al., 2014). Laser absorption spectrometry may also
lowski, 1994; Majoube, 1971). be faster than the IRMS method because laser-based analyzers do not
The different temperature sensitivities of ∆′ 17O and d-excess are most require complex sample conversion from liquid water to gaseous O2 and
noticeable at low temperatures (low δ18O values, compare Fig. 3b and d) the absorption analysis only takes a few minutes. However, high-quality
and can provide complementary information to decouple the isotopic laser absorption isotope data require long integration times (Schauer
effects of temperature, relative humidity, and supersaturation on et al., 2016; Steig et al., 2014) or many injections of the same sample,
meteoric waters (Landais et al., 2012a, 2012b; Landais et al., 2008; which can cumulatively take longer than a single IRMS analysis (Berman
Uemura et al., 2010; Winkler et al., 2012). et al., 2013), and additional work is needed to directly compare δ18O,
δ17O, and ∆′ 17O data from laser-based and IRMS systems. An example of
7. Analytical methods and considerations the workflow required to achieve ~10 per meg ∆′ 17O precision on a
Picarro water isotope analyzer is described in Schauer et al. (2016).
7.1. Analytical methods to measure water ∆′ 17O

Triple oxygen isotope ratios are measured with dual inlet isotope 7.2. Analytical recommendations and best practices
ratio mass spectrometry or laser absorption spectrometry (Table 8).
With careful analysis, both methods can achieve high quality 17O/16O 7.2.1. VSMOW-SLAP normalization
measurements and similar (~ 10 per meg) precision for the ∆′ 17O We recommend that triple oxygen isotope compositions be normal­
parameter. Typically, triple oxygen isotope data are measured in ized to the VSMOW-SLAP scale following the approach described by
analytical sessions and data corrections (VSMOW-SLAP normalization, Schoenemann et al. (2013). We provide code (Supplement 7) and an
drift corrections, etc.) are applied over a full session (Thompson, 2012; example data file (Supplement 8) to show how to do this normalization.
Werner and Brand, 2001). For IRMS systems, an analytical session is This normalization technique improves the accuracy of isotope mea­
defined for each reactor (~ 200 injections) and lasts approximately 2–4 surements, simplifies inter-lab data comparisons (Brand and Coplen,
2001; Coplen, 1988; Gonfiantini, 1978; Meijer et al., 2000; Paul et al.,

Table 8
Water triple oxygen isotope analysis methods.
IRMS Picarro Los Gatos Research (LGR)

Analysis method Dual-inlet isotope ratio mass spectrometry (IRMS) Cavity ring-down spectroscopy Cavity-enhanced laser absorption spectroscopy
(CRDS, laser-based)
Instrument Various (Delta Plus, MAT 253, Nu Perspective) L2140-i Triple Water Isotope Analyzer (TWIA)
Sample preparation CoF3 reaction H2O(l) → H2O(g) H2O(l) → H2O(g)
Analyte O2 gas Water vapor Water vapor
Analysis time 2 to 3 hours 1.5 to 3 hours < 1 hour to 7+ hours
∆′ 17O precision ~ 10 per meg < 8 per meg ~ 10 per meg
Lab footprint 10s m2 1-2 m2 1-2 m2
Method development Barkan and Luz, 2005 Steig et al., 2014; Schauer et al., 2016 Berman et al., 2013

16
P.G. Aron et al. Chemical Geology 565 (2021) 120026

Table 9
Isotopic composition of common standards and reference waters.
δ18O (‰) δ17O (‰) δ2H (‰) ∆′ 17O (per meg)
a b
VSMOW , VSMOW2 0 0 0 0
SLAPa, SLAP2b –55.5 –29.6968c –428.0 0
GISP − 24.78 ± 0.09e –13.16 ± 0.05c − 189.7 ± 1.0e 22 ± 11c
–13.12 ± 0.05d 23 ± 10d
–13.5 ± 0.3f 24 ± 12f
USGS 45 –2.238 ± 0.011e –1.19 ± 0.3d –10.3 ± 0.4e 12 ± 1d
–1.1 ± 0.3f 13 ± 7f
e
USGS 46 –29.80 ±0.03 –15.85 ± 0.02d –235.8 ± 0.7e 20 ± 2d
–15.7 ± 0.2 f 19 ± 11f
USGS 47 –19.80 ± 0.02e –10.47 ± 0.02d –150.2 ± 0.5e 40 ± 1d
–10.4 ± 0.4f 32 ± 9f
USGS 48 –2.224 ± 0.012e –1.15 ± 0.01d –2.0 ± 0.4e 26 ± 3d
–1.1 ± 0.2f 31 ± 6f
USGS 49 –50.55 ± 0.04e –27.8 ± 0.5f –394.7 ± 0.4e 13 ± 8f
USGS 50 4.95 ± 0.02e 2.7 ± 0.2f 32.8 ± 0.4e –10 ± 7f

Notes: All isotopic data are normalized to the VSMOW-SLAP scale.

a
Gonfiantini, 1978; Barkan and Luz, 2005.
b
Lin et al., 2010.
c
Schoenemann et al., 2013.
d
Berman et al., 2013 (isotopic data are weighted by the inverse of precision; uncertainties are the averages of three experimental methods weighted by the number
of measurements).
e
USGS Reports of Stable Isotopic Composition Reference Materials (uncertainty is the 95% confidence interval).
f
This study (uncertainty is one unweighted standard deviation). The ∆′ 17O of USGS47 reported in this study is substantially lower than the value from Berman et al.
(2013). Our measured δ18O and δ17O values are nearly identical to the accepted (USGS) or previously published (Berman et al., 2013) values, and we do not have a
compelling analytical explanation for this one low ∆′ 17O value.

2007; Schoenemann et al., 2013), and corrects for analytical in­ sample waters span a large (> 10‰) range in δ18O or d-excess. Sec­
consistencies such as pressure-baseline offsets (Yeung et al., 2018). ondary reference analyses should be distributed evenly throughout
At a minimum, VSMOW and SLAP (or VSMOW2 and SLAP2, which every analytical session to monitor instrument drift.
are isotopically indistinguishable from VSMOW and SLAP (Lin et al., Finally, the isotopic composition of standards and reference waters
2010)), should be measured the beginning and end of every analytical should bracket the expected isotopic composition of unknowns. This is
session. Ideally, VSMOW or VSMOW2 and SLAP or SLAP2 should also be typically not an issue for δ18O, δ17O, and δ2H because USGS reference
analyzed within each session because instrument nonlinearities can waters span the common range of natural waters (Fig. 4a and c, Table 9).
evolve through time. We recommend that the value of δ17OSLAP However, the range of ∆′ 17O and d-excess of standards and reference
(–29.6968‰) be calculated directly from the defined values of δ18OSLAP waters (Table 9) is much smaller than observed ∆′ 17O (Fig. 3b) and d-
(–55.5‰), ∆′ 17OSLAP (0.00‰), and λref (0.528) (Gonfiantini, 1978; excess (Fig. 3e) variability. Therefore, we recommend that laboratories
Schoenemann et al., 2013) because this approach results in excellent develop additional internal reference waters to expand isotopic ranges.
inter-lab ∆′ 17O reproducibility and mathematically is the same approach New reference waters can be developed from evaporated snow or cre­
that most labs already use to correct δ18O data (Berman et al., 2013; ative collections of combustion water (e.g., condensate from a home
Kaiser, 2009; Schoenemann et al., 2013). furnace) that have very low (< –350 per meg) ∆′ 17O values that inherit
oxygen from atmospheric O2 (∆′ 17O ~ = –400 to –500 per meg) (Barkan
7.2.2. Secondary reference standards and Luz, 2011; Wostbrock et al., 2020; Yeung et al., 2012; Young et al.,
We recommend regular analysis of commercially available second­ 2014).
ary water standards (e.g., GISP and USGS reference waters) to ensure
proper calibration to the VSMOW-SLAP scale, evaluate the accuracy and 7.2.3. Analytical sanity checks
precision of isotopic measurements, and monitor analytical drift. The We recommend laboratories perform evaporation experiments and
δ18O and δ2H values of GISP and USGS reference water are readily develop mixing curves to monitor analytical performance. Laboratory-
available (Araguas-Araguas and Rozanski, 1995; Brand et al., 2014; controlled evaporation and mixing experiments can result in substan­
Gonfiantini, 1984); presently, values of δ17O and ∆′ 17O are only reported tial isotopic variation (far above analytical precision) and are well-
by individual laboratories. Isotopic compositions (δ18O, δ17O, δ2H, and predicted by common isotope models (for example, Eqs. (10) and
∆′ 17O) of international and secondary reference waters are summarized (11)). These quality checks are relatively simple and should be included
in Table 9. Additional information about the USGS reference waters in as part of regular analytical maintenance and upkeep. Code to calculate
Table 9 are available in USGS reports on the isotopic composition of isotopic variation during mixing and evaporation experiments is pro­
reference materials (Report of Stable Isotopic Composition Reference vided in Supplements 5 and 6, respectively.
Material USGS45, 2014; Report of Stable Isotopic Composition Refer­
ence Material USGS46, 2014; Report of Stable Isotopic Composition 7.2.4. Analytical ∆′ 17O precision
Reference Material USGS47, 2014; Report of Stable Isotopic Composi­ Typical reported precision for ∆′ 17O (~ 10 per meg) is orders of
tion Reference Material USGS48, 2014; Report of Stable Isotopic magnitude better than analytical errors in δ17O and δ18O (~ 0.1 to 1‰).
Composition Reference Material USGS49, 2015; Report of Stable Iso­ One major source of analytical error in both IRMS and laser absorption
topic Composition Reference Material USGS50, 2015). spectrometers is physical fractionation of water vapor during sample
In our experience, ~ 5–10% of samples within an analytical session handling (i.e., injection, sample conversion, and transport of the vapor
should be secondary reference waters. This proportion is necessary for a to optical cavities or through O2 prep lines). However, this fractionation
meaningful assessment of the accuracy and precision of data, although is presumably mass-dependent and δ18O and δ17O errors are very highly
more USGS and/or GISP analyses may be necessary when unknown correlated along a line with a slope close to 0.528, resulting in very

17
P.G. Aron et al. Chemical Geology 565 (2021) 120026

precise ∆′ 17O measurements (Barkan and Luz, 2007; Landais et al., 2006; (Barkan and Luz, 2007; Luz and Barkan, 2010), we recommend that
Schauer et al., 2016; Schoenemann et al., 2013; Steig et al., 2014). In ∆′ 17O be calculated with a value of 0.528 for λref. This value maintains
other words, if δ18O and δ17O vary during sample handling, they do so consistency with previous work (Table 2), clearly distinguishes equi­
along a line that is parallel to λref (0.528) and ∆′ 17O errors are largely librium (θeq = 0.529) and kinetic (θdiff = 0.518) fractionation effects
independent of the precision for δ18O and δ17O. (Fig. 14), and removes most ∆′ 17O effects from Rayleigh distillation.
Achieving sufficiently high (~10 per meg or better) ∆′ 17O precision Finally, we recommend that triple oxygen water isotope studies
can be an analytical challenge, but is necessary in hydrologic triple include both ∆′ 17O and d-excess data when possible. Combining ∆′ 17O
oxygen isotope studies because the natural ∆′ 17O variability of most and d-excess can reveal information about hydrologic cycling (Section
meteoric waters (Fig. 3b) is not much greater than analytical ∆′ 17O 6) and the parameters (e.g., α, δ, and λ values) that drive isotope frac­
precision. On IRMS systems, sufficient ∆′ 17O precision is achieved by tionation. Adding δ2H measurements is straightforward for studies that
careful sample preparation and well-tuned isotopic analysis as described use laser absorption spectrometers and is a worthwhile additional
in Section 8.2. With laser absorption spectrometers, high ∆′ 17O precision measurement for those that use IRMS techniques.
can be achieved with a long (up to 20 minutes) integration time on each
injection (Schauer et al., 2016; Steig et al., 2014) or many (> 20) repeat 7.2.6. Memory effects
analyses of the same sample (e.g., Berman et al., 2013). Regardless of analytical method (IRMS or laser absorption spec­
Currently, there is no universal way to calculate and report the troscopy), we recommend analyzing samples in order of increasing or
analytical precision and accuracy of triple oxygen isotope measure­ decreasing δ18O and using preparatory injections to minimize memory
ments. For example, some studies use the pooled standard deviation of effects. As needed, USGS reference waters can help bridge large δ18O
secondary reference waters to report analytical precision (e.g., Landais gaps between sequential samples. In Picarro and LGR analyzers, the
et al., 2010; Li et al., 2015) while others use the root mean square error isotopic compositions of water from preparatory injections are analyzed
of replicate analyses to report analytical precision and accuracy (e.g., but the data are typically ignored (e.g., Bailey et al., 2013; Berman et al.,
Schauer et al., 2016; Bershaw et al., 2020). The pooled standard devi­ 2013; Schauer et al., 2016; Steig et al., 2014; Tian et al., 2018). In IRMS
ation (σp) is: systems, memory effects are concentrated in the CoF3 reactor (Barkan
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
∑k ̅ and Luz, 2005) and can be cleared with preparatory injections. Product
i=1 (ni − 1)si
2
gases (O2 and HF) from these injections should be pumped away without
σp = ∑ k (17)
i=1 (ni − 1) purification or analysis.
In our reactors at the University of Michigan, we use one preparatory
where si is the standard deviation and ni is the number of replicate injection when sequential δ18O values are within 5‰, and two prepa­
measurements of the i-th sample and k is the total number of samples. ratory injections when sequential δ18O values differ by more than 5‰. In
The root mean square error (RMSE) is: our experience, typically no more than two preparatory injections are
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ necessary to clear IRMS memory effects, but we encourage each lab to
∑k ( )2
i=1 yi − ̂ yi independently determine best practices to minimize memory effects
RMSE = (18)
k− 1 from individual reactors.

where yi is the observed isotopic composition, ̂ y i is the accepted isotopic 7.2.7. Sample selection
composion, and k is the total number of samples. Preferences among We recommend using existing δ18O and d-excess data to select
these (and other) statistical reporting techniques vary among labora­ samples for triple oxygen isotope analysis. For example, at the Univer­
tories and researchers, so we highlight the most important principles in sity of Michigan, we use a Picarro water isotope analyzer to relatively
∆′ 17O error reporting and leave the particular approach or statistical test quickly measure δ18O and δ2H values, and then systematically select a
to the discretion of individual authors. subset of samples for the much more laborious and time intensive IRMS
At a minimum, triple oxygen isotope papers should report analytical δ18O and δ17O analyses. Initial δ18O data should also be used to deter­
precision and accuracy, explain any drift and/or memory corrections mine the requisite number of preparatory injections and to arrange
that were used to correct δ18O or δ17O data, state which standards were analytical order to minimize memory effects on ∆′ 17O.
used to normalize the data and the number of replicate analyses of both Sample selection should also consider the expected range of ∆′ 17O
standards and unknowns, and clearly articulate the value of λref used to variation and be sure to select samples that are likely to result in sta­
calculate ∆′ 17O. For example, we report this analytical information in tistically significant ∆′ 17O variation (i.e., greater than ~ 10 per meg).
Section 8.2 alongside our analytical methods, state the value of λref in For example, assuming non-steady-state evaporation, the maximum
Section 4.1 as part of our description of the new surface water dataset, possible slope between d-excess and ∆′ 17O is ~ 2 per meg ‰− 1 (e.g.,
and include the number of replicate analyses in the Supplements 1–3. Barkan and Luz, 2007; Li et al., 2015). Therefore, statistically significant
Finally, we note the important distinction between analytical error ∆′ 17O variation is most probable among samples with more than 5‰
(reported as the pooled standard deviation, root mean square error, etc.) variation in d-excess. Datasets with only a few per mil variability in δ18O
and the uncertainty of individual samples (typically reported as one and/or d-excess typically result in ∆′ 17O variation within analytical
standard deviation for replicate analyses of unknowns). It is important precision. We encourage researchers to explore the expected ∆′ 17O
to differentiate these measures of uncertainty because the analytical variability on a case-by-case basis with code in Supplement 4.
error represents the performance of the analytical system while the
standard deviation of analyses represents error on each unknown. 8. Analytical methods for the new surface water data

7.2.5. Isotopic reporting recommendations As part of this review, we report a new, near-global dataset of surface
First and foremost, it is imperative that triple oxygen isotope data are water triple oxygen isotope data (Figs. 2–4 and 6). A brief summary of
reported to three decimal places to calculate ∆′ 17O and facilitate data our results is in Section 4. Here, we explain our sample collection and
comparisons. For every sample, data can be reported as δ18O and δ17O analytical methods to give a sense of the IRMS workflow to make ac­
from individual analyses, both to three decimal places, or as average curate and precise ∆′ 17O measurements.
δ18O and average ∆′ 17O from multiple analyses, both to three decimal
places (Schoenemann et al., 2013). Supplements 1–3 are included as 8.1. Sample collection
templates to report unknown, standard, and reference water data.
Following precedent set by early triple oxygen isotope studies We organized a crowdsource effort to collect over 1,600 water

18
P.G. Aron et al. Chemical Geology 565 (2021) 120026

samples from around the world for isotope analysis. Water was collected cups are 2 × 108, 3 × 1011, and 1 × 1011 Ω, respectively. Analysis in the
in 2 dram glass vials (Ace Glass 8779-20) or 20 ml HDPE plastic vials mass spectrometer takes approximately two hours.
(Wheaton 986716). Samples collected in plastic vials were transferred Triple oxygen isotope data are normalized to the VSMOW-SLAP scale
into glass vials within a few months of collection so we do not expect any following the approach described by Schoenemann et al. (2013). We
fractionation with the sample containers (Spangenberg, 2012). Vials analyze VSMOW2 and SLAP2 in at least triplicate at the beginning, end,
were capped with PolyCone caps to prevent leaks or evaporation, and middle of every reactor, and routinely analyze six USGS reference
filtered (0.45 μm, VWR 28145-493), and stored in a dark environment waters (USGS45, 46, 47, 48, 49, 50) to determine long-term, external
before isotopic analysis. precision of our system, to monitor isotopic drift, and to ensure
analytical accuracy. The root mean square error of replicate triple ox­
ygen isotope analyses of USGS reference waters in our lab is 0.3‰ for
8.2. Isotopic analysis
δ17O, 0.9‰ for δ18O, and 10 per meg for ∆′ 17O.
We used a Picarro L2130-i cavity ringdown spectrometer with a
9. Applications and directions of future work
high-precision vaporizer (A0211) and attached autosampler to measure
the δ18O and δ2H values of every freshwater sample collected (over
9.1. Modern applications
1,500 samples). The L2130-i does not measure δ17O. Each sample was
analyzed nine times; we use the average of the last four analyses. We
Studies of ∆′ 17O variation in the hydrosphere have two primary ap­
used the Picarro ChemCorrect software to monitor samples for organic
plications: as a complement to d-excess and as an analog to the geologic
contamination and normalized measured δ18O and δ2H to the VSMOW-
record. In modern hydrologic studies, ∆′ 17O and d-excess can provide
SLAP scale with USGS reference waters (USGS45, 46, 49, and 50) and
complementary information about moisture transport and moisture
four in-house liquid standards. Isotopic drift and precision were moni­
source conditions because ∆′ 17O is less sensitive to temperature than d-
tored using the Picarro L2130-i Drift and Precision Test worksheet,
excess (Fig. 14). For example, a combination of ∆′ 17O and d-excess can
which is available for download from the Picarro community support
decouple equilibrium and kinetic fractionation effects and reconstruct
forum (https://www.picarro.com/support/community). Precision of
both temperature and relative humidity at a moisture source (e.g.,
repeat analyses of deionized water was better than 0.1‰ and 0.3‰ for
Landais et al., 2012a; Uechi and Uemura, 2019).
δ18O and δ2H, respectively.
It is also important to continue to study modern meteoric water ∆′ 17O
Using the Picarro data, we selected 104 samples (rivers and lakes)
variability in order to improve and expand its use. Geologic and pale­
from the crowdsourced dataset for triple oxygen isotope analysis.
oclimate applications of ∆′ 17O are very appealing, but it is important to
Samples were selected from 17 Köppen climate classes across 6 conti­
first understand hydrologic ∆′ 17O variability, especially of unevaporated
nents and span 30‰ in δ18O and 50‰ in d-excess.
waters. Future hydrologic triple oxygen isotope studies should focus on
Triple oxygen isotopes were analyzed with a dual inlet Nu Perspec­
rivers and/or amount-weighted monthly precipitation to evaluate
tive isotope ratio mass spectrometer. We convert liquid water to O2 gas
δ′ 18O–δ′ 17O regression lines and refine our understanding of spatio­
with cobalt(III) fluoride and a custom-built fluorination line based on
temporal ∆′ 17O variability. Tap waters were a useful starting point to
the method outlined by Baker et al. (2002) and refined by Luz and
understand ∆′ 17O variation (Li et al., 2015; Tian et al., 2019), but these
Barkan (2010). Our analytical methods have been described previously
data can be complicated by non-local isotopic signals so future work
(Li et al., 2017; Li et al., 2015; Passey et al., 2014), although these
should prioritize natural meteoric waters (e.g., precipitation, surface
measurements were made with a different mass spectrometer (Thermo
and subsurface water, or snow and ice).
253) and fluorination line previously at Johns Hopkins University. Our
In addition to ∆′ 17O studies from long-term precipitation collections,
methods have changed only slightly since the laboratory was relocated
there is a particular need for studies that focus on understanding ∆′ 17O
to the University of Michigan.
variation in convective precipitation. This work is critical to explain
Briefly, we inject ~ 2 μL of water through a septum port into a
∆′ 17O variation of mid- and low-latitude precipitation (e.g., Landais
360–370◦ C CoF3 nickel reactor to convert liquid water to O2 gas and
et al., 2010; Li et al., 2015). For example, ∆′ 17O values of convective
gaseous hydrofluoric acid (HF) (Eq. (16)). Helium gas carries O2 gas
precipitation may decrease due to mixing and/or sub-cloud evaporation
through a nickel trap immersed in liquid nitrogen (− 196◦ C) to remove
or may increase if evaporated vapor re-condenses. Currently, the bal­
HF. O2 gas is further purified by passing through a custom-built stainless
ance of these effects is not clear in observational (Landais et al., 2010) or
steel column (~ 1 m, 1/8′′ OD) that is packed with a 5 Å molecular sieve
modeling (Risi et al., 2010, 2013) studies because there has been rela­
(Strem Chemicals, CAS#69912-79-4) and immersed in a methanol/dry
tively little work in this area and the frequency of sample collection
ice slush (− 80◦ C). After purification, the O2 gas collects in a − 196◦ C
(sub-event, event, daily, monthly, etc.) is inconsistent.
trap that is packed with a 5 Å molecular sieve. This process takes ~ 15
Future work is also needed in lake systems to constrain the param­
minutes. After the O2 is collected, helium gas is pumped away (14 mi­
eters in isotopic evaporation models (Eqs. (10) and (11)). Observations
nutes), liquid nitrogen is replaced by a − 80◦ C methanol/dry ice slush,
of water vapor ∆′ 17O will be especially important for this because iso­
and the O2 is transferred to a − 180◦ C cold finger (12 minutes) that is
topic models of evaporation are very sensitive to the isotopic composi­
part of the dual inlet system of the Nu mass spectrometer. The cold finger
tion of water vapor (e.g., Gázquez et al., 2018; Gonfiantini et al., 2018;
has a few pellets of 5 Å molecular sieve to ensure the O2 gas remains in
Passey and Ji, 2019), but very little is known about vapor ∆′ 17O. Lake
the cold finger. Finally, the cold finger is heated (9 minutes) to 90◦ C to
studies in humid and seasonally dry regions will also be helpful because
release O2 from the molecular sieve, and the sample is introduced to the
to date most ∆′ 17O lake work has focused on hyperarid climates where
mass spectrometer. In total, sample preparation takes just over an hour.
lakes are very evaporated (e.g., western US (Passey and Ji, 2019), the
The O2 gas is analyzed in dual inlet mode for m/z 32, 33, and 34. To
Atacama Desert, and Sistant Basin in eastern Iran (Surma et al., 2018,
minimize analytical error, each analysis consists of 40 cycles during
Surma et al., 2015, respectively)). Finally, we encourage data-model
which the ratio of sample to reference gas (99.999% compressed oxy­
comparisons now that many state of the art isotope-enabled general
gen, with approximate values of δ17OVSMOW = 10.3‰, δ18OVSMOW =
circulation models include δ17O (e.g., Brady et al., 2019). This type of
20.3‰) is determined. Each cycle consists of 50 seconds of integration
work can fill in missing gaps from the observational record and improve
time on the sample gas or reference gas and 20 seconds of idle time
our understanding of kinetic fractionation and ∆′ 17O variability (Risi
between integrations. Resistances on the m/z 32, 33, and 34 Faraday

19
P.G. Aron et al. Chemical Geology 565 (2021) 120026

et al., 2013; Schoenemann and Steig, 2016; Wong et al., 2017). water types remain understudied, and future triple oxygen isotope
Future modern triple oxygen isotope studies should also expand to measurements should focus on surface waters and amount-weighted
include water types that have not yet been studied. These include, but monthly precipitation. This future work will further evaluate meteoric
are not limited to, 1) water vapor, which is an important component of water δ′ 18O–δ′ 17O relationships and expand the utility of ∆′ 17O in hy­
isotopic models of evaporation and may affect precipitation ∆′ 17O in drologic and paleoclimate applications.
regions with convective precipitation or a high degree of moisture
recycling; 2) soil water, which frequently undergoes extensive frac­ Data availability
tionation in the upper soil layers (Barnes and Allison, 1984); 3)
groundwater, which can integrate information about seasonal recharge, All new isotope data associated with this review are included in
local and regional water tables, and paleoclimate conditions (Jasechko, Supplements 1–3 and are available from the University of Utah Water
2019); and 4) seawater. Generally it is assumed that the isotopic Isotope Database (https://wateriso.utah.edu/waterisotopes/). R scripts
composition of seawater is invariant and similar to that of VSMOW (Luz (Supplements 4, 5, 6, and 7) and the example raw data file (Supplement
and Barkan, 2010; Zakharov et al., 2019), but this idea is largely un­ 8) that accompanies Supplement 7 can be downloaded from https://gith
tested for triple oxygen isotopes. ub.com/phoebearon/17O.

9.2. Paleoclimate and geologic applications

Declaration of Competing Interest
The most appealing application of triple oxygen isotopes in the
geologic record is the ability to track aridity and constrain the isotopic The authors declare that they have no known competing financial
effects of evaporation. This has long been a challenge in isotope interests or personal relationships that could have appeared to influence
geochemistry because very few minerals contain both oxygen and the work reported in this paper.
hydrogen, and there has not been a d-excess equivalent in geologic
materials. Now, the addition of ∆′ 17O to geologic δ18O records can help Acknowledgements
identify effects of evaporation and aridity (Alexandre et al., 2019;
Gázquez et al., 2018; Passey et al., 2014; Surma et al., 2018), add con­ This study was supported by the University of Michigan Rackham
straints on diagenesis and formation conditions of sedimentary records Predoctoral Fellowship and Department of Earth and Environmental
(Levin et al., 2014), and reconstruct the isotopic composition of ancient Sciences Turner Award (Aron), NSF Tectonics Program Award 1550101
meteoric and ocean waters (e.g., Gehler et al., 2011; Herwartz et al., (Poulsen), NSF-EAR Postdoctoral Fellowship Award 1725621 (Beverly),
2015; Liljestrand et al., 2020; Passey and Ji, 2019). For example, ∆′ 17O and start-up funds from the University of Michigan Department of Earth
and ‘clumped isotopes’ (∆47) in lake carbonates can be used to recon­ and Environmental Sciences (Levin, Passey). We thank Ryan Horwitz for
struct the isotopic composition of unevaporated paleo-water (Passey and assistant with triple oxygen isotope analyses, Sarah Katz for her keen
Ji, 2019). Similarly, ∆′ 17O from gypsum hydration water offers new editorial eye, and the many individuals who participated in our
insights into changes in relative humidity across glacial-interglacial crowdsourced sample collection. We also thank Editor-in-Chief Michael
cycles that are impossible to discern with records of δ18O alone E. Böttcher, Eric Steig, and two anonymous reviewers for providing
(Gázquez et al., 2018). helpful and constructive reviews that improved this manuscript.
To date, ∆′ 17O has been measured in carbonates, sulfates, phos­
phates, nitrates, silicates, and oxides to answer an array of paleoclimate Appendix A. Supplementary data
questions (Bao et al., 2016). Additional work is still needed to calibrate
solid ∆′ 17O standards (Barkan et al., 2019; Wostbrock et al., 2020), Supplementary data to this article can be found online at https://doi.
define fractionation factors (Bergel et al., 2020; Cao and Liu, 2011; Guo org/10.1016/j.chemgeo.2020.120026.
and Zhou, 2019; Sharp et al., 2016; Voarintsoa et al., 2020), and refine
analytical methods (Affek and Barkan, 2018; Fosu et al., 2020; Sakai References
et al., 2017), but the potential of triple oxygen isotopes to separate
equilibrium and kinetic fractionation effects in paleoclimate applica­ Aemisegger, F., Pfahl, S., Sodemann, H., Lehner, I., Seneviratne, S.I., Wernli, H., 2014.
Deuterium excess as a proxy for continental moisture recycling and plant
tions is immense. Additional work to understand the modern variation
transpiration. Atmos. Chem. Phys. 14, 4029–4054. https://doi.org/10.5194/acp-14-
in ∆′ 17O of both evaporated and unevaporated waters will help expand 4029-2014.
triple oxygen isotopes as tool for paleoclimate reconstructions. Affek, H.P., Barkan, E., 2018. A new method for high-precision measurements of 17O/16O
ratios in H2O. Rapid Commun. Mass Spectrom. 32, 2096–2097. https://doi.org/
10.1002/rcm.8290.
10. Conclusion Affolter, S., Häuselmann, A.D., Fleitmann, D., Häuselmann, P., Leuenberger, M., 2015.
Triple isotope (δD, δ17O, δ18O) study on precipitation, drip water and speleothem
Measurements of small, mass-dependent triple oxygen isotope vari­ fluid inclusions for a Western Central European cave (NW Switzerland). Quat. Sci.
Rev. 127, 73–89. https://doi.org/10.1016/j.quascirev.2015.08.030.
ations have come a long way in less than two decades. Deviations from a Alexandre, A., Webb, E., Landais, A., Piel, C., Devidal, S., Sonzogni, C., Couapel, M.,
δ′ 18O–δ′ 17O mass-dependent reference line that were once considered Mazur, J.-C., Pierre, M., Prié, F., Vallet-Coulomb, C., Roy, J., 2019. Effects of grass
unimportant and too difficult to measure are now well characterized and leaf anatomy, development and light/dark alternation on the triple oxygen isotope
signature of leaf water and phytoliths: insights for a new proxy of continental
can separate equilibrium and kinetic fractionation effects. Building upon atmospheric humidity. Biogeosciences 16, 4613–4625. https://doi.org/10.5194/bg-
previous work, this review presents a compilation of new and published 2019-73.
meteoric water triple oxygen isotope data. These data do not fit a single Angert, A., Cappa, C.D., DePaolo, D.J., 2004. Kinetic 17O effects in the hydrologic cycle:
Indirect evidence and implications. Geochim. Cosmochim. Acta 68, 3487–3495.
δ′ 18O–δ′ 17O global meteoric water line, which means that triple oxygen https://doi.org/10.1016/j.gca.2004.02.010.
isotopes may be sensitive to hydrologic processes that do not dominate Araguas-Araguas, L., Rozanski, K., 1995. Interlaboratory comparison for deuterium and
equally across the globe. In hydrologic applications, ∆′ 17O complements oxygen-18 analysis of precipitation samples. Vienna.
Bailey, A., Toohey, D., Noone, D., 2013. Characterizing moisture exchange between the
d-excess and adds information about moisture transport, rainout, and
Hawaiian convective boundary layer and free troposphere using stable isotopes in
evaporation that is impossible to quantify with δ18O alone. In paleo­ water. J. Geophys. Res. Atmos. 118, 8208–8221. https://doi.org/10.1002/
climate applications, ∆′ 17O can constrain evaporation and help recon­ jgrd.50639.
struct relative humidity or the isotopic composition of paleo-waters. Baker, L., Franchi, I.A., Maynard, J., Wright, I.P., Pillinger, C.T., 2002. A technique for
the determination of 18O/16O and 17O/16O Isotopic ratios in water from small liquid
Uncertainties remain about the effects of convective precipitation on and solid samples. Anal. Chem. 74, 1665–1673. https://doi.org/10.1021/
∆′ 17O and the spatiotemporal variability of ∆′ 17O. Additionally, many ac010509s.

20
P.G. Aron et al. Chemical Geology 565 (2021) 120026

Bao, H., Fairchild, I.J., Wynn, P.M., Spotl, C., 2009. Stretching the Envelope of Past Dütsch, M., Pfahl, S., Sodemann, H., 2017. The impact of nonequilibrium and
Surface Environments: Neoproterozoic Glacial Lakes from Svalbard. Science 323, equilibrium fractionation on two different deuterium excess definitions. J. Geophys.
119–122. Res. Atmos. 122, 12732–12746. https://doi.org/10.1002/2017JD027085.
Bao, H., Cao, X., Hayles, J.A., 2016. Triple oxygen isotopes: fundamental relationships Farquhar, J., Thiemens, M.H., 2000. Oxygen cycle of the Martian atmosphere-regolith
and applications. Annu. Rev. Earth Planet. Sci. 44, 463–492. https://doi.org/ system: of secondary phases in Nakhla and Lafayette. J. Geophys. Res. 105, 991–997.
10.1146/annurev-earth-060115-012340. https://doi.org/10.1029/1999je001194.
Barkan, E., Luz, B., 2005. High precision measurements of 17O/16O and 18O/16O ratios in Fosu, B.R., Subba, R., Peethambaran, R., Bhattacharya, S.K., Ghosh, P., 2020. Technical
H2O. Rapid Commun. Mass Spectrom. 19, 3737–3742. https://doi.org/10.1002/ note: developments and applications in triple oxygen isotope analysis of carbonates.
rcm.2250. ACS Earth Sp. Chem. 4, 702–710. https://doi.org/10.1021/
Barkan, E., Luz, B., 2007. Diffusivity fractionations of H16 17 16 18
2 O/H2 O and H2 O/H2 O in air acsearthspacechem.9b00330.
and their implications for isotope hydrology. Rapid Commun. Mass Spectrom. 21, Franz, P., Röckmann, T., Franz, P., High-precision, T.R., 2005. High-precision isotope
2999–3005. https://doi.org/10.1002/rcm.3180. measurements of H16 17 18 17
2 O, H2 O, H2 O, and the ∆ O-anomaly of water vapor in the
Barkan, E., Luz, B., 2011. The relationships among the three stable isotopes of oxygen in southern lowermost stratosphere. Atmos. Chem. Phys. 5, 5373–5403.
air, seawater and marine photosynthesis. Rapid Commun. Mass Spectrom. 25, Galewsky, J., Steen-Larsen, H.C., Field, R.D., Worden, J., Risi, C., Schneider, M., 2016.
2367–2369. https://doi.org/10.1002/rcm.5125. Stable isotopes in atmospheric water vapor and applications to the hydrologic cycle.
Barkan, E., Affek, H., Luz, B., Bergel, S.J., Voarintsoa, N.R.G., Musan, I., 2019. Rev. Geophys. 54, 809–865. https://doi.org/10.1002/2015RG000512.
Calibration of δ17O and 17Oexcess values of three international standards: IAEA-603, Gat, J., 1996. Oxygen and hydrogen isotopes in the hydrologic cycle. Annu. Rev. Earth
NBS19 and NBS18. Rapid Commun. Mass Spectrom. 33, 737–740. https://doi.org/ Planet. Sci. 24, 225–262. https://doi.org/10.1146/annurev.earth.24.1.225.
10.1002/rcm.8391. Gázquez, F., Calafora, J.M., Evans, N.P., Hodell, D.A., 2017. Using stable isotopes (δ17O,
Barnes, C.J., Allison, G., 1984. The distribution of deuterium and 18O in dry soils: 3. δ18O and δD) of gypsum hydration water to ascertain the role of water condensation
Theory for non-isothermal water movement. J. Hydrol. 74, 119–135. https://doi. in the formation of subaerial gympsum speleothems. Chem. Geol. 452, 34–46.
org/10.1016/0022-1694(84)90144-6. https://doi.org/10.1016/j.chemgeo.2017.01.021.
Bergel, S.J., Barkan, E., Stein, M., Affek, H.P., 2020. Carbonate 17Oexcess as a paleo- Gázquez, F., Morellón, M., Bauska, T., Herwartz, D., Surma, J., Moreno, A.,
hydrology proxy: Triple oxygen isotope fractionation between H2O and biogenic Staubwasser, M., Valero-garcés, B., Delgado-huertas, A., Hodell, D.A., 2018. Triple
aragonite, derived from freshwater mollusks. Geochim. Cosmochim. Acta. 275, oxygen and hydrogen isotopes of gypsum hydration water for quantitative paleo-
36–47. https://doi.org/10.1016/j.gca.2020.02.005. humidity reconstruction. Earth Planet. Sci. Lett. 481, 177–188. https://doi.org/
Berman, E.S.F., Levin, N.E., Landais, A., Li, S., Owano, T., 2013. Measurement of δ18O, 10.1016/j.epsl.2017.10.020.
δ17O, and 17O-excess in Water by Off-Axis Integrated Cavity Output Spectroscopy Gehler, A., Tütken, T., Pack, A., 2011. Triple oxygen isotope analysis of bioapatite as
and Isotope Ratio Mass Spectrometry. Anal. Chem. https://doi.org/10.1021/ tracer for diagenetic alteration of bones and teeth. Palaeogeogr. Palaeoclimatol.
ac402366t. Palaeoecol. 310, 84–91. https://doi.org/10.1016/j.palaeo.2011.04.014.
Bershaw, J., Hansen, D.D., Schauer, A.J., 2020. Deuterium excess and 17O-excess Gonfiantini, R., 1978. Standards for stable isotope measurements in natural compounds.
variability in meteoric water across the Pacific Northwest, USA. Tellus B Chem. Phys. Nature 271, 534–536. https://doi.org/10.1038/271534a0.
Meteeorol. 72 (1), 1–17. https://doi.org/10.1080/16000889.2020.1773722. Gonfiantini, R., 1984. Advisory group meeting on stable isotope reference samples for
Bhattacharya, S.K., Savarino, J., Thiemens, M.H., 2000. A new class of oxygen isotopic geochemical and hydrochemical investigations. Vienna.
fractionation in photodissociation of carbon dioxide: Potential implications for Gonfiantini, R., Wassenaar, L.I., Araguas-Araguas, L., Aggarwal, P.K., 2018. A unified
atmospheres of Mars and Earth. Geophys. Res. Lett. 27, 1459–1462. https://doi.org/ Craig-Gordon isotope model of stable hydrogen and oxygen isotope fractionation
10.1029/1999GL010793. during fresh or saltwater evaporation. Geochim. Cosmochim. Acta 235, 224–236.
Blunier, T., Barnett, B., Bender, M.L., Hendricks, M.B., 2002. Biological oxygen https://doi.org/10.1016/j.gca.2018.05.020.
productivity during the last 60,000 years from triple oxygen isotope measurements. Guo, W., Zhou, C., 2019. Triple oxygen isotope fractionation in the DIC-H2O-CO2 system:
Global Biogeochem. Cycles 16, 1–15. https://doi.org/10.1029/2001gb001460. A numerical framework and its implications. Geochim. Cosmochim. Acta 246,
Blunier, T., Bender, M.L., Barnett, B., Von Fischer, J.C., 2012. Planetary fertility during 541–564. https://doi.org/10.1016/j.gca.2018.11.018.
the past 400 ka based on the triple isotope composition of O2 in trapped gases from Herwartz, D., Pack, A., Krylov, D., Xiao, Y., Muehlenbachs, K., Sengupta, S., Di Rocco, T.,
the Vostok ice core. Clim. Past 8, 1509–1526. https://doi.org/10.5194/cp-8-1509- 2015. Revealing the climate of snowball Earth from δ17O systematics of
2012. hydrothermal rocks. Proc. Natl. Acad. Sci. U. S. A. 112, 5337–5341. https://doi.org/
Bowen, G.J., 2010. Isoscapes: spatial pattern in isotopic biogeochemistry. Annu. Rev. 10.1073/pnas.1422887112.
Earth Planet. Sci. 38, 161–187. https://doi.org/10.1146/annurev-earth-040809- Horita, J., Wesolowski, D.J., 1994. Liquid-vapor fractionation of oxygen and hydrogen
152429. isotopes of water from the freezing to the critical temperature. Geochim.
Bowen, G.J., Cai, Z., Fiorella, R.P., Putman, A.L., 2019. Isotopes in the water cycle: Cosmochim. Acta 58, 3425–3437.
regional- to global-scale patterns and applications. Annu. Rev. 47, 453–479. https:// Hulston, J.R., Thode, H.G., 1965. Variations in the S33, S34, and S36 Contents of
doi.org/10.1146/annurev-earth-053018- 060220. Meteorites and Their Relation to Chemical and Nuclear Effects. J. Geophys. Res. 70,
Brady, E., Stevenson, S., Bailey, D., Liu, Z., Noone, D., Nusbaumer, J., 2019. The 3475–3484.
connected isotopic water cycle in the community earth system model version 1. Jasechko, S., 2019. Global Isotope Hydrogeology—Review. Rev. Geophys. 57, 835–965.
J. Adv. Model. Earth Syst. 11, 2547–2566. https://doi.org/10.1029/ https://doi.org/10.1029/2018RG000627.
2019MS001663. Joussaume, S., Jouzel, J., Sadourny, R., 1984. A general circulation model of water
Brand, W.A., Coplen, T.B., 2001. An interlaboratory study to test instrument performance isotope cycles in the atmosphere. Nature 311, 24–29. https://doi.org/10.1038/
of hydrogen dual-inlet isotope-ratio mass spectrometers. Fresenius J. Anal. Chem. 311680a0.
370, 358–362. https://doi.org/10.1007/s002160100814. Jouzel, J., Merlivat, L., 1984. Deuterium and Oxygen 18 in Precipitation: Modeling of the
Brand, W.A., Coplen, T.B., Vogl, J., Rosner, M., Prohaska, T., 2014. Assessment of Isotopic Effects During Snow Formation. J. Geophys. Res. 89, 11749–11757. https://
international reference materials for isotope-ratio analysis (IUPAC technical report). doi.org/10.1029/JD089iD07p11749.
Pure Appl. Chem. 86, 425–467. https://doi.org/10.1515/pac-2013-1023. Kaiser, J., 2009. Reformulated 17O correction of mass spectrometric stable isotope
Brooks, J.R., Barnard, H.R., Coulombe, R., McDonnell, J.J., 2010. Ecohydrologic measurements in carbon dioxide and a critical appraisal of historic ‘absolute’ carbon
separation of water between trees and streams in a Mediterranean climate. Nat. and oxygen isotope ratios. Geochim. Cosmochim. Acta 72, 1312–1334. https://doi.
Geosci. 3, 100–104. https://doi.org/10.1038/NGEO722. org/10.1016/j.gca.2007.12.011.
Cao, X., Liu, Y., 2011. Equilibrium mass-dependent fractionation relationships for triple Kaseke, K.F., Wang, L., Wanke, H., Tian, C., Lanning, M., Jiao, W., 2018. Precipitation
oxygen isotopes. Geochim. Cosmochim. Acta 75, 7435–7445. https://doi.org/ Origins and Key Drivers of Precipitation Isotope (18O, 2H, and 17O) Compositions
10.1016/j.gca.2011.09.048. Over Windhoek. J. Geophys. Res. Atmos. 123, 7311–7330. https://doi.org/10.1029/
Cernusak, L.A., Barbour, M.M., Arndt, S.K., Cheesman, A.W., English, N.B., Feild, T.S., 2018JD028470.
Helliker, B.R., Holloway-Phillips, M.M., Holtum, J.A.M., Kahmen, A., Mcinerney, F. Kendall, C., Coplen, T.B., 2001. Distribution of oxygen-18 and deuteriun in river waters
A., Munksgaard, N.C., Simonin, K.A., Song, X., Stuart-Williams, H., West, J.B., across the United States. Hydrol. Process. 15, 1363–1393. https://doi.org/10.1002/
Farquhar, G.D., 2016. Stable isotopes in leaf water of terrestrial plants. Plant Cell hyp.217.
Environ. 39, 1087–1102. https://doi.org/10.1111/pce.12703. Landais, A., Barkan, E., Yakir, D., Luz, B., 2006. The triple isotopic composition of
Coplen, T.B., 1988. Normalization of oxygen and hydrogen isotope data. Chem. Geol. oxygen in leaf water. Geochim. Cosmochim. Acta 70, 4105–4115. https://doi.org/
Isot. Geosci. Sect. 72, 293–297. 10.1016/j.gca.2006.06.1545.
Craig, H., 1961. Isotopic Variations in Meteoric Waters. Science 133, 1702–1703. Landais, A., Barkan, E., Luz, B., 2008. Record of δ18O and 17O-excess in ice from Vostok
https://doi.org/10.1126/science.133.3465.1702. Antarctica during the last 150,000 years. Geophys. Res. Lett. 35, 1–5. https://doi.
Craig, H., Gordon, L.I., 1965. Deuterium and oxygen-18 variations in the ocean and the org/10.1029/2007GL032096.
marine atmosphere. In: Tongiorgi, E. (Ed.), Proceedings of a Conference on Stable Landais, A., Risi, C., Bony, S., Vimeux, F., Descroix, L., Falourd, S., Bouygues, A., 2010.
Isotopes in Oceanographic Studies and Paleotemperatures. Spoleto, Italy, pp. 9–130. Combined measurements of 17O-excessand d-excess in African monsoon
Criss, R.E., 1999. Principles of Stable Isotope Distrubution. Oxford University Press, New precipitation: Implications for evaluating convective parameterizations. Earth
York. Planet. Sci. Lett. 298, 104–112. https://doi.org/10.1016/j.epsl.2010.07.033.
Criss, R.E., Farquhar, J., 2008. Abundance, Notation, and Fractionation of Light Stable Landais, A., Ekaykin, A., Barkan, E., Winkler, R., Luz, B., 2012a. Seasonal variations of
17
Isotopes. Rev. Mineral. Geochemistry 68, 15–30. https://doi.org/10.2138/ O-excess and d-excess in snow precipitation at Vostok station, East Antarctica.
rmg.2008.68.3. J. Glaciol. 58, 725–733. https://doi.org/10.3189/2012JoG11J237.
Dansgaard, W., 1964. Stable isotopes in precipitation. Tellus 16, 436–468. https://doi. Landais, A., Steen-Larsen, H.C., Guillevic, M., Masson-Delmotte, V., Vinther, B.,
org/10.3402/tellusa.v16i4.8993. Winkler, R., 2012b. Triple isotopic composition of oxygen in surface snow and water

21
P.G. Aron et al. Chemical Geology 565 (2021) 120026

vapor at NEEM (Greenland). Geochim. Cosmochim. Acta 77, 304–316. https://doi. Desert and the formation of the Central Andean rain-shadow. Earth Planet. Sci. Lett.
org/10.1016/j.gca.2011.11.022. 506, 184–194. https://doi.org/10.1016/j.epsl.2018.10.040.
Levin, N.E., Raub, T.D., Dauphas, N., Eiler, J.M., 2014. Triple oxygen isotope variations Report of Stable Isotopic Composition Reference Material USGS45. Reston, VA. https://is
in sedimentary rocks. Geochim. Cosmochim. Acta 139, 173–189. https://doi.org/ otopes.usgs.gov/lab/referencematerials/USGS45.pdf.
10.1016/j.gca.2014.04.034. Report of Stable Isotopic Composition Reference Material USGS46. Reston, VA. https://is
Li, S., Levin, N.E., Chesson, L.A., 2015. Continental scale variation in 17O-excess of otopes.usgs.gov/lab/referencematerials/USGS46.pdf.
meteoric waters in the United States. Geochim. Cosmochim. Acta 164, 110–126. Report of Stable Isotopic Composition Reference Material USGS47. Reston, VA. https://is
https://doi.org/10.1016/j.gca.2015.04.047. otopes.usgs.gov/lab/referencematerials/USGS47.pdf.
Li, S., Levin, N.E., Soderberg, K., Dennis, K.J., Caylor, K.K., 2017. Triple oxygen isotope Report of Stable Isotopic Composition Reference Material USGS48. Reston, VA. https://is
composition of leaf waters in Mpala, central Kenya. Earth Planet. Sci. Lett. 468, otopes.usgs.gov/lab/referencematerials/USGS48.pdf.
38–50. https://doi.org/10.1016/j.epsl.2017.02.015. Report of Stable Isotopic Composition Reference Material USGS49. Reston, VA. https://is
Liljestrand, F.L., Knoll, A.H., Tosca, N.J., Cohen, P.A., Macdonald, F.A., Peng, Y., otopes.usgs.gov/lab/referencematerials/USGS49.pdf.
Johnston, D.T., 2020. The triple oxygen isotope composition of Precambrian chert. Report of Stable Isotopic Composition Reference Material USGS50. Reston, VA. https://is
Earth Planet. Sci. Lett. 537, 1–10. https://doi.org/10.1016/J.EPSL.2020.116167. otopes.usgs.gov/lab/referencematerials/USGS50.pdf.
Lin, Y., Clayton, R.N., Groning, M., 2010. Calibration of δ17O and δ18O of international Risi, C., Landais, A., Bony, S., Jouzel, J., Masson-Delmotte, V., Vimeux, F., 2010.
measurement standards – VSMOW, VSMOW2, SLAP, and SLAP2. Rapid Commun. Understanding the 17O excess glacial-interglacial variations in Vostok precipitation.
Mass Spectrom. 24, 773–776. https://doi.org/10.1002/rcm.4449. J. Geophys. Res. Atmos. 115, 1–15. https://doi.org/10.1029/2008JD011535.
Luz, B., Barkan, E., 2010. Variations of 17O/16O and 18O/16O in meteoric waters. Risi, C., Landais, A., Winkler, R., Vimeux, F., 2013. Can we determine what controls the
Geochim. Cosmochim. Acta 74, 6276–6286. https://doi.org/10.1016/j. spatio-temporal distribution of d-excess and 17O-excess in precipitation using the
gca.2010.08.016. LMDZ general circulation model? Clim. Past 9, 2173–2193. https://doi.org/
Luz, B., Barkan, E., Yam, R., Shemesh, A., 2009. Fractionation of oxygen and hydrogen 10.5194/cp-9-2173-2013.
isotopes in evaporating water. Geochim. Cosmochim. Acta 73, 6697–6703. https:// Roscoe, H.K., Fowler, C.L., Shanklin, J.D., Hill, J.G.T., 2004. Possible long-term changes
doi.org/10.1016/j.gca.2009.08.008. in stratospheric circulation: Evidence from total ozone measurements at the edge of
Majoube, M., 1971. Oxygen-18 and deuterium fractionation between water and steam. the Antarctic vortex in early winter. Q. J. R. Meteorol. Soc. 130, 1123–1135. https://
J. Chem. Phys. 68, 1432–1436. doi.org/10.1256/qj.03.70.
Marrero, T.R., Mason, E.A., 1972. Gaseous diffusion coefficients. J. Phys. Chem. Rowley, D.B., Garzione, C.N., 2007. Stable isotope-based paleoaltimetry. Annu. Rev.
Ref. Data 1, 3–118. https://doi.org/10.1063/1.3253094. Earth Planet. Sci. 35, 463–508. https://doi.org/10.1146/annurev.
Matsuhisa, Y., Goldsmith, J.R., Clayton, R.N., 1978. Mechanisms of hydrothermal earth.35.031306.140155.
crystallization of quartz at 250◦ C and 15 kbar. Geochim. Cosmochim. Acta 42, Rozanski, K., Araguás-Araguás, L., Gonfiantini, R., 1993. Isotopic Patterns in Modern Global
173–182. https://doi.org/10.1016/0016-7037(78)90130-8. Precipitation. In: Swart, P.K., Lohmann, K.C., Mckenzie, J., Savin, S. (Eds.), Climate
McKeegan, K.D., Leshin, L.A., 2001. Stable Isotope Variations in Extraterrestrial Change in Continental Isotopic Records Geophysical Monograph Series, 78. American
Materials. Rev. Mineral. Geochemistry 43, 279–318. https://doi.org/10.2138/ Geophysical Union, Washington, DC, pp. 1–36. https://doi.org/10.1029/GM078p0001.
gsrmg.43.1.279. Rumble, D., Miller, M.F., Franchi, I.A., Greenwood, R.C., 2007. Oxygen three-isotope
McKinney, C.R., McCrea, J.M., Epstein, S., Allen, H.A., Urey, H.C., 1950. Improvements fractionation lines in terrestrial silicate minerals: An inter-laboratory comparison of
in mass spectrometers for the measurement of small differences in isotope hydrothermal quartz and eclogitic garnet. Geochim. Cosmochim. Acta 71,
abundance ratios. Rev. Sci. Instrum. 21, 724–730. https://doi.org/10.1063/ 3592–3600. https://doi.org/10.1016/j.gca.2007.05.011.
1.1745698. Sakai, S., Matsuda, S., Hikida, T., Shimono, A., Mcmanus, J.B., Zahniser, M., Nelson, D.,
Meijer, H.A.J., Li, W.J., 1998. The Use of Electrolysis for Accurate δ17O and δ18O Isotope Dettman, D.L., Yang, D., Ohkouchi, N., 2017. High-Precision Simultaneous 18O/16O,
13
Measurements in Water. Isot. Evironmental Heal. Stud. 34, 349–369. https://doi. C/ 12C, and 17O/16O Analyses for Microgram Quantities of CaCO3 by Tunable
org/10.1080/10256019808234072. Infrared Laser Absorption Spectroscopy. Anal. Chem. 89, 11846–11852. https://doi.
Meijer, H.A.J., Neubert, R.E.M., Visser, G.H., 2000. Cross contamination in dual inlet org/10.1021/acs.analchem.7b03582.
isotope ratio mass spectrometers. Int. J. Mass Spectrom. 198, 45–61. https://doi. Salati, E., Dall’Olio, A., Matsui, E., Gat, J.R., 1979. Recycling of water in the Amazon
org/10.1016/s1387-3806(99)00266-3. Basin: An isotopic study. Water Resour. Res. 15, 1250–1258. https://doi.org/
Merlivat, L., 1978. Molecular diffusivities of H16 16 18
2 O, HD O, and H2 O in gases. J. Chem. 10.1029/WR015i005p01250.
Phys. 69, 2864–2871. https://doi.org/10.1063/1.436884. Schauer, A.J., Schoenemann, S.W., Steig, E.J., 2016. Routine high-precision analysis of
Miller, M.F., 2002. Isotopic fractionation and the quantification of 17O anomalies in the triple water-isotope ratios using cavity ring-down spectroscopy. Rapid Commun.
oxygen three-isotope system: an appraisal and geochemical significance. Geochimica Mass Spectrom. 30, 2059–2069. https://doi.org/10.1002/rcm.7682.
et Cosmochimica Acta 66 (11), 1889–1991. Schoenemann, S.W., Steig, E.J., 2016. Seasonal and spatial variations of 17Oexcess and
Miller, M.F., 2018. Precipitation regime influence on oxygen triple-isotope distributions dexcess in Antarctic precipitation: Insights from an intermediate complexity isotope
in Antarctic precipitation and ice cores. Earth Planet. Sci. Lett. 481, 316–327. model. J. Geophys. Res. Atmos. 121, 11215–11247. https://doi.org/10.1002/
https://doi.org/10.1016/j.epsl.2017.10.035. 2016JD025117.Received.
New, M., Lister, D., Hulme, M., Makin, I., 2002. A high-resolution data set of surface Schoenemann, S.W., Schauer, A.J., Steig, E.J., 2013. Measurement of SLAP2 and GISP
climate over global land areas. Clim. Res. 21, 1–25. δ17O and proposed VSMOW-SLAP normalization for δ17O and 17Oexcess. Rapid
Noone, D., Risi, C., Bailey, A., Berkelhammer, M., Brown, D.P., Buenning, N., Gregory, S., Commun. Mass Spectrom. 27, 582–590. https://doi.org/10.1002/rcm.6486.
Nusbaumer, J., Schneider, D., Sykes, J., Vanderwende, B., Wong, J., Meillier, Y., Schoenemann, S.W., Steig, E.J., Ding, Q., Markle, B.R., Schauer, A.J., 2014. Triple water-
Wolfe, D., 2013. Determining water sources in the boundary layer from tall tower isotopologue record from WAIS Divide, Antarctica: Controls on glacial-interglacial
profiles of water vapor and surface water isotope ratios after a snowstorm in changes in 17Oexcess of precipitation. J. Geophys. Res. Atmos. 119, 8741–8763.
Colorado. Atmos. Chem. Phys. 13, 1607–1623. https://doi.org/10.5194/acp-13- https://doi.org/10.1002/2014JD021770.Received.
1607-2013. Sharp, Z.D., Gibbons, J.A., Maltsev, O., Atudorei, V., Pack, A., Sengupta, S., Shock, E.L.,
Pack, A., Herwartz, D., 2014. The triple oxygen isotope composition of the Earth mantle Knauth, L.P., 2016. A calibration of the triple oxygen isotope fractionation in the
and understanding ∆17O variations in terrestrial rocks and minerals. Earth Planet. SiO2-H2O system and applications to natural samples. Geochim. Cosmochim. Acta
Sci. Lett. 390, 138–145. https://doi.org/10.1016/j.epsl.2014.01.017. 186, 105–119. https://doi.org/10.1016/j.gca.2016.04.047.
Pang, H., Hou, S., Landais, A., Masson-Delmotte, V., Prie, F., Steen-Larsen, H.C., Risi, C., Sharp, Z.D., Wostbrock, J.A.G., Pack, A., 2018. Mass-dependent triple oxygen isotope
Li, Y., Jouzel, J., Wang, Y., He, J., Minster, B., Falourd, S., 2015. Spatial distribution variations in terrestrial materials. Geochem. Perspect. Lett. 7, 27–31. https://doi.
of 17O-excess in surface snow along a traverse from Zhongshan station to Dome A, org/10.7185/geochemlet.1815.
East Antarctica. Earth Planet. Sci. Lett. 414, 126–133. https://doi.org/10.1016/j. Spangenberg, J.E., 2012. Caution on the storage of waters and aqueous solutions in
epsl.2015.01.014. plastic containers for hydrogen and oxygen stable isotope analysis. Rapid Commun.
Pang, H., Hou, S., Landais, A., Delmotte, V.M., Jouzel, J., 2019. Influence of Summer Mass Spectrom. 26, 2627–2636. https://doi.org/10.1002/rcm.6386.
Sublimation on δD, δ18O, and δ17O in Precipitation, East Antarctica, and Steig, E.J., Gkinis, V., Schauer, A.J., Schoenemann, S.W., Samek, K., Hoffnagle, J.,
Implications for Climate Reconstruction From Ice Cores. J. Geophys. Res. Atmos. Dennis, K.J., Tan, S.M., 2014. Calibrated high-precision 17O-excess measurements
124, 7339–7358. https://doi.org/10.1029/2018JD030218. using cavity ring-down spectroscopy with laser-current-tuned cavity resonance.
Passey, B.H., Ji, H., 2019. Triple oxygen isotope signatures of evaporation in lake waters Atmos. Meas. Tech. 7, 2421–2435. https://doi.org/10.5194/amt-7-2421-2014.
and carbonates: A case study from the western United States. Earth Planet. Sci. Lett. Surma, J., Assonov, S., Bolourchi, M.J., Staubwasser, M., 2015. Triple oxygen isotope
518, 1–12. https://doi.org/10.1016/j.epsl.2019.04.026. signatures in evaporated water bodies from the Sistan Oasis, Iran. Geophys. Res. Lett.
Passey, B.H., Hu, H., Ji, H., Montanari, S., Li, S., Henkes, G.A., Levin, N.E., 2014. Triple 42, 8456–8462. https://doi.org/10.1002/2015GL066475.
oxygen isotopes in biogenic and sedimentary carbonates. Geochim. Cosmochim. Surma, J., Assonov, S., Herwartz, D., Voigt, C., Staubwasser, M., 2018. The evolution of
17
Acta 141, 1–25. https://doi.org/10.1016/j.gca.2014.06.006. O-excess in surface water of the arid environment during recharge and
Paul, D., Skrzypek, G., Forizs, I., 2007. Normalization of measured stable isotopic evaporation. Sci. Rep. 8, 1–10. https://doi.org/10.1038/s41598-018-23151-6.
compositions to isotope reference scales – a review. Rapid Commun. Mass Spectrom. Thiemens, M.H., 2006. History and applications of mass-independent isotope effects.
21, 3006–3014. https://doi.org/10.1002/rcm.3185. Annu. Rev. Earth Planet. Sci. 34, 217–262. https://doi.org/10.1146/annurev.
Poulsen, C.J., Ehlers, T.A., Insel, N., 2010. Onset of convective rainfall during gradual earth.34.031405.125026.
late miocene rise of the central andes. Science 328, 490–493. https://doi.org/ Thiemens, M.H., Heidenreich, J.E., 1983. The mass-independent fractionation of oxygen:
10.1126/science.1185078. A novel isotope effect and its possible cosmochemical implications. Science 219,
Rech, J.A., Currie, B.S., Jordan, T.E., Riquelme, R., Lehmann, S.B., Kirk-Lawlor, N.E., 1073–1075. https://doi.org/10.1126/science.219.4588.1073.
Li, S., Gooley, J.T., 2019. Massive middle Miocene gypsic paleosols in the Atacama Thiemens, M.H., Jackson, T.L., Brenninkmeijer, C.A.M., 1995. Observation of a mass
independent oxygen isotopic composition in terrestrial stratospheric CO2, the link to

22
P.G. Aron et al. Chemical Geology 565 (2021) 120026

ozone chemistry, and the possible occurrence in the Martian atmosphere. Geophys. Werner, R.A., Brand, W.A., 2001. Referencing strategies and techniques in stable isotope
Res. Lett. 22, 255–257. https://doi.org/10.1029/94GL02996. ratio analysis. Rapid Commun. Mass Spectrom. 15, 501–519. https://doi.org/
Thiemens, M.H., Chakraborty, S., Dominguez, G., 2012. The physical chemistry of mass- 10.1002/rcm.258.
independent isotope effects and their observation in nature. Annu. Rev. Phys. Chem. Winkler, R., Landais, A., Sodemann, H., Dümbgen, L., Prié, F., Masson-Delmotte, V.,
63, 155–177. https://doi.org/10.1146/annurev-physchem-032511-143657. Stenni, B., Jouzel, J., 2012. Deglaciation records of 17O-excess in East Antarctica:
Thompson, M., 2012. Precision in chemical analysis: A critical survey of uses and abuses. Reliable reconstruction of oceanic normalized relative humidity from coastal sites.
Anal. Methods 4, 1598–1611. https://doi.org/10.1039/c2ay25083g. Clim. Past 8, 1–16. https://doi.org/10.5194/cp-8-1-2012.
Tian, C., Wang, L., 2019. Data Descriptor: Stable isotope variations of daily precipitation Wong, T.E., Nusbaumer, J., Noone, D.C., 2017. Evaluation of modeled land-atmosphere
from 2014–2018 in the central United States. Nature Publishing Group. 6, 1–8. exchanges with a comprehensive water isotope fractionation scheme in version 4 of
https://doi.org/10.1038/sdata.2019.18. the Community Land Model. J. Adv. Model. Earth Syst. 9, 978–1001. https://doi.
Tian, C., Wang, L., Kaseke, K.F., Bird, B.W., 2018. Stable isotope compositions (δ2H, δ18O org/10.1002/2016MS000842.
and δ17O) of rainfall and snowfall in the central United States. Sci. Rep. 8, 1–15. Wostbrock, J.A.G., Cano, E.J., Sharp, Z.D., 2020. An internally consistent triple oxygen
https://doi.org/10.1038/s41598-018-25102-7. isotope calibration of standards for silicates, carbonates and air relative to VSMOW2
Tian, C., Wang, L., Tian, F., Zhao, S., Jiao, W., 2019. Spatial and temporal variations of and SLAP2. Chem. Geol. 533, 119432 https://doi.org/10.1016/j.
tap water 17O-excess in China. Geochim. Cosmochim. Acta 260, 1–14. https://doi. chemgeo.2019.119432.
org/10.1016/j.gca.2019.06.015. Yeung, L.Y., Young, E.D., Schauble, E.A., 2012. Measurements of 18O18O and 17O18O in
Touzeau, A., Landais, A., Stenni, B., Uemura, R., Fukui, K., Fujita, S., Guilbaud, S., the atmosphere and the role of isotope-exchange reactions. J. Geophys. Res. Atmos.
Ekaykin, A., Casado, M., Barkan, E., Luz, B., Magand, O., Teste, G., Le Meur, E., 117, 1–20. https://doi.org/10.1029/2012JD017992.
Baroni, M., Savarino, J., Bourgeois, I., Risi, C., 2016. Acquisition of isotopic Yeung, L.Y., Hayles, J.A., Hu, H., Ash, J.L., Sun, T., 2018. Scale distortion from pressure
composition for surface snow in East Antarctica and the links to climatic parameters. baselines as a source of inaccuracy in triple-isotope measurements. Rapid Commun.
The Cryosphere 10 (2), 837–852. https://doi.org/10.5194/tc-10-837-2016. Mass Spectrom. 32, 1811–1821. https://doi.org/10.1002/rcm.8247.
Uechi, Y., Uemura, R., 2019. Dominant influence of the humidity in the moisture source Young, E.D., Galy, A., 2004. The isotope geochemistry and cosmochemistry of
region on the 17O-excess in precipitation on a subtropical island. Earth Planet. Sci. magnesium. Re 55, 197–230. https://doi.org/10.2138/gsrmg.55.1.197.
Lett. 513, 20–28. https://doi.org/10.1016/j.epsl.2019.02.012. Young, E.D., Galy, A., Nagahara, H., 2002. Kinetic and equilibrium mass-dependent
Uemura, R., Barkan, E., Abe, O., Luz, B., 2010. Triple isotope composition of oxygen in isotope fractionation laws in nature and their geochemical and cosmochemical
atmospheric water vapor. Geophys. Res. Lett. 37, 1–4. https://doi.org/10.1029/ significance. Geochim. Cosmochim. Acta 66, 1095–1104. https://doi.org/10.1016/
2009GL041960. S0016-7037(01)00832-8.
Uemura, R., Masson-Delmotte, V., Jouzel, J., Landais, A., Motoyama, H., Stenni, B., Young, E.D., Yeung, L.Y., Kohl, I.E., 2014. On the δ17O budget of atmospheric O2.
2012. Ranges of moisture-source temperature estimated from Antarctic ice cores Geochim. Cosmochim. Acta 135, 102–125. https://doi.org/10.1016/j.
stable isotope records over glacial-interglacial cycles. Clim. Past 8, 1109–1125. gca.2014.03.026.
https://doi.org/10.5194/cp-8-1109-2012. Zachos, J., 2001. Trends, Rhythms, and Aberrations in Global Climate 65 Ma to Present,
Voarintsoa, N.R.G., Barkan, E., Bergel, S., Vieten, R., Affek, H.P., 2020. Triple oxygen 292, pp. 686–693. https://doi.org/10.1126/science.1059412.
isotope fractionation between CaCO3 and H2O in inorganically precipitated calcite Zakharov, D.O., Bindeman, I.N., Tanaka, R., Friðleifsson, G.Ó., Reed, M.H., Hampton, R.
and aragonite. Chem. Geol. 539, 119500 https://doi.org/10.1016/j. L., 2019. Triple oxygen isotope systematics as a tracer of fluids in the crust: A study
chemgeo.2020.119500. from modern geothermal systems of Iceland. Chem. Geol. 530, 119312 https://doi.
org/10.1016/j.chemgeo.2019.119312.

23

View publication stats